https://scholar4ai.com/ideas 2026-10-26T00:00:00+09:00 Scholar4AI https://scholar4ai.com/ideas/who-gets-ev-chargers 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Electric vehicle charging appears on 89 recent papers across operations research, economics and AI, mostly on where to put chargers to serve demand; only 8 carry an equity, income or rural keyword. Public money now pays for much of the network. Are poorer and rural areas left with charging deserts, and does that slow electric vehicle adoption there?

<h2>What prompted this</h2> <p>Since October 2025, 89 papers carry a charging-station, charging-infrastructure or EV-charging keyword (36 in operations research journals, 32 in economics, 20 in AI): "electric vehicle charging" on 21, "charging infrastructure" on 19, "charging stations" on 8, "charging station location" on 5. Only 8 also carry an equity, income, disadvantaged, rural, justice, access or inequality keyword.</p> <h2>Questions</h2> <ul> <li>Per resident or per car, how unequal is public charger coverage by neighbourhood income and between urban and rural areas?</li> <li>Did subsidy programmes with equity goals (such as those requiring a share of funds in disadvantaged areas) change where chargers went?</li> <li>Does a new public charger raise electric vehicle registrations nearby, and is the effect larger where few people can charge at home?</li> </ul> <h2>Data that could help</h2> <p>Charger locations and installation dates (national charger registries), vehicle registrations by area, census income and housing data, and subsidy award records.</p> <h2>Next step</h2> <p>Relate charger installations to subsequent electric vehicle registrations by census tract, separately for tracts with mostly apartments and mostly houses.</p> https://scholar4ai.com/ideas/what-key-audit-matters-tell-investors 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Key and critical audit matters appear on only 9 recent papers, although auditors of thousands of listed firms now describe the riskiest parts of each audit in their reports. Do these paragraphs say anything firm-specific after a few years, and does anyone react to them?

<h2>What prompted this</h2> <p>Since October 2025, 9 papers carry a key-audit-matter, critical-audit-matter or expanded-audit-report keyword (8 in accounting journals): "key audit matters" on 5, "critical audit matters" on 4, "audit reports" on 2. Expanded audit reports have been required in the UK and EU since 2016–2017 and for large US filers since 2019, so years of reports now exist.</p> <h2>Questions</h2> <ul> <li>Has the wording of audit matters become more boilerplate over time, and can a language model measure how specific each one is?</li> <li>Do specific, unusual audit matters predict later restatements, impairments or going-concern problems?</li> <li>Do investors, analysts or lenders react when a new audit matter appears or an old one disappears?</li> </ul> <h2>Data that could help</h2> <p>The text of audit reports (Audit Analytics, annual reports), restatements and impairments, and market reactions around filing dates.</p> <h2>Next step</h2> <p>Measure the year-on-year similarity of audit matter text for the same firms since 2019 and test whether drops in similarity (new content) predict later restatements.</p> https://scholar4ai.com/ideas/what-digital-leadership-measures 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Digital leadership appears on 17 recent papers and leadership keywords on almost 300, and 70 leadership papers carry a digital or AI keyword. New leadership constructs keep appearing, each with its own survey scale. Is digital leadership something new, or transformational leadership measured with new words?

<h2>What prompted this</h2> <p>Since October 2025, 296 papers carry a leadership keyword (184 in management journals), among them "leadership" on 26, "women in leadership" on 23, "digital leadership" on 17, "ethical leadership" on 14 and "strategic leadership" on 12. Seventy of them also carry a digital or AI keyword (digital transformation 26, digital leadership 17). Like the <a href="/ideas/is-the-technology-acceptance-model-still-useful">technology acceptance model</a>, the leadership scales are mostly measured with employee surveys.</p> <h2>Questions</h2> <ul> <li>How strongly do digital leadership scores correlate with transformational leadership scores for the same leaders, and does digital leadership predict anything once transformational leadership is controlled for?</li> <li>Do the outcomes digital leadership is said to drive (digital transformation, innovation) show up in data other than the same survey?</li> <li>How many different "digital leadership" scales are in use, and do they agree?</li> </ul> <h2>Data that could help</h2> <p>Survey datasets that include several leadership scales, the published scales themselves, and firm outcomes from outside the survey (patents, IT spending, digital revenue).</p> <h2>Next step</h2> <p>Collect the item wording of every digital leadership scale used since 2020 and compare it item by item with standard transformational leadership scales.</p> https://scholar4ai.com/ideas/what-carbon-capture-costs 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Carbon capture and storage, carbon removal and direct air capture appear on about 34 recent papers, mostly in economics, while governments now pay large subsidies for them. How much does a tonne captured actually cost in the projects that run, and which uses are worth the subsidy?

<h2>What prompted this</h2> <p>Since October 2025, about 34 papers carry a carbon-capture, carbon-removal or direct-air-capture keyword (25 in economics journals): "carbon capture and storage" on 19, "carbon capture utilization and storage" on 8, "carbon dioxide removal" on 6, "direct air capture" on 2. Subsidies such as the US 45Q credit and the first commercial storage hubs in Europe have turned capture from a modelling assumption into a set of real projects with costs.</p> <h2>Questions</h2> <ul> <li>In operating projects, how much CO2 is captured relative to design capacity, and what is the realized cost per tonne stored?</li> <li>Which sources (ethanol plants, gas processing, cement, power) capture most cheaply, and does the subsidy go mostly to the cheapest, which might have been captured anyway?</li> <li>How do the climate models that assume large-scale capture by 2050 compare with the build-out actually under way?</li> </ul> <h2>Data that could help</h2> <p>Project capacity and injection reports (regulatory filings, the Global CCS Institute database), subsidy claims, and cost disclosures from operators.</p> <h2>Next step</h2> <p>Compile design and reported capture rates for every operating capture project and compare realized costs per tonne with the assumptions in widely used climate scenarios.</p> https://scholar4ai.com/ideas/unemployment-insurance-for-the-ai-transition 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Unemployment insurance and job search appear on 86 recent papers, but only 20 of them carry an AI, automation or algorithm keyword. If language models displace office workers, the benefit systems built for factory layoffs will meet a different kind of job loser. What should change: benefit length, wage insurance or retraining?

<h2>What prompted this</h2> <p>Since October 2025, 86 papers carry an unemployment-insurance, unemployment-benefit, job-search or displaced-worker keyword (66 in economics journals): "unemployment insurance" on 18, "unemployment benefits" on 10, "job search" on 7. Only 20 of them also carry an automation, AI or algorithm keyword (automation 7, artificial intelligence 5, generative AI 4).</p> <h2>Questions</h2> <ul> <li>Do workers who lose jobs exposed to language models search longer and take larger pay cuts than other displaced workers with similar wages?</li> <li>Would wage insurance (topping up a lower-paid new job) get them back to work faster than longer benefits?</li> <li>Do job seekers who use AI to write applications get more interviews, and does the flood of applications lengthen everyone's search?</li> </ul> <h2>Data that could help</h2> <p>Administrative unemployment spells with prior occupation (registers in Scandinavia, Germany's IAB data), occupational AI-exposure scores, and job-board application data.</p> <h2>Next step</h2> <p>Compare unemployment durations and re-employment wages of workers displaced from high and low AI-exposure occupations since 2023 in one country's register data.</p> https://scholar4ai.com/ideas/trade-credit-when-rates-rise 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Trade credit appears on 112 recent papers, but only 3 of them carry an interest-rate or monetary-policy keyword. When bank credit gets dearer, firms lean on their suppliers and delay paying them. Does trade credit carry monetary tightening down the supply chain, and who ends up financing whom?

<h2>What prompted this</h2> <p>Since October 2025, 112 papers carry a trade-credit, accounts-payable, supplier-credit or payment-terms keyword (68 in finance journals, 16 in operations research), with "trade credit" alone on 90. Of these, 28 also carry a financial-constraint, bank-credit, crisis or shock keyword. Only 3 mention interest rates or monetary easing, although "monetary policy transmission" appears on 119 papers elsewhere in the archive.</p> <h2>Questions</h2> <ul> <li>After the 2022–2023 rate rises, did large buyers lengthen their payment terms, and did their small suppliers borrow more from banks to fill the gap?</li> <li>Does trade credit soften monetary tightening for constrained firms (they borrow from suppliers) or pass it on to the weakest firm in the chain?</li> <li>Did payment-practice disclosure rules (UK, EU proposals) shorten payment terms, or only make them visible?</li> </ul> <h2>Data that could help</h2> <p>Firm-level accounts payable and receivable (Compustat, Orbis, CSMAR), supplier–customer links, payment-practice reports and policy-rate changes.</p> <h2>Next step</h2> <p>Compare the change in days payable of large firms with the change in days receivable and bank debt of their listed suppliers across the 2022–2023 tightening.</p> https://scholar4ai.com/ideas/text-to-sql-for-business-data 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Text-to-SQL, table question answering and semantic parsing appear on only 12 recent papers, all in AI journals, none in business ones. Firms are rolling out assistants that turn managers' questions into database queries. How often do the queries return the wrong numbers, and would a manager notice?

<h2>What prompted this</h2> <p>Since October 2025, 12 papers carry a text-to-SQL, table-question-answering, semantic-parsing or natural-language-interface keyword, all in AI journals: "text-to-SQL" on 4, "table question answering" on 4, "semantic parsing" on 4. None appears in the management, accounting or operations research journals, where the users of business data publish.</p> <h2>Questions</h2> <ul> <li>On real company databases (many tables, ambiguous column names, business definitions such as "active customer"), how often do language models write queries that run but return the wrong answer?</li> <li>Which errors are most common: wrong joins, wrong filters, wrong definitions of a business term?</li> <li>Do managers who receive a number from an assistant check it, and does showing the query or a plain-language explanation change that?</li> </ul> <h2>Data that could help</h2> <p>Public benchmarks (Spider, BIRD) and, more importantly, query logs and gold answers from a company data warehouse, plus user experiments.</p> <h2>Next step</h2> <p>Build a test set of 200 business questions with gold answers on an anonymized company database and measure how often several models return a wrong but plausible number.</p> https://scholar4ai.com/ideas/synthetic-control-in-practice 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

The synthetic control method appears on 29 recent papers, mostly in economics, and its use is falling (4 in the last quarter against about 8 expected), while difference-in-differences is the most-used method on this site. Do published synthetic control studies report what readers need to trust them: pre-period fit, donor pools and placebo tests?

<h2>What prompted this</h2> <p>Since October 2025, "synthetic control method" appears on 29 papers (24 in economics journals, 5 in finance), and "synthetic difference-in-differences" on 2. Only 4 were published in July–September 2026 against about 8 expected. On the <a href="/rankings/equations">equation ranking</a>, <a href="/equations/difference-in-differences">difference-in-differences</a> is mentioned far more often, often in settings with a single treated unit where a synthetic control would fit better.</p> <h2>Questions</h2> <ul> <li>In recent synthetic control studies, how many report the donor pool, the pre-treatment fit, the weights and placebo-based inference?</li> <li>How sensitive are their estimates to the choice of donors and of the pre-treatment period, when re-run with the authors' data?</li> <li>In difference-in-differences papers with one treated region or country, would a synthetic control give a different answer?</li> </ul> <h2>Data that could help</h2> <p>Replication packages of published synthetic control and single-unit difference-in-differences studies.</p> <h2>Next step</h2> <p>Code the reporting practices of every synthetic control paper in the archive since 2025 and re-run a sample with alternative donor pools.</p> https://scholar4ai.com/ideas/speaking-up-about-ai-mistakes 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Psychological safety and employee voice appear on 21 recent papers, almost all in management journals, and only 2 of them carry an AI keyword. As AI tools enter everyday work, someone has to notice and report when they are wrong. Do teams with psychological safety catch AI errors that others let pass?

<h2>What prompted this</h2> <p>Since October 2025, 21 papers carry a psychological-safety, employee-voice, speaking-up or employee-silence keyword (17 in management journals): "employee voice" on 11, "psychological safety" on 8. Only 2 of them also carry an AI keyword. Studies of AI at work focus on adoption and productivity, not on who flags the errors.</p> <h2>Questions</h2> <ul> <li>In teams using AI assistants, does psychological safety predict how often members question or correct AI output, and does that show up in error rates?</li> <li>Do people stay silent about AI errors when a manager championed the tool?</li> <li>Does using AI change voice in the other direction, with employees blaming the tool rather than raising their own concerns?</li> </ul> <h2>Data that could help</h2> <p>Field data from firms that log AI suggestions and human overrides, team surveys on psychological safety, and experiments with planted AI errors.</p> <h2>Next step</h2> <p>Run an experiment in teams that review AI-generated work with planted errors, varying a manipulation of psychological safety, and count how many errors are reported.</p> https://scholar4ai.com/ideas/short-selling-in-chinas-a-shares 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Short selling and short-selling constraints appear on 56 recent papers, but only 10 of them carry a China keyword, although China restricted securities lending repeatedly in 2023 and 2024 to support prices. What did the restrictions do to price efficiency, liquidity and the overvaluation of hard-to-short stocks?

<h2>What prompted this</h2> <p>Since October 2025, 56 papers carry a short-selling, securities-lending or margin-trading keyword (47 in finance journals): "short-selling constraints" on 23, "short selling" on 21, "short-selling ban" on 5, "securities lending" on 5, "margin trading" on 4. Only 10 also carry a China or Chinese-market keyword, although China's securities lending rules changed several times in 2023 and 2024.</p> <h2>Questions</h2> <ul> <li>After securities lending was restricted, did stocks that had been heavily shorted become more overvalued relative to fundamentals, and did their prices react more slowly to bad news?</li> <li>Did the restrictions reduce volatility in falling markets, as intended, or only move bearish trading to index futures and options?</li> <li>Did quantitative funds that relied on short positions shrink or move to other markets?</li> </ul> <h2>Data that could help</h2> <p>Daily securities lending and margin balances by stock, the dates of rule changes, stock prices, earnings announcements and futures positions (CSMAR, Wind).</p> <h2>Next step</h2> <p>Compare post-earnings drift and price reactions to bad news in stocks with high and low pre-restriction short interest, before and after each rule change.</p> https://scholar4ai.com/ideas/screening-foreign-investment-for-national-security 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Foreign direct investment appears on 198 recent papers, but national security on 11 and investment screening on a handful. Since 2018, most rich countries have widened the review of foreign takeovers on security grounds. Which deals do the reviews stop, and what does screening do to investment flows and target firms?

<h2>What prompted this</h2> <p>Since October 2025, 16 papers carry a national-security, investment-screening or FDI-screening keyword (6 in management journals, 5 in economics), with "national security" on 11 (one of them about the US agency of that name). "Foreign direct investment" appears on 198 papers. Screening regimes have spread from the US to the EU, the UK, Japan and Australia, and now cover outbound investment too.</p> <h2>Questions</h2> <ul> <li>After a country widens its screening (more sectors, lower thresholds), do inflows from the targeted countries fall, shift to minority stakes or go to unscreened sectors?</li> <li>Do target firms' valuations fall when a deal is blocked or withdrawn, and do the firms later find other buyers?</li> <li>Are outbound investment rules (on AI and semiconductors in China) changing where venture capital invests?</li> </ul> <h2>Data that could help</h2> <p>Screening decisions and notifications (the EU's annual screening reports, CFIUS reports), deal data with acquirer nationality (SDC, Orbis M&amp;A), and target firm outcomes.</p> <h2>Next step</h2> <p>Compare acquisitions by Chinese and other foreign buyers in sectors added to screening lists before and after each country's reform.</p> https://scholar4ai.com/ideas/routing-home-health-care 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Home health care, nurse scheduling and workforce scheduling appear on only 15 recent papers in the operations research and AI journals. Ageing countries are moving care from hospitals and nursing homes into homes, which turns nurses into a vehicle fleet. How much travel and overtime can better routing save, and do the plans respect continuity of care?

<h2>What prompted this</h2> <p>Since October 2025, 15 papers carry a nurse-scheduling, home-health-care, workforce-scheduling or staff-rostering keyword (11 in operations research journals, 4 in AI): "workforce scheduling" on 7, "home health care" on 5, "nurse scheduling" on 5. The <a href="/ideas/operating-room-scheduling">operating room</a> and <a href="/ideas/ambulance-dispatch-and-emergency-care">ambulance</a> problems get more attention than care delivered at home, which is where demand is growing fastest.</p> <h2>Questions</h2> <ul> <li>In real home care operations, how much travel time and overtime do optimized routes save compared with planners' routes?</li> <li>How much does keeping the same nurse with the same patient (continuity of care) cost in efficiency, and does it improve outcomes?</li> <li>Do the models capture what makes plans fail in practice: cancellations, traffic, visits that run long?</li> </ul> <h2>Data that could help</h2> <p>Visit logs with planned and actual times from home care providers, nurse rosters, and patient outcomes such as readmissions.</p> <h2>Next step</h2> <p>Partner with a home care provider and re-plan one month of visits with a published routing model, measuring travel savings and changes in continuity of care.</p> https://scholar4ai.com/ideas/process-mining-for-audit-and-controls 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Process mining and robotic process automation appear on only 25 recent papers, mostly in AI and operations research, with 3 in accounting. Event logs from enterprise systems record every step of purchasing, invoicing and payment. Can auditors test whole populations of transactions instead of samples, and what do they find?

<h2>What prompted this</h2> <p>Since October 2025, 25 papers carry a process-mining, robotic-process-automation or business-process keyword (14 in AI journals, 5 in operations research, 3 in accounting): "process mining" on 8, "robotic process automation" on 5, "business process modeling" on 4. Audit firms advertise full-population testing with these tools, but evidence on what it changes is scarce.</p> <h2>Questions</h2> <ul> <li>When auditors analyse the full event log of the purchase-to-pay process, how many control deviations do they find that sampling would have missed, and how many are real problems?</li> <li>Do firms whose auditors use process mining have fewer later restatements or material weaknesses?</li> <li>Do automated processes (bots) reduce control deviations, or create new ones that nobody reviews?</li> </ul> <h2>Data that could help</h2> <p>Anonymized enterprise system event logs (public process mining challenge datasets), audit firm data on tool use, and restatements and control weaknesses (Audit Analytics).</p> <h2>Next step</h2> <p>Apply standard conformance checking to a public purchase-to-pay event log and classify the deviations found by type and likely materiality.</p> https://scholar4ai.com/ideas/pricing-water-in-dry-years 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Water scarcity, irrigation and groundwater appear on 72 recent papers, with 21 in the last quarter against about 16 expected, but water pricing itself on only 4. Droughts are forcing cities and farm districts to ration water. Would prices ration it better, and who would pay?

<h2>What prompted this</h2> <p>Since October 2025, 72 papers carry a water-scarcity, water-pricing, irrigation, groundwater or water-use keyword (44 in economics journals, 17 in AI): "water scarcity" on 22, "irrigation" on 6, "groundwater pumping" on 5, "water pricing" on 4, "household water use" on 4. The quarter to September 2026 had 21 such papers against about 16 expected. The AI papers mostly forecast groundwater levels; the economics of allocating water is thin.</p> <h2>Questions</h2> <ul> <li>When cities raised water prices in droughts (Cape Town, California, Spain), how much did use fall, and how did the burden split between rich and poor households?</li> <li>Do water markets among farmers (Australia's Murray–Darling, California) move water to higher-value crops, and what happens to farm communities that sell?</li> <li>Does free or cheap electricity for pumps drive groundwater depletion, as in India, and would metering change it?</li> </ul> <h2>Data that could help</h2> <p>Household billing data around tariff changes, water market trade records, groundwater level monitoring and electricity use by farm pumps.</p> <h2>Next step</h2> <p>Estimate the price elasticity of residential water demand by income group from billing data around a drought-time tariff increase.</p> https://scholar4ai.com/ideas/parcel-lockers-and-the-last-mile 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Last-mile delivery appears on 79 recent papers, and the topic is growing (23 papers in the last quarter against about 18 expected), but parcel lockers on only 6. Lockers and pickup points promise fewer failed deliveries and fewer van stops. Do customers use them, and do they cut cost and emissions once the trips customers make to fetch parcels are counted?

<h2>What prompted this</h2> <p>Since October 2025, 83 papers carry a last-mile, parcel-locker or pickup-point keyword (69 in operations research journals), with "last-mile delivery" on 79. The quarter to September 2026 had 23 such papers against about 18 expected. Only 13 carry a locker, pickup or collection keyword, and "parcel lockers" appears on 6.</p> <h2>Questions</h2> <ul> <li>Where lockers were installed, what share of parcels moved from home delivery, and how far do customers travel to collect them, by car or on foot?</li> <li>Do shared (carrier-neutral) lockers save more than each carrier running its own network?</li> <li>How should prices or delivery fees be set to steer customers to lockers without losing them to competitors?</li> </ul> <h2>Data that could help</h2> <p>Carrier data on deliveries and failed attempts by address, locker usage logs, and surveys or mobility data on how customers collect parcels.</p> <h2>Next step</h2> <p>Compare failed-delivery rates and van stops per parcel in neighbourhoods before and after locker installation, using carrier data.</p> https://scholar4ai.com/ideas/non-gaap-earnings-in-the-ai-boom 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Non-GAAP earnings appear on only 6 recent papers. Technology firms report profits that exclude stock-based pay and, increasingly, the costs of AI investment, while AI start-ups report "adjusted" margins of their own design. Which adjustments mislead, and do investors see through them?

<h2>What prompted this</h2> <p>Since October 2025, "non-GAAP earnings" appears on 6 papers (4 in accounting journals, 2 in finance), and no other keyword in the archive names adjusted or pro forma earnings. At the same time, the gap between reported and adjusted profits has grown at the largest technology firms, through stock-based compensation, restructuring charges and, now, the depreciation of AI data centres.</p> <h2>Questions</h2> <ul> <li>How large is the gap between GAAP and non-GAAP earnings at AI-exposed firms, and which items explain it?</li> <li>Do firms that exclude stock-based compensation have higher future dilution than investors expect?</li> <li>Do the new depreciation policies for AI servers (longer useful lives) make GAAP earnings themselves less comparable, so that the adjustments go both ways?</li> </ul> <h2>Data that could help</h2> <p>Non-GAAP reconciliations from earnings releases (8-K exhibits), stock-based compensation and share counts (Compustat), depreciation policy notes and analysts' street earnings (I/B/E/S).</p> <h2>Next step</h2> <p>Collect the non-GAAP reconciliations of the 50 largest technology firms for 2019–2026 and decompose the gap by type of adjustment.</p> https://scholar4ai.com/ideas/mini-grids-and-rural-electrification 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Electricity access, energy access and rural electrification appear on 106 recent papers, mostly in economics, and the topic is fading (21 papers in the last quarter against about 28 expected). Yet the push to connect Africa's unconnected relies on mini-grids and solar home systems. Which works better, for whom, and at what cost per household?

<h2>What prompted this</h2> <p>Since October 2025, 106 papers carry an electricity-access, energy-access, rural-electrification, mini-grid or off-grid keyword (94 in economics journals): "energy access" on 39, "electricity access" on 34, "rural electrification" on 27, "mini-grids" on 5. The quarter to September 2026 had 21 such papers against about 28 expected, just as large programmes to connect hundreds of millions of people in Africa get under way.</p> <h2>Questions</h2> <ul> <li>Per household, what do grid extension, mini-grids and solar home systems cost, and how much electricity do households actually use from each after five years?</li> <li>Do mini-grids raise firm activity and incomes in villages more than solar home systems, which mostly power lights and phones?</li> <li>What happens to mini-grids when the national grid arrives: are they absorbed, stranded or kept as backup?</li> </ul> <h2>Data that could help</h2> <p>Programme data on connections and costs, utility billing or smart meter data from mini-grid operators, household surveys and night-time lights.</p> <h2>Next step</h2> <p>Use the staggered roll-out of a national electrification programme to compare village incomes and enterprise counts by connection type.</p> https://scholar4ai.com/ideas/mental-health-and-the-labour-market 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Mental health, depression and anxiety appear on 238 recent papers, but most are economics papers on causes or AI papers that detect depression; only 22 carry an employment, earnings or labour-market keyword. Mental-health conditions are now a leading cause of work incapacity in many countries. What do they cost in jobs and earnings, and what brings people back?

<h2>What prompted this</h2> <p>Since October 2025, 238 papers carry a mental-health, depression, anxiety or psychological-well-being keyword (103 in economics journals, 82 in AI, most of them on detecting depression from text or signals): "mental health" on 50, "depression" on 23, "anxiety" on 22. Only 22 also carry an employment, earnings, productivity, absenteeism, unemployment or labour-market keyword.</p> <h2>Questions</h2> <ul> <li>After a first diagnosis of depression or anxiety, how much do employment and earnings fall over five years, and how much do they recover with treatment?</li> <li>Do faster access to therapy or employment-support programmes (individual placement and support) bring people back to work, and at what cost per job?</li> <li>Is the rise in disability claims for mental health among young adults a change in health, in diagnosis, or in the rules?</li> </ul> <h2>Data that could help</h2> <p>Linked health and employment registers (Scandinavia, the Netherlands), disability claim records, and trial data from employment-support programmes.</p> <h2>Next step</h2> <p>Use linked register data to trace employment and earnings around first mental-health diagnoses, compared with people diagnosed a few years later.</p> https://scholar4ai.com/ideas/measuring-energy-poverty 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Energy poverty appears on 107 recent papers, almost all in economics, but only 12 energy-poverty papers carry a measurement or index keyword. The measures in use (spending shares, inability to heat, multidimensional indices) classify different households as poor. How much do the findings depend on the measure chosen?

<h2>What prompted this</h2> <p>Since October 2025, 135 papers carry an energy-poverty, fuel-poverty or energy-affordability keyword (123 in economics journals): "energy poverty" on 107, "energy affordability" on 24, "fuel poverty" on 5. Only 12 also carry a measurement, index, indicator or multidimensional keyword (multidimensional energy poverty index 4, energy poverty indicators 4), although the topic's conclusions rest on how the poor are counted.</p> <h2>Questions</h2> <ul> <li>In the same household survey, how much do the common measures (10% spending share, twice the median share, low-income-high-cost, self-reported inability to keep warm, multidimensional indices) overlap?</li> <li>Do the papers' main results (effects of digital finance, green credit or energy prices on energy poverty) survive a change of measure?</li> <li>Which measure best predicts outcomes that matter (cold-related health problems, arrears, disconnections)?</li> </ul> <h2>Data that could help</h2> <p>Household surveys with energy spending, income and self-reported comfort (EU-SILC, the UK's English Housing Survey, China's CFPS), and health and arrears records.</p> <h2>Next step</h2> <p>Compute five common energy poverty measures in one survey and tabulate how many households each classifies as poor and how many all of them agree on.</p> https://scholar4ai.com/ideas/llms-as-qualitative-coders 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Qualitative methods appear on 46 recent papers, mostly in management, and only 7 of them carry an AI or language-model keyword. Researchers are starting to let language models code interviews and open-ended answers. Do the codes agree with trained human coders, and what does the method section need to report?

<h2>What prompted this</h2> <p>Since October 2025, 46 papers carry a qualitative-research, qualitative-analysis, grounded-theory, interview, focus-group or qualitative-comparative-analysis keyword (26 in management journals): "fuzzy-set qualitative comparative analysis" on 25, "qualitative research" on 4, "expert interviews" on 4, "grounded theory" on 2. Only 7 also carry an AI or language-model keyword. A <a href="/ideas/language-models-as-survey-respondents">related note</a> asks whether models can stand in for survey respondents; this one is about models reading what real respondents said.</p> <h2>Questions</h2> <ul> <li>On interview transcripts already coded by trained researchers, how closely do language-model codes agree, by code and by how abstract the code is?</li> <li>Do models miss what is unusual (the rare theme that changes the story) more than what is common?</li> <li>What should authors report so readers can judge model-assisted coding: prompts, model versions, agreement statistics, human checks?</li> </ul> <h2>Data that could help</h2> <p>Published qualitative datasets with codebooks and human coding (qualitative data archives such as the UK Data Service), and journal guidelines on AI use.</p> <h2>Next step</h2> <p>Re-code a publicly archived interview study with two language models using the original codebook and compare agreement with the original coders code by code.</p> https://scholar4ai.com/ideas/islamic-finance-and-the-green-transition 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Islamic banking, Islamic finance and sukuk appear on 65 recent papers, but only 16 of them carry a green, sustainability, ESG or climate keyword. Green sukuk issuance has grown in the Gulf and Southeast Asia. Is Sharia-compliant finance a natural fit for sustainable investment, or just another label?

<h2>What prompted this</h2> <p>Since October 2025, 65 papers carry an Islamic-finance, sukuk or Sharia keyword (43 in finance journals, 10 in management): "Islamic banking" on 25, "Islamic finance" on 19, "sukuk" on 12, "sharia compliance" on 10. Only 16 also carry a green, sustainability, ESG or climate keyword (sustainable finance 4, green bonds 3), while green bonds as a whole are one of the larger topics in the finance journals.</p> <h2>Questions</h2> <ul> <li>Do green sukuk price differently from conventional green bonds of the same issuer, and is there a "double label" premium?</li> <li>Are the projects financed by green sukuk different from those of conventional green bonds, or are the frameworks copied?</li> <li>Do Sharia screens (which exclude some sectors and high leverage) already tilt Islamic funds toward lower-carbon portfolios?</li> </ul> <h2>Data that could help</h2> <p>Sukuk and green bond issuance with prices (Bloomberg, Refinitiv), issuer frameworks and allocation reports, and Islamic fund holdings with carbon data.</p> <h2>Next step</h2> <p>Match green sukuk to conventional green bonds from the same sovereign or corporate issuers and compare yields at issue and in the secondary market.</p> https://scholar4ai.com/ideas/is-bitcoin-an-inflation-hedge 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Bitcoin appears on 164 recent papers, and 53 crypto papers carry a hedging, safe-haven or inflation keyword, but "inflation hedging" itself appears on just 6. The 2021–2023 inflation was the first real test of the "digital gold" claim, and spot ETFs have since changed who holds Bitcoin. What did the test show?

<h2>What prompted this</h2> <p>Since October 2025, 411 papers carry a Bitcoin, cryptocurrency or crypto-asset keyword (285 in finance journals, 75 in economics), with "Bitcoin" on 164. The quarter to September 2026 had 104 such papers against about 93 expected. Of the 411, 53 carry a hedging, safe-haven or inflation keyword (safe-haven assets 20, hedging 15, hedging effectiveness 11), and "inflation hedging" appears on only 6.</p> <h2>Questions</h2> <ul> <li>During the inflation surge, did Bitcoin rise with inflation surprises and inflation expectations, or fall with rate hikes like a technology stock?</li> <li>Did the arrival of spot ETFs in 2024 make Bitcoin move more with equities, as institutional ownership grew?</li> <li>In high-inflation countries (Argentina, Turkey, Nigeria), did local demand for Bitcoin and stablecoins rise with inflation, even if global prices did not?</li> </ul> <h2>Data that could help</h2> <p>Bitcoin prices, inflation releases and surprises, breakeven inflation, ETF flows, and local exchange volumes and premiums in high-inflation countries.</p> <h2>Next step</h2> <p>Estimate Bitcoin's response to US inflation surprises and policy-rate surprises separately for 2019–2021, 2021–2023 and the ETF period.</p> https://scholar4ai.com/ideas/institutional-investors-in-rental-housing 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Institutional investors appear on 151 recent papers, almost all about stock ownership; only 6 institutional-investor papers carry a housing or rent keyword. Funds now own large numbers of single-family homes and apartments in the US, Spain and elsewhere, and several governments want to restrict them. Do they raise rents, improve maintenance, or both?

<h2>What prompted this</h2> <p>Since October 2025, 167 papers carry an institutional-investor keyword (116 in finance journals), with "institutional investors" on 151, mostly in studies of stock ownership and corporate governance. Only 6 of them also carry a housing, rent, home or real-estate keyword. "Rental housing" appears on 7 papers and "housing market" on 53.</p> <h2>Questions</h2> <ul> <li>In neighbourhoods where funds bought many homes, did rents and evictions rise faster than elsewhere, controlling for demand?</li> <li>Do institutional landlords maintain homes better (or worse) than small landlords, and do tenants stay longer?</li> <li>Would proposed bans or taxes on institutional purchases of homes lower prices for first-time buyers, or only change who rents them out?</li> </ul> <h2>Data that could help</h2> <p>Property transaction records with buyer type (deeds data, CoreLogic, land registries), listing rents, eviction filings and code-violation records.</p> <h2>Next step</h2> <p>Match purchases by large institutional buyers to listing rents and eviction filings at the address level and compare with nearby homes bought by individuals.</p> https://scholar4ai.com/ideas/gender-gaps-in-economics-and-finance-academia 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Gender gaps and gender bias appear on 60 recent papers, and board gender diversity on 67 more, but only 13 gender papers concern academia or research. Women remain a minority among economics and finance professors and authors in top journals. Where in the pipeline do they drop out, and how do refereeing and citations treat their work?

<h2>What prompted this</h2> <p>Since October 2025, 185 papers carry a gender-gap, gender-bias, gender-diversity or women-in-science keyword (54 in economics journals, 51 in finance): "board gender diversity" on 67, "gender diversity" on 33, "gender gap" on 32, "gender bias" on 28. Only 13 concern academia, publication or research (women in science 3, research productivity 3, academic promotion 2), although the journals tracked here are themselves the stage.</p> <h2>Questions</h2> <ul> <li>In the tracked journals, what share of authors are women, by field and journal, and has it changed since 2015?</li> <li>Are papers by women cited less within the same journal and year, after accounting for topic?</li> <li>Do women's papers spend longer in review, and does double-blind review change that?</li> </ul> <h2>Data that could help</h2> <p>Author names with inferred gender (with known limits) from OpenAlex, citation counts, journals' editorial data where shared, and survey data on promotion.</p> <h2>Next step</h2> <p>Estimate the share of female authors by journal and year in this site's archive and compare citation counts of comparable papers.</p> https://scholar4ai.com/ideas/gamification-from-trading-apps-to-workplaces 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Gamification appears on only 20 recent papers, split across management, finance and accounting. Confetti on a trading app, points for saving, badges at work: the same design tricks are used to make people trade, save, learn and work more. Which uses help the people they are aimed at, and which only help the platform?

<h2>What prompted this</h2> <p>Since October 2025, "gamification" appears on 20 papers (11 in management journals, 4 in finance, 3 in accounting). Half of them concern trading, investing, financial literacy, learning or work engagement. Regulators have questioned "digital engagement practices" on trading apps, while employers adopt gamified training and sales targets.</p> <h2>Questions</h2> <ul> <li>When a trading app removes game-like features, do its users trade less, take less risk and earn more?</li> <li>Do gamified savings and pension apps raise saving for people who saved too little, or mostly for those who already saved?</li> <li>At work, does gamification raise effort for long, or does it wear off and crowd out intrinsic motivation?</li> </ul> <h2>Data that could help</h2> <p>Brokerage account data around design changes, savings app data with randomized features, and workplace experiments with gamified incentives.</p> <h2>Next step</h2> <p>Find a trading or savings app design change (a feature removed or added) and compare user behaviour before and after with a difference-in-differences design.</p> https://scholar4ai.com/ideas/fintech-entry-and-bank-competition 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Bank competition appears on 68 recent papers, but only 19 bank-competition papers carry a fintech, digital or platform keyword. Fintech lenders and digital banks were supposed to erode banks' market power. Did loan and deposit margins actually fall where they entered, and for whom?

<h2>What prompted this</h2> <p>Since October 2025, 75 papers carry a bank-competition, bank-market-power or banking-market-structure keyword (53 in finance journals, 20 in economics), with "bank competition" on 68. Only 19 of them also carry a fintech, digital, big-tech, platform or mobile keyword (fintech 7, digital finance 4, digital banking 4), although "fintech" appears on 189 papers in the archive.</p> <h2>Questions</h2> <ul> <li>In regions where fintech lenders gained share, did banks' loan rates and deposit margins narrow, or did fintech take only the borrowers banks did not want?</li> <li>Did big-tech platforms (Ant, Tencent, Mercado Pago) raise competition, or build new market power of their own?</li> <li>Does fintech competition make banks take more risk, as the old competition–fragility view would predict?</li> </ul> <h2>Data that could help</h2> <p>Regional fintech lending volumes (HMDA, platform disclosures, China's digital finance index), bank-level margins and risk measures, and the timing of fintech entry.</p> <h2>Next step</h2> <p>Relate the change in county-level bank loan spreads to fintech mortgage and personal-loan market shares over 2015–2025.</p> https://scholar4ai.com/ideas/family-firm-succession-in-china 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Family business appears on 124 recent papers and generational succession on 18. The founders who built China's private firms in the 1980s and 1990s are now handing over. Do the heirs keep the firms, sell them or bring in professional managers, and what happens to performance and investment?

<h2>What prompted this</h2> <p>Since October 2025, 162 papers carry a family-business, family-ownership or succession keyword (74 in management journals, 63 in finance): "family business" on 124, "family ownership" on 26, "generational succession" on 18, "family business succession" on 11, "CEO succession" on 10. "Chinese family firms" appears on 11. The volume is steady (35 papers in July–September 2026 against about 40 expected), but few papers follow the wave of handovers now under way in China.</p> <h2>Questions</h2> <ul> <li>When a founder hands over to a child, do investment, R&amp;D and risk-taking change, and does the answer depend on whether the heir studied abroad?</li> <li>How many founders sell or hire professional managers instead, and are those firms the better or the worse performers?</li> <li>Do political connections, which founders built personally, survive the handover?</li> </ul> <h2>Data that could help</h2> <p>Chairman and CEO changes in Chinese listed firms with family relationships (CSMAR's private-firm and family modules), firm accounts and patents, and heirs' education from annual reports.</p> <h2>Next step</h2> <p>Identify founder-to-child successions among Chinese listed private firms since 2010 and compare them with founder-to-professional transitions on investment and performance.</p> https://scholar4ai.com/ideas/emotion-ai-in-marketing 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Emotion recognition and affective computing appear on 325 recent papers, almost all in AI journals; only 19 carry a consumer, marketing or ethics keyword. The EU now bans emotion recognition at work and in schools, but not in shops or advertising. Does reading faces and voices tell marketers anything reliable, and should it be allowed?

<h2>What prompted this</h2> <p>Since October 2025, 325 papers carry an emotion-recognition, affective-computing, facial-expression or eye-tracking keyword (301 in AI journals): "emotion recognition" on 76, "multimodal emotion recognition" on 60, "affective computing" on 53, "facial expression recognition" on 29. Only 19 also carry a consumer, marketing, advertising, retail, privacy or ethics keyword (neuromarketing 4, AI ethics 3).</p> <h2>Questions</h2> <ul> <li>Do emotion scores from faces or voices predict purchases or ad recall better than simply asking people, out of the lab and across cultures?</li> <li>How accurate are commercial emotion AI tools for groups underrepresented in their training data?</li> <li>Since the EU banned emotion recognition in workplaces and education, have vendors moved to retail and advertising, and do consumers know they are being read?</li> </ul> <h2>Data that could help</h2> <p>Field experiments in advertising testing with and without emotion measures, vendor accuracy audits by demographic group, and vendor product and market data before and after the EU rules.</p> <h2>Next step</h2> <p>Compare how well facial emotion scores and short self-reports predict choices among the same participants in an online ad-testing experiment.</p> https://scholar4ai.com/ideas/deposit-pricing-in-the-rate-cycle 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Deposit rates appear on only 31 recent papers, although the gap between policy rates and what banks paid savers was one of the largest transfers of the 2022–2024 tightening. How fast did deposit rates follow policy rates this time, and what made some banks pay more?

<h2>What prompted this</h2> <p>Since October 2025, 31 papers carry a deposit-rate, deposit-pricing, bank-deposit or uninsured-deposit keyword (22 in finance journals, 8 in economics): "deposit interest rates" on 17, "bank deposits" on 8, "uninsured deposits" on 3. The quarter to September 2026 had 8 such papers against about 7 expected. Bank runs and deposit insurance draw far more attention than the everyday pricing of deposits.</p> <h2>Questions</h2> <ul> <li>How much of each policy-rate rise did banks pass on to savers (the deposit beta), and was it lower than in the 2004–2006 cycle, as the deposits channel predicts?</li> <li>Did online banks and money market funds force incumbents to pay more, and did the 2023 bank failures change that?</li> <li>Now that rates are falling, do banks cut deposit rates faster than they raised them?</li> </ul> <h2>Data that could help</h2> <p>Branch- and product-level deposit rates (RateWatch or central bank statistics), bank balance sheets, money market fund flows and policy-rate paths.</p> <h2>Next step</h2> <p>Estimate deposit betas by bank for the 2022–2024 rises and the following cuts, and relate them to local market concentration and the bank's share of online deposits.</p> https://scholar4ai.com/ideas/activist-investors-and-the-energy-transition 2026-10-26T00:00:00+09:00 2026-10-26T00:00:00+09:00

Shareholder activism appears on only 11 recent papers, and 8 activism papers carry a climate or ESG keyword. Activist funds have pushed oil companies in opposite directions: toward climate targets in 2021, away from renewable spending more recently. Which campaigns win, and what happens to the companies' strategies and emissions afterwards?

<h2>What prompted this</h2> <p>Since October 2025, 16 papers carry a shareholder-activism, activist or shareholder-engagement keyword (7 in finance journals, 6 in management): "shareholder activism" on 11, "shareholder engagement" on 2. Eight of them also carry a climate, ESG, environmental or carbon keyword. The campaigns themselves, from Engine No. 1 at ExxonMobil to funds urging BP and others to cut low-carbon spending, are much better documented in the press than in the journals.</p> <h2>Questions</h2> <ul> <li>Do activists who push for more fossil investment and those who push for climate targets target different firms (by valuation, by past returns)?</li> <li>After a campaign succeeds, do capital spending plans, emission targets and returns change in the direction the activist wanted?</li> <li>Do index funds' votes decide these contests, and did their voting shift as ESG became politically contested?</li> </ul> <h2>Data that could help</h2> <p>Activism campaign databases (13D filings, SharkRepellent), proxy votes, company capital spending plans and emission targets before and after campaigns.</p> <h2>Next step</h2> <p>Code every activist campaign at oil, gas and utility companies since 2015 by the direction of its climate demand and compare outcomes.</p> https://scholar4ai.com/ideas/zoning-reform-and-housing-supply 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Housing supply, land use regulation and upzoning appear on 48 recent papers in the economics journals. Cities from Auckland to Minneapolis and states from California to Oregon have loosened zoning in the last decade. How many extra homes did the reforms produce, where, and did rents respond?

<h2>What prompted this</h2> <p>Since October 2025, 48 papers carry a zoning, land-use-regulation, housing-supply, building-regulation or upzoning keyword (37 in economics journals): "housing supply" on 19, "land use regulation" on 12, "upzoning" on 7, "building regulations" on 7. Housing prices appear on 123 papers, so the demand side is far better covered than the supply rules.</p> <h2>Questions</h2> <ul> <li>After upzoning, how much of the new building is on the rezoned parcels, and how long before permits turn into completed homes?</li> <li>Do rents fall in the rezoned neighbourhoods, in the city as a whole, or neither within five years?</li> <li>Which reforms worked: allowing more units per lot, cutting parking requirements, speeding approvals, or taxing undeveloped land?</li> </ul> <h2>Data that could help</h2> <p>Parcel-level zoning maps before and after reforms, building permits and completions, rents from listing data, and the dates of reforms.</p> <h2>Next step</h2> <p>Compare completions on upzoned and non-upzoned parcels in Auckland and Minneapolis in the years after their reforms.</p> https://scholar4ai.com/ideas/zombie-firms-after-cheap-money 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Capital reallocation is a fast-rising keyword in the economics journals. After a decade of near-zero rates, interest rates rose almost everywhere, and Japan ended negative rates in 2024. Did higher rates finally clear out zombie firms and move capital to productive ones?

<h2>What prompted this</h2> <p>In economics, "capital reallocation" had 6 papers in July–September 2026 against about 1 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, "zombie firms" appears on 21 papers, "capital reallocation" on 20 and "resource misallocation" on 43. Of the 24 papers with a zombie keyword, 8 also carry an interest-rate, monetary or credit keyword.</p> <h2>Questions</h2> <ul> <li>Did the share of zombie firms (firms that cannot cover interest from profits for years) fall after rates rose, and did it fall through exit, restructuring or recovery?</li> <li>Did banks keep lending to them, especially banks with weak capital, as Japanese banks did in the 1990s?</li> <li>Did capital and workers move to more productive firms, and did aggregate productivity respond?</li> </ul> <h2>Data that could help</h2> <p>Firm accounts (Orbis, Compustat Global and national data such as Japanese listed-firm data), bank-firm loan links, and policy-rate changes by country.</p> <h2>Next step</h2> <p>Compute the zombie share by country and year for 2015–2025 with a standard interest-coverage definition, and plot it against policy rates.</p> https://scholar4ai.com/ideas/zero-day-options-and-intraday-volatility 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The VIX and market liquidity are rising keywords in the finance journals, while options that expire on the day they trade have become a large share of US index options volume. Yet none of the 192 recent option and VIX papers carries a zero-day or 0DTE keyword. Do same-day options make intraday markets calmer or more fragile?

<h2>What prompted this</h2> <p>In the finance journals, "VIX" had 13 papers in July–September 2026 against about 8 expected from its earlier share, and "market liquidity" had 13 against 7 (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 192 papers carry a VIX, option-pricing, options-market or implied-volatility keyword (137 in finance journals). Only 4 of them carry a retail or intraday keyword, and none carries a zero-day or 0DTE keyword.</p> <h2>Questions</h2> <ul> <li>Does dealers' hedging of same-day options dampen or amplify intraday index moves, depending on whether dealers are long or short gamma?</li> <li>Do retail investors lose money on these options on average, and more than on older short-dated options?</li> <li>Has the VIX, which is built from 30-day options, become less informative as trading moves to same-day contracts?</li> </ul> <h2>Data that could help</h2> <p>Intraday trades and quotes of index options, estimates of dealer positioning, intraday index returns, and measures of retail order flow.</p> <h2>Next step</h2> <p>Relate estimated dealer gamma in same-day index options to intraday S&amp;P 500 volatility on each trading day of 2023–2025.</p> https://scholar4ai.com/ideas/youth-unemployment-in-china 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Youth employment and unemployment appear on 25 recent papers, mostly in economics, while China's youth unemployment rate reached record highs and the statistic itself was revised. What are graduates doing instead, how do families absorb the cost, and does the mismatch come from degrees or from the jobs?

<h2>What prompted this</h2> <p>Since October 2025, 25 papers carry a youth-unemployment, youth-employment, graduate-unemployment or college-graduate keyword (18 in economics journals): "youth employment" on 10, "youth unemployment" on 7, "recent college graduates" on 2. Six were published in July–September 2026, in line with earlier quarters, despite the size of the problem.</p> <h2>Questions</h2> <ul> <li>Where did the graduates of 2022–2025 go: further study, civil service exams, gig work, or back to their parents' homes, and for how long?</li> <li>Do fields of study with the largest enrolment growth have the worst outcomes, and did the private tutoring crackdown and property slump remove the jobs those graduates expected?</li> <li>Do families cut spending and saving to support unemployed graduates, and does that show up in consumption data?</li> </ul> <h2>Data that could help</h2> <p>Graduate employment surveys, household panels with young adults' status, civil service exam applicant counts, and job posting data by education requirement.</p> <h2>Next step</h2> <p>Track the employment status of 2022–2025 graduates in the China Family Panel Studies waves and compare fields of study.</p> https://scholar4ai.com/ideas/women-entrepreneurs-and-finance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Women entrepreneurs appear on 32 recent papers in the management and economics journals, and 14 of 43 such papers carry a funding, finance or credit keyword. Women-led startups still receive a small share of venture capital, and women-owned small firms pay more for loans in several studies. Is the gap demand, supply, or the way applications are read?

<h2>What prompted this</h2> <p>Since October 2025, 43 papers carry a women-entrepreneur or women-owned-business keyword (23 in management journals, 18 in economics): "women entrepreneurs" on 32, "women's entrepreneurship" on 7, "women-owned businesses" on 5. Fourteen of them also carry a funding, finance, venture, loan, credit, investor or crowdfunding keyword; "access to finance" appears on 65 papers in the archive, mostly about firms in general.</p> <h2>Questions</h2> <ul> <li>In loan and venture applications with the same business plan and financials, do women founders receive worse terms, and does the gap close when the evaluator is a model rather than a person?</li> <li>Do women ask for less, and if so, is that a response to expected rejection?</li> <li>Do funds and lenders that set targets for women-led firms change their deal flow, or only their reporting?</li> </ul> <h2>Data that could help</h2> <p>Loan application data with decisions, venture deal data with founder gender (PitchBook, Crunchbase), and audit studies with matched applications.</p> <h2>Next step</h2> <p>Run a correspondence study sending matched loan applications that differ only in the founder's gender to online lenders.</p> https://scholar4ai.com/ideas/wildfire-evacuation-planning 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Forest fires, wildfire risk and smoke appear on 48 recent papers, split among economics, AI and operations research, and evacuation planning on 4. Fires now reach cities, and evacuations have failed with deadly results. Do evacuation models fit how people actually leave, and where should the roads, shelters and warnings be?

<h2>What prompted this</h2> <p>Since October 2025, 48 papers carry a wildfire, forest-fire, bushfire or evacuation keyword (19 in economics journals, 15 in AI, 10 in operations research): "forest fires" on 13, "wildfire smoke" on 7, "wildfire risk" on 6, "wildfire mitigation" on 4, "evacuation planning" on 4, "wildfire suppression" on 3. The economics papers measure damages, the AI papers detect fires, and the few evacuation models are rarely validated against real evacuations.</p> <h2>Questions</h2> <ul> <li>In recent evacuations, when did people leave relative to the orders, which routes did they take, and where did congestion form, compared with what the models assume?</li> <li>Can mobile phone traces calibrate departure-time distributions by neighbourhood and warning type?</li> <li>Which investments (road capacity, staged warnings, shelters in place) reduce expected deaths most per dollar in the highest-risk towns?</li> </ul> <h2>Data that could help</h2> <p>Mobile phone mobility data during evacuations, evacuation order timings, road network data and fire spread records.</p> <h2>Next step</h2> <p>Reconstruct the departure times and routes of one recent evacuation from mobility data and compare them with the output of a standard evacuation model.</p> https://scholar4ai.com/ideas/who-writes-the-annual-report-now 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Annual reports, 10-K filings, narrative disclosure and readability appear on 99 recent papers in the finance and accounting journals, but only 9 of them carry a generative-AI, language-model or AI-generated keyword. If language models now draft the management discussion, do the text measures researchers rely on still mean anything?

<h2>What prompted this</h2> <p>Since October 2025, 99 papers carry an annual-report, 10-K, MD&amp;A, narrative-disclosure, readability or textual-analysis keyword (60 in finance journals, 19 in accounting): "10-K filings" on 27, "narrative disclosure" on 13, "readability" on 10. Only 9 of the 99 also carry a generative-AI, ChatGPT, language-model or AI-generated keyword.</p> <h2>Questions</h2> <ul> <li>Can AI-written passages in filings be detected, and what share of recent filings contains them?</li> <li>Did readability, tone and boilerplate measures shift after 2023 in a way that reflects the drafting tool rather than the firm?</li> <li>Do investors react differently to AI-drafted and human-drafted disclosures, once they can tell them apart?</li> </ul> <h2>Data that could help</h2> <p>The text of 10-K filings from SEC EDGAR for 2018–2025, AI-text detectors, and market reactions to filings.</p> <h2>Next step</h2> <p>Run an AI-text detector on the MD&amp;A sections of all 10-K filings for 2021–2025 and track the share flagged by year and by firm size.</p> https://scholar4ai.com/ideas/who-pays-for-carbon-pricing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Income redistribution, welfare losses and social welfare are rising keywords in the economics journals, and carbon taxes and emissions trading are large topics. Yet only 34 of 422 recent carbon-pricing papers carry a distributional or household keyword. As carbon prices reach heating and transport fuels, who bears the cost, and does returning the revenue make it fair?

<h2>What prompted this</h2> <p>In economics, "income redistribution" (7 papers in July–September 2026 against about 2 expected), "welfare loss" (10 against 5) and "social welfare" (20 against 12) are growing (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, 422 papers carry a carbon-tax, carbon-pricing or emissions-trading keyword (260 in economics journals, 80 in finance, 46 in operations research). Only 34 of them carry a distribution, household, inequality, regressivity or redistribution keyword, and "revenue recycling" appears on 12.</p> <h2>Questions</h2> <ul> <li>How regressive are carbon prices on heating and motor fuel, across income groups and between cities and the countryside?</li> <li>Which ways of returning the revenue (equal dividends, lower income taxes, targeted transfers) leave most households better off?</li> <li>Does visible compensation raise public support for higher carbon prices?</li> </ul> <h2>Data that could help</h2> <p>Household expenditure surveys, energy use by fuel, carbon price schedules (including the EU's new emissions trading for buildings and road transport), and survey experiments on public support.</p> <h2>Next step</h2> <p>Compute the cost of a given carbon price for each income decile from a national household expenditure survey, with and without an equal per-person dividend.</p> https://scholar4ai.com/ideas/whistleblower-rewards 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Whistleblowing appears on 16 recent papers across the tracked journals, mostly on whether employees speak up. Reward programmes at the SEC and now the EU and other regulators pay tipsters a share of penalties. Do rewards bring forward the discovery of fraud, and do they change what happens inside firms?

<h2>What prompted this</h2> <p>Since October 2025, 16 papers carry a whistleblowing keyword (7 in accounting journals, 4 in management): "whistleblowing" on 10, "whistleblower protection" on 4, "whistleblower rewards" on 1. "Corporate fraud" appears on 15 papers. Fraud detection as an AI topic is far larger, but the human channel through which most frauds are found is barely studied.</p> <h2>Questions</h2> <ul> <li>After a reward programme starts, are frauds discovered earlier in their life and with smaller investor losses?</li> <li>Do firms respond with better internal reporting channels, or with stricter confidentiality agreements?</li> <li>Do the rewards attract tips about the largest frauds, or many small ones?</li> </ul> <h2>Data that could help</h2> <p>Enforcement actions with their discovery dates and tip sources, whistleblower award announcements, and restatements.</p> <h2>Next step</h2> <p>Compare the time between fraud start and discovery for enforcement cases before and after the SEC's reward programme, using disclosed case timelines.</p> https://scholar4ai.com/ideas/where-chinese-firms-list-now 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Initial public offerings appear on 48 recent papers, but only 7 of 91 IPO papers carry a listing-venue, cross-listing, Hong Kong, STAR Market or delisting keyword. Chinese firms have moved from New York to Hong Kong and Shanghai's STAR Market, and some have delisted from the US. Does the venue change who invests, how firms are valued and how they behave?

<h2>What prompted this</h2> <p>Since October 2025, 91 papers carry an IPO keyword (51 in finance journals), with "initial public offerings" on 48 and "IPO underpricing" on 19. Only 7 of the 91 also carry a listing-venue, cross-listing, Hong Kong, STAR Market, delisting or overseas-listing keyword; "STAR Market" appears on 8 papers in all.</p> <h2>Questions</h2> <ul> <li>Are firms that list in Hong Kong rather than the US valued lower, and does the difference reflect investor base, disclosure or politics?</li> <li>What happened to the firms that delisted from US exchanges: did they relist elsewhere, go private or shrink?</li> <li>Does the STAR Market's registration-based system produce IPO pricing closer to that of Hong Kong or of the old approval system?</li> </ul> <h2>Data that could help</h2> <p>IPO and listing records across exchanges, delisting announcements, valuation multiples and ownership data (CSMAR, Wind, Bloomberg).</p> <h2>Next step</h2> <p>Build a list of Chinese companies that listed abroad or delisted since 2019 and compare their valuation multiples with matched firms on domestic exchanges.</p> https://scholar4ai.com/ideas/when-one-customer-dominates 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Supply chain concentration and supply chain relationships are among the fastest-growing keywords across the tracked journals. Finance papers treat a dominant customer as risk; management papers treat it as dependence, through resource dependence theory. The two rarely meet. Does depending on one big customer make suppliers more innovative or more fragile?

<h2>What prompted this</h2> <p>Across all journals, "supply chain concentration" had 18 papers in July–September 2026 against about 7 expected and "supply chain relationships" 17 against 5; in management, "resource dependence theory" grew too (10 against 5); see the <a href="/trends">trends page</a>. Since October 2025, 75 papers carry a customer-concentration or supply-chain-concentration keyword (43 in finance journals, 15 in economics, 10 in management), and only 3 of them carry the resource-dependence keyword.</p> <h2>Questions</h2> <ul> <li>Do suppliers with a dominant customer invest more in relationship-specific innovation, and does the customer share the gains?</li> <li>What happens to them when the customer leaves, is acquired or moves production abroad?</li> <li>Which responses reduce the risk: long-term contracts, cross-shareholdings or diversification?</li> </ul> <h2>Data that could help</h2> <p>Customer disclosures in filings (Compustat segment data; the top-five customer disclosures of Chinese listed firms in CSMAR), patents and supply-chain links.</p> <h2>Next step</h2> <p>Use the top-five customer disclosures of Chinese listed firms to find suppliers that lost a major customer, and compare their patenting and sales before and after.</p> https://scholar4ai.com/ideas/what-chinas-green-patents-invent 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Green innovation is one of the largest topics in the finance and economics journals, much of it counting the green patents of Chinese firms. Only 10 of 641 recent papers on it carry a patent-quality, citation or novelty keyword. Is the boom in green patents a boom in green technology?

<h2>What prompted this</h2> <p>Since October 2025, 641 papers carry a green-innovation or green-patent keyword (271 in finance journals, 215 in economics, 134 in management). Only 10 of them carry a patent-quality, citation, breakthrough, novelty or strategic-patenting keyword. "Green patents" alone appears on 49 papers.</p> <h2>Questions</h2> <ul> <li>How many green patents of Chinese listed firms are utility models rather than inventions, and how many are granted, cited or renewed?</li> <li>Do subsidies and policies that reward patent counts raise counts more than quality?</li> <li>Do the headline findings of the green-innovation literature hold when only granted, cited inventions count?</li> </ul> <h2>Data that could help</h2> <p>Patent records with type, grant status, citations and renewals (CNIPA, PATSTAT), green technology classes (the Y02 tags), and firm data (CSMAR).</p> <h2>Next step</h2> <p>For one year of green patents filed by Chinese listed firms, report the share of invention patents, the grant rate and citations, and compare them with green patents filed elsewhere.</p> https://scholar4ai.com/ideas/wealth-taxes-return 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Wealth and inheritance taxes appear on 19 recent papers in the economics journals, with 8 in the last quarter against about 4 per quarter before. Several countries are debating or reintroducing taxes on large fortunes. What did the earlier wealth taxes collect, how much wealth left, and what do the new designs change?

<h2>What prompted this</h2> <p>Since October 2025, 19 papers carry a wealth-tax, inheritance-tax or estate-tax keyword, all in economics journals: "wealth tax" on 12, "inheritance tax" on 4. Eight of them were published in July–September 2026 against about 4 per quarter earlier. "Wealth inequality" appears on 41 papers.</p> <h2>Questions</h2> <ul> <li>In countries that kept or reintroduced wealth taxes (Switzerland, Spain, Norway, Colombia), how much did revenue and reported wealth respond, and how much of the response was avoidance rather than emigration?</li> <li>Do minimum taxes on very high incomes, as proposed in several countries, raise more than wealth taxes with the same rate?</li> <li>Do inheritance tax changes alter gifts and saving years before death?</li> </ul> <h2>Data that could help</h2> <p>Tax records with wealth and migration information (Scandinavian registers, Spanish regional data), rich lists, and the dates of tax changes.</p> <h2>Next step</h2> <p>Compare reported wealth and migration of top wealth holders in Norway around its 2022 wealth tax increase with those in Sweden, which has no wealth tax.</p> https://scholar4ai.com/ideas/weak-supervision-for-financial-documents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Pseudo-labeling, label noise and weakly supervised learning appear on nearly 500 recent papers, almost all in AI journals, and only 7 of them concern financial, legal or business documents. Banks and auditors have millions of unlabelled contracts, invoices and filings and few labelled ones. Can weak supervision from rules and models replace hand labelling, and what errors does it introduce?

<h2>What prompted this</h2> <p>Since October 2025, 489 papers carry a weak-supervision, pseudo-label, noisy-label or label-noise keyword (487 in AI journals): "pseudo-labeling" on 215, "label noise" on 99, "weakly supervised learning" on 62, "pseudo-label refinement" on 36. Only 7 of the 489 also carry a finance, document, text-classification, legal, contract, invoice, economics, accounting or news keyword.</p> <h2>Questions</h2> <ul> <li>For tasks such as classifying contract clauses, flagging unusual invoices and tagging risk factors in filings, how close does weak supervision from rules and large models get to hand-labelled accuracy, at what cost?</li> <li>Are the errors of weakly supervised models concentrated in the rare cases that matter most (fraud, non-standard clauses)?</li> <li>Do language models used as labellers reproduce their own biases in the labels, and how can that be audited?</li> </ul> <h2>Data that could help</h2> <p>Public financial document datasets with partial labels (contract clause datasets, SEC filings), and a partner firm's documents with a small gold-labelled set.</p> <h2>Next step</h2> <p>Label a contract clause dataset with rules and a language model, train on the weak labels, and compare accuracy on a gold set with a model trained on the same number of hand labels.</p> https://scholar4ai.com/ideas/warehouse-robots-and-workers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Warehouse automation, operations and order picking appear on 63 recent papers in the operations research and AI journals, but only 10 of them carry a labour, worker, human, injury or wage keyword. Warehouses are where robots meet the most people. Do robots raise pace and injuries for the remaining workers, and do wages follow productivity?

<h2>What prompted this</h2> <p>Since October 2025, 63 papers carry a warehouse, order-picking, fulfilment-centre or distribution-centre keyword (43 in operations research journals, 18 in AI): "warehouse automation" on 13, "warehouse operations" on 10, "order picking" on 9, "distribution centers" on 8. Only 10 of them also carry a labour, worker, human, ergonomics, injury, wage or employment keyword, most of those about human-robot interaction.</p> <h2>Questions</h2> <ul> <li>After a warehouse installs mobile robots, do picks per worker-hour rise, and do injury rates rise with them, as some public injury data suggest?</li> <li>Do wages at automated warehouses rise with productivity, or do the jobs that remain pay the same for more pace?</li> <li>Do robot installations reduce total warehouse employment in a region, or shift it to larger sites?</li> </ul> <h2>Data that could help</h2> <p>Facility-level injury filings (US OSHA data), warehouse employment and wages by establishment, and robot installation announcements.</p> <h2>Next step</h2> <p>Match facility-level injury filings to warehouse automation announcements and compare injury rates before and after installation with non-automated sites.</p> https://scholar4ai.com/ideas/voices-in-earnings-calls 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Earnings calls are a staple of finance and accounting research, read almost always as transcripts. In the AI journals, multimodal fusion is one of the biggest topics. Do voice and video carry information about firms that the words miss?

<h2>What prompted this</h2> <p>Since October 2025, 50 papers in the archive carry an earnings-call keyword (20 in finance journals, 17 in accounting). None of them carries a multimodal, audio, voice, video or facial keyword. Meanwhile "multimodal fusion" appears on 502 papers, 495 of them in AI journals, and is the 11th most frequent keyword of the last 30 days. Earnings call transcripts are also among the most-mentioned data sources in finance on the <a href="/rankings/datasets">dataset ranking</a>.</p> <h2>Questions</h2> <ul> <li>Do executives' vocal cues (pitch, pauses, speech rate) predict earnings surprises, restatements or crash risk beyond what they say?</li> <li>Does the market price these cues during the call, or only later?</li> <li>Do managers learn to control them once analysts and machines listen?</li> </ul> <h2>Data that could help</h2> <ul> <li>Call recordings, which many firms post on their investor-relations sites, with timestamped transcripts.</li> <li>Speech-emotion models pretrained on public corpora such as IEMOCAP or RAVDESS, and intraday prices.</li> <li>A text-only baseline such as the Loughran-McDonald tone measure.</li> </ul> <p>Voice analysis touches on privacy, so the study should use public recordings and report firm-level results only.</p> <h2>Next step</h2> <p>Collect audio for one quarter of S&amp;P 500 calls and test whether a pretrained speech-emotion score adds anything to a text-only tone measure.</p> https://scholar4ai.com/ideas/virtual-power-plants 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Demand response, distributed energy resources, virtual power plants and peer-to-peer energy trading appear on 94 recent papers, mostly in economics and AI, and the topic is fading, with 17 in the last quarter against about 23 per quarter before. Aggregators now bid household batteries and heat pumps into power markets. How much flexibility do households really deliver, and what do they get for it?

<h2>What prompted this</h2> <p>Since October 2025, 94 papers carry a virtual-power-plant, demand-response, microgrid, distributed-energy-resource or peer-to-peer-energy keyword (51 in economics journals, 32 in AI, 9 in operations research): "demand response" on 41, "distributed energy resources" on 11, "virtual power plants" on 10, "peer-to-peer energy trading" on 8, "microgrids" on 7. Seventeen were published in July–September 2026 against about 23 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When an aggregator calls on household batteries and heat pumps, what share of the contracted flexibility is actually delivered, and does it decline as households override the controls?</li> <li>What do households earn per year, and is it enough to keep them enrolled after the sign-up bonus?</li> <li>Do virtual power plants reduce the need for peak capacity, or mostly shift payments from utilities to aggregators?</li> </ul> <h2>Data that could help</h2> <p>Aggregator dispatch and delivery records, household enrolment and churn data, and capacity market results.</p> <h2>Next step</h2> <p>Obtain dispatch and delivered-flexibility data from one virtual power plant programme and measure the delivery ratio and household churn over two years.</p> https://scholar4ai.com/ideas/virtual-influencers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Virtual influencers, avatars and virtual humans appear on 33 recent papers in the management and AI journals, with 9 in the last quarter against about 7 per quarter before. Brands now run computer-generated personalities with millions of followers. Do they sell, do followers care that no one is there, and what should disclosure rules require?

<h2>What prompted this</h2> <p>Since October 2025, 33 papers carry a virtual-influencer, virtual-human, digital-human, avatar or virtual-idol keyword (17 in management journals, 11 in AI): "virtual influencers" on 12, "avatars" on 5, "virtual humans" on 4, "virtual idols" on 2. The quarter to September 2026 had 9 such papers against about 7 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In matched campaigns, do virtual influencers deliver the same engagement and sales as human influencers at lower cost and risk?</li> <li>Does disclosure that an influencer is synthetic reduce trust, and does the effect fade as audiences get used to them?</li> <li>Who is responsible when a virtual influencer makes a false claim or endorses a product that harms someone?</li> </ul> <h2>Data that could help</h2> <p>Campaign performance data from brands and agencies, follower surveys, and platform disclosure policies.</p> <h2>Next step</h2> <p>Compare engagement and conversion for matched product promotions by virtual and human influencers on one platform, with and without synthetic-disclosure labels.</p> https://scholar4ai.com/ideas/venture-capital-in-the-ai-boom 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Venture capital is one of the fastest-growing keywords in the finance journals, and startups are a big topic in management and economics, but only 18 of these papers touch artificial intelligence. As more venture money goes to AI companies, are other startups crowded out, or does AI lower the cost of starting up?

<h2>What prompted this</h2> <p>In the finance journals, "venture capital" had 13 papers in July–September 2026 against about 5 expected from its earlier share (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 250 papers carry a venture-capital or startup keyword (106 in management journals, 80 in economics, 56 in finance). Only 18 of them also carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>As AI deals take a growing share of venture funding, do non-AI startups raise less, later or at lower valuations?</li> <li>Do AI tools let founders build more with less capital, so that seed rounds shrink while later rounds grow?</li> <li>Do AI-heavy funds earn returns that differ from those of earlier waves such as the internet, mobile and clean tech?</li> </ul> <h2>Data that could help</h2> <p>Deal data from PitchBook, Crunchbase, Preqin or VentureXpert (the M&amp;A and private capital category of the <a href="/rankings/datasets">dataset ranking</a>), with AI exposure classified from company descriptions.</p> <h2>Next step</h2> <p>Classify startups as AI or non-AI from their descriptions and compare round sizes and the time between rounds before and after 2023.</p> https://scholar4ai.com/ideas/vaccine-allocation-for-the-next-pandemic 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Vaccine distribution, allocation, uptake and hesitancy appear on 55 recent papers, and 18 of them, mostly in operations research, carry a supply-chain or allocation keyword. The COVID-19 roll-out produced data on who got vaccinated when, and at what cost. What did the allocation models get right, and what would they change for next time?

<h2>What prompted this</h2> <p>Since October 2025, 55 papers carry a vaccine, vaccination or immunization keyword (25 in economics journals, 16 in operations research, 10 in AI): "vaccine distribution" on 6, "COVID-19 vaccination" on 5, "childhood immunization" on 4, "vaccine uptake" on 4, "vaccine hesitancy" on 4, "vaccine allocation" on 3. Eighteen of the 55 also carry a supply-chain, distribution, allocation, cold-chain, logistics, inventory or optimization keyword.</p> <h2>Questions</h2> <ul> <li>Comparing the 2021 allocation rules of different countries (by age, by occupation, by region) with their outcomes, which rules saved the most lives per dose, and did the optimization models of the time predict that?</li> <li>How much did hesitancy and no-shows, which the models ignored, change the optimal allocation?</li> <li>Would an allocation that accounts for distribution capacity in poor countries have changed global outcomes more than one that optimizes within rich countries?</li> </ul> <h2>Data that could help</h2> <p>Country-level vaccination timelines and mortality data, allocation rules by country, wastage and no-show records, and the models published in 2020–2021.</p> <h2>Next step</h2> <p>Re-run three published allocation models with the actual uptake and delivery data from 2021 and compare their recommendations with what countries did.</p> https://scholar4ai.com/ideas/urban-air-mobility 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Urban air mobility, eVTOL aircraft and vertiports appear on 24 recent papers in the operations research and AI journals, with 10 in the last quarter against about 4 per quarter before. The first commercial air taxi services are being certified. Where do the routing and scheduling models get their demand numbers, and would anyone pay the fares the models assume?

<h2>What prompted this</h2> <p>Since October 2025, 24 papers carry an urban-air-mobility, eVTOL, air-taxi or vertiport keyword (20 in operations research journals, 4 in AI): "urban air mobility" on 20, "eVTOL" on 5, "eVTOL aircraft" on 4, "vertiports" on 2. The quarter to September 2026 had 10 such papers against about 4 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>What demand assumptions do the vertiport location and fleet scheduling models use, and how do they compare with stated-preference surveys and the fares the operators have announced?</li> <li>At realistic battery, pilot and vertiport costs, which city pairs break even, and how many passengers per day do they need?</li> <li>Do the models account for noise and airspace constraints that have blocked earlier urban aviation?</li> </ul> <h2>Data that could help</h2> <p>Operators' published fares and routes, stated-preference surveys on willingness to pay, helicopter service data as a benchmark, and cost estimates from certification filings.</p> <h2>Next step</h2> <p>Collect the demand and cost assumptions from the recent urban air mobility optimization papers and compare them with the fares and costs disclosed by operators preparing to launch.</p> https://scholar4ai.com/ideas/university-spin-offs-in-the-ai-era 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

University spin-offs and university entrepreneurship are among the fastest-growing keywords in the management journals, while AI research shifts from universities to companies. Only 2 of 37 recent papers on academic entrepreneurship mention AI. Do universities still turn AI research into startups, or does industry absorb the talent first?

<h2>What prompted this</h2> <p>In the management journals, "university spin-offs" had 6 papers in July–September 2026 against about 1 expected, "university entrepreneurship" 5 against 1, and "technology commercialization" 5 against 2 (see the <a href="/categories/management">management dashboard</a>). Since October 2025, 37 papers carry a spin-off, university-entrepreneurship or technology-commercialization keyword (29 in management journals); only 2 of them carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>Are AI spin-offs from universities fewer, later or smaller than spin-offs in other fields, given how fast AI researchers move to industry?</li> <li>Do universities with more computing resources keep more AI faculty and spin off more companies?</li> <li>Do technology-transfer offices treat AI, which often comes as open code and models, differently from patents?</li> </ul> <h2>Data that could help</h2> <p>Startup databases (Crunchbase, PitchBook) linked to universities, faculty moves to industry, and university patents and software licences.</p> <h2>Next step</h2> <p>Link university-founded startups in Crunchbase to their fields and count AI spin-offs per AI faculty member, by university, for 2015–2025.</p> https://scholar4ai.com/ideas/treasury-market-liquidity-and-the-basis-trade 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Treasury yields and bonds appear on 39 recent papers in the finance and economics journals, but only 2 of them carry a liquidity, dealer, hedge-fund or market-functioning keyword. Hedge funds' leveraged basis trades and dealers' balance sheet limits have been blamed for Treasury market stress. How fragile is the market now, and would central clearing help?

<h2>What prompted this</h2> <p>Since October 2025, 39 papers carry a Treasury keyword (28 in finance journals, 9 in economics): "Treasury yields" on 14, "Treasury bonds" on 4, "treasury bills" on 3. Only 2 of the 39 also carry a liquidity, basis-trade, hedge-fund, dealer, market-functioning or auction keyword, while "market liquidity" appears on 51 papers about other markets.</p> <h2>Questions</h2> <ul> <li>How large is the cash-futures basis trade, and how fast would it unwind if repo rates or futures margins jumped?</li> <li>Do dealers' balance sheet constraints explain why Treasury liquidity worsens in stress, and did the 2023 leverage-ratio relief change that?</li> <li>Would mandatory central clearing, now being phased in, reduce the stress or move it elsewhere?</li> </ul> <h2>Data that could help</h2> <p>Treasury transaction data (TRACE for Treasuries), futures positioning reports, repo market statistics and dealer inventories from the Federal Reserve.</p> <h2>Next step</h2> <p>Estimate the size of the basis trade from futures positioning and repo data for 2019–2025 and relate it to liquidity measures around stress episodes.</p> https://scholar4ai.com/ideas/transfer-learning-for-credit-across-countries 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Transfer learning, domain adaptation and domain generalization appear on nearly 700 recent papers, almost all in AI journals, and only 10 of them concern credit, lending, banking or markets. Lenders entering new countries, and credit bureaus in countries with little data, want to reuse models trained elsewhere. How much transfers, and what breaks?

<h2>What prompted this</h2> <p>Since October 2025, 679 papers carry a transfer-learning, domain-adaptation or domain-generalization keyword (664 in AI journals): "domain adaptation" on 182, "domain generalization" on 178, "transfer learning" on 162, "unsupervised domain adaptation" on 76. Only 10 of the 679 also carry a credit, loan, bank, finance, default, microfinance, emerging-market, fraud or stock keyword.</p> <h2>Questions</h2> <ul> <li>Does a credit model trained in one country, adapted with a few thousand local loans, beat a model trained from scratch on the local data, and how does the answer depend on how different the countries' borrowers and institutions are?</li> <li>Which features transfer (payment behaviour, mobile phone data) and which do not (income, employment categories)?</li> <li>Does transferred scoring import the source country's biases against groups that do not exist in the target?</li> </ul> <h2>Data that could help</h2> <p>Loan datasets from several countries (public lending datasets, microfinance data, partner lenders), and borrower features that are available in both.</p> <h2>Next step</h2> <p>Adapt a credit model trained on one public lending dataset to another country's dataset with small local samples and compare it with local-only training at each sample size.</p> https://scholar4ai.com/ideas/the-missing-economics-of-medical-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Clinical decision support appears on 94 recent papers and medical image segmentation on 188, among more than 1,300 medical AI papers in the AI journals, but only 65 of all medical papers carry a cost, reimbursement, deployment or workflow keyword, and cost-effectiveness appears on 27 papers in the whole archive. Which medical AI tools pay for themselves, and for whom?

<h2>What prompted this</h2> <p>Since October 2025, 1,607 papers carry a medical, clinical, diagnosis, radiology, healthcare or patient keyword (1,320 in AI journals): "medical image segmentation" on 188, "clinical decision support" on 94, "medical image classification" on 73, "computer-aided diagnosis" on 46. Only 65 of the 1,607 also carry a cost-effectiveness, reimbursement, economic-evaluation, deployment, implementation, adoption, cost-saving, workflow or clinical-practice keyword; "cost-effectiveness" appears on 27 papers in the archive and "clinical workflow" on 7.</p> <h2>Questions</h2> <ul> <li>For the medical AI tools that have been reimbursed (for example stroke detection and diabetic retinopathy screening), what did the cost per quality-adjusted life year turn out to be in practice?</li> <li>Do tools that save clinician time reduce costs, or do clinicians fill the time with more cases and the hospital bills more?</li> <li>Which of the hundreds of published segmentation and diagnosis models have reached any clinic, and what separated them?</li> </ul> <h2>Data that could help</h2> <p>Reimbursement decisions and health technology assessments, hospital adoption records, claims data around adoption, and the literature's own deployment reports.</p> <h2>Next step</h2> <p>Trace the twenty most-cited medical AI models of 2020–2022 to any clinical deployment and record the economic evidence that accompanied each.</p> https://scholar4ai.com/ideas/the-language-divide-in-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Low-resource languages are a fast-growing topic in the AI journals, and the digital divide is rising in economics. Language models work worse in languages with little training text, which are spoken mostly in poorer countries. Does that slow AI adoption, and its economic gains, where those languages are spoken?

<h2>What prompted this</h2> <p>In the AI journals, "low-resource languages" had 16 papers in July–September 2026 against about 10 expected; in economics, "digital divide" grew too (11 against 4); see the <a href="/trends">trends page</a>. Since October 2025, all 46 papers with the low-resource-languages keyword are in AI journals, while "digital divide" appears on 70 papers (25 in economics, 23 in finance, 11 in management, 9 in AI). Only 8 of the digital-divide papers carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>How much worse do common language models perform in each language, and does that predict AI use by country once income and internet access are held fixed?</li> <li>When a model gets better at a language, do firms and workers who use that language adopt AI faster?</li> <li>Do translation layers close the gap, or do they cost accuracy where it matters most, such as in health, law and finance?</li> </ul> <h2>Data that could help</h2> <p>Multilingual benchmark scores by model and language, AI usage by country (app downloads, search interest), and data on the languages spoken in each country.</p> <h2>Next step</h2> <p>Match multilingual benchmark scores to countries by main language and correlate them with AI app adoption, controlling for income and internet use.</p> https://scholar4ai.com/ideas/the-four-day-week 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Working hours appear on 41 recent papers in the economics journals and working time reduction on 4, but the four-day week, now being trialled by hundreds of firms and by some governments, appears under no keyword of its own. Do firms that cut hours without cutting pay keep output, and who gains the hours?

<h2>What prompted this</h2> <p>Since October 2025, 52 papers carry a working-hours, working-time or work-week keyword (45 in economics journals): "working hours" on 41, "working time reduction" on 4, "flexible working hours" on 3. No keyword in the archive names the four-day week, although trials in the UK, Iceland, Germany, Japan and Portugal have reported results.</p> <h2>Questions</h2> <ul> <li>In firm trials, did output per week hold up because of fewer meetings and less idle time, or because workers compressed the same hours?</li> <li>Did turnover, sick leave and hiring costs fall by enough to pay for the fifth day?</li> <li>Which jobs cannot compress, and would a general shift widen the gap between office and shift workers?</li> </ul> <h2>Data that could help</h2> <p>Trial data from the pilot organizers, firm productivity and turnover records, and time-use surveys of participants.</p> <h2>Next step</h2> <p>Collect the published results of the national and firm-level trials since 2021 and compare productivity and retention outcomes by sector.</p> https://scholar4ai.com/ideas/the-compute-bill-of-ai-papers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Inference latency and resource-constrained deployment are among the fastest-growing keywords in the AI journals, yet only 15 of 226 such papers carry an energy, carbon or green-AI keyword. How much computing do the methods in AI journals cost to train and run, and would reporting it change which methods win?

<h2>What prompted this</h2> <p>In the AI journals, "inference latency" had 12 papers in July–September 2026 against about 5 expected and "resource-constrained deployment" 10 against 4 (see the <a href="/categories/ai">AI dashboard</a>). Since October 2025, 226 papers carry an inference-latency, resource-constrained or edge-deployment keyword (199 in AI journals). Only 15 of them carry an energy, carbon, green-AI or power-consumption keyword, and "green AI" itself appears on 6.</p> <h2>Questions</h2> <ul> <li>What share of recent AI-journal papers report training compute, inference time or energy, and do they report them for the baselines too?</li> <li>If results were compared per unit of computing, how often would the winner change?</li> <li>Would a short reporting checklist (hardware, hours, energy) be adopted, as code release was?</li> </ul> <h2>Possible approach</h2> <p>Code a sample of recent papers for what they report about computing, and re-rank the reported results by accuracy per GPU-hour where both are given.</p> <h2>Next step</h2> <p>Code 100 recent papers from the AI journals for whether they report hardware, training time, inference time and energy use.</p> https://scholar4ai.com/ideas/tariff-pass-through-in-the-2025-trade-war 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tariffs and the trade war appear on 184 recent papers, mostly in the economics journals, but only 47 carry a price, pass-through, incidence or inflation keyword, and "tariff pass-through" on 3. The 2025 tariff rounds were larger and broader than those of 2018. Did foreign exporters cut prices this time, or did consumers pay again?

<h2>What prompted this</h2> <p>Since October 2025, 184 papers carry a tariff or trade-war keyword (119 in economics journals, 43 in finance): "US-China trade war" on 48, "tariffs" on 39, "import tariffs" on 18, "trade war" on 12. Only 47 of them also carry a pass-through, consumer-price, price, incidence, inflation or welfare keyword, and "tariff pass-through" appears on 3.</p> <h2>Questions</h2> <ul> <li>Did import prices at the border fall after the 2025 tariffs, as they mostly did not in 2018, and did the answer differ by exporter country and product?</li> <li>How much of the tariff reached retail prices, and how quickly?</li> <li>Did importers reroute through third countries, and did that limit the pass-through?</li> </ul> <h2>Data that could help</h2> <p>Monthly import price indices and customs data by product and origin, retail scanner prices, and the tariff schedules with their effective dates.</p> <h2>Next step</h2> <p>Estimate border price changes by product for the 2025 tariff rounds using monthly customs unit values and compare them with the 2018 estimates.</p> https://scholar4ai.com/ideas/tabular-foundation-models 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tabular data appears on 53 recent papers and tabular classification on 13, almost all in AI journals, and 16 tabular papers involve pretrained or language models. Most business data is tables: loans, transactions, patients, sales. Do pretrained tabular models beat gradient boosting on real business tables, and when?

<h2>What prompted this</h2> <p>Since October 2025, 77 papers carry a tabular keyword (73 in AI journals): "tabular data" on 53, "tabular data classification" on 13, "tabular data generation" on 4. Sixteen of them also carry a foundation-model, pretraining, transformer, language-model, in-context or zero-shot keyword. Thirteen were published in July–September 2026 against about 20 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>On credit, fraud, churn and insurance tables with their real sizes and missingness, do pretrained tabular models beat tuned gradient boosting, and by how much per unit of computing?</li> <li>Do the gains hold under the temporal distribution shift typical of business data, or only on random splits?</li> <li>Can a model pretrained on public tables transfer to a bank's private tables without leaking anything from the public ones?</li> </ul> <h2>Data that could help</h2> <p>Public business tables with time stamps (lending, fraud, insurance datasets), and private tables from a partner firm with temporal splits.</p> <h2>Next step</h2> <p>Benchmark two pretrained tabular models against tuned gradient boosting on five public business datasets using temporal rather than random splits.</p> https://scholar4ai.com/ideas/synthetic-financial-data-from-diffusion-models 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Diffusion models are among the largest topics in the AI journals, and synthetic data generation is growing with them. Yet only 6 of 119 recent synthetic-data papers carry a finance keyword. Could synthetic transaction and loan data let banks share data for research without exposing customers, and how close is it to the real thing?

<h2>What prompted this</h2> <p>Since October 2025, "diffusion models" appears on 361 papers (357 in AI journals), and 119 papers carry a synthetic-data keyword (114 in AI journals); only 6 of them carry a finance, stock, bank, trading, credit or fraud keyword. "Differential privacy" appears on 72 papers.</p> <h2>Questions</h2> <ul> <li>Do models trained on synthetic loan or transaction data perform on real data as well as models trained on the real thing?</li> <li>Do synthetic datasets keep the rare events that matter most, such as defaults, fraud and crashes?</li> <li>How well do they protect privacy against attacks that try to find real customers in the synthetic data?</li> </ul> <h2>Data that could help</h2> <p>Public financial datasets (LendingClub loans, credit card transactions, stock returns) to train and test the generators, and standard privacy-attack benchmarks.</p> <h2>Next step</h2> <p>Train a tabular diffusion model on LendingClub loans and compare default models trained on the synthetic data with models trained on the real data.</p> https://scholar4ai.com/ideas/sustainability-linked-loans 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Green credit and green loans appear on 58 recent papers, most of them about China's green credit policy, while sustainability-linked loans, whose interest rate moves with the borrower's ESG targets, appear on only 3. Are the targets ambitious, are they met, and does the rate step actually change behaviour?

<h2>What prompted this</h2> <p>Since October 2025, 58 papers carry a green-credit, green-loan or sustainability-linked keyword (34 in finance journals, 17 in economics): "green credit" on 27, "green loans" on 15, "green credit policy" on 12, "sustainability-linked loans" on 3 and "sustainability-linked bonds" on 2. The quarter to September 2026 had 15 such papers against about 13 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Are the key performance targets in sustainability-linked loans set below what the borrower was already on track to achieve?</li> <li>When targets are missed, is the rate step large enough to matter, and do lenders waive it?</li> <li>Do borrowers with these loans cut emissions faster than similar borrowers with ordinary loans?</li> </ul> <h2>Data that could help</h2> <p>Loan terms from syndicated loan data (DealScan) and bond prospectuses, borrowers' reported emissions and targets, and later target outcomes.</p> <h2>Next step</h2> <p>Collect the targets and rate steps of 200 sustainability-linked loans and compare the targets with the borrowers' pre-loan emission trends.</p> https://scholar4ai.com/ideas/surge-pricing-and-fairness 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Dynamic pricing appears on 51 recent papers and price discrimination on 28, mostly in operations research and economics; 29 of 91 such papers touch consumer welfare, but price fairness appears on 3. Retailers and venues have faced backlash over surge prices, and some have retreated. Where is the line that consumers will not accept, and what does it cost firms to cross it?

<h2>What prompted this</h2> <p>Since October 2025, 91 papers carry a dynamic-pricing, surge-pricing, personalized-pricing, price-discrimination or algorithmic-pricing keyword (36 in operations research journals, 34 in economics, 14 in finance, 5 in management): "dynamic pricing" on 51, "price discrimination" on 28, "personalized pricing" on 6, "surge pricing" on 4. Twenty-nine of them carry a consumer, welfare, fairness or trust keyword, but "price fairness" appears on only 3.</p> <h2>Questions</h2> <ul> <li>After a visible pricing backlash (concert tickets, fast food surge pricing, ride-hailing in emergencies), do sales and brand sentiment fall for long, and do rivals gain?</li> <li>Which forms of variation do consumers accept (time of day, demand) and which do they reject (personal data, emergencies)?</li> <li>Do laws against price gouging and personalized pricing change what firms do, or only what they call it?</li> </ul> <h2>Data that could help</h2> <p>Price and sales data from firms that introduced dynamic pricing, social media sentiment around backlash episodes, and survey experiments on acceptance.</p> <h2>Next step</h2> <p>Measure sales and search interest for a retailer before and after a publicized surge-pricing episode, and compare with a competitor.</p> https://scholar4ai.com/ideas/supply-chain-finance-after-greensill 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Supply chain finance appears on 101 recent papers, mostly in the finance and operations research journals, but only 32 of them carry a risk, default, disclosure, accounting or fintech keyword, and reverse factoring on one. Since the Greensill collapse, firms must disclose these programmes. Who bears the risk, and does disclosure change pricing?

<h2>What prompted this</h2> <p>Since October 2025, "supply chain finance" appears on 101 papers (59 in finance journals, 25 in operations research), and "reverse factoring" on 1. Only 32 of the 101 also carry a risk, default, bankruptcy, disclosure, accounting, fintech, blockchain or platform keyword, and most of those are about blockchain and credit risk models rather than about the programmes' failure modes.</p> <h2>Questions</h2> <ul> <li>Since the 2023 disclosure rules for supplier finance programmes, did firms that disclose large programmes pay more for debt, or see their payables reclassified?</li> <li>When a buyer's finance programme stops, how fast do its suppliers run into distress?</li> <li>Do fintech-run programmes price supplier risk, or the buyer's risk alone?</li> </ul> <h2>Data that could help</h2> <p>The new supplier-finance disclosures in annual reports, supplier-customer links, suppliers' payment terms and distress events.</p> <h2>Next step</h2> <p>Hand-collect the supplier finance disclosures of large US and European firms for 2023–2025 and relate programme size to suppliers' days of receivables.</p> https://scholar4ai.com/ideas/supply-chain-due-diligence-laws 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Child labour, forced labour, modern slavery and human rights due diligence appear on 41 recent papers, with 13 in the last quarter against about 9 per quarter before, and 12 of them carry a supply-chain or supplier keyword. Germany's supply chain law, the EU directive and import bans on forced-labour goods now require firms to check suppliers. Do the checks find anything, and do suppliers change?

<h2>What prompted this</h2> <p>Since October 2025, 41 papers carry a due-diligence, modern-slavery, forced-labour, human-rights or child-labour keyword (19 in economics journals, 6 in management, 6 in AI, 5 in finance): "child labor" on 13, "human rights" on 7, "modern slavery" on 6, "forced labor" on 5, "human rights due diligence" on 3. The quarter to September 2026 had 13 such papers against about 9 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Do social audits required by the laws detect forced labour, or do they certify what suppliers show on audit day?</li> <li>After import bans on goods from regions with forced labour, did trade reroute through third countries, and did the targeted goods' prices rise?</li> <li>Do firms covered by the laws cut suppliers in high-risk countries, and what happens to the workers there?</li> </ul> <h2>Data that could help</h2> <p>Firms' due diligence reports, customs detention records under import bans, trade data by product and origin, and audit reports where available.</p> <h2>Next step</h2> <p>Compare US imports of cotton goods and solar components by origin before and after the 2022 forced-labour import ban, and look for rerouting.</p> https://scholar4ai.com/ideas/subscription-churn 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Customer retention and churn appear on 33 recent papers and subscription pricing and models on 10, across the AI, operations research and finance journals. Subscriptions now cover software, cars, media and razors, and regulators are forcing easy cancellation. How much of subscription revenue depends on people not cancelling, and what happens when they can?

<h2>What prompted this</h2> <p>Since October 2025, 52 papers carry a subscription, churn or customer-retention keyword, spread across AI (13), operations research (12), finance (9), economics (9) and management (7): "customer retention" on 18, "customer churn" on 8, "customer churn prediction" on 7, "subscription pricing" on 6, "subscription models" on 4. Most churn papers are prediction exercises; the economics of inertia is rarely their subject.</p> <h2>Questions</h2> <ul> <li>When "click to cancel" rules took effect, how much did cancellations rise, and did firms respond with lower prices or more retention offers?</li> <li>How much of subscription revenue comes from subscribers who no longer use the service, by category?</li> <li>Do firms that make cancelling hard lose customers' later purchases and referrals?</li> </ul> <h2>Data that could help</h2> <p>Transaction panels with recurring charges and usage, the dates of cancellation rules by jurisdiction, and firm disclosures of churn rates.</p> <h2>Next step</h2> <p>Use bank transaction data to estimate the share of recurring charges with no matching usage and compare cancellation rates before and after a click-to-cancel rule.</p> https://scholar4ai.com/ideas/student-loan-repayment-restart 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Student loans and student debt appear on only 12 recent papers, none of them in the last quarter, even though US repayments restarted in 2023 after a three-year pause and delinquencies then jumped. Did the pause change borrowers' finances for good, and what happened when payments resumed?

<h2>What prompted this</h2> <p>Since October 2025, 12 papers carry a student-loan or student-debt keyword (7 in economics journals, 3 in finance): "student loans" on 8, "student debt" on 7. None of the 12 was published in July–September 2026, against about 4 per quarter earlier, while household debt as a whole remains an active topic.</p> <h2>Questions</h2> <ul> <li>During the pause, did borrowers pay down other debt, save, or borrow more, and did those who kept paying differ?</li> <li>When payments resumed, how many borrowers fell behind, and did the new income-driven plans and their legal reversals change that?</li> <li>Did the pause and the forgiveness debates change enrolment or borrowing by new students?</li> </ul> <h2>Data that could help</h2> <p>Credit bureau panels with student loan balances and payment status, the Department of Education's repayment statistics, and enrolment data.</p> <h2>Next step</h2> <p>Use a credit bureau panel to follow borrowers' balances and delinquencies on all accounts from 2019 through 2025.</p> https://scholar4ai.com/ideas/stock-connect-ten-years-on 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Foreign investors and capital market liberalization are growing keywords in the finance journals, with 13 papers in July–September 2026 against about 7 per quarter before. More than ten years after Shanghai-Hong Kong Stock Connect opened, has foreign ownership changed how Chinese stocks are priced, and what happened when foreign money left?

<h2>What prompted this</h2> <p>Since October 2025, 36 papers carry a Stock Connect, capital-market-liberalization, capital-account-liberalization, financial-opening or foreign-investor keyword (26 in finance journals). "Foreign investors" appears on 17, "capital account liberalization" on 8, "capital market liberalization" on 7 and the two Stock Connect programmes on 7. The quarter to September 2026 had 13 such papers against about 7 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Are stocks eligible for Stock Connect priced more in line with their fundamentals, with less of the retail-driven noise found in A-shares?</li> <li>When foreign investors sold heavily in 2022–2024, did eligible stocks fall more than ineligible ones, and did domestic institutions step in?</li> <li>Did eligibility change firms' disclosure, dividends or governance?</li> </ul> <h2>Data that could help</h2> <p>Daily northbound and southbound flows, eligibility lists and their changes, A-share and H-share prices, and firm data (CSMAR).</p> <h2>Next step</h2> <p>Compare the return and volatility behaviour of stocks added to and removed from the eligible list around each list revision.</p> https://scholar4ai.com/ideas/state-space-models-for-limit-order-books 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Mamba and other selective state-space models are among the most frequent new architectures in the AI journals, built for long sequences. Limit order books produce exactly such sequences. Yet only 4 of 329 recent Mamba and state-space papers carry a finance keyword, all of them in AI journals. Can these models forecast short-term price moves better, and is the gain tradable after costs?

<h2>What prompted this</h2> <p>Since October 2025, "Mamba" appears on 293 papers, 290 of them in AI journals, and 329 papers carry a Mamba or state-space-model keyword; only 4 of them carry a finance, stock, trading, market or asset keyword. "Limit order book" appears on 13 papers and "high-frequency trading" on 21, mostly in finance journals.</p> <h2>Questions</h2> <ul> <li>On standard limit order book data, do state-space models predict mid-price moves better than convolutional and Transformer models at the same computing cost?</li> <li>Do the gains survive realistic trading costs and latency?</li> <li>Do they capture order-flow patterns that last longer than other models can see?</li> </ul> <h2>Data that could help</h2> <p>Public limit order book benchmarks (for example the FI-2010 dataset or LOBSTER samples) and published baseline models.</p> <h2>Next step</h2> <p>Train a Mamba model on a public limit order book benchmark, and compare its forecasts and simulated trading results with published baselines.</p> https://scholar4ai.com/ideas/stablecoins-and-monetary-policy 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Stablecoins and monetary policy transmission are both among the fastest-growing keywords, but the papers that join them sit only in finance journals. Now that issuers hold large amounts of Treasury bills, does policy pass through stablecoins into crypto credit, and do stablecoin flows move bill yields?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "stablecoins" is among the ten keywords that grew most across all journals: 22 papers in July–September 2026, against about 8 expected from its earlier share. "Monetary policy transmission" grew as well (39 against 24). Since October 2025, 51 papers in the archive carry a stablecoin keyword (42 in finance journals, 9 in economics). Nine of them also carry a monetary-policy, interest-rate or Treasury keyword, and all nine are in finance journals.</p> <h2>Questions</h2> <ul> <li>When policy rates move, do stablecoin issuance and on-chain lending rates follow, the way deposits and loan rates do at banks?</li> <li>In the other direction, do large stablecoin inflows push down short-term Treasury bill yields?</li> <li>Does the pass-through differ between issuers with audited reserves and the rest, and has new stablecoin regulation changed it?</li> </ul> <h2>Data that could help</h2> <ul> <li>Daily supply by issuer from on-chain records and market data (the cryptocurrency category of the <a href="/rankings/datasets">dataset ranking</a>), plus issuers' reserve reports.</li> <li>Monetary policy surprises around central-bank announcements, and Treasury bill yields.</li> <li>A vector autoregression or local projections around the surprises (see the <a href="/rankings/equations">equation ranking</a>).</li> </ul> <h2>Next step</h2> <p>Read the nine finance papers that combine both topics and note which direction of transmission each one tests.</p> https://scholar4ai.com/ideas/sports-scheduling-and-analytics 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sports betting, sports analytics, football and tournament design appear on 120 recent papers across the AI, economics, finance and operations research journals. Leagues sell schedules to broadcasters, players are traded on expected value, and betting markets price every match. Do the optimized schedules and ratings survive the money, and do betting markets correct the analytics?

<h2>What prompted this</h2> <p>Since October 2025, 120 papers carry a sports, football, soccer, basketball, baseball, Olympic, athlete or esports keyword (43 in AI journals, 30 in economics, 25 in finance, 13 in operations research): "sports betting" on 8, "sports analytics" on 7, "football" on 5, "European football leagues" on 4, "tournament design" on 4, "sports scheduling" on 3. The scheduling, analytics and betting strands rarely appear together.</p> <h2>Questions</h2> <ul> <li>When leagues let broadcasters shape the schedule, how much fairness (rest days, travel) do teams lose, and does it show in results?</li> <li>Do player ratings from tracking data predict transfer fees and contracts better than the market does, or has the market caught up?</li> <li>Do betting markets incorporate analytics quickly enough that published models no longer beat the odds?</li> </ul> <h2>Data that could help</h2> <p>League schedules and broadcast contracts, player tracking and transfer data, and historical betting odds.</p> <h2>Next step</h2> <p>Measure the forecasting edge of published match-prediction models against closing betting odds for the last five seasons of two leagues.</p> https://scholar4ai.com/ideas/sports-betting-legalization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sports betting, gambling and lotteries appear on 42 recent papers in the finance and economics journals. Since 2018, most US states have legalized online sports betting, and app-based betting is spreading elsewhere. What has it done to household finances, bankruptcies and the stock trading of the same people?

<h2>What prompted this</h2> <p>Since October 2025, 42 papers carry a sports-betting, gambling, lottery, casino or betting keyword (20 in finance journals, 16 in economics): "sports betting" on 8, "gambling culture" on 5, "lotteries" on 6, "betting odds" on 4. Household debt and consumer credit are large topics, but the state-by-state legalization is an experiment that few of these papers use.</p> <h2>Questions</h2> <ul> <li>After a state legalizes online betting, do credit card balances, overdrafts and bankruptcies rise, and in which households?</li> <li>Does betting substitute for stock trading and lottery play, or add to them?</li> <li>Do tax revenues cover the costs that show up in credit markets and problem-gambling services?</li> </ul> <h2>Data that could help</h2> <p>Credit bureau panels by state, legalization dates, betting handle and revenue reports, and brokerage trading data.</p> <h2>Next step</h2> <p>Use the staggered legalization across US states to estimate changes in credit card delinquencies and bankruptcies in the two years after launch.</p> https://scholar4ai.com/ideas/spare-parts-and-3d-printing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Additive manufacturing and 3D printing appear on 31 recent papers, split among operations research, AI and management, and spare parts management on 2. Printing parts on demand could replace warehouses of slow-moving inventory. Which parts are worth printing, where should the printers be, and do certification rules allow it?

<h2>What prompted this</h2> <p>Since October 2025, 31 papers carry an additive-manufacturing, 3D-printing or spare-parts keyword (14 in operations research journals, 11 in AI, 5 in management): "additive manufacturing" on 13, "3D printing" on 11, "metal additive manufacturing" on 3, "spare parts management" on 2, "spare parts provisioning" on 1. Six were published in July–September 2026 against about 8 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>For a given fleet (aircraft, ships, rail, machinery), which share of spare parts could be printed at acceptable cost and lead time, and how much inventory would that remove?</li> <li>Where should printing capacity sit, centrally or near the fleet, given demand uncertainty and the cost of idle printers?</li> <li>How much do certification and liability rules restrict which safety-relevant parts can be printed, and how are they changing?</li> </ul> <h2>Data that could help</h2> <p>Spare parts demand and inventory records from a fleet operator, printing cost and lead time data by part type, and certification rules.</p> <h2>Next step</h2> <p>Classify the spare parts catalogue of one fleet operator by printability and compute the inventory reduction from printing the eligible slow movers.</p> https://scholar4ai.com/ideas/sovereign-defaults-with-china-as-creditor 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sovereign default and sovereign debt crises appear on 42 recent papers in the economics and finance journals, but only one of them carries a creditor, China or Common Framework keyword. Recent restructurings in Zambia, Ghana and Sri Lanka involved China as a major lender. Do defaults take longer and resolve differently when the creditors are new?

<h2>What prompted this</h2> <p>Since October 2025, 42 papers carry a sovereign-default, sovereign-debt-crisis, debt-restructuring or debt-relief keyword (21 in economics journals, 20 in finance). Only one of them also carries a China, Common Framework, official-creditor, bilateral or creditor keyword. "Sovereign default" appears on 14 papers and "sovereign debt crisis" on 11.</p> <h2>Questions</h2> <ul> <li>Have restructurings since 2020 taken longer, and given smaller haircuts, than earlier ones, and does the share of Chinese lending explain the difference?</li> <li>Do collateralized and confidential loans change how private bondholders are treated?</li> <li>Did the G20 Common Framework shorten the process, or add a step?</li> </ul> <h2>Data that could help</h2> <p>Restructuring timelines and terms (the IMF and academic restructuring databases), creditor composition from AidData's Chinese lending data, and bond prices around key events.</p> <h2>Next step</h2> <p>Compile the timeline of each restructuring since 2020 with its creditor composition and compare durations with restructurings of the 2000s.</p> https://scholar4ai.com/ideas/social-media-and-adolescents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Social media appears on 402 recent papers across the tracked journals, mostly as text data in AI and management research; only 18 carry a mental-health, adolescent, well-being or children keyword. Several countries are banning social media for under-16s. What does the evidence actually show, and what will the bans reveal?

<h2>What prompted this</h2> <p>Since October 2025, 402 papers carry a social-media, smartphone or screen-time keyword (180 in AI journals, 117 in management, 52 in finance). Only 18 of them also carry a mental-health, adolescent, teen, youth, well-being, depression, suicide, children or student keyword; "social media use" appears on 13 papers and "adolescent mental health" on 7.</p> <h2>Questions</h2> <ul> <li>Did adolescent mental health indicators change after Australia's under-16 ban and similar laws, compared with countries without them?</li> <li>Does the effect of use depend on what is used for what (passive scrolling versus messaging friends), and can that be measured without self-reports?</li> <li>Do the bans reduce use, or move it to platforms outside the law?</li> </ul> <h2>Data that could help</h2> <p>Health surveys and clinical records by age and country, app usage data from device studies, and the dates and scope of bans.</p> <h2>Next step</h2> <p>Compare trends in adolescent mental health and app usage in Australia with those in New Zealand and Canada around the ban's start.</p> https://scholar4ai.com/ideas/social-enterprises-and-b-corps 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Social enterprises, social entrepreneurship and hybrid organizations appear on 41 recent papers, mostly in management, and the topic is fading, with 7 papers in the last quarter against about 10 per quarter before. Grant funding and impact investment have shrunk. Which social enterprises survive without them, and do certified B Corporations behave differently from the rest?

<h2>What prompted this</h2> <p>Since October 2025, 41 papers carry a social-enterprise, social-entrepreneurship, hybrid-organization, benefit-corporation or B Corp keyword (31 in management journals, 9 in economics): "social enterprises" on 19, "social entrepreneurship" on 16, "hybrid organizations" on 5, "benefit corporations" on 2, "B Corp certification" on 1. Seven were published in July–September 2026 against about 10 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When grants and impact funds dry up, do social enterprises close, drop their social activities, or find commercial revenue, and which ones manage the last?</li> <li>Do certified B Corporations pay workers more, pollute less or give more than uncertified peers of the same size, or does certification follow behaviour that was already there?</li> <li>Do customers pay a premium for certification, and in which markets?</li> </ul> <h2>Data that could help</h2> <p>B Corp certification records with scores, firm accounts and employment data, and funding data from impact investors.</p> <h2>Next step</h2> <p>Match B Corp certification records to firm accounts and compare wage levels and growth before and after certification with similar uncertified firms.</p> https://scholar4ai.com/ideas/small-sellers-under-platform-algorithms 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Business growth is a rising keyword across the tracked journals, and small and medium-sized enterprises are a large topic, but only 19 of 431 recent SME papers carry a platform, e-commerce or algorithm keyword. Millions of small firms now depend on marketplace search rankings. What happens to them when the algorithm changes?

<h2>What prompted this</h2> <p>"Business growth" grew across all journals (25 papers in July–September 2026 against 15 expected) and in management (see the <a href="/trends">trends page</a>). Since October 2025, 431 papers carry an SME or small-business keyword (157 in management journals, 121 in economics, 117 in finance); only 19 of them carry a platform, e-commerce, algorithm or marketplace keyword, and "platform dependence" appears on 3.</p> <h2>Questions</h2> <ul> <li>When a marketplace changes its ranking or fee rules, how much do small sellers' sales move, and do they leave or adapt?</li> <li>Are sellers that sell through several channels more resilient?</li> <li>Do rules such as the EU's Digital Markets Act change how platforms treat small sellers?</li> </ul> <h2>Data that could help</h2> <p>Seller-level sales and rankings from marketplace data providers, announced changes to algorithms and fees, and SME surveys.</p> <h2>Next step</h2> <p>Collect announced ranking or fee changes on one large marketplace and track the sales ranks of a panel of small sellers around them.</p> https://scholar4ai.com/ideas/small-models-for-finance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Knowledge distillation, model compression and small language models appear on more than 700 recent papers in AI journals, and only 12 of them concern finance, banking, legal or business tasks. Banks cannot send customer data to the largest models, and regulators want models they can inspect. How small can a model be and still do financial work, and what is lost in the shrinking?

<h2>What prompted this</h2> <p>Since October 2025, 763 papers carry a knowledge-distillation, model-compression, small-language-model, quantization, pruning or parameter-efficient keyword (748 in AI journals): "knowledge distillation" on 345, "model compression" on 114, "parameter-efficient fine-tuning" on 111, "small language models" on 17. Only 12 of the 763 also carry a finance, bank, stock, fraud, credit, economics, legal, accounting, business or customer keyword.</p> <h2>Questions</h2> <ul> <li>On tasks such as sentiment, entity extraction from filings and customer intent classification, how much accuracy is lost when a large model is distilled to one that runs on a bank's own servers?</li> <li>Do small models fine-tuned on a firm's data beat large general models on that firm's tasks, and for how long before the data drifts?</li> <li>Do compressed models fail in different places than their teachers, in ways that matter for compliance?</li> </ul> <h2>Data that could help</h2> <p>Public financial NLP benchmarks (Financial PhraseBank, FiQA, FinQA), open models at several sizes, and a partner firm's task data.</p> <h2>Next step</h2> <p>Distil a large model into three smaller ones on financial sentiment and filing extraction tasks, and map where the small models' errors differ from the teacher's.</p> https://scholar4ai.com/ideas/sin-taxes 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Alcohol consumption, tobacco control and excise taxes appear on 46 recent papers in the economics journals, with 17 in the last quarter against about 9 per quarter before. Sugar taxes have spread to more than 100 jurisdictions and alcohol taxes are being rewritten. Do the taxes change health, and who bears them?

<h2>What prompted this</h2> <p>Since October 2025, 46 papers carry a sugar-tax, alcohol, tobacco or excise-tax keyword (39 in economics journals): "alcohol consumption" on 16, "tobacco control" on 7, "excise tax" on 4, "sugar-sweetened beverage tax" on 2. The quarter to September 2026 had 17 such papers against about 9 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Do sugar taxes change consumption enough to show up in weight and dental health, or do firms reformulate and consumers switch within the taxed category?</li> <li>Do minimum unit prices for alcohol, as in Scotland, reduce harm more than proportional taxes, and do they hit low-income drinkers hardest?</li> <li>Do large tobacco tax increases now mostly shift smokers to vaping, and what does that mean for the tax base?</li> </ul> <h2>Data that could help</h2> <p>Scanner purchase data, health outcome records, product reformulation data and the dates and sizes of tax changes by jurisdiction.</p> <h2>Next step</h2> <p>Compare sugar content and sales of soft drinks in jurisdictions with tiered sugar taxes and with flat volume taxes, before and after each tax.</p> https://scholar4ai.com/ideas/shadow-banks-and-macroprudential-leakage 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Macroprudential policy and financial intermediation are both rising in the finance journals, but papers that follow credit out of the regulated banking system are rare, and none of them is in an economics journal. When regulators tighten on banks, how much lending simply moves to shadow banks and private credit funds?

<h2>What prompted this</h2> <p>In the <a href="/categories/finance">finance dashboard</a> trends, "macroprudential policy" had 23 papers in July–September 2026 against about 14 expected from its earlier share, and "financial intermediation" had 13 against about 3. Since October 2025, 129 papers carry a macroprudential keyword (90 in finance journals, 35 in economics). Only 7 of them also carry a shadow-banking, private-credit or non-bank keyword, all in finance journals. "Private credit" itself appears on just 4 papers.</p> <h2>Questions</h2> <ul> <li>After loan-to-value limits, countercyclical capital buffers or loan caps tighten, does credit to the same borrowers rise at non-banks?</li> <li>Is the credit that leaks out riskier, more expensive or shorter-term than the bank credit it replaces?</li> <li>Does the leakage undo the policy's effect on house prices and leverage, or only part of it?</li> </ul> <h2>Data that could help</h2> <ul> <li>A macroprudential policy database (for example the IMF's iMaPP) for the timing of measures.</li> <li>Syndicated loans (DealScan), private credit deals (Preqin, PitchBook) and bank lending from supervisory filings.</li> </ul> <h2>Next step</h2> <p>Find borrowers in DealScan that borrow from both banks and non-banks, and compare the two kinds of borrowing around national macroprudential tightenings.</p> https://scholar4ai.com/ideas/second-life-or-recycling-for-ev-batteries 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Remanufacturing and closed-loop supply chains are growing in the operations research journals, and the first large wave of electric-vehicle batteries is reaching the end of its first life. Should a used battery go to stationary storage, to remanufacturing or straight to recycling, and who should decide?

<h2>What prompted this</h2> <p>In the operations research journals, "remanufacturing" had 8 papers in July–September 2026 against about 4 expected and "closed-loop supply chain" 9 against 5 (see the <a href="/categories/operations-research">operations research dashboard</a>). Since October 2025, 107 papers carry a remanufacturing, closed-loop or reverse-logistics keyword (69 in operations research journals). Twelve of them carry a battery keyword, and "second-life batteries" appears on just one paper.</p> <h2>Questions</h2> <ul> <li>Given a battery's state of health, when is second-life storage worth more than recycling, once testing, repurposing and the risk of failure are priced in?</li> <li>How do mineral prices and recycling mandates, such as the recycled-content targets of the EU battery regulation, change that choice?</li> <li>Who does better as the owner of used batteries: carmakers, battery makers or third parties?</li> </ul> <h2>Data that could help</h2> <p>Battery ageing data such as the NASA and CALCE datasets (in the <a href="/rankings/datasets">dataset ranking</a>'s list), mineral prices and storage-market prices.</p> <h2>Next step</h2> <p>Build a decision model that routes used batteries by state of health, and compare its value under 2023 and 2026 lithium prices.</p> https://scholar4ai.com/ideas/school-choice-and-matching 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

School choice, market design, stable matching and deferred acceptance appear on 89 recent papers, mostly in economics. Dozens of cities now assign school places with matching algorithms. Do families understand the rules well enough to benefit, and do the assignments change segregation and achievement?

<h2>What prompted this</h2> <p>Since October 2025, 89 papers carry a school-choice, matching-market, stable-matching, deferred-acceptance, market-design or assignment-mechanism keyword (72 in economics journals): "school choice" on 21, "market design" on 18, "stable matching" on 9, "matching markets" on 9, "deferred acceptance" on 8. Seventeen were published in July–September 2026 against about 21 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In cities using deferred acceptance, what share of families still submit lists that are not in their true order, and who loses from it?</li> <li>Did the switch from older mechanisms change school segregation by income, and did it change achievement for students who got their preferred school?</li> <li>Do information interventions (ranked lists with admission chances) change choices and outcomes?</li> </ul> <h2>Data that could help</h2> <p>Application lists and assignments from city school districts, student outcomes, and the dates of mechanism changes.</p> <h2>Next step</h2> <p>Use application data from a district that switched to deferred acceptance to measure changes in list length, strategic ranking and segregation.</p> https://scholar4ai.com/ideas/sanctions-evasion-through-third-countries 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sanctions appear on 85 recent papers, with 23 in the last quarter against about 20 per quarter before, and 35 of them carry a Russia, enforcement or effectiveness keyword. Trade data show exports from Europe to Central Asia and the Caucasus surging after 2022. How much of the sanctioned trade simply took a detour, and what stops it?

<h2>What prompted this</h2> <p>Since October 2025, 85 papers carry a sanctions keyword (45 in economics journals, 21 in finance, 11 in management): "sanctions" on 18, "economic sanctions" on 15, "financial sanctions" on 11, "trade sanctions" on 9, "sanctions enforcement" on 6. Thirty-five of the 85 also carry an evasion, circumvention, third-country, re-export, Russia, enforcement or effectiveness keyword.</p> <h2>Questions</h2> <ul> <li>How much of the fall in direct exports of sanctioned goods to Russia reappears as exports to its neighbours, product by product?</li> <li>Which firms do the rerouting: new shell exporters, or existing distributors?</li> <li>Did secondary sanctions on banks in third countries in 2024 cut the detours, and at what cost to legitimate trade?</li> </ul> <h2>Data that could help</h2> <p>Monthly bilateral trade data by product (UN Comtrade, Eurostat), customs records with exporter identities where available, and the dates of sanctions packages.</p> <h2>Next step</h2> <p>Compare the growth of EU exports of dual-use goods to Kazakhstan, Kyrgyzstan, Armenia and the UAE with Russia's imports of the same goods since 2022.</p> https://scholar4ai.com/ideas/robots-and-industrial-upgrading 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Industrial upgrading is a large topic in the finance and economics journals, and China installs more industrial robots than any other country. Yet of 397 recent papers with a robot keyword, only 2 carry an industrial-upgrading keyword. Do robots move Chinese firms up the value chain, or just replace workers in the same tasks?

<h2>What prompted this</h2> <p>Since October 2025, "industrial upgrading" appears on 117 papers (56 in finance journals, 48 in economics) and "industrial robots" on 46 (26 in economics). Of the 397 papers with any robot keyword (272 of them in AI journals), only 2 carry an industrial-upgrading or industrial-structure keyword. In the management journals, "industrial upgrading" is rising (5 papers in July–September 2026 against about 2 expected).</p> <h2>Questions</h2> <ul> <li>Do cities and firms that adopt more robots shift toward higher value-added products and exports?</li> <li>Do they hire more engineers and fewer production workers, and what happens to wages?</li> <li>Do robot subsidies speed up upgrading, or only the automation of existing lines?</li> </ul> <h2>Data that could help</h2> <p>Robot installations by industry and country (International Federation of Robotics), customs data on export unit values, firm data (Chinese industrial enterprise data, CSMAR) and job postings.</p> <h2>Next step</h2> <p>Build a city-level robot exposure measure from industry robot data and local industry shares, and relate it to changes in export unit values for 2010–2023.</p> https://scholar4ai.com/ideas/robotaxis-and-ride-hailing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Ride-hailing and ride-sharing appear on 62 recent papers, mostly in operations research, and 11 of them carry an autonomous-vehicle keyword, with robotaxis on 3. Driverless taxis now operate commercially in several cities. What happens to prices, waiting times and drivers' earnings where they compete with human-driven rides?

<h2>What prompted this</h2> <p>Since October 2025, 62 papers carry a ride-hailing, ride-sharing, mobility-on-demand or taxi keyword (37 in operations research journals, 13 in economics): "ride-hailing" on 24, "ride-hailing platforms" on 14, "ride-sharing" on 13, "taxi drivers" on 3, "robotaxis" on 3. Eleven of the 62 also carry an autonomous, driverless, self-driving or robotaxi keyword; "autonomous vehicles" appears on 64 papers in the archive, mostly in AI journals.</p> <h2>Questions</h2> <ul> <li>In cities where robotaxis operate at scale, did ride prices and waiting times fall, and did human drivers' hours and earnings decline?</li> <li>Do robotaxis take the dense, profitable trips and leave the rest to drivers, and does that change the economics of the platforms?</li> <li>How do the fleet-sizing and dispatch models that assume all-autonomous fleets perform in mixed markets?</li> </ul> <h2>Data that could help</h2> <p>Trip-level data released by city regulators (as in San Francisco and Chicago), robotaxi service areas and launch dates, and driver earnings surveys.</p> <h2>Next step</h2> <p>Compare ride-hailing trip prices, waiting times and driver hours in the parts of a city where robotaxis operate with the parts where they do not, before and after launch.</p> https://scholar4ai.com/ideas/robo-voting-and-proxy-advisors 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Fund families now vote at thousands of shareholder meetings a year, and many follow proxy advisors' recommendations almost automatically. Yet only 9 recent papers in the archive carry a shareholder-voting or proxy-advisor keyword, none of them about funds. Does robo-voting decide close votes, and has pass-through voting changed that?

<h2>What prompted this</h2> <p>Since October 2025, only 9 papers carry a proxy-advisor, shareholder-voting or proxy-voting keyword (4 in economics journals, 2 in accounting, 2 in finance, 1 in management), and none of them carries a passive-fund, index-fund, institutional-investor, mutual-fund or robo keyword. Board composition, by contrast, is one of the fastest-growing keywords across all journals (<a href="/trends">trends page</a>).</p> <h2>Questions</h2> <ul> <li>How often do funds vote exactly as a proxy advisor recommends, within hours of the recommendation, and does that decide close votes?</li> <li>Do companies change their proposals to win the advisors' support?</li> <li>Has pass-through voting, which lets fund investors choose how their shares are voted, changed outcomes?</li> </ul> <h2>Data that could help</h2> <p>Fund voting records (Form N-PX filings, ISS Voting Analytics), proxy advisors' recommendations, vote results and meeting dates.</p> <h2>Next step</h2> <p>From N-PX filings, compute for each fund family the share of votes cast as ISS recommended and the time between the recommendation and the vote.</p> https://scholar4ai.com/ideas/robo-advisors-after-the-llm 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Robo-advisors appear on 12 recent papers, and the topic is fading: 1 paper in July–September 2026 against about 5 per quarter before. Meanwhile people ask chatbots for financial advice. Does conversational advice change portfolios more than the old questionnaire-based robo-advice did, and for better or worse?

<h2>What prompted this</h2> <p>Since October 2025, 16 papers carry a robo-advice keyword (8 in finance journals), with "robo-advisors" on 12. Only one of them was published in July–September 2026, against about 5 per quarter over the nine months before. "Household portfolio choice" appears on 15 papers. Generative AI, by contrast, is rising in almost every category on the <a href="/trends">trends page</a>.</p> <h2>Questions</h2> <ul> <li>When households ask a chatbot about investing, do they get advice that fits their situation, and do they act on it more than on a robo-advisor's allocation?</li> <li>Does conversational advice reduce known mistakes (under-diversification, high fees, panic selling) or add new ones (overconfidence, chasing stories)?</li> <li>Should the same suitability rules that apply to advisers apply to chatbots?</li> </ul> <h2>Data that could help</h2> <p>Experiments in which participants receive advice from a robo-advisor questionnaire, a chatbot or a human, with incentivized portfolio choices; brokerage data on users who adopt in-app AI assistants.</p> <h2>Next step</h2> <p>Run an online experiment in which participants allocate a real bonus after advice from one of the three sources, and measure diversification and fees.</p> https://scholar4ai.com/ideas/return-to-office-mandates 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Remote and hybrid work appear on 53 recent papers, and 18 of them carry a productivity, turnover, retention or performance keyword. Large firms have ordered workers back to the office, some for five days a week. Do the mandates raise output, or mostly lose the workers with other options?

<h2>What prompted this</h2> <p>Since October 2025, 53 papers carry a remote-work, working-from-home, hybrid-work or return-to-office keyword (25 in economics journals, 17 in management): "remote work" on 38, "working from home" on 18, "hybrid work" on 13. Eighteen of them also carry a turnover, retention, quit, attrition, productivity, mandate, performance or innovation keyword; "employee turnover" appears on 17 papers in the archive.</p> <h2>Questions</h2> <ul> <li>After a return-to-office mandate, who quits: senior staff, women, caregivers, the best-rated employees?</li> <li>Do measured output and promotion rates change for those who stay, and do the firms hire replacements at higher pay?</li> <li>Do firms that mandate five days differ from those that settle on two or three, and do stock markets react to the announcements?</li> </ul> <h2>Data that could help</h2> <p>Worker-level tenure and exit data (Revelio Labs, LinkedIn), mandate announcement dates, firm reviews (Glassdoor) and job postings with location requirements.</p> <h2>Next step</h2> <p>Compare exit rates by seniority and gender at firms that announced five-day mandates with matched firms that kept hybrid policies, over the following year.</p> https://scholar4ai.com/ideas/research-integrity-and-paper-mills 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Research integrity, plagiarism, research misconduct and retractions appear on 28 recent papers across the AI, finance and economics journals. Retractions have reached record numbers, and paper mills sell authorship and citations, with language models making fabrication cheaper. How much of the recent literature in these fields is affected, and can it be detected?

<h2>What prompted this</h2> <p>Since October 2025, 28 papers carry a research-integrity, plagiarism, research-misconduct, retraction, paper-mill, predatory or citation-manipulation keyword (13 in AI journals, 7 in finance, 4 in economics): "research integrity" on 10, "plagiarism" on 5, "research misconduct" on 3, "retraction" on 3. The tracked journals publish more than 40,000 papers a year between them, so the integrity literature is tiny relative to the output it concerns.</p> <h2>Questions</h2> <ul> <li>What share of recent papers in the tracked journals show paper-mill signals (tortured phrases, impossible author affiliations, citation rings), by journal and country?</li> <li>Do special issues and guest-edited collections carry more such signals, and did journals that cut special issues see fewer?</li> <li>Can AI-fabricated results in empirical finance and economics be detected from their statistics, for example from the distribution of reported test statistics?</li> </ul> <h2>Data that could help</h2> <p>The full text and metadata of recent papers, retraction records (Retraction Watch), and the published detectors for tortured phrases and citation anomalies.</p> <h2>Next step</h2> <p>Run tortured-phrase and citation-anomaly detectors over the last two years of papers in the tracked journals and compare rates across journals and special issues.</p> https://scholar4ai.com/ideas/renewable-forecasts-and-market-bidding 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Wind power, photovoltaic power and solar irradiance forecasting appear on 52 recent papers, almost all in AI journals, and only 8 of them carry a market, bidding, price, imbalance or trading keyword. Forecasts are judged by error metrics, but wind and solar operators pay for forecast errors in imbalance charges. Which forecast improvements pay, and which do not?

<h2>What prompted this</h2> <p>Since October 2025, 52 papers carry a wind-power, photovoltaic-power, solar-power or renewable-energy forecasting keyword (46 in AI journals, 5 in economics): "wind power forecasting" on 22, "photovoltaic power forecasting" on 13, "wind speed forecasting" on 11, "solar irradiance forecasting" on 5, "renewable energy forecasting" on 6. Only 8 of the 52 also carry a market, bidding, price, imbalance, trading, economics, cost or dispatch keyword.</p> <h2>Questions</h2> <ul> <li>For a wind or solar plant selling into a day-ahead market with imbalance charges, how much revenue does a given reduction in forecast error add, and does the value depend on the direction of errors?</li> <li>Do the deep learning forecasts in recent papers improve revenue relative to standard ensemble weather forecasts, or only improve metrics that do not map to revenue?</li> <li>Does better forecasting by many plants reduce system balancing costs, or shift them?</li> </ul> <h2>Data that could help</h2> <p>Plant-level production and forecast data, day-ahead and imbalance prices (ENTSO-E, Nord Pool), and the forecast models of published papers.</p> <h2>Next step</h2> <p>Compute the imbalance cost of forecast errors for a set of wind plants under standard and published deep learning forecasts using one year of market prices.</p> https://scholar4ai.com/ideas/remote-work-and-city-centres 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Remote and hybrid work appear on 53 recent papers across the economics and management journals, but only 12 of them carry a city, housing, commuting, office or rent keyword. Office vacancies are at record highs in many downtowns. Where did spending, rents and tax bases move, and are the shifts permanent?

<h2>What prompted this</h2> <p>Since October 2025, 53 papers carry a remote-work, working-from-home, hybrid-work or telework keyword (25 in economics journals, 17 in management): "remote work" on 38, "working from home" on 18, "hybrid work" on 13. Only 12 of them also carry a city, urban, housing, commuting, downtown, office, rent or real-estate keyword.</p> <h2>Questions</h2> <ul> <li>How much of the fall in downtown spending moved to suburbs and how much disappeared, and did restaurants and shops follow the workers?</li> <li>Have office-to-residential conversions and rent falls in city centres brought residents back?</li> <li>What happens to city budgets that depend on office property taxes and commuter spending?</li> </ul> <h2>Data that could help</h2> <p>Card transaction data by location, office vacancy and rent data, mobile phone mobility data, and municipal tax revenues.</p> <h2>Next step</h2> <p>Compare the recovery of card spending in downtown and suburban areas across US cities by their share of remote-capable jobs, from 2020 to 2025.</p> https://scholar4ai.com/ideas/remittances-and-instant-payments 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Remittances appear on 50 recent papers, mostly in economics, but only 12 of them carry a cost, fee, digital, mobile, fintech or stablecoin keyword. Instant payment systems, mobile money links and stablecoins are cutting the cost of sending money home. Does cheaper sending raise the amount received, and who switches?

<h2>What prompted this</h2> <p>Since October 2025, 50 papers carry a remittance keyword (33 in economics journals, 10 in finance), with "remittances" on 39 and "international remittances" on 5. Only 12 of the 50 also carry a cost, fee, instant-payment, digital, mobile, fintech, crypto or stablecoin keyword. The global average cost of sending remittances has fallen but remains above the 3% target.</p> <h2>Questions</h2> <ul> <li>When a corridor gets a cheaper channel (a mobile money link or a stablecoin route), do remittances rise by more than the fee saved, and does the money shift from informal channels?</li> <li>Who uses the new channels: younger migrants, those with bank accounts, those sending small amounts?</li> <li>Do recipients spend remittances differently when they arrive in digital wallets rather than cash?</li> </ul> <h2>Data that could help</h2> <p>Remittance prices by corridor (the World Bank's Remittance Prices Worldwide), central bank remittance statistics, mobile money operator data and household surveys in recipient countries.</p> <h2>Next step</h2> <p>Match corridor-level price changes since 2019 to remittance flows and test whether flows respond more where digital channels entered.</p> https://scholar4ai.com/ideas/related-party-transactions-and-tunneling 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Related-party transactions appear on 23 recent papers and tunneling on 17, almost all in finance journals and mostly about Chinese business groups. As rules on these transactions tighten and disclosure moves to machine-readable formats, can tunneling be detected at scale, and did the rules reduce it?

<h2>What prompted this</h2> <p>Since October 2025, 41 papers carry a related-party or tunneling keyword (27 in finance journals, 6 in economics, 4 in accounting): "related-party transactions" on 23 and "tunneling" on 17. Most study a single country, and few use the transaction-level disclosures that listed firms now file.</p> <h2>Questions</h2> <ul> <li>Can anomalous related-party transactions (prices far from market, timing before earnings announcements) be detected automatically from disclosures, and do the flagged firms later face enforcement?</li> <li>Did stricter approval rules for related-party transactions in China and India reduce their volume, or move them into unreported forms?</li> <li>Do minority shareholders price the expected tunneling into group-affiliated firms?</li> </ul> <h2>Data that could help</h2> <p>Transaction-level related-party disclosures (CSMAR for China, stock exchange filings elsewhere), enforcement records and stock returns.</p> <h2>Next step</h2> <p>Build a detector for abnormal related-party prices from Chinese listed firms' transaction disclosures and check whether flagged firms are later sanctioned.</p> https://scholar4ai.com/ideas/refugee-and-immigrant-entrepreneurs 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Immigrant entrepreneurs and refugee entrepreneurship appear on 13 recent papers, while immigration as a topic is rising fast in the economics journals. Millions of recent refugees and migrants have started businesses. Do their firms grow, hire natives and survive, and which policies (work permits, credit, recognition of qualifications) make the difference?

<h2>What prompted this</h2> <p>Since October 2025, "immigrant entrepreneurs" appears on 6 papers, "immigrant entrepreneurship" on 4 and "refugee entrepreneurship" on 3, while papers with any immigration keyword rose to 42 in July–September 2026 from about 30 per quarter earlier. Sixteen of the 73 papers with an immigrant or diaspora keyword also carry an entrepreneurship, startup, self-employment or founder keyword.</p> <h2>Questions</h2> <ul> <li>Do businesses started by recent refugees (Ukrainians in Europe, Syrians in Turkey and Germany, Venezuelans in Latin America) survive and grow at the same rate as natives' businesses of the same type?</li> <li>Does faster access to work permits and bank accounts raise business formation, and does it reduce informal work?</li> <li>Do immigrant-founded firms hire natives, and in which sectors?</li> </ul> <h2>Data that could help</h2> <p>Business registers with founder nationality, refugee registration data, and employer-employee records.</p> <h2>Next step</h2> <p>Compare business registrations by Ukrainian nationals across EU countries with different speeds of work permit access since 2022.</p> https://scholar4ai.com/ideas/recommender-systems-and-welfare 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Recommender systems appear on 111 recent papers and personalized recommendation on 54, almost all in AI journals; 45 recommendation papers carry a diversity, fairness or popularity-bias keyword, and welfare, consumer surplus and addiction appear in almost none. Recommendation decides what billions of people see and buy. Who gains, and by how much?

<h2>What prompted this</h2> <p>Since October 2025, 513 papers carry a recommender or recommendation keyword (440 in AI journals): "recommender systems" on 111, "personalized recommendation" on 54, "sequential recommendation" on 47, "multimodal recommendation" on 38. Forty-five of them also carry a filter-bubble, diversity, polarization, fairness, welfare, popularity-bias or well-being keyword, led by "popularity bias" on 19 and "recommendation diversity" on 13.</p> <h2>Questions</h2> <ul> <li>In platform experiments, how much consumer surplus do recommendations create relative to search alone, and how is it split between consumers, sellers and the platform?</li> <li>Do recommendations that optimize engagement reduce the variety of what people consume over months, and does the variety loss show up in satisfaction?</li> <li>When platforms are required to offer a non-personalized feed, as the EU now does, who switches, and what changes?</li> </ul> <h2>Data that could help</h2> <p>Platform experiments with randomized recommendation policies, consumption logs over long horizons, and the EU's transparency reports on feed options.</p> <h2>Next step</h2> <p>Run or obtain a platform experiment that randomizes recommendation strength and measure purchases, variety and self-reported satisfaction over three months.</p> https://scholar4ai.com/ideas/reading-tables-and-charts 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Document understanding, table question answering and chart understanding appear on 17 recent papers, all in AI journals, with 7 in the last quarter against about 3 per quarter before. Financial information lives in tables, footnotes and charts that models still misread. How accurate are current models on real filings, and which errors would mislead an analyst?

<h2>What prompted this</h2> <p>Since October 2025, 17 papers carry a document-understanding, table-question-answering, chart-understanding, document-layout or visually-rich-document keyword, all in AI journals: "document understanding" on 5, "table question answering" on 4, "document image analysis" on 3, "chart understanding" on 2. Seven were published in July–September 2026 against about 3 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>On questions that require combining a table, a footnote and a chart from the same annual report, how often do multimodal models get the number right, and how often do they confidently get it wrong?</li> <li>Which document features cause errors (merged cells, units in headers, negative numbers in parentheses, restated columns)?</li> <li>Would a layer that extracts tables into structured data before the model reads them remove most of the errors?</li> </ul> <h2>Data that could help</h2> <p>Annual reports with XBRL-tagged values as ground truth, public financial document question sets (FinQA, TAT-QA), and chart datasets built from earnings presentations.</p> <h2>Next step</h2> <p>Build 500 questions whose answers are XBRL-tagged values in annual reports and measure the accuracy of three multimodal models on the PDF pages.</p> https://scholar4ai.com/ideas/rd-tax-incentives-and-relabeling 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tax incentives appear on 47 recent papers and R&D investment on 267, but R&D tax incentives on only 5, and 26 of 75 incentive papers touch innovation. Governments spend heavily on R&D tax relief. How much of the reported increase in R&D is real research, and how much is ordinary spending relabeled?

<h2>What prompted this</h2> <p>Since October 2025, 75 papers carry a tax-incentive, tax-credit, R&amp;D-tax or super-deduction keyword (47 in economics journals, 16 in finance); "tax incentives" appears on 47 and "R&amp;D tax incentives" on 5. Twenty-six of the 75 also carry an R&amp;D, innovation or patent keyword. "R&amp;D investment" itself appears on 267 papers, one of the most frequent topics in the archive.</p> <h2>Questions</h2> <ul> <li>When R&amp;D tax relief becomes more generous, do patents, new products and researcher hiring rise in proportion to reported R&amp;D, or does reported R&amp;D rise alone?</li> <li>Which firms relabel: those near the eligibility threshold, those with weak auditors, those in sectors where the line between R&amp;D and engineering is blurry?</li> <li>Did the US requirement to amortize R&amp;D from 2022 change reported R&amp;D the other way?</li> </ul> <h2>Data that could help</h2> <p>Firm R&amp;D spending and tax relief claims, patents and product launches, researcher employment from job postings, and the dates of policy changes.</p> <h2>Next step</h2> <p>Compare the growth of reported R&amp;D with the growth of R&amp;D job postings for firms around a change in the generosity of relief.</p> https://scholar4ai.com/ideas/rail-freight-modal-shift 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Rail freight, railway operations and intermodal freight transport appear on 55 recent papers, with 14 in the last quarter against about 12 per quarter before. Governments want freight moved from trucks to rail for climate reasons, yet rail's share keeps falling in most of Europe and the US. Which bottlenecks (terminals, reliability, pricing) matter, and what would change shippers' choices?

<h2>What prompted this</h2> <p>Since October 2025, 55 papers carry a railway, rail-freight, railroad, intermodal or rail-transport keyword (27 in operations research journals, 13 in economics, 11 in AI): "rail freight" on 6, "railway operations" on 5, "intermodal freight transport" on 4, "railway safety" on 3, "China Railway Express" on 2. The quarter to September 2026 had 14 such papers against about 12 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>For which origin-destination pairs and goods does rail lose to trucks on cost and time, and how much of the gap is terminal handling rather than line haul?</li> <li>Do shippers respond to carbon prices on trucking, or do they need reliability improvements first?</li> <li>Did the China-Europe rail services survive the collapse of their subsidies and the war in Ukraine, and what did they carry?</li> </ul> <h2>Data that could help</h2> <p>Freight flow data by mode (Eurostat, US Commodity Flow Survey), rail punctuality records, terminal capacity data and shipper surveys.</p> <h2>Next step</h2> <p>Estimate a mode-choice model on recent freight flow data with terminal time and reliability as separate variables, and simulate a carbon price on trucking.</p> https://scholar4ai.com/ideas/rag-for-finance-and-tax 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Retrieval-augmented generation appears on 165 recent papers, 171 of 178 such papers in AI journals, and only 12 touch finance, legal, tax, accounting, audit or compliance questions. Banks, tax advisers and auditors are deploying it over filings, rules and contracts. How often does it cite the right source, and what happens when the source changes?

<h2>What prompted this</h2> <p>Since October 2025, 178 papers carry a retrieval-augmented keyword (171 in AI journals, 5 in management): "retrieval-augmented generation" on 165, "graph-based retrieval-augmented generation" on 8. Only 12 of the 178 also carry a finance, bank, legal, regulation, accounting, tax, audit, compliance, economics or business keyword, and those few include legal question answering and tax return filing.</p> <h2>Questions</h2> <ul> <li>On questions about accounting standards, tax rules and loan agreements, how often does a retrieval system answer from the right passage, from a superseded version, or from nothing?</li> <li>When rules change, how long do deployed systems keep answering from the old text, and who notices?</li> <li>Do professionals check the retrieved sources, and does showing them reduce errors or only slow the work?</li> </ul> <h2>Data that could help</h2> <p>Question sets with gold answers and sources built from accounting standards, tax codes and standard contracts, including versions over time.</p> <h2>Next step</h2> <p>Build a test set of 300 questions over the current and prior versions of one accounting standard and measure how often retrieval systems cite the wrong version.</p> https://scholar4ai.com/ideas/quiet-quitting-and-engagement 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Job satisfaction appears on 39 recent papers in the management and economics journals, employee engagement on 17, work engagement on 9 and quiet quitting on 2. Engagement surveys drive billions in consulting fees. Does measured engagement predict anything that firms can observe, such as output, errors or exits, and did it actually fall?

<h2>What prompted this</h2> <p>Since October 2025, 88 papers carry an engagement, job-satisfaction or disengagement keyword (56 in management journals, 21 in economics): "job satisfaction" on 39, "employee engagement" on 17, "work engagement" on 9, "quiet quitting" on 2. Almost all measure engagement with surveys and relate it to other surveys.</p> <h2>Questions</h2> <ul> <li>Does survey engagement predict later output, errors, absences and exits in administrative records, and how much of the link survives once pay and job type are controlled for?</li> <li>Did engagement fall after 2020, as the quiet-quitting story claims, in data that are not surveys, such as hours logged and voluntary tasks?</li> <li>Do engagement interventions (recognition programmes, flexibility) change behaviour or only survey answers?</li> </ul> <h2>Data that could help</h2> <p>Firms' engagement survey results linked to productivity, absence and exit records; time-use and software activity logs where available.</p> <h2>Next step</h2> <p>Partner with a firm that runs regular engagement surveys and test whether survey scores predict exits and performance ratings over the following year.</p> https://scholar4ai.com/ideas/quantum-optimization-benchmarks 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Quantum computing, quantum machine learning and quantum annealing appear on 115 recent papers, mostly in AI journals, and 33 of them concern optimization. Firms in finance and logistics run quantum pilots. Do the quantum results beat the best classical solvers on the same instances, or only the classical baselines the papers chose?

<h2>What prompted this</h2> <p>Since October 2025, 115 papers carry a quantum keyword (81 in AI journals, 15 in economics, 13 in operations research): "quantum machine learning" on 22, "quantum computing" on 14, "quantum neural networks" on 9, "hybrid quantum-classical models" on 8. Thirty-three of the 115 also carry an optimization, annealing, combinatorial, portfolio, scheduling, logistics or finance keyword; "quantum annealing" appears on 3 and "quantum optimization" on 2.</p> <h2>Questions</h2> <ul> <li>On portfolio, routing and scheduling instances used in the quantum papers, how do the reported solutions compare with Gurobi, CPLEX or a tuned metaheuristic given the same time?</li> <li>Do the hybrid methods' gains come from the quantum part or from the classical pre- and post-processing?</li> <li>At what instance sizes, if any, do current devices beat classical solvers on energy use rather than solution quality?</li> </ul> <h2>Data that could help</h2> <p>The instances and results published in recent quantum optimization papers, and standard classical solvers.</p> <h2>Next step</h2> <p>Re-solve the benchmark instances of ten recent quantum optimization papers with a commercial solver and a tuned local search under the same time limit.</p> https://scholar4ai.com/ideas/public-procurement-and-corruption-in-the-data 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Public procurement appears on 53 recent papers, with 40 procurement papers in the last quarter against about 26 per quarter before, and 33 of 123 carry a corruption, digital, e-procurement or bidding keyword. Open contracting data now cover millions of tenders. Can red flags in the data predict corruption, and do governments act on them?

<h2>What prompted this</h2> <p>Since October 2025, 123 papers carry a procurement keyword (52 in economics journals, 20 in finance, 20 in management, 20 in operations research): "public procurement" on 53, "procurement strategy" on 11, "green public procurement" on 4, "e-procurement" on 3. The quarter to September 2026 had 40 such papers against about 26 per quarter earlier. Thirty-three of the 123 also carry a corruption, e-procurement, digital, transparency, competition or bidding keyword; "corruption" alone appears on 87 papers.</p> <h2>Questions</h2> <ul> <li>Do red flags computed from open contracting data (single bidders, short deadlines, repeated winners) predict later audit findings and prosecutions?</li> <li>When tenders move to electronic platforms, do single-bid rates and prices fall, and do the effects fade as bidders learn the system?</li> <li>Do firms that win contracts with red flags perform worse on the contracts?</li> </ul> <h2>Data that could help</h2> <p>Open contracting data (national procurement portals, the Open Contracting Partnership), audit and court records, and contract performance data.</p> <h2>Next step</h2> <p>Compute standard red-flag indicators for one country's procurement portal and test whether they predict audit findings in the following years.</p> https://scholar4ai.com/ideas/public-debt-with-higher-rates 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Public debt, fiscal rules and debt sustainability appear on 215 recent papers in the economics and finance journals, but only 41 of them carry an interest-rate, debt-service or monetary keyword. For a decade, low rates made debt look free. With interest costs now the fastest-growing budget item in several countries, which governments adjust, and how?

<h2>What prompted this</h2> <p>Since October 2025, 215 papers carry a public-debt, debt-sustainability, government-debt, fiscal-sustainability, debt-to-GDP or fiscal-rule keyword (121 in economics journals, 87 in finance): "public debt" on 98, "fiscal rules" on 32, "fiscal sustainability" on 31, "debt sustainability" on 12. Only 41 of them also carry an interest-rate, debt-service, interest-burden or monetary keyword.</p> <h2>Questions</h2> <ul> <li>As interest costs rise, do governments cut investment, raise taxes, or let deficits widen, and does the response depend on fiscal rules?</li> <li>Do bond markets now punish deficits faster, as the UK's 2022 episode suggested, and which countries are most exposed?</li> <li>How much do rising rates on domestic debt redistribute from taxpayers to bondholders?</li> </ul> <h2>Data that could help</h2> <p>Government budget data by category, interest payments and debt maturity (IMF, OECD), sovereign yields and the dates of fiscal rule changes.</p> <h2>Next step</h2> <p>Decompose the change in primary balances across OECD countries since 2022 into spending and revenue responses, and relate them to each country's interest cost increase.</p> https://scholar4ai.com/ideas/prompt-injection-in-financial-agents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Prompt injection, jailbreak attacks and LLM security appear on 37 recent papers, all in AI journals. Banks and brokers are deploying agents that read emails, documents and web pages and then act. A hidden instruction in a document can make such an agent move money or leak data. How exposed are deployed financial agents, and which defences hold?

<h2>What prompted this</h2> <p>Since October 2025, 37 papers carry a prompt-injection, jailbreak, adversarial-prompt, LLM-security, agent-security or red-teaming keyword, all in AI journals: "prompt injection" on 11, "jailbreak attacks" on 10, "LLM security" on 6, "adversarial prompts" on 5, "red teaming" on 4. Agentic AI and LLM agents are among the rising keywords in the AI journals, and none of the security papers concerns finance.</p> <h2>Questions</h2> <ul> <li>In a sandbox with a financial assistant that reads documents and can call payment and trading tools, what share of injected instructions in invoices, emails and web pages succeed?</li> <li>Which defences (tool permission layers, instruction hierarchies, human confirmation thresholds) reduce success without making the agent useless?</li> <li>How would an attack look in audit logs, and could existing fraud controls catch it?</li> </ul> <h2>Data that could help</h2> <p>Open agent frameworks, a corpus of documents with injected instructions, and the control frameworks that banks use for payment authorization.</p> <h2>Next step</h2> <p>Build a sandboxed financial agent with payment tools and measure the success rate of injected instructions in 1,000 documents under three defence configurations.</p> https://scholar4ai.com/ideas/product-returns-and-reverse-logistics 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Reverse logistics, return policies and product returns appear on 53 recent papers, almost all in operations research, with 15 in the last quarter against about 12 per quarter before. Online return rates are several times those of stores, and retailers are starting to charge for returns. Which policies cut returns without cutting sales, and what happens to the returned goods?

<h2>What prompted this</h2> <p>Since October 2025, 53 papers carry a product-return, return-policy, returns-management or reverse-logistics keyword (48 in operations research journals): "reverse logistics" on 30, "return policy" on 10, "product returns" on 8, "returns management" on 3. The quarter to September 2026 had 15 such papers against about 12 per quarter earlier. Most model the network; few use data on what consumers do when policies change.</p> <h2>Questions</h2> <ul> <li>When a retailer introduces a return fee or shortens the return window, how much do returns fall, and how much do sales and repeat purchases fall with them?</li> <li>What share of returned items is resold as new, resold as used, or destroyed, and how does that vary by category?</li> <li>Do virtual try-on tools and better size information reduce returns more cheaply than fees?</li> </ul> <h2>Data that could help</h2> <p>Retailer transaction and return data around policy changes, disposition records for returned goods, and consumer survey data.</p> <h2>Next step</h2> <p>Compare return rates and repeat purchase rates at retailers that introduced return fees in 2023–2024 with those that did not, using consumer panel data.</p> https://scholar4ai.com/ideas/private-equity-in-nursing-homes 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Private equity is a small but rising keyword in the finance journals, and only 2 of 22 recent private-equity papers touch health care. Private equity firms now own a large share of nursing homes and physician practices in several countries. Do their cost cuts show up in quality of care, and who pays?

<h2>What prompted this</h2> <p>Since October 2025, 22 papers carry a private-equity keyword (16 in finance journals), with 8 in July–September 2026 against about 5 per quarter over the nine months before. Only 2 of them also carry a health, hospital, nursing or care keyword. "Nursing homes" appears on 4 papers.</p> <h2>Questions</h2> <ul> <li>After a private equity buyout, do staffing levels, inspection deficiencies and patient outcomes in nursing homes change?</li> <li>Do buyouts raise prices for insurers and patients, or cut costs that others were overpaying?</li> <li>Do the effects differ between funds that hold for a few years and those that hold for longer?</li> </ul> <h2>Data that could help</h2> <p>Ownership records of nursing homes and practices (public ownership files in the US, company registers elsewhere), inspection and staffing data, and claims or cost data.</p> <h2>Next step</h2> <p>Match nursing home ownership changes to staffing and inspection records in one country and compare homes bought by private equity with those bought by other operators.</p> https://scholar4ai.com/ideas/privacy-rules-and-small-advertisers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Privacy concerns, GDPR, data protection and consumer privacy appear on 81 recent papers, mostly about what consumers feel. Few study what the rules did to the firms that advertise. When tracking became harder, large platforms kept their data and small advertisers lost theirs. Did privacy rules raise concentration?

<h2>What prompted this</h2> <p>Since October 2025, 81 papers carry a GDPR, privacy-regulation, data-protection, cookie, app-tracking or consumer-privacy keyword (28 in management journals, 23 in AI, 12 in operations research, 11 in finance): "privacy concerns" on 32, "GDPR" on 9, "data protection" on 9, "consumer privacy" on 8. Thirty-three of them carry an advertising, marketing, targeting, competition or platform keyword, but almost none concerns the firms that buy advertising.</p> <h2>Questions</h2> <ul> <li>After app tracking restrictions and cookie consent rules, did small advertisers' customer acquisition costs rise more than large advertisers', and did some stop advertising online?</li> <li>Did the rules move advertising spending toward the platforms with the most first-party data?</li> <li>Did new firms find it harder to launch products, and did venture investment in consumer apps fall?</li> </ul> <h2>Data that could help</h2> <p>Advertising spending by firm size from advertising platforms and panels, customer acquisition costs from e-commerce data providers, and the dates of privacy changes.</p> <h2>Next step</h2> <p>Compare the advertising cost per acquisition of small and large e-commerce sellers before and after the 2021 app tracking change, using data from an e-commerce analytics provider.</p> https://scholar4ai.com/ideas/pricing-cyber-insurance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Cybersecurity risks and cyberattacks appear on 226 recent papers, mostly in AI journals, and 56 of them carry a risk-management, insurance, investment or cost keyword, with cyber insurance on 9. Insurers struggle to price a risk that is correlated across firms and shaped by the insured's own choices. What do losses look like, and does insurance change security?

<h2>What prompted this</h2> <p>Since October 2025, 226 papers carry a cybersecurity or cyber keyword (112 in AI journals, 49 in finance, 32 in management): "cybersecurity risks" on 39, "cybersecurity" on 35, "cyberattack detection" on 11, "cyberattacks" on 11, "cyber insurance" on 9, "cyber risk management" on 5, "cybersecurity investment" on 3. Fifty-six of the 226 carry an investment, game, insurance, optimization, cost or risk-management keyword.</p> <h2>Questions</h2> <ul> <li>How fat-tailed and correlated are cyber losses across firms, and how much capital does an insurer need to cover a widespread event such as a cloud outage or a supply chain attack?</li> <li>Do insured firms invest more in security to get coverage, or less because they are covered, and do insurers' security requirements change behaviour?</li> <li>Did ransomware claims and the exclusions written after them change who is insurable?</li> </ul> <h2>Data that could help</h2> <p>Claims data from insurers or brokers, breach notification records, firms' security postures from external scans, and policy terms.</p> <h2>Next step</h2> <p>Fit a loss distribution to public breach and ransomware records with firm size and sector, and estimate the capital required for a portfolio with a common-cause event.</p> https://scholar4ai.com/ideas/price-setting-in-live-streaming-commerce 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Live-streaming e-commerce and price competition are among the fastest-growing keywords in operations research, and influencer marketing is rising in management. OR papers model live commerce as a pricing game, management papers as persuasion. Who captures the margin: the platform, the streamer or the brand?

<h2>What prompted this</h2> <p>On the <a href="/categories/operations-research">operations research dashboard</a>, "price competition" (11 papers in July–September 2026 against about 2 expected), "live streaming e-commerce" (7 against 2), "digital platforms" and "Stackelberg game" are all growing; in management, "influencer marketing" is too (10 against 5). Since October 2025, 48 papers carry a live-streaming keyword (23 in operations research, 17 in management). Thirteen of them also carry a game or pricing keyword: 11 in operations research and none in management.</p> <h2>Questions</h2> <ul> <li>When top streamers win "lowest price" deals, do brands lose margin, or win it back through volume and awareness?</li> <li>Do commissions and prices match Stackelberg models with the platform, the streamer or the brand as the leader?</li> <li>How do the often high return rates of live commerce change the optimal price?</li> </ul> <h2>Data that could help</h2> <ul> <li>Prices of the same products in live streams and in regular store listings, from third-party sales trackers.</li> <li>Financial reports of listed consumer brands that rely on live commerce.</li> </ul> <h2>Next step</h2> <p>Collect paired live-stream and store prices for a few hundred products and compare the gap with what published pricing models predict.</p> https://scholar4ai.com/ideas/precision-agriculture-adoption 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Precision and smart agriculture, plant disease detection and agricultural technology adoption appear on 148 recent papers, split between AI and economics; 47 of them carry an adoption, farmer, cost or income keyword. AI papers detect disease on leaves; economics papers study adoption of older technologies. Do the new tools reach farms, and do they raise income?

<h2>What prompted this</h2> <p>Since October 2025, 148 papers carry a precision-agriculture, smart-agriculture, plant-disease, crop-yield, agricultural-technology or digital-agriculture keyword (90 in AI journals, 45 in economics): "precision agriculture" on 28, "smart agriculture" on 26, "plant disease detection" on 18, "agricultural technology adoption" on 11, "crop yield prediction" on 3. Forty-seven of them also carry an adoption, farmer, smallholder, economics, cost, profit, income, deployment, welfare or policy keyword, almost all in the economics journals.</p> <h2>Questions</h2> <ul> <li>Of the plant disease and yield models published in AI journals, which have been deployed in apps that farmers use, and does their field accuracy match the published accuracy?</li> <li>For smallholders, do app-based diagnoses change input use and yields, and does the benefit cover a smartphone and data?</li> <li>Do large farms' gains from precision application widen the gap with small farms, and do subsidies narrow it?</li> </ul> <h2>Data that could help</h2> <p>Usage data from agricultural advisory apps, farm surveys with technology use and yields, and field validation studies of published models.</p> <h2>Next step</h2> <p>Validate three published plant disease models on images collected by farmers in the field through an advisory app, and compare accuracy with the published figures.</p> https://scholar4ai.com/ideas/populism-and-trade 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Populism appears on 26 recent papers, mostly in the economics journals, with 7 in the last quarter against about 5 per quarter before. Only 2 of them carry a trade, globalization, tariff or automation keyword, although the "China shock" is the best-known economic explanation of populist votes. Do the newer shocks (inflation, immigration, AI) now matter more?

<h2>What prompted this</h2> <p>Since October 2025, 26 papers carry a populism keyword (15 in economics journals, 6 in management): "populism" on 15, "right-wing populism" on 6, "populist parties" on 4. Only 2 of the 26 also carry a trade, globalization, import, China-shock, tariff, protection, inequality or automation keyword. "Trade policy" and "globalization" appear on 46 and 41 papers each, but not in the same papers as populism.</p> <h2>Questions</h2> <ul> <li>In the elections of 2024–2026, did regions hit hardest by import competition still vote more populist, or did inflation and immigration exposure take over as predictors?</li> <li>Do populist governments deliver the trade protection they promise, and does protection then reduce their support?</li> <li>Does exposure to AI-related job losses already predict voting, as import competition did?</li> </ul> <h2>Data that could help</h2> <p>Regional election results, trade exposure measures by region, local inflation and immigration data, and AI exposure by occupation.</p> <h2>Next step</h2> <p>Regress regional vote shares in recent European and US elections on import competition, local inflation and AI exposure, and compare the coefficients with those from the 2010s.</p> https://scholar4ai.com/ideas/pollution-havens-along-the-belt-and-road 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Environmental degradation is one of the fastest-rising keywords in the economics journals, and the Belt and Road Initiative is a large topic in finance and economics. But the pollution haven hypothesis appears on only 14 recent papers. Did Chinese investment abroad move dirty production to countries with weaker rules?

<h2>What prompted this</h2> <p>In economics, "environmental degradation" had 13 papers in July–September 2026 against about 5 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, the Belt and Road Initiative appears on 48 papers and outward foreign direct investment on 52, but the pollution haven hypothesis and its variants on only 14, of which 5 also carry a Belt and Road or outward-investment keyword.</p> <h2>Questions</h2> <ul> <li>Do Chinese-financed plants abroad (power, steel, cement) emit more per unit of output than similar plants financed by others in the same countries?</li> <li>Did China's 2021 pledge to stop building new coal power plants abroad change the mix of projects?</li> <li>Does local enforcement, more than national rules, decide where dirty projects go?</li> </ul> <h2>Data that could help</h2> <p>Databases of overseas development finance (for example Boston University's China's Global Power Database), plant-level emissions and satellite pollution measures, and indices of environmental regulation by country.</p> <h2>Next step</h2> <p>Match Chinese-financed power plants abroad to satellite measures of nearby air pollution, and compare them with other plants of the same type and age.</p> https://scholar4ai.com/ideas/political-connections-after-the-anti-corruption-campaign 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Political connections are a large topic across the finance, economics and management journals, and a rising one in management. Yet only 10 of 106 recent papers on them carry a corruption keyword. More than a decade after China's anti-corruption campaign began, are connections still worth anything, and to whom?

<h2>What prompted this</h2> <p>In the management journals, "political connections" had 8 papers in July–September 2026 against about 4 expected (see the <a href="/categories/management">management dashboard</a>). Since October 2025, 106 papers carry a political-connection keyword (46 in finance journals, 28 in economics, 23 in management, 8 in accounting). Only 10 of them carry an anti-corruption or corruption keyword, and "anti-corruption campaign" appears on 16 papers in all.</p> <h2>Questions</h2> <ul> <li>Did the value of connections, measured by stock reactions to officials' promotions or downfalls, fall after 2012, and has it recovered?</li> <li>Did firms swap personal ties for other channels, such as hiring retired officials as directors?</li> <li>Do connections still help with bank loans and subsidies, or mainly with avoiding scrutiny?</li> </ul> <h2>Data that could help</h2> <p>Biographies of officials and executives (CSMAR executive data, CNRDS), the dates of announced investigations, and the bank loans and subsidies of listed firms.</p> <h2>Next step</h2> <p>Build an event list of investigations of provincial officials since 2013 and measure the stock reactions of firms whose executives had ties to them.</p> https://scholar4ai.com/ideas/policing-technology 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Policing, law enforcement and crime appear on 63 recent papers, with 18 in the last quarter against about 15 per quarter before, but only 12 carry a technology, surveillance, camera or algorithm keyword, and "predictive policing" on 3. Cameras, licence plate readers and risk scores are spreading fast. Do they cut crime, and at what cost to whom?

<h2>What prompted this</h2> <p>Since October 2025, 63 papers carry a police, policing, law-enforcement, crime-prevention or criminal-justice keyword (33 in economics journals, 16 in AI): "law enforcement" on 13, "crime rates" on 8, "crime prevention" on 7, "policing" on 5, "predictive policing" on 3. Only 12 of the 63 also carry a technology, surveillance, camera, algorithm or AI keyword.</p> <h2>Questions</h2> <ul> <li>Do licence plate readers and camera networks reduce crime, displace it, or mainly raise clearance rates?</li> <li>Do algorithmic risk scores change who is stopped and charged, and do they reduce or reproduce disparities?</li> <li>Did body cameras change complaints and use of force in the long run, once the novelty wore off?</li> </ul> <h2>Data that could help</h2> <p>Crime and arrest records at the beat level, dates of technology deployments by police department, and complaint records.</p> <h2>Next step</h2> <p>Use the staggered installation of licence plate reader networks across US cities to estimate changes in vehicle theft and its displacement to neighbouring areas.</p> https://scholar4ai.com/ideas/place-based-policy-and-chip-clusters 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Industrial policy and place-based policy are both rising in the economics journals, and governments now subsidize chip plants in specific towns, from Arizona to Kumamoto. Yet only 4 of 127 recent papers on such policies mention semiconductors. Do these plants create lasting local clusters, or expensive islands?

<h2>What prompted this</h2> <p>In economics, "industrial policy" had 17 papers in July–September 2026 against 11 expected and "place-based policy" 10 against 4; across all journals "innovation policy" grew too (30 against 16); see the <a href="/trends">trends page</a>. Since October 2025, 90 papers carry an industrial-policy keyword (59 in economics journals), and 13 of them carry a place-based, cluster or regional keyword. Of the 127 papers with either kind of keyword, only 4 carry a semiconductor or chip keyword.</p> <h2>Questions</h2> <ul> <li>Around new subsidized fabs, do local suppliers, wages and new firms grow beyond the plant itself, and for how long?</li> <li>Do housing costs and labour shortages eat the gains, as in some earlier large-plant cases?</li> <li>Do the subsidies change where the next plants and suppliers locate?</li> </ul> <h2>Data that could help</h2> <p>Plant announcements and subsidies (CHIPS Act awards in the US, Japan's support for fabs in Kumamoto and Hokkaido), local employment and business registers, land prices and patents.</p> <h2>Next step</h2> <p>Compare the municipalities around the TSMC-led plant in Kumamoto with similar municipalities, before and after its 2021 announcement, on employment, new firms and land prices.</p> https://scholar4ai.com/ideas/phillips-curve-after-the-pandemic 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The Phillips curve, monetary policy rules and labour market tightness are all rising keywords in the economics journals, after an inflation surge that flat Phillips curves did not predict. Only 7 of 27 recent Phillips-curve papers carry a pandemic, tightness or nonlinearity keyword. Is the curve steep when labour markets are tight, and what does that mean for policy rules?

<h2>What prompted this</h2> <p>In economics, "Phillips curve" (7 papers in July–September 2026 against about 2 expected), "monetary policy rules" (7 against 2) and "labor market tightness" (6 against 1) are all growing (see the <a href="/categories/economics">economics dashboard</a>), and the New Keynesian Phillips curve is on the <a href="/rankings/equations">equation ranking</a>'s list. Since October 2025, 27 papers carry a Phillips-curve keyword (20 in economics journals); 7 of them carry a labour-market-tightness, pandemic or nonlinear keyword.</p> <h2>Questions</h2> <ul> <li>Measured with vacancies per unemployed worker rather than unemployment, is the curve steep when labour markets are tight and flat otherwise?</li> <li>Does the same pattern hold across countries and across US cities, where tightness varied?</li> <li>Would policy rules that respond to tightness have kept inflation steadier?</li> </ul> <h2>Data that could help</h2> <p>Vacancies and unemployment by country and metro area, inflation (FRED, OECD) and inflation expectation surveys.</p> <h2>Next step</h2> <p>Estimate a Phillips curve with the vacancy-to-unemployment ratio and a tightness threshold on US metro data for 2000–2025.</p> https://scholar4ai.com/ideas/pension-funds-and-private-markets 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Pension funds appear on only 19 recent papers, and 2 of them touch infrastructure, private equity or illiquid assets, although pension money is the main source of capital for private markets. As defined-contribution plans are allowed to hold private assets, are the returns worth the fees and the lock-up?

<h2>What prompted this</h2> <p>Since October 2025, 19 papers carry a pension-fund keyword (12 in finance journals), with "pension funds" on 10. Only 2 of them also carry an infrastructure, private-market, private-asset, illiquid, private-equity, private-credit or alternative keyword. "Infrastructure investment" appears on 35 papers, mostly in economics, about public spending rather than investors.</p> <h2>Questions</h2> <ul> <li>After fees, do pension funds' private equity and infrastructure holdings beat listed alternatives over full cycles, and do reported returns smooth the losses?</li> <li>When private assets are opened to defined-contribution savers, who chooses them, and do they understand the lock-up?</li> <li>Do funds that must sell private assets to pay benefits sell at discounts in downturns?</li> </ul> <h2>Data that could help</h2> <p>Pension fund holdings and performance reports, private fund cash flows (Preqin, PitchBook), and secondary-market transaction prices for fund stakes.</p> <h2>Next step</h2> <p>Compare the reported and the cash-flow-based returns of private holdings in the annual reports of large public pension funds for 2015–2025.</p> https://scholar4ai.com/ideas/pay-transparency-laws 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The gender pay gap appears on 55 recent papers, mostly in economics, while pay transparency appears on 2. The EU pay transparency directive takes effect in 2026, and several US states now require salary ranges in job postings. Does transparency narrow the gap, compress pay, or move it into bonuses?

<h2>What prompted this</h2> <p>Since October 2025, 65 papers carry a gender-pay-gap, pay-gap or pay-transparency keyword (53 in economics journals): "gender pay gap" on 55, "executive-employee pay gap" on 6, "pay transparency" on 2. The policy is arriving faster than the research.</p> <h2>Questions</h2> <ul> <li>After salary ranges became mandatory in job postings, did posted ranges narrow, did firms post wider ranges to say nothing, and did the gender gap among new hires change?</li> <li>Under the EU directive's reporting and pay-audit rules, do firms with large gaps raise women's pay, slow men's pay growth or restructure jobs?</li> <li>Do transparency rules reduce individual bargaining, and does that help or hurt the workers who bargained least?</li> </ul> <h2>Data that could help</h2> <p>Job postings with salary ranges (Lightcast), firm pay gap reports (UK and EU), and linked employer-employee data.</p> <h2>Next step</h2> <p>Compare the gender gap in starting pay for jobs posted in US states with and without salary range laws, before and after each law.</p> https://scholar4ai.com/ideas/passive-flows-and-price-impact 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Price impact and fund flows are among the fastest-growing keywords in the finance journals, yet none of the price-impact papers carries a fund-flow or passive-investing keyword. Now that passive funds hold so much of the market, how much do their flows move prices, and are markets as inelastic as some theories claim?

<h2>What prompted this</h2> <p>In the finance journals, "fund flows" had 7 papers in July–September 2026 against about 1 expected from its earlier share, and "price impact" had 12 against about 5 (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 30 papers carry a price-impact keyword (27 in finance journals), and none of them carries a fund-flow, passive, index-fund or ETF keyword.</p> <h2>Questions</h2> <ul> <li>How much does a dollar of passive inflow raise the prices of the stocks it buys, and does the effect revert?</li> <li>Has the price impact of flows grown with the passive share, as the inelastic markets hypothesis predicts?</li> <li>Who sells to passive funds when they buy: active funds, households or the firms themselves?</li> </ul> <h2>Data that could help</h2> <p>Fund holdings and flows (CRSP Mutual Fund Database, Morningstar), ETF creations and redemptions, index reconstitution events and stock returns (CRSP).</p> <h2>Next step</h2> <p>Estimate the price response to flow-induced trading around annual index reconstitutions, year by year from 2010 to 2025.</p> https://scholar4ai.com/ideas/parametric-insurance-and-cat-bonds 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Reinsurance, catastrophe bonds and parametric insurance appear on only 25 recent papers in the finance and operations research journals, although catastrophe bond issuance has reached records and parametric covers are spreading to poorer countries. Do payouts arrive when they are needed, and who bears the basis risk?

<h2>What prompted this</h2> <p>Since October 2025, 25 papers carry a catastrophe-bond, insurance-linked, parametric-insurance or reinsurance keyword (20 in finance journals, 3 in operations research): "reinsurance" on 9, "parametric insurance" on 5, "catastrophe bonds" on 5. Natural disasters, by contrast, appear on 82 papers, and climate risk on 242.</p> <h2>Questions</h2> <ul> <li>How often do parametric triggers miss real losses (no payout after damage, or payout without damage), and does the basis risk fall on governments or households?</li> <li>Do catastrophe bond spreads price climate trends, or only the last few years of losses?</li> <li>Does the record issuance of catastrophe bonds lower reinsurance prices for primary insurers, and does that reach policyholders?</li> </ul> <h2>Data that could help</h2> <p>Catastrophe bond terms and spreads (public issuance data), parametric programme payouts (for example the Caribbean and African risk pools), and loss data from disaster databases such as EM-DAT.</p> <h2>Next step</h2> <p>Match the payouts of regional parametric risk pools since 2015 to reported disaster losses and compute the share of events that triggered no payout.</p> https://scholar4ai.com/ideas/overtourism-and-housing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tourism appears on 130 recent papers across the finance, economics and management journals, but only 11 of them carry an overtourism, housing, short-term-rental, resident or tax keyword. Cities from Barcelona to Kyoto are capping short-term rentals and taxing visitors. Do the measures lower rents, and what do they cost in jobs?

<h2>What prompted this</h2> <p>Since October 2025, 130 papers carry a tourism keyword (39 in finance journals, 36 in economics, 36 in management): "tourism" on 15, "sustainable tourism" on 10, "overtourism" on 5, "short-term rentals" on 4. Only 11 of the 130 also carry an overtourism, housing, resident, crowding, short-term-rental, Airbnb or tax keyword.</p> <h2>Questions</h2> <ul> <li>After a city bans or caps short-term rentals, how much housing returns to the long-term market, and do rents fall in tourist neighbourhoods?</li> <li>Do tourist taxes reduce visitor numbers, or only raise revenue, and where does the revenue go?</li> <li>Which residents gain and lose: tenants, owners who rented to tourists, hospitality workers?</li> </ul> <h2>Data that could help</h2> <p>Short-term rental listings over time (scraped data such as Inside Airbnb), long-term rents, visitor statistics and the dates of regulations by city.</p> <h2>Next step</h2> <p>Track short-term rental listings and long-term rents in Barcelona around the 2024 announcement that tourist flat licences will not be renewed.</p> https://scholar4ai.com/ideas/operating-room-scheduling 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Operating room scheduling and planning appear on 16 recent papers in the operations research journals, among 52 surgery-related papers, most of which are AI work on surgical images. Hospitals rarely implement the optimized schedules. What stops them, and do the schedules survive surgeons' preferences and emergencies?

<h2>What prompted this</h2> <p>Since October 2025, 52 papers carry a surgery or operating-room keyword (32 in AI journals, 14 in operations research): "operating room scheduling" on 13, "surgical navigation" on 7, "surgical planning" on 4, "operating room planning" on 3. The scheduling papers are methods papers; implementation and its obstacles are rarely their subject.</p> <h2>Questions</h2> <ul> <li>In hospitals that implemented optimized block schedules, how much did utilization and overtime change, and how much of the gain survived after a year?</li> <li>How do surgeon preferences, emergency cases and turnover times, which the models simplify, erode the plans?</li> <li>Would models that schedule for robustness against duration uncertainty be adopted more readily than those that maximize utilization?</li> </ul> <h2>Data that could help</h2> <p>Operating room logs with planned and actual times from hospitals that adopted scheduling tools, and interviews with surgical managers.</p> <h2>Next step</h2> <p>Partner with a hospital that adopted a scheduling tool and compare planned and actual utilization in the year before and after adoption.</p> https://scholar4ai.com/ideas/open-weight-models-and-open-source 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Open-source software appears on 13 recent papers and open-source language models on 6, while named open-weight models such as DeepSeek, Llama and Qwen appear on dozens of AI papers as tools; 78 of 126 open-source papers involve language models. Firms give away models worth hundreds of millions to train. Who adopts them, and what do the givers get?

<h2>What prompted this</h2> <p>Since October 2025, 126 papers carry an open-source, open-weight or named open model keyword (96 in AI journals, 13 in economics, 8 in management): "DeepSeek" on 12, "DeepSeek-R1" on 11, "Llama 3" on 10, "Qwen2.5" on 6, "open source software" on 13, "open-source LLMs" on 6. Seventy-eight of the 126 also carry a language-model, foundation-model or fine-tuning keyword. The economics of open source, on 41 papers with 9 touching firms or investment, has not caught up with models.</p> <h2>Questions</h2> <ul> <li>Which firms deploy open-weight models rather than paid APIs, and do they do so for cost, privacy, control or regulation?</li> <li>Do the firms that release open models gain in cloud revenue, hiring, standards or valuation, and did the release of strong open models from China change the strategy of US firms?</li> <li>Does the availability of open models lower prices for closed ones, as open-source software did for proprietary software?</li> </ul> <h2>Data that could help</h2> <p>Model download and fine-tuning statistics (Hugging Face), job postings naming models, API price histories and firm disclosures.</p> <h2>Next step</h2> <p>Track API prices of closed models around the release dates of major open-weight models and measure the price response.</p> https://scholar4ai.com/ideas/open-innovation-with-generative-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Open innovation and crowdsourcing appear on 71 recent papers, mostly in management, but only 3 of them carry an AI keyword. Firms ran idea contests and crowdsourcing platforms to get ideas from outside. If a language model can generate thousands of ideas in an hour, what is the crowd still for, and how do firms tell good ideas from fluent ones?

<h2>What prompted this</h2> <p>Since October 2025, 71 papers carry an open-innovation or crowdsourcing keyword (39 in management journals, 11 in economics): "open innovation" on 42, "crowdsourcing" on 13, "open innovation ecosystems" on 5. Only 3 of the 71 also carry an artificial-intelligence, generative-AI, language-model or machine-learning keyword.</p> <h2>Questions</h2> <ul> <li>In idea contests, how do ideas from language models compare with crowd ideas on novelty, feasibility and what firms actually implement?</li> <li>Do participants in crowdsourcing platforms now submit AI-generated ideas, and does that raise the volume and lower the diversity of submissions?</li> <li>Does the value of open innovation shift from generating ideas to testing them with outside users?</li> </ul> <h2>Data that could help</h2> <p>Submissions and outcomes from innovation contests and platforms, with timestamps around 2023, and experiments that mix model-generated and human ideas for blind evaluation.</p> <h2>Next step</h2> <p>Run a blind evaluation in which managers rate a mix of crowd-sourced and model-generated ideas for a real problem, and track which ones they implement.</p> https://scholar4ai.com/ideas/open-data-mandates 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Open data, data sharing and open science appear on 121 recent papers, with 36 in the last quarter against about 26 per quarter before, spread across every tracked field. Funders and journals now require data and code, and open government data portals have multiplied. Does the shared data get reused, and do mandates change what researchers study?

<h2>What prompted this</h2> <p>Since October 2025, 121 papers carry an open-data, data-sharing, open-science, data-availability or replication-package keyword (32 in AI journals, 29 in economics, 27 in finance, 20 in management, 12 in operations research): "open data" on 35, "data sharing" on 24, "open data portals" on 9, "open science" on 9, "data availability" on 7. The quarter to September 2026 had 36 such papers against about 26 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Of the datasets and code deposited under journal mandates, what share can be downloaded and run years later, and what share is cited or reused?</li> <li>Do mandates push researchers toward questions that can be answered with shareable data and away from proprietary data such as bank records and firm microdata?</li> <li>Do open government data portals get used by firms and researchers in proportion to what they cost, and which datasets generate the use?</li> </ul> <h2>Data that could help</h2> <p>Journal data repositories and their download and citation logs, portal usage statistics, and the data sources named in published papers over time.</p> <h2>Next step</h2> <p>Sample 300 replication packages deposited under mandates in 2019–2021 and record whether each still downloads, runs and has been cited.</p> https://scholar4ai.com/ideas/oil-prices-inflation-and-electric-vehicles 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Oil prices are a rising keyword in the economics journals, and oil-price shocks remain a standard driver in inflation models. But none of the 189 recent oil-price papers carries an electric-vehicle keyword. As electric vehicles spread, is oil's grip on inflation weakening, and where first?

<h2>What prompted this</h2> <p>In economics, "oil prices" had 16 papers in July–September 2026 against about 10 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, 189 papers carry an oil-price keyword (99 in economics journals, 78 in finance); 26 of them also carry an inflation keyword, and none carries an electric-vehicle keyword, while "electric vehicle adoption" appears on 20 papers.</p> <h2>Questions</h2> <ul> <li>Has the pass-through of oil-price shocks to consumer prices fallen where electric vehicles took a large share of new car sales, such as Norway, China or California?</li> <li>Does the change show up first in fuel spending, then in transport services, then in core inflation?</li> <li>What does it mean for how central banks should read oil-price spikes?</li> </ul> <h2>Data that could help</h2> <p>Oil prices, national and regional consumer price indices, the electric share of car sales and of the car fleet, and household fuel spending.</p> <h2>Next step</h2> <p>Estimate the pass-through of oil-price changes to consumer prices by country and year for 2010–2025, and relate it to the electric share of the car fleet.</p> https://scholar4ai.com/ideas/nuclear-restarts-and-small-reactors 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Nuclear energy appears on 26 recent papers and small modular reactors on 4, mostly in the economics journals. Technology firms now sign deals to restart closed plants and fund new reactors for data centres, and Japan has restarted reactors idle since 2011. What do these deals cost, who bears the risk, and does public acceptance follow?

<h2>What prompted this</h2> <p>Since October 2025, 46 papers carry a nuclear-energy keyword (34 in economics journals, 7 in finance): "nuclear energy" on 26, "small modular reactors" on 4, "nuclear energy policy" on 3. Twenty-two of them carry an acceptance, cost, investment or public-opinion keyword, but the new financing model, in which a single large buyer underwrites a plant, appears in none.</p> <h2>Questions</h2> <ul> <li>What price per megawatt-hour do data-centre operators pay in restart and new-build agreements, and how does it compare with renewables plus storage?</li> <li>Who bears cost overruns in these contracts, and do they shift risk to ratepayers through capacity markets?</li> <li>Did the restarts in Japan change local acceptance, and did electricity prices in those regions fall?</li> </ul> <h2>Data that could help</h2> <p>Published power purchase agreements and regulatory filings for restarts, construction cost records for small modular reactors, regional electricity prices and public opinion surveys.</p> <h2>Next step</h2> <p>Collect the disclosed terms of nuclear power purchase agreements signed by technology firms since 2023 and compare their prices with contemporaneous renewable contracts.</p> https://scholar4ai.com/ideas/nowcasting-with-alternative-data 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

High-frequency data, big data and alternative data appear on 236 recent papers across the finance, AI and economics journals, but only 10 of them carry a card-transaction, mobility, search, satellite or electricity keyword, and nowcasting appears on 9. Banks, card networks and job boards hold daily measures of the economy. Which of them nowcast GDP, inflation and employment best, and could statistical offices use them?

<h2>What prompted this</h2> <p>Since October 2025, 236 papers carry a nowcasting, alternative-data, high-frequency-data, big-data or real-time-data keyword (66 in finance journals, 62 in AI, 54 in economics): "high-frequency data" on 67, "big data" on 50, "big data analytics" on 28, "alternative data" on 11, "nowcasting" on 9. Only 10 of the 236 also carry a card-transaction, credit-card, mobility, Google Trends, search, payment, satellite, electricity or job-posting keyword, although "Google Trends" appears on 22 papers in the archive.</p> <h2>Questions</h2> <ul> <li>In real time, with the data available on each day, which alternative sources improved nowcasts of GDP, inflation and unemployment during 2020–2025, and did the gains survive after the pandemic?</li> <li>Do card transaction and job posting data nowcast regional and sectoral activity that official statistics report with long lags?</li> <li>What would it cost statistical offices to license the data, and do the providers' incentives bias what they release?</li> </ul> <h2>Data that could help</h2> <p>Daily card spending indices published by banks and card networks, job posting counts, mobility data, electricity consumption and the official release calendars.</p> <h2>Next step</h2> <p>Build a real-time nowcasting exercise for quarterly GDP in two countries with and without card spending and job posting data for 2019–2025.</p> https://scholar4ai.com/ideas/neural-solvers-for-continuous-time-finance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Continuous-time models built on the Hamilton-Jacobi-Bellman equation are rising in the finance journals, while neural operators and physics-informed networks that solve such equations are growing in AI. All five recent papers that pair the two are in AI journals. Can these solvers make richer portfolio and asset-pricing models usable?

<h2>What prompted this</h2> <p>In the finance journals, "Hamilton-Jacobi-Bellman equation" had 8 papers in July–September 2026 against about 2 expected (see the <a href="/categories/finance">finance dashboard</a>); the Bellman equation is also on the <a href="/rankings/equations">equation ranking</a>'s list. In the AI journals, "neural operators" grew (9 against 3), and "physics-informed neural networks" appears on 84 papers since October 2025. Of the 70 papers with an HJB, portfolio-choice or optimal-investment keyword, 5 also carry a neural-network or machine-learning keyword, and all 5 are in AI journals.</p> <h2>Questions</h2> <ul> <li>Can neural solvers handle portfolio problems with many assets, frictions or stochastic volatility that grid methods cannot?</li> <li>How accurate are they where closed forms exist, and do the errors matter more for policies than for values?</li> <li>Can a solver trained once (an operator) let researchers estimate structural models that need to be solved thousands of times?</li> </ul> <h2>Data that could help</h2> <p>Benchmark problems with known solutions, and calibrations from standard asset-pricing models.</p> <h2>Next step</h2> <p>Solve a portfolio problem with stochastic volatility that has a known solution using a physics-informed network, and compare the optimal policies.</p> https://scholar4ai.com/ideas/net-metering-reform-and-rooftop-solar 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Rooftop solar, prosumers and net metering appear on 50 recent papers, mostly in economics, and the topic is fading, with 7 papers in the last quarter against about 14 per quarter before. California and other markets have cut the payment for exported power. Did installations collapse, and did batteries take their place?

<h2>What prompted this</h2> <p>Since October 2025, 50 papers carry a rooftop-solar, residential-solar, net-metering, prosumer or distributed-generation keyword (39 in economics journals): "rooftop solar" on 15, "prosumers" on 13, "distributed generation" on 13, "net metering" on 6. Seven were published in July–September 2026 against about 14 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>After the payment for exported power was cut, how far did new installations fall, and did the share with batteries rise?</li> <li>Did the reform reduce the cost shift from solar owners to other customers, as its supporters claimed, and by how much?</li> <li>Do solar owners change when they use electricity once exports pay less?</li> </ul> <h2>Data that could help</h2> <p>Interconnection records with installation dates and battery pairing, utility tariff changes, and smart meter data.</p> <h2>Next step</h2> <p>Compare monthly residential solar installations and battery attachment rates in California before and after the 2023 net billing change with those in neighbouring states.</p> https://scholar4ai.com/ideas/negative-electricity-prices-and-storage 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Energy markets are one of the rising keywords in the finance and economics journals, and wholesale power prices now often turn negative when solar output peaks. Batteries profit from that spread, but they also erase it. How much is storage worth as it scales, and who should pay for it?

<h2>What prompted this</h2> <p>"Energy markets" grew across all journals (16 papers in July–September 2026 against about 8 expected) and in finance (11 against about 5); see the <a href="/trends">trends page</a>. Since October 2025, "electricity prices" appears on 39 papers, "battery energy storage" on 38 (31 in economics journals) and "renewable energy integration" on 69. Of the 93 papers with an electricity-price keyword, 17 also carry a storage or battery keyword.</p> <h2>Questions</h2> <ul> <li>How fast does the arbitrage value of batteries fall as more storage enters a market?</li> <li>Do capacity payments or other support for storage pay off for consumers, or mainly for owners?</li> <li>How should operations research models of battery dispatch account for the price effect of all batteries acting together?</li> </ul> <h2>Data that could help</h2> <p>Hourly wholesale prices and generation by source (ENTSO-E, US grid operators, Nord Pool), battery capacity by market, and capacity-market results.</p> <h2>Next step</h2> <p>For one market with fast battery growth, regress the daily spread between peak and off-peak prices on installed storage capacity, controlling for solar output.</p> https://scholar4ai.com/ideas/narcissistic-ceos-from-text 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Narcissism, CEO characteristics and CEO personality appear on 42 recent papers, mostly in the management and accounting journals, with measures built from signatures, photos and pronoun counts. Language models can now score personality from thousands of hours of speech. Do the old measures and the new ones agree, and does personality predict decisions out of sample?

<h2>What prompted this</h2> <p>Since October 2025, 42 papers carry a narcissism, CEO-personality, CEO-trait, hubris or CEO-characteristics keyword (19 in management journals, 10 in accounting): "narcissism" on 14, "CEO characteristics" on 12, "CEO narcissism" on 7, "CEO personality" on 3. Eleven were published in July–September 2026, in line with earlier quarters.</p> <h2>Questions</h2> <ul> <li>Do narcissism scores from signatures, photo size and pronoun use agree with scores that language models assign from interviews and earnings calls, and which predicts acquisitions, risk taking and fraud better?</li> <li>Do the published effects of CEO narcissism replicate in new samples with the new measures?</li> <li>Do boards hire different personalities after a scandal, and can that be seen in the data?</li> </ul> <h2>Data that could help</h2> <p>Earnings call transcripts and interviews, proxy statements with signatures and photos, deal and accounting data (SDC, Compustat, Audit Analytics).</p> <h2>Next step</h2> <p>Score the same 500 CEOs with the signature-based measure and a language-model measure from their earnings calls, and compare the two against acquisition behaviour.</p> https://scholar4ai.com/ideas/multiple-testing-in-research-on-chinese-listed-firms 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

"Chinese listed companies" is the third most frequent keyword across the tracked journals over the past year, and most of these papers draw on the same few databases. Yet none of the 1,392 recent papers with such a keyword carries a replication, multiple-testing or publication-bias keyword. How many of their findings would survive a correction for everyone testing on the same data?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "Chinese listed companies" is the third most frequent keyword over the last 12 complete months, after "China" and "large language models". Since October 2025, 1,392 papers carry a Chinese-listed-company keyword (928 in finance journals, 216 in economics, 172 in management, 54 in accounting). None of them carries a multiple-testing, p-hacking, replication, publication-bias, data-mining or false-discovery keyword.</p> <h2>Questions</h2> <ul> <li>If every published test on the same panel of firms counts as one of many, how many reported effects clear a stricter threshold, such as a t-statistic of 3?</li> <li>Do the results hold out of sample, for example in later years or in firms listed after a paper's sample ends?</li> <li>Which outcomes, such as green innovation, ESG scores or digital transformation, attract the most tests?</li> </ul> <h2>Possible approach</h2> <p>Collect the main reported t-statistics and sample periods from a large set of papers that use CSMAR or Wind data, and estimate the false discovery rate with the methods developed for the "factor zoo" in asset pricing.</p> <h2>Next step</h2> <p>Extract the main coefficient and t-statistic from 200 recent papers on Chinese listed firms and plot their distribution around 1.96.</p> https://scholar4ai.com/ideas/multi-tier-supply-chain-visibility 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Supply chain transparency, traceability and visibility appear on 83 recent papers, and the topic is fading, with 16 in the last quarter against about 19 per quarter before. Most disruptions and most forced-labour and emissions problems sit beyond the first tier of suppliers, where firms rarely look. How far can visibility be built, and does it pay?

<h2>What prompted this</h2> <p>Since October 2025, 83 papers carry a supply-chain-visibility, supply-chain-mapping, multi-tier, supply-chain-transparency or traceability keyword (24 in management journals, 20 in AI, 17 in operations research): "supply chain transparency" on 35, "traceability" on 15, "supply chain traceability" on 8, "supply chain visibility" on 5. Sixteen were published in July–September 2026 against about 19 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When firms map their suppliers' suppliers, how often do they find shared sub-tier suppliers that make their alternative sources less alternative than they thought?</li> <li>Do firms that mapped deeper before 2020 recover faster from disruptions, and do investors reward disclosed mapping?</li> <li>Can shipping manifests and customs data reconstruct sub-tier networks well enough for risk management, without suppliers' cooperation?</li> </ul> <h2>Data that could help</h2> <p>Bill-of-lading and customs data (for example from Panjiva or ImportGenius), supplier disclosure lists, and disruption event records.</p> <h2>Next step</h2> <p>Reconstruct the two-tier supply network of a set of electronics firms from shipping manifests and measure how concentrated the second tier is.</p> https://scholar4ai.com/ideas/moving-relief-before-the-disaster 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Humanitarian logistics is an active topic in operations research, and climate shocks a rising one in economics. Aid agencies now release money and move supplies when a forecast crosses a threshold, before a flood or drought hits. Only 11 of 103 recent humanitarian-logistics papers carry a forecast, early-warning or climate keyword. When does acting early pay off?

<h2>What prompted this</h2> <p>Since October 2025, 103 papers carry a humanitarian-logistics, disaster-relief, emergency-logistics or relief keyword (48 in operations research journals). Only 11 of them carry a forecast, anticipatory, early-warning, climate or pre-positioning keyword. "Climate shocks" appears on 44 papers, 32 of them in economics.</p> <h2>Questions</h2> <ul> <li>Given how good the forecasts are, at what probability of a flood or drought should supplies be moved, and how much is lost to false alarms?</li> <li>Does acting early reduce deaths, distress sales of assets and hunger more than the same money spent after the disaster?</li> <li>Where should warehouses be when forecasts are unsure where the shock will land?</li> </ul> <h2>Data that could help</h2> <p>Archived forecasts and outcomes (for example river-flood forecasts and rainfall records), aid agencies' data on early actions, and household surveys after disasters.</p> <h2>Next step</h2> <p>Build a two-stage stochastic model that moves supplies according to archived flood forecasts for one river basin, and compare it with responding after the flood.</p> https://scholar4ai.com/ideas/moderators-and-mediators-everywhere 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

"Moderating effect" and "mediating effect" are among the most frequent keywords in finance and management papers, and mediation models rank high on the equation ranking. How robust are these interaction and mediation results, and what would good practice look like?

<h2>What prompted this</h2> <p>Since October 2025, "moderating effect" appears on 430 papers (219 in finance journals, 146 in management) and "mediating effect" on 267 (171 in finance); 29 carry both. "Moderating effect" is the 36th most frequent keyword of the last 30 days, and it is growing in the accounting journals (8 papers in July–September 2026 against about 4 expected). On the <a href="/rankings/equations">equation ranking</a>, the mediation model is mentioned by 62 papers in 30 days, 29 of them in finance.</p> <h2>Questions</h2> <ul> <li>How many moderators does a typical paper test, and does it correct for testing many?</li> <li>How many mediation claims rest on the three-step regressions of Baron and Kenny, which causal mediation analysis has questioned, and how many would survive its assumptions?</li> <li>Are interactions reported with marginal effects and enough statistical power?</li> </ul> <h2>Possible approach</h2> <p>Code a sample of recent papers for the number of moderators and mediators tested, the inference method, corrections for multiple testing and the availability of replication data.</p> <h2>Next step</h2> <p>Draw the finance papers with a "moderating effect" keyword from <a href="/data/papers.json">papers.json</a> and record how many moderators each one tests.</p> https://scholar4ai.com/ideas/model-collapse-and-synthetic-data 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Synthetic data generation appears on 77 recent papers and synthetic data on 18, almost all in AI journals, while model collapse, the degradation of models trained on their predecessors' output, appears on 3. The web is filling with generated text and images that will train the next models. How fast does quality decay, and what does that do to the value of human data?

<h2>What prompted this</h2> <p>Since October 2025, 117 papers carry a synthetic-data, model-collapse, self-consuming or recursive-training keyword (111 in AI journals): "synthetic data generation" on 77, "synthetic data" on 18, "synthetic data augmentation" on 10, "model collapse" on 3. Eleven of the 117 also carry a collapse, degradation, recursive, distribution-shift or contamination keyword, and "data quality" appears on 30 papers in the archive.</p> <h2>Questions</h2> <ul> <li>When models are trained on mixtures of human and generated data at the shares now seen on the web, how fast do rare patterns and tails disappear, and does filtering by quality stop it?</li> <li>Does the decay differ for text, images and tabular data used in finance and economics?</li> <li>If human-written data becomes scarce, what is it worth, and who owns the archives that have it?</li> </ul> <h2>Data that could help</h2> <p>Web crawl samples by year with AI-text detection, controlled training runs on mixed data, and licensing deals for data archives.</p> <h2>Next step</h2> <p>Train a small model on successive generations of mixed human and generated data at web-realistic shares and measure the loss of tail vocabulary and rare facts per generation.</p> https://scholar4ai.com/ideas/mobile-money-as-climate-insurance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Financial inclusion is a large topic in the finance journals, and climate shocks a growing one in economics, but only 14 of 367 recent inclusion and payments papers carry a climate or disaster keyword. When droughts and floods hit, does mobile money bring help fast enough to keep households from selling assets or taking children out of school?

<h2>What prompted this</h2> <p>Since October 2025, 367 papers carry a financial-inclusion, digital-payment or mobile-money keyword (229 in finance journals, 88 in economics), and only 14 of them carry a climate, weather, drought, flood, disaster or shock keyword. "Climate shocks" appears on 44 papers, 32 of them in economics. In the management journals, "inclusive growth" (6 papers in July–September 2026 against about 1 expected) and "developing countries" are rising.</p> <h2>Questions</h2> <ul> <li>After droughts or floods, do households with mobile money receive more remittances, and faster?</li> <li>Does that protect their consumption, their assets (livestock, land) and their children's schooling?</li> <li>Does the effect hold where agent networks are thin, or does the shock knock out the network itself?</li> </ul> <h2>Data that could help</h2> <p>Geocoded household surveys (LSMS, DHS), rainfall and flood data (ERA5, satellite flood maps), and mobile money agent locations and transactions where available.</p> <h2>Next step</h2> <p>Merge household panels from one East African country with rainfall shocks and the households' mobile money use, and compare how much consumption falls after droughts.</p> https://scholar4ai.com/ideas/mixture-of-experts-as-regime-switching 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Mixture-of-experts models are among the fastest-growing topics in the AI journals, and regime-switching models have a long history in finance and economics. A gate that picks experts works much like a regime switch, yet no finance or economics paper in the archive carries the keyword. Can it forecast better across market regimes?

<h2>What prompted this</h2> <p>"Mixture of experts" is among the top growing keywords across all journals on the <a href="/trends">trends page</a> (47 papers in July–September 2026 against 31 expected). Since October 2025 it appears on 152 papers: 150 in AI journals, 2 in operations research, and none in finance or economics. Regime switching appears on 70 papers (33 in finance, 27 in economics), and only one of them also carries a deep-learning, neural-network or mixture-of-experts keyword. The Markov-switching model is on the <a href="/rankings/equations">equation ranking</a>, half of its papers in finance.</p> <h2>Questions</h2> <ul> <li>In volatility or inflation forecasting, does a mixture of experts with a learned gate beat a Markov-switching model, and do its experts look like regimes such as calm and crisis?</li> <li>Can a Markov-switching likelihood serve as a structured gate, giving a hybrid of the two?</li> <li>Do the regimes it finds line up with known events?</li> </ul> <h2>Data that could help</h2> <p>Realized volatility from intraday returns, the VIX, and monthly macro series such as FRED-MD.</p> <h2>Next step</h2> <p>Fit a two-expert mixture and a two-state Markov-switching model to the daily realized volatility of the S&amp;P 500, and compare their forecasts and inferred regimes.</p> https://scholar4ai.com/ideas/mixed-ownership-reform 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Non-state-owned enterprises are a rising keyword in the economics journals, and state ownership is one of the most common comparisons in research on Chinese firms. Yet "mixed ownership reform" appears on only 9 recent papers. When private investors buy into state firms, do governance and performance change, or only the shareholder list?

<h2>What prompted this</h2> <p>In economics, "non-state-owned enterprises" had 12 papers in July–September 2026 against 7 expected (see the <a href="/categories/economics">economics dashboard</a>). Over the last 12 complete months "state-owned enterprises" appeared on 666 papers, and since October 2025 "non-state-owned enterprises" on 169 (107 in finance journals). "Mixed ownership reform" appears on just 9 papers and "mixed ownership" on 2.</p> <h2>Questions</h2> <ul> <li>When private investors take board seats in state firms, do investment efficiency, pay structures and profitability change?</li> <li>Does it matter whether the new investor is a strategic partner, a fund or another state-linked firm?</li> <li>Do the firms' banks and suppliers treat them differently afterwards?</li> </ul> <h2>Data that could help</h2> <p>Ownership and board data of listed firms (CSMAR, CNRDS), announcements of mixed-ownership deals, and firm accounts.</p> <h2>Next step</h2> <p>Collect announced mixed-ownership deals of listed state firms since 2015, and compare board composition and investment efficiency before and after.</p> https://scholar4ai.com/ideas/minimum-wages-and-automation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The minimum wage appears on 34 recent papers in the economics and finance journals, but only 2 of 44 minimum-wage papers carry an automation, robot or technology keyword. Self-checkouts, kiosks and warehouse robots are cheapest where wages are highest. Do minimum wage increases speed up automation, and over what horizon?

<h2>What prompted this</h2> <p>Since October 2025, 44 papers carry a minimum-wage keyword (35 in economics journals, 8 in finance), with "minimum wage" on 34. Only 2 of them also carry an automation, robot, technology, capital-labour or self-service keyword, while "labor demand" appears on 32 papers and industrial robots are a rising topic elsewhere.</p> <h2>Questions</h2> <ul> <li>After large minimum wage increases, do firms in affected areas buy more automation equipment and post fewer low-wage jobs, with a lag of years rather than months?</li> <li>Which tasks go first, and do the remaining jobs pay more or require more skill?</li> <li>Do the employment effects found in short-run studies understate long-run effects that run through automation?</li> </ul> <h2>Data that could help</h2> <p>Minimum wage changes by state and country, equipment investment and job postings by establishment, and robot or kiosk adoption from supplier data.</p> <h2>Next step</h2> <p>Relate state minimum wage changes in the US to the share of restaurant and retail job postings that mention self-service technology over the following three years.</p> https://scholar4ai.com/ideas/micromobility-after-the-boom 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Bike sharing, electric scooters and shared micromobility appear on 19 recent papers, mostly in operations research, and the topic is fading, with 2 papers in the last quarter against about 6 per quarter before. Several operators have failed and cities have banned scooters. Which systems survived, and did they replace car trips or walking?

<h2>What prompted this</h2> <p>Since October 2025, 19 papers carry a bike-sharing, micromobility, e-scooter or shared-mobility keyword (14 in operations research journals): "bike sharing" on 11, "electric scooters" on 3, "shared micromobility" on 2, "dockless bike-sharing" on 2. Only 2 were published in July–September 2026 against about 6 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Across cities, which system designs (docked, dockless, publicly subsidized, operator-funded) survived the operator failures of 2022–2024, and at what cost per trip?</li> <li>What share of micromobility trips replaced car trips, and did the share change as novelty wore off?</li> <li>Did the scooter bans in Paris and elsewhere shift trips to bikes, transit or cars?</li> </ul> <h2>Data that could help</h2> <p>Trip data published by operators and cities, operator entries and exits by city, and travel surveys before and after bans.</p> <h2>Next step</h2> <p>Compile the entry and exit of micromobility operators across 50 cities since 2018 and relate survival to the system design and subsidy model.</p> https://scholar4ai.com/ideas/microfinance-meets-digital-lenders 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Microfinance and microcredit appear on 35 recent papers, and only 2 of them carry a regulation, digital, fintech or competition keyword. App-based lenders now compete with microfinance institutions in Africa and Asia, and several countries have capped rates. Do borrowers end up with cheaper credit, or more of it?

<h2>What prompted this</h2> <p>Since October 2025, 35 papers carry a microfinance or microcredit keyword (18 in finance journals, 10 in economics); "microfinance" appears on 17 and "microfinance institutions" on 14. Only 2 of the 35 also carry an interest-rate-cap, regulation, digital, fintech, competition or over-indebtedness keyword. Financial inclusion as a whole is a large topic, so the gap is specific to how microfinance institutions compete now.</p> <h2>Questions</h2> <ul> <li>When digital lenders enter a market, do microfinance institutions lose their best borrowers, cut rates or move up-market?</li> <li>Do interest rate caps push lenders out of rural areas, as earlier caps did, or do digital lenders' lower costs change that?</li> <li>Do borrowers with both app loans and microfinance loans become over-indebted?</li> </ul> <h2>Data that could help</h2> <p>Microfinance institution financials (the MIX Market successor datasets), credit bureau data where digital lenders report, and the dates of rate caps by country.</p> <h2>Next step</h2> <p>Compare microfinance institutions' loan growth and portfolio quality in Kenya, where digital lending is widespread, before and after the 2021 digital credit regulations.</p> https://scholar4ai.com/ideas/metaverse-retail-after-the-hype 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Virtual reality, the metaverse, augmented reality and immersive technologies appear on 123 recent papers, but the topic is fading, with 21 papers in the last quarter against about 33 per quarter before; 23 of the papers concern retail or consumers. Which immersive applications survived the hype cycle, and what do their users actually buy?

<h2>What prompted this</h2> <p>Since October 2025, 123 papers carry a virtual-reality, metaverse, augmented-reality, immersive or extended-reality keyword (60 in AI journals, 44 in management): "virtual reality" on 37, "metaverse" on 34, "augmented reality" on 30, "immersive technologies" on 13. Twenty-one were published in July–September 2026 against about 33 per quarter earlier. Twenty-three of the 123 carry a retail, shopping, consumer, purchase, marketing, brand or tourism keyword.</p> <h2>Questions</h2> <ul> <li>Which uses (virtual try-on, showrooms, training, tourism previews) produced measurable sales or savings, and which were abandoned?</li> <li>Do consumers who try products in augmented reality return fewer goods?</li> <li>Did firms that invested heavily in metaverse projects in 2021–2022 write them off, and did markets punish or reward the retreat?</li> </ul> <h2>Data that could help</h2> <p>Retailer data on augmented-reality feature use and returns, firms' project announcements and later disclosures, and app usage statistics.</p> <h2>Next step</h2> <p>Compare return rates for products bought with and without virtual try-on at one retailer, controlling for product and customer.</p> https://scholar4ai.com/ideas/merger-review-and-deal-outcomes 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Mergers and acquisitions appear on 228 recent papers, but only 31 of them carry a merger-control, competition-policy or enforcement keyword. Merger enforcement tightened in the US and Europe after 2021 and loosened again in places since. Did the threat of review stop deals, reshape them, or simply delay them?

<h2>What prompted this</h2> <p>Since October 2025, 228 papers carry a merger, acquisition or takeover keyword (111 in finance journals, 68 in economics, 24 in management), with "mergers and acquisitions" on 122. Only 31 of them also carry an antitrust, merger-control, competition-policy, enforcement, competition-law or market-power keyword; "merger control" appears on 11 and "competition policy" on 34.</p> <h2>Questions</h2> <ul> <li>After the US agencies' tougher guidelines in 2023, did the number of large horizontal deals fall, or did firms switch to deals below reporting thresholds?</li> <li>Do deals that are challenged and then completed with remedies deliver the price effects that the remedies were meant to prevent?</li> <li>Did acquirers' stock reactions to deal announcements change with the perceived enforcement regime?</li> </ul> <h2>Data that could help</h2> <p>Deal data (SDC Platinum, Zephyr), enforcement actions and second requests, and product prices in affected markets.</p> <h2>Next step</h2> <p>Count announced deals by size relative to the reporting threshold before and after the 2023 guideline change and look for bunching just below it.</p> https://scholar4ai.com/ideas/measuring-overconfidence-from-executive-speech 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

CEO and managerial overconfidence appear on 98 recent papers, mostly measured from option-exercise behaviour or press coverage, as in the 2000s. Only 6 of them carry a text, language or speech keyword, and 4 of those are in AI journals. Can a language model score overconfidence from earnings calls, and does it predict the same decisions?

<h2>What prompted this</h2> <p>Since October 2025, 98 papers carry an overconfidence keyword (44 in finance journals, 24 in AI, 18 in economics): "overconfidence" on 51, "CEO overconfidence" on 22, "managerial overconfidence" on 15. Only 6 of the 98 also carry a text, language, machine-learning, speech, tone or linguistic keyword, and 4 of those 6 are in AI journals.</p> <h2>Questions</h2> <ul> <li>Does a language model's reading of CEOs' answers on earnings calls (certainty, dismissal of risks, forecast ranges) agree with the option-based measure, and where do they disagree?</li> <li>Which measure better predicts overinvestment, overpriced acquisitions and later restatements?</li> <li>Can the text measure be computed for private firms and non-US firms, where option data do not exist?</li> </ul> <h2>Data that could help</h2> <p>Earnings call transcripts, option exercise data (ExecuComp), deal and investment data (Compustat, SDC).</p> <h2>Next step</h2> <p>Score 1,000 earnings calls with a language model for overconfidence and compare the scores with the option-based measure for the same CEOs.</p> https://scholar4ai.com/ideas/measuring-digital-transformation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Digital transformation is among the most frequent keywords in the finance and management journals, much of it about Chinese listed firms, and studies often score it by counting digital terms in annual reports. Do those scores track what firms actually invest and hire?

<h2>What prompted this</h2> <p>Since October 2025, "digital transformation" appears on 544 papers (211 in management journals, 204 in finance, 74 in economics), and "enterprise digital transformation" on 80 more; it is the 10th most frequent keyword of the last 30 days. Only 7 of the 634 papers with either keyword also carry a text-analysis keyword, although indices built from word frequencies in annual reports are a common way to measure it.</p> <h2>Questions</h2> <ul> <li>How well do text scores correlate with hard measures such as software and IT investment, digital job postings, cloud contracts and digital patents?</li> <li>How much of a score is boilerplate copied from year to year, or a response to policy campaigns?</li> <li>Do headline results on productivity, ESG or innovation hold when hard measures replace the text score?</li> </ul> <h2>Data that could help</h2> <ul> <li>Annual reports of Chinese listed companies, and the ready-made digital indices in CSMAR and CNRDS (the China financial and firm databases on the <a href="/rankings/datasets">dataset ranking</a>).</li> <li>Online job postings, CNIPA patents, and the software items in fixed-asset notes.</li> </ul> <h2>Next step</h2> <p>For one year of annual reports, correlate a common word-count index with each firm's digital job postings and software assets.</p> https://scholar4ai.com/ideas/marriage-markets-and-sex-ratios 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Marriage, the marriage market, sex ratios at birth and bride prices appear on 59 recent papers, almost all in the economics journals. In China and India, decades of skewed sex ratios have left millions of men unable to marry. What does that do to saving, crime, migration and the price of a wedding?

<h2>What prompted this</h2> <p>Since October 2025, 59 papers carry a sex-ratio, marriage-market, gender-imbalance, bride-price or marriage keyword (52 in economics journals): "marriage" on 12, "sex ratio at birth" on 9, "marriage market" on 8, "marriage timing" on 6, "delayed marriage" on 4, "bride price" on 4. The cohorts born under the most skewed ratios are now of marriage age.</p> <h2>Questions</h2> <ul> <li>In counties with the most skewed sex ratios, how much more do families with sons save, and does the saving show up in housing prices and bride prices?</li> <li>Do unmarried men migrate more, commit more crime, or take riskier jobs?</li> <li>As the ratio at birth normalizes, do bride prices and saving rates fall with it?</li> </ul> <h2>Data that could help</h2> <p>Census and population registers with sex ratios by cohort and county, household panels with marriage and saving data (China Family Panel Studies, China Household Finance Survey), and local crime statistics.</p> <h2>Next step</h2> <p>Relate county-level sex ratios of the 1990–2000 birth cohorts to current household saving rates and marriage payments in Chinese household survey data.</p> https://scholar4ai.com/ideas/markups-and-inflation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Market power, price setting and markups appear on 186 recent papers, but only 8 of them carry an inflation keyword. The claim that firms used the inflation surge to raise margins ("greedflation") shaped policy debates. Did markups rise, in which sectors, and did they fall back?

<h2>What prompted this</h2> <p>Since October 2025, 186 papers carry a markup, profit-margin, price-setting or market-power keyword (107 in economics journals, 35 in finance): "market power" on 87, "price setting" on 34, "markups" on 29. Only 8 of them also carry an inflation keyword, although "inflation" appears on 211 papers in the archive.</p> <h2>Questions</h2> <ul> <li>Using firm-level data, did markups rise during 2021–2023 beyond what cost pass-through explains, and did they fall when costs eased?</li> <li>Were margin increases concentrated in sectors with supply shortages, with few competitors, or with fast price changes?</li> <li>Did consumers' expectations of price increases make it easier for firms to raise prices in step?</li> </ul> <h2>Data that could help</h2> <p>Firm accounts with cost and revenue (Compustat, Orbis), producer and consumer price indices by sector, and estimates of markups from production functions.</p> <h2>Next step</h2> <p>Estimate quarterly markups for listed firms by sector for 2019–2025 and compare their path with input cost indices.</p> https://scholar4ai.com/ideas/market-manipulation-detection 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Market manipulation, stock market manipulation, price manipulation and pump-and-dump schemes appear on about 20 recent papers, split between the AI and finance journals. Spoofing, wash trading and coordinated pumps now happen in milliseconds and across venues, including crypto. Can detectors trained on known cases find new ones, and what do regulators miss?

<h2>What prompted this</h2> <p>Since October 2025, "market manipulation" appears on 10 papers, "stock market manipulation" on 5, "price manipulation" on 5, "pump-and-dump schemes" on 2 and "wash trading" on 1, split between AI journals (which build detectors) and finance journals (which study episodes). High-frequency trading and limit order books appear on 34 papers between them.</p> <h2>Questions</h2> <ul> <li>Trained on prosecuted spoofing and pump-and-dump cases, how well do detectors flag new cases, and how many false alarms do they raise per true case at market scale?</li> <li>How much wash trading is there on crypto exchanges, by exchange, and does it respond to enforcement?</li> <li>Do coordinated pumps organized on social media leave signatures in order flow before prices move?</li> </ul> <h2>Data that could help</h2> <p>Enforcement case records with dates and instruments, order book data (LOBSTER, exchange feeds), crypto trade data and social media archives.</p> <h2>Next step</h2> <p>Train a detector on the order book patterns of prosecuted spoofing cases and measure its alarm rate on a month of ordinary trading in the same stocks.</p> https://scholar4ai.com/ideas/maritime-decarbonization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Port operations, maritime logistics, container shipping and port congestion appear on dozens of recent papers in the operations research journals, while ship emissions and maritime decarbonization appear on 9 and 8. Shipping now pays for carbon in Europe and faces a global fuel standard. Do slower ships and new fuels cut emissions, or shift them to ports and queues?

<h2>What prompted this</h2> <p>Since October 2025, "port operations" appears on 19 papers, "maritime logistics" on 11, "container shipping" on 10, "container terminals" on 10 and "port congestion" on 10, mostly in operations research journals. "Ship emissions" appears on 9 papers, "maritime decarbonization" on 8, "emission control areas" on 5 and "port decarbonization" on 4. The operations and the emissions literatures mostly sit in different papers.</p> <h2>Questions</h2> <ul> <li>Since shipping entered the EU emissions trading system in 2024, did ships slow down, reroute to transshipment ports just outside the covered area, or pay?</li> <li>Does slow steaming reduce total emissions once the extra ships needed for the same cargo are counted?</li> <li>Where will ammonia and methanol bunkering be available first, and how do routing models change when fuel is available at only a few ports?</li> </ul> <h2>Data that could help</h2> <p>Ship position and speed data (AIS), port call records, emissions trading data for maritime, and bunkering infrastructure announcements.</p> <h2>Next step</h2> <p>Compare average speeds and port calls of container ships on Europe-Asia routes before and after January 2024 using AIS data.</p> https://scholar4ai.com/ideas/management-guidance-in-uncertain-times 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Management earnings forecasts appear on 11 recent papers in the accounting and finance journals. Firms withdrew guidance in the pandemic and again around the 2025 tariff shocks. What do investors and analysts do when guidance disappears, and does the practice recover?

<h2>What prompted this</h2> <p>Since October 2025, 12 papers carry a management-forecast or management-guidance keyword (6 in accounting journals, 6 in finance), with "management earnings forecasts" on 11. Four of them were published in July–September 2026, against about 2 per quarter earlier. Economic policy uncertainty and trade policy uncertainty are among the most frequent keywords in the finance journals.</p> <h2>Questions</h2> <ul> <li>When firms withdraw guidance in a shock, does analyst forecast dispersion and stock volatility rise more for those firms than for similar firms that keep guiding?</li> <li>Do firms that stopped guiding during the pandemic ever restart, and are those that restarted different?</li> <li>Did the 2025 tariff shocks lead firms to switch from point forecasts to ranges or scenarios?</li> </ul> <h2>Data that could help</h2> <p>Guidance records (I/B/E/S Guidance), analyst forecasts, stock volatility and the dates of shocks.</p> <h2>Next step</h2> <p>Count guidance withdrawals by quarter for 2019–2026 and compare the analyst dispersion of withdrawing firms with matched firms.</p> https://scholar4ai.com/ideas/long-term-care-workforce 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Elderly care, long-term care insurance and informal care appear on 75 recent papers, with 22 in the last quarter against about 17 per quarter before, and 24 carry a labour, worker or immigration keyword. Ageing countries need millions more care workers. Does paying more bring them, does immigration fill the gap, and what do robots change?

<h2>What prompted this</h2> <p>Since October 2025, 75 papers carry a long-term-care, elderly-care, nursing-home or caregiving keyword (41 in economics journals, 16 in AI): "elderly care" on 19, "long-term care insurance" on 14, "long-term care" on 9, "informal care" on 8. Twenty-four of them also carry a labour, worker, wage, workforce, immigration, robot or informal-care keyword. The quarter to September 2026 had 22 such papers against about 17 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When care wages rise, as in Japan's and Germany's recent increases, how many workers enter from other sectors, and does turnover fall?</li> <li>Do care-worker visas raise the supply without lowering wages for existing workers?</li> <li>Do families cut informal care when formal care becomes available, and does that raise women's employment?</li> </ul> <h2>Data that could help</h2> <p>Care sector employment and wages, long-term care insurance claims (Japan, Germany, Korea), visa statistics and household surveys with caregiving questions.</p> <h2>Next step</h2> <p>Compare care-worker employment growth across Japanese prefectures after the national pay increases, using differences in the local wage gap to other sectors.</p> https://scholar4ai.com/ideas/long-run-costs-of-pandemic-school-closures 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

"Long-term effects" is a rising keyword in the economics journals, and COVID-19 is still one of the most frequent keywords overall. Yet only 22 of 778 recent pandemic papers carry a schooling or human-capital keyword. Five years on, do the students who lost the most school time show it in exams, graduation and early careers?

<h2>What prompted this</h2> <p>In economics, "long-term effects" had 7 papers in July–September 2026 against about 2 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, 778 papers carry a COVID-19 or pandemic keyword (318 in finance journals, 296 in economics). Only 22 of them carry a school-closure, learning-loss, educational-attainment or human-capital keyword, 16 of those in economics.</p> <h2>Questions</h2> <ul> <li>Did students whose schools stayed closed longer score lower years later, graduate less often or go to weaker colleges?</li> <li>Did tutoring and catch-up programmes close the gap, and for whom?</li> <li>What do the first labour market outcomes of these cohorts show?</li> </ul> <h2>Data that could help</h2> <p>School-level closure data, standardized test scores (state assessments, NAEP, PISA), college enrolment and early earnings records.</p> <h2>Next step</h2> <p>Relate district-level weeks of remote schooling in 2020–2021 to the change in test scores through 2025 in US state assessment data.</p> https://scholar4ai.com/ideas/loan-officers-versus-algorithms 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Soft information and relationship lending appear on 26 recent papers in the finance and economics journals, but only 5 of them carry an algorithm, machine-learning, fintech or automation keyword. As banks automate small-business lending, what is lost when the loan officer's judgment goes, and can models learn it?

<h2>What prompted this</h2> <p>Since October 2025, 26 papers carry a loan-officer, relationship-lending or soft-information keyword (15 in finance journals, 7 in economics): "soft information" on 19, "relationship lending" on 6. Only 5 of them also carry an algorithm, machine-learning, fintech, automation, digital or hard-information keyword.</p> <h2>Questions</h2> <ul> <li>Do loans approved by loan officers against the model's recommendation default more or less often than loans the model approved?</li> <li>Does automation cut lending to borrowers whose strengths are not in the data (new firms, immigrants, informal businesses), and does it reduce discrimination?</li> <li>Can text from loan applications and site visits give models the soft information officers use?</li> </ul> <h2>Data that could help</h2> <p>Loan application files with both officer decisions and model scores from a lender, and loan performance.</p> <h2>Next step</h2> <p>Partner with a lender that records officer overrides of model scores and compare the default rates of overridden and non-overridden loans.</p> https://scholar4ai.com/ideas/llm-pricing-agents-and-collusion 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Price competition is rising steeply in the operations research journals and AI governance leads the AI journals, while collusion remains a small topic that is rarely tied to algorithms. When sellers let language-model agents set prices, do prices drift up without any agreement?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "price competition" rises more steeply than any other growing keyword in the operations research journals (11 papers in July–September 2026 against about 2 expected), and "AI governance" tops the growing keywords of the AI journals. Since October 2025, 46 papers carry a collusion keyword (28 in economics); only 7 also carry an algorithm, reinforcement-learning or AI keyword, and "algorithmic collusion" appears on 5. Q-learning, the algorithm behind the best-known simulations of algorithmic collusion, shows up on the <a href="/rankings/equations">equation ranking</a> almost only in AI journals.</p> <h2>Questions</h2> <ul> <li>In repeated pricing games, do language-model agents reach prices above the competitive level, as Q-learning agents do in simulations?</li> <li>Does it depend on the prompt, on memory, or on seeing rivals' prices?</li> <li>What tests could a competition authority run on such agents before they are deployed?</li> </ul> <h2>Possible approach</h2> <p>Simulate repeated Bertrand competition with differentiated products, with language-model agents facing each other and facing Q-learning agents, and compare their prices with the Nash equilibrium and the joint-profit benchmark.</p> <h2>Next step</h2> <p>Run a two-firm simulation with off-the-shelf models and record average prices over many rounds.</p> https://scholar4ai.com/ideas/llm-confidence-calibration 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Confidence calibration and uncertainty calibration appear on 54 recent papers, mostly in AI journals, and 35 calibration papers involve language or vision-language models, with 94 calibration-related papers in the last quarter against about 57 per quarter before. When a model answers a financial or legal question, its stated confidence rarely matches its accuracy. Can calibration be fixed, and do users read it?

<h2>What prompted this</h2> <p>Since October 2025, 280 papers carry a calibration or overconfidence keyword (176 in AI journals, 53 in finance, 28 in economics): "confidence calibration" on 28, "uncertainty calibration" on 26, "model calibration" on 19, besides the behavioural overconfidence of executives and investors studied in finance. Thirty-five of the 280 also carry a language-model or vision-language keyword. The quarter to September 2026 had 94 such papers against about 57 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>On financial, accounting and legal questions, how far are models' stated confidences from their accuracy, and does the gap grow for harder questions?</li> <li>Which methods (verbalized confidence, sampling consistency, trained calibrators) give usable confidence for decisions, and do they survive fine-tuning?</li> <li>When users are shown calibrated confidence, do they rely on the model more where it is right and less where it is wrong?</li> </ul> <h2>Data that could help</h2> <p>Question sets with gold answers in finance and law, model outputs with confidence under several methods, and user experiments with confidence displays.</p> <h2>Next step</h2> <p>Measure the calibration of three models on a financial analysis question set under verbalized and consistency-based confidence, and test whether showing it changes user reliance.</p> https://scholar4ai.com/ideas/llm-built-uncertainty-indices 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Economic, climate, trade and monetary policy uncertainty indices appear on 423 recent papers in the finance and economics journals, almost all built from counts of newspaper words. Only 5 of them use language models, embeddings or text mining beyond word counts. Would an index that reads the articles agree with the word counts, and would it predict better?

<h2>What prompted this</h2> <p>Since October 2025, 423 papers carry a policy-uncertainty, uncertainty-index or text-based keyword (243 in finance journals, 144 in economics): "economic policy uncertainty" on 161, "climate policy uncertainty" on 106, "policy uncertainty" on 68, "trade policy uncertainty" on 48, "monetary policy uncertainty" on 32. Only 5 of the 423 also carry a language-model, BERT, transformer, machine-learning, deep-learning, embedding, text-mining or natural-language-processing keyword.</p> <h2>Questions</h2> <ul> <li>Do indices built by a language model reading the full text (asking whether the article describes policy uncertainty and of what kind) track the word-count indices, and where do they diverge?</li> <li>Do the model-built indices predict investment, hiring and asset prices better, or does the simpler index already capture what matters?</li> <li>Can the same approach build indices for countries and languages where the newspaper word lists were never validated?</li> </ul> <h2>Data that could help</h2> <p>Newspaper archives (Factiva, LexisNexis), the published uncertainty indices, and macro and market data.</p> <h2>Next step</h2> <p>Score a year of articles from the newspapers used by one uncertainty index with a language model and compare the resulting monthly index with the published one.</p> https://scholar4ai.com/ideas/llm-as-a-judge-reliability 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Evaluation metrics appear on 56 recent papers and LLM-as-a-judge on 16, all in AI journals, with 25 evaluation papers in the last quarter in line with earlier quarters. Models now grade other models' answers, and firms use them to score customer conversations and reports. How reliable and how biased are the model judges, and how far do the biases propagate?

<h2>What prompted this</h2> <p>Since October 2025, 106 papers carry an LLM-as-a-judge, automatic-evaluation, evaluation-metric or human-evaluation keyword (97 in AI journals): "evaluation metrics" on 56, "LLM-as-a-judge" on 16, "LLM evaluation" on 9, "benchmark evaluation" on 5, "human evaluation" on 3. The quarter to September 2026 had 25 such papers against about 26 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>On tasks with expert ground truth in finance, accounting and law, how well do model judges agree with experts, and do they favour longer, more confident or more similar-sounding answers?</li> <li>When a model is trained on rewards from a model judge, do the judge's biases compound in the trained model?</li> <li>In business uses (scoring customer service calls, grading analyst notes), do the scores change behaviour in the direction the firm wants?</li> </ul> <h2>Data that could help</h2> <p>Expert-graded answer sets in professional domains, judge outputs from several models, and training runs with model-judged rewards.</p> <h2>Next step</h2> <p>Compare model-judge scores with expert grades on 300 answers to accounting and legal questions and measure length and self-preference biases.</p> https://scholar4ai.com/ideas/llm-agents-and-the-bullwhip-effect 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Agent-based modelling is rising in the operations research journals, and firms are starting to let AI agents place orders. The bullwhip effect, where small swings in demand grow larger up the supply chain, appears on only 9 recent papers, and just one of them carries an agent or AI keyword. Do AI ordering agents dampen the bullwhip, or amplify it when they all react to the same signals?

<h2>What prompted this</h2> <p>In the operations research journals, "agent-based modeling" had 6 papers in July–September 2026 against about 3 expected (see the <a href="/categories/operations-research">operations research dashboard</a>). Since October 2025, "agent-based modeling" appears on 81 papers (40 in economics journals, 16 in operations research) and "LLM agents" on 44, while "bullwhip effect" appears on 9, only one of which carries a language-model, agent or AI keyword.</p> <h2>Questions</h2> <ul> <li>In the classic beer distribution game, do language-model agents order more steadily than human players, or overreact?</li> <li>What happens when every tier of the supply chain uses the same model?</li> <li>Does sharing point-of-sale data with the agents remove the bullwhip, as it does for people?</li> </ul> <h2>Possible approach</h2> <p>Run the beer game with a language-model agent at each tier, varying the information the agents share and whether all tiers use the same model, and measure how much the variance of orders grows from tier to tier.</p> <h2>Next step</h2> <p>Implement a four-tier beer game with language-model agents and compare the order variance at each tier with published results for human players.</p> https://scholar4ai.com/ideas/learning-to-decompose 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Benders decomposition, column generation, branch-and-cut and related methods appear on 240 recent papers, almost all in the operations research journals, and the topic is fading, with 42 papers in the last quarter against about 64 per quarter before. Only 16 of them use learning. Can learned cut selection, column pricing and branching make these methods faster on real problems?

<h2>What prompted this</h2> <p>Since October 2025, 240 papers carry a Benders, column-generation, branch-and-price, branch-and-cut, cutting-plane or Lagrangian-relaxation keyword (218 in operations research journals, 22 in AI): "Benders decomposition" on 65, "column generation" on 46, "branch-and-cut" on 35, "logic-based Benders decomposition" on 30, "branch-and-price" on 28. Forty-two were published in July–September 2026 against about 64 per quarter earlier. Only 16 of the 240 also carry a machine-learning, reinforcement-learning, neural or data-driven keyword.</p> <h2>Questions</h2> <ul> <li>Does a model trained to select Benders cuts or price columns on past instances of a firm's problem cut solution times on new instances by more than careful engineering of the classical method?</li> <li>How much training data does it need, and does it break when the problem's data distribution shifts?</li> <li>Which parts of the pipeline (branching, cut selection, primal heuristics, column pricing) give the largest gains per unit of engineering?</li> </ul> <h2>Data that could help</h2> <p>Standard instance libraries (MIPLIB, routing and scheduling benchmarks) and firms' instance histories for a repeated planning problem.</p> <h2>Next step</h2> <p>Train a cut-selection model on one firm's repeated network design instances and compare solution times with a tuned classical Benders implementation on held-out instances.</p> https://scholar4ai.com/ideas/language-models-as-survey-respondents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Survey experiments are a staple of economics, and some researchers now simulate respondents with language models ("silicon samples"). Only 6 of 179 recent survey papers carry a language-model keyword, 5 of them in AI journals. Where do simulated answers match real people, and where would they mislead economic research?

<h2>What prompted this</h2> <p>Since October 2025, 179 papers carry a survey, questionnaire or respondent keyword (117 in economics journals, 37 in finance). Only 6 of them carry a language-model, ChatGPT or generative-AI keyword, and 5 of those are in AI journals. "Survey experiment" appears on 15 papers, 14 of them in economics.</p> <h2>Questions</h2> <ul> <li>For published survey experiments in economics, do simulated respondents reproduce the direction and size of the treatment effects?</li> <li>Do they capture differences between groups (income, age, country), or flatten them?</li> <li>Could they help design and size real surveys, even if they cannot replace them?</li> </ul> <h2>Possible approach</h2> <p>Replicate published survey experiments with language-model respondents given the demographics of the original sample, and compare the effects with the published ones.</p> <h2>Next step</h2> <p>Pick ten published survey experiments in economics, simulate them with two language models, and compare the estimated treatment effects.</p> https://scholar4ai.com/ideas/language-models-as-heuristic-designers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Matheuristics and local search are among the fastest-growing keywords in the operations research journals, and recent systems let language models write and evolve heuristic code. Only 2 of the 252 heuristics papers in OR journals carry a language-model keyword. Can generated heuristics beat hand-designed ones on standard routing and scheduling instances?

<h2>What prompted this</h2> <p>In the operations research journals, "matheuristic" had 14 papers in July–September 2026 against about 8 expected, and "local search" 15 against 7 (see the <a href="/categories/operations-research">operations research dashboard</a>). Since October 2025, 547 papers carry a heuristic, local-search or matheuristic keyword (254 in AI journals, 252 in operations research). Only 13 of them carry a language-model keyword, 2 of those in operations research journals.</p> <h2>Questions</h2> <ul> <li>On standard instances (Solomon, CVRPLIB, Taillard), do heuristics that language models write and refine beat a well-tuned adaptive large neighborhood search?</li> <li>Are the generated heuristics new ideas or recombinations of known operators, and can people read them?</li> <li>How much computing do they cost compared with what they gain?</li> </ul> <h2>Data that could help</h2> <p>Benchmark instances from the optimization category of the <a href="/rankings/datasets">dataset ranking</a>, with their published best-known solutions.</p> <h2>Next step</h2> <p>Let a language model evolve destroy and repair operators for an adaptive large neighborhood search on the Solomon instances, and compare the gaps to the best-known solutions.</p> https://scholar4ai.com/ideas/knowledge-sharing-with-ai-at-work 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Knowledge transfer, sharing and management appear on 259 recent papers across the management and AI journals, but only 48 of them carry an AI or digital keyword, and knowledge hiding on 3. Firms are building assistants trained on employees' documents and chats. Do employees share less when sharing feeds a machine?

<h2>What prompted this</h2> <p>Since October 2025, 259 papers carry a knowledge-transfer, knowledge-sharing, knowledge-management or knowledge-hiding keyword (112 in management journals, 105 in AI): "knowledge transfer" on 114, "knowledge sharing" on 56, "knowledge management" on 43, "knowledge hiding" on 3. Forty-eight of them also carry an AI, generative-AI, language-model or digital keyword, most of those about tools rather than about people's willingness to share.</p> <h2>Questions</h2> <ul> <li>When a firm announces that an internal assistant will be trained on employees' work, does documentation and answering of colleagues' questions rise or fall?</li> <li>Do experts whose know-how is most codifiable hide more, and do incentives for sharing offset that?</li> <li>Does the assistant reduce the value of expert employees in pay and promotions, and do they see it coming?</li> </ul> <h2>Data that could help</h2> <p>Internal collaboration logs (documents, chat answers, wiki edits) from firms that deployed assistants, with deployment dates, and employee surveys.</p> <h2>Next step</h2> <p>Compare documentation and question-answering activity in internal tools before and after an AI assistant announcement at a firm that shares its logs.</p> https://scholar4ai.com/ideas/knowledge-graphs-for-compliance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Knowledge graphs appear on 141 recent papers and neuro-symbolic AI on 14, almost all in AI journals; only 34 of 481 such papers concern accounting, tax, legal, regulatory, financial or audit uses. Compliance is a domain of explicit rules and entities, which language models alone handle badly. Do graphs and symbolic rules make them reliable enough for audits and tax?

<h2>What prompted this</h2> <p>Since October 2025, 481 papers carry a knowledge-graph, neuro-symbolic, symbolic-reasoning, rule-based, ontology, formal-verification or logical-reasoning keyword (451 in AI journals): "knowledge graphs" on 141, "knowledge graph completion" on 41, "temporal knowledge graphs" on 38, "neuro-symbolic AI" on 14, "symbolic reasoning" on 8. Only 34 of the 481 also carry an accounting, tax, legal, regulatory, compliance, finance, contract, audit, law, bank, fraud or anti-money-laundering keyword, although "regulatory compliance" appears on 60 papers in the archive.</p> <h2>Questions</h2> <ul> <li>On tasks such as checking a transaction against sanctions lists and ownership chains, or applying a tax rule to a set of facts, do systems that combine a language model with a knowledge graph and explicit rules make fewer errors than the model alone, and are their errors explainable?</li> <li>How much effort does it take to keep the rules and graphs current when regulations change, and who maintains them?</li> <li>Would auditors and regulators accept outputs with symbolic justifications where they reject model-only outputs?</li> </ul> <h2>Data that could help</h2> <p>Public regulatory texts and sanctions lists, company ownership graphs (Open Ownership, Orbis), and test cases with known answers from tax and compliance training materials.</p> <h2>Next step</h2> <p>Build a benchmark of 200 compliance cases (sanctions screening with ownership chains, tax rule application) and compare a language model alone with the same model backed by a graph and rules.</p> https://scholar4ai.com/ideas/japan-wage-growth-and-inflation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Japan appears on 184 recent papers, with 56 in the last quarter against about 41 per quarter before, but only 21 of them carry a wage, inflation or deflation keyword. Japan's annual wage rounds have produced the largest raises in decades, and the Bank of Japan has ended negative rates. Is the wage-price cycle real, and does it reach small firms?

<h2>What prompted this</h2> <p>Since October 2025, 184 papers carry a Japan keyword (117 in economics journals, 49 in finance): "Japan" on 81, "Japanese companies" on 18, "Bank of Japan" on 11, "Japanese yen" on 8. The quarter to September 2026 had 56 such papers against about 41 per quarter earlier. Only 21 of the 184 also carry a wage, labour-shortage, real-wage, inflation or deflation keyword; "wage growth" appears on 14 papers and "deflation" on 13.</p> <h2>Questions</h2> <ul> <li>Did the large raises at big firms in the 2024–2026 wage rounds pass to their suppliers and to small firms, where most people work?</li> <li>Have real wages turned positive for long enough to change household inflation expectations and spending?</li> <li>Did the labour shortage, rather than the central bank, end deflation, and what does that imply for countries now ageing as fast as Japan?</li> </ul> <h2>Data that could help</h2> <p>Monthly labour survey wages by firm size, the wage round results, household expectation surveys and price data by sector.</p> <h2>Next step</h2> <p>Compare wage growth at firms with fewer than 30 employees with that at large firms in Japan's monthly labour survey for 2022–2026.</p> https://scholar4ai.com/ideas/is-the-technology-acceptance-model-still-useful 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The technology acceptance model is one of the fastest-growing keywords across the tracked journals, and a quarter of recent TAM and UTAUT papers now study AI. But the model was built for workplace software decades ago. Do perceived usefulness and ease of use predict who actually uses generative AI, and what do they miss?

<h2>What prompted this</h2> <p>"Technology acceptance model" grew across all journals (19 papers in July–September 2026 against about 10 expected) and in finance (8 against 2); see the <a href="/trends">trends page</a>. Since October 2025, 109 papers carry a TAM or UTAUT keyword (67 in management journals, 24 in finance); 26 of them also carry a generative-AI, ChatGPT, language-model or AI keyword.</p> <h2>Questions</h2> <ul> <li>Do intentions measured with TAM surveys predict actual use, as logged by the tools, weeks later?</li> <li>Do things the model leaves out, such as trust in accuracy, fear of replacement or the tool doing the whole task, explain more of who uses it?</li> <li>Do results replicate across countries and occupations, or reflect the student samples many studies use?</li> </ul> <h2>Possible approach</h2> <p>Pair a standard TAM survey with usage logs from an AI tool deployed in a firm or university, and compare its predictive power with and without the added factors.</p> <h2>Next step</h2> <p>Survey the users of one AI tool with a standard TAM questionnaire and match their answers to their usage over the following month.</p> https://scholar4ai.com/ideas/is-biodiversity-risk-priced 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sustainable investment is rising across the tracked journals, and biodiversity risk has become a finance keyword. Yet few of these papers carry a keyword about returns, bonds or the cost of capital. Do markets price firms' exposure to nature loss, and has nature-related disclosure changed that?

<h2>What prompted this</h2> <p>"Sustainable investment" grew across all journals (19 papers in July–September 2026 against about 11 expected; see the <a href="/trends">trends page</a>). Since October 2025, "biodiversity risk" appears on 35 papers (28 in finance journals), and 116 papers carry some biodiversity keyword (45 in economics, 44 in finance). Only 5 of the 116 also carry a keyword about stock returns, asset pricing, the cost of capital, bonds, equity or portfolios.</p> <h2>Questions</h2> <ul> <li>Do firms more exposed to nature loss (land use, water, supply chains in sensitive areas) earn higher expected returns, or pay more for credit?</li> <li>Did investors start pricing it after the Kunming-Montreal biodiversity framework (2022) or the nature-related disclosure recommendations (TNFD, 2023)?</li> <li>Can biodiversity exposure be measured from firms' locations and supply chains rather than from what they report?</li> </ul> <h2>Data that could help</h2> <p>Facility locations of listed firms, protected-area and species-range maps, text-based exposure measures, and stock and bond returns.</p> <h2>Next step</h2> <p>Build a location-based exposure measure from facility locations and protected-area maps, and test whether it predicts returns or credit spreads.</p> https://scholar4ai.com/ideas/investment-scams-on-social-media 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Financial fraud appears on 48 recent papers, but most study fraud by firms or fraud detection in transactions; only 6 carry a social-media, crypto, elderly or victim keyword, and "fraud victimization" on 5. Investment scams run through social platforms now cost households billions. Who falls for them, and what stops them?

<h2>What prompted this</h2> <p>Since October 2025, 48 papers carry a scam, investment-fraud, financial-fraud, Ponzi or fraud-victim keyword (18 in finance journals, 16 in AI); "financial fraud" appears on 22 and "fraud victimization" on 5. Only 6 of the 48 also carry a social-media, crypto, elderly, older-adult, romance or victim keyword. "Investor protection" appears on 49 papers, mostly about securities law.</p> <h2>Questions</h2> <ul> <li>Which households lose money to investment scams, and do the losses cluster among people with high financial literacy and confidence rather than low?</li> <li>Do warnings from regulators and platforms reduce losses, or do scams simply move?</li> <li>How much scam money flows through crypto exchanges, and does exchange compliance change the flow?</li> </ul> <h2>Data that could help</h2> <p>Consumer complaint databases (for example the FTC Consumer Sentinel and FBI IC3 reports), bank transaction data with fraud flags, and platform takedown records.</p> <h2>Next step</h2> <p>Analyse the complaint records of one regulator for 2020–2025 by age, scam type and channel, and relate losses to platform enforcement announcements.</p> https://scholar4ai.com/ideas/interest-rate-risk-on-bank-balance-sheets 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Interest rate risk appears on just 11 recent papers in the finance journals, only one of them in July–September 2026, although unrealized losses on bank bond holdings triggered the 2023 bank failures. Which banks still carry large unrealized losses, and does supervision now see them?

<h2>What prompted this</h2> <p>Since October 2025, 13 papers carry an interest-rate-risk, held-to-maturity or unrealized-loss keyword (9 in finance journals, 3 in economics), and only one of them was published in July–September 2026. "Bank failures" appears on 12 papers. Bank runs and deposit insurance are rising topics, but the balance-sheet side of the 2023 failures is barely studied.</p> <h2>Questions</h2> <ul> <li>How much of the banking system's unrealized losses remain after rates stayed high, and are they concentrated in banks with uninsured deposits?</li> <li>Did banks lengthen or shorten the duration of their bond holdings after 2023, and did the use of hedges change?</li> <li>Would reporting held-to-maturity securities at fair value have changed depositors' behaviour?</li> </ul> <h2>Data that could help</h2> <p>Call Reports and FR Y-9C filings, which report securities by maturity bucket and accounting classification, deposit composition and bank stock prices.</p> <h2>Next step</h2> <p>Estimate each US bank's unrealized losses from its Call Report maturity buckets for 2022–2025 and relate them to deposit outflows and equity prices.</p> https://scholar4ai.com/ideas/insurers-retreat-from-climate-risk 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Insurance appears on more than 400 recent papers in the finance and economics journals, but only 52 of them carry a climate or disaster keyword, and property insurance barely appears. As insurers withdraw from wildfire- and hurricane-prone regions, who holds the risk, and what happens to house prices and mortgages?

<h2>What prompted this</h2> <p>Since October 2025, 408 papers carry an insurance keyword (196 in finance journals, 174 in economics), the most frequent being health, life, unemployment and deposit insurance. Only 52 of them also carry a wildfire, flood, climate, disaster, hurricane or catastrophe keyword, and "property insurance" appears on a single paper. "Natural disasters" appears on 82 papers.</p> <h2>Questions</h2> <ul> <li>When private insurers stop writing policies in a region, how much do house prices and mortgage lending fall, and who buys the homes?</li> <li>Do state-backed insurers of last resort, which are growing fast, price risk, or hide it?</li> <li>Does the retreat change where people and businesses move, and how quickly?</li> </ul> <h2>Data that could help</h2> <p>Insurer market-exit filings and non-renewal rates by county, state insurer-of-last-resort enrolment, house transactions and mortgage originations (HMDA).</p> <h2>Next step</h2> <p>Build a county panel of insurer exits and non-renewals for California and Florida since 2018, and relate it to house price growth and mortgage volumes.</p> https://scholar4ai.com/ideas/insider-trading-plans 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Insider trading appears on 40 recent papers in the finance and accounting journals, and 17 of 45 carry a regulation or enforcement keyword, but "trading plan" appears on one. Since 2023, US executives' pre-scheduled trading plans must have cooling-off periods and be disclosed. Did the plans stop being a cover for informed trades?

<h2>What prompted this</h2> <p>Since October 2025, 45 papers carry an insider-trading keyword (32 in finance journals, 7 in accounting); "insider trading" appears on 40 and "insider trading regulation" on 5. Of the 45, 17 carry a plan, enforcement, regulation, detection or machine-learning keyword, but the keyword "trading plan" appears on a single paper.</p> <h2>Questions</h2> <ul> <li>Before the 2023 rule, did trades under pre-scheduled plans earn abnormal returns, especially plans adopted shortly before trades?</li> <li>After the cooling-off period and disclosure requirements, did the returns to plan trades fall, and did executives trade less under plans?</li> <li>Did firms with the most profitable plan trades before the rule change their behaviour most?</li> </ul> <h2>Data that could help</h2> <p>Insider trading filings (Form 4) with plan indicators, plan adoption and modification disclosures from 10-Q and 10-K filings, and stock returns.</p> <h2>Next step</h2> <p>Compare abnormal returns of plan and non-plan insider sales in the two years before and after the 2023 rule.</p> https://scholar4ai.com/ideas/inheritance-and-mobility 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Intergenerational mobility and transmission appear on 142 recent papers, mostly in economics, and 33 of them carry a wealth, inheritance, housing or parental keyword. As house prices and inheritances grow relative to wages, does where you end up depend more on what your parents own than on what you earn?

<h2>What prompted this</h2> <p>Since October 2025, 142 papers carry an intergenerational-mobility, social-mobility or intergenerational-transmission keyword (122 in economics journals): "intergenerational transmission" on 26, "intergenerational mobility" on 18, "social mobility" on 12. Thirty-three of them also carry a wealth, inheritance, housing, gift, parental or bequest keyword, and "intergenerational wealth transfer" appears on 4.</p> <h2>Questions</h2> <ul> <li>For cohorts born since 1980, how much of the correlation between parents' and children's wealth runs through inheritances and gifts, and how much through housing help at the time of purchase?</li> <li>Has the share of first-time buyers who received family help risen, and does help predict where they live and whether they move for work?</li> <li>Do countries with large inheritance taxes show higher wealth mobility across generations?</li> </ul> <h2>Data that could help</h2> <p>Linked parent-child wealth records (Scandinavian registers, the Panel Study of Income Dynamics), surveys on home purchase financing, and inheritance tax data.</p> <h2>Next step</h2> <p>Estimate the share of first-home down payments financed by family gifts by cohort from household surveys that record the source of funds.</p> https://scholar4ai.com/ideas/informal-economy-from-above 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Aerial imagery is rising steeply in the AI journals, and economists already use night-time lights to measure growth. Yet none of the papers on the informal economy uses either. Can image models measure informal activity that surveys miss?

<h2>What prompted this</h2> <p>In the <a href="/categories/ai">AI dashboard</a> trends, "aerial imagery" had 9 papers in July–September 2026 against about 1 expected from its earlier share. In economics, night-time lights are an established proxy for activity: 26 papers carry the keyword since October 2025 (24 in economics journals), none of them with a machine-learning keyword. And of the 55 papers with an informal-economy, informal-employment or informal-sector keyword, none carries a satellite, night-light, aerial or remote-sensing keyword.</p> <h2>Questions</h2> <ul> <li>Can models trained on high-resolution imagery (market stalls, rooftops, vehicles, building materials) predict informal employment where surveys exist, and fill in where they do not?</li> <li>Do such measures change growth or inequality estimates in countries with large informal sectors?</li> <li>Do the models carry over from one country to another? Cross-dataset generalization is another rising AI keyword.</li> </ul> <h2>Data that could help</h2> <ul> <li>Free Sentinel-2 and Landsat imagery, VIIRS night-time lights, and commercial imagery for a few cities.</li> <li>Geocoded household surveys (LSMS, DHS) as ground truth; both are in the <a href="/rankings/datasets">dataset ranking</a>'s list.</li> </ul> <h2>Next step</h2> <p>Pick one country with geocoded survey clusters and test whether image features add to night-time lights in predicting informal employment shares.</p> https://scholar4ai.com/ideas/industry-4-for-smes 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Industry 4.0, Industry 5.0 and smart manufacturing appear on 121 recent papers, but only 26 of them carry a small-firm, adoption, barrier or maturity keyword. Large manufacturers have automated and connected their plants; most small ones have not. What holds them back, and which support programmes have moved the needle?

<h2>What prompted this</h2> <p>Since October 2025, 121 papers carry an Industry 4.0, Industry 5.0, smart-manufacturing or smart-factory keyword (37 in management journals, 31 in operations research, 28 in AI): "Industry 4.0" on 54, "smart manufacturing" on 35, "Industry 5.0" on 32. Only 26 of them also carry a small-firm, SME, adoption, barrier, readiness or maturity keyword.</p> <h2>Questions</h2> <ul> <li>Among small manufacturers, is the binding constraint financing, skills, the absence of an engineer on staff, or the lack of a clear return?</li> <li>Do subsidized adoption programmes (Germany's and Japan's digitalization grants, US manufacturing extension) raise productivity, or mostly fund equipment that would have been bought anyway?</li> <li>Do small suppliers adopt when a large customer requires data connections, and do they capture any of the gains?</li> </ul> <h2>Data that could help</h2> <p>Firm surveys on technology adoption, programme participation records, firm accounts and supplier-customer links.</p> <h2>Next step</h2> <p>Compare the productivity growth of small manufacturers that received digitalization grants with similar applicants that were rejected or waitlisted.</p> https://scholar4ai.com/ideas/index-insurance-for-farmers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Agricultural and index insurance appear on 20 recent papers, split between the finance and economics journals. Weather index insurance was meant to protect smallholders cheaply, but take-up stayed low and basis risk high. With satellite data and mobile payments, is it finally working, and where?

<h2>What prompted this</h2> <p>Since October 2025, 20 papers carry an index-insurance, crop-insurance, agricultural-insurance or weather-insurance keyword (9 in finance journals, 7 in economics): "agricultural insurance" on 16, "index insurance" on 3, "crop insurance" on 2. Climate shocks, by contrast, are one of the fastest-growing keywords in the economics journals.</p> <h2>Questions</h2> <ul> <li>Do satellite-based indices (vegetation, soil moisture) track farm losses better than rainfall at a weather station, and does better tracking raise take-up?</li> <li>Do bundled products (insurance with credit, seed or mobile money) reach farmers who would not buy insurance alone?</li> <li>When insured farmers are paid after a drought, do they invest more in the next season than uninsured neighbours?</li> </ul> <h2>Data that could help</h2> <p>Insurance programme enrolment and payout records, satellite indices, farm surveys with yields and investment, and mobile money data.</p> <h2>Next step</h2> <p>Compare the correlation of farm-level losses with rainfall-based and satellite-based indices in one programme that switched from one to the other.</p> https://scholar4ai.com/ideas/immigration-and-housing-costs 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Immigration and migrant workers appear on 141 recent papers, mostly in economics, with 42 in July–September 2026 against about 30 per quarter before. Only 18 carry a housing, rent or house-price keyword. Housing costs are now the loudest argument in immigration politics. How much of rent growth do arrivals explain, and how much is a lack of building?

<h2>What prompted this</h2> <p>Since October 2025, 141 papers carry an immigration or migrant keyword (116 in economics journals): "immigration" on 19, "migrant workers" on 12, "immigration policy" on 8. The quarter to September 2026 had 42 such papers against about 30 per quarter over the nine months before. Only 18 of the 141 also carry a housing, rent, house-price or home-price keyword.</p> <h2>Questions</h2> <ul> <li>In cities with large recent inflows (Canada, Australia, Germany, the UK), how much of rent growth is attributable to immigration once construction, interest rates and incomes are held fixed?</li> <li>Do immigrants raise rents mainly where housing supply is constrained by regulation?</li> <li>Do policies that cut immigration to ease housing, such as Canada's 2024 caps, lower rents, and at what cost to construction labour?</li> </ul> <h2>Data that could help</h2> <p>Regional migration flows, rents and house prices, building permits and housing supply elasticity estimates.</p> <h2>Next step</h2> <p>Estimate the rent response to immigration inflows across regions, interacted with measures of housing supply constraints, for 2015–2025.</p> https://scholar4ai.com/ideas/human-ai-collaboration-in-the-audit 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Human-AI collaboration is rising in the management and AI journals, while audit quality remains a core accounting topic. Few papers connect them. When auditors work with AI tools, does audit quality improve, and who answers for the judgments the tools shape?

<h2>What prompted this</h2> <p>"Human-AI collaboration" is among the fastest-growing keywords across all journals on the <a href="/trends">trends page</a> (36 papers in July–September 2026 against about 20 expected), and in management. Since October 2025, 335 papers carry an audit keyword (135 in accounting journals). 32 of them also carry an AI keyword, but only 8 of those are in accounting journals.</p> <h2>Questions</h2> <ul> <li>After audit firms roll out AI tools, do restatements, missed going-concern opinions or audit fees change?</li> <li>Does AI pull auditors' attention toward what it flags and away from what it misses?</li> <li>How should audit committees and regulators assign responsibility for AI-assisted judgments?</li> </ul> <h2>Data that could help</h2> <ul> <li>Restatements, opinions and fees from Audit Analytics, and audit inspection reports.</li> <li>Dates of AI rollouts at large audit firms, from announcements and job postings, as a staggered treatment.</li> <li>Experiments with practising auditors on misstatements that the AI flags and ones it does not.</li> </ul> <h2>Next step</h2> <p>Build a timeline of AI tool rollouts at the largest audit firms and check whether it can serve as a staggered treatment.</p> https://scholar4ai.com/ideas/hukou-reform 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Household registration and hukou reform appear on 22 recent papers in the economics and finance journals, and rural-urban migration on 15. Chinese cities have loosened residence rules, and many smaller cities now grant local registration freely. Do migrants settle when they can, and do cities gain workers or welfare costs?

<h2>What prompted this</h2> <p>Since October 2025, 22 papers carry a hukou or household-registration keyword (14 in economics journals, 7 in finance): "household registration" on 14, "hukou reform" on 6. "Rural-urban migration" appears on 15 papers. The reforms differ sharply by city size, which few papers exploit.</p> <h2>Questions</h2> <ul> <li>In cities that removed registration limits, did migrants bring their families, buy homes and enrol children in local schools, and did their consumption rise?</li> <li>Did large cities that kept points-based systems lose workers to cities that opened, and did wages respond?</li> <li>What did local governments spend on schools and health care for new residents, and did the fiscal transfers follow the people?</li> </ul> <h2>Data that could help</h2> <p>City-level reform dates and rules, migrant surveys (the China Migrants Dynamic Survey), housing transactions and municipal budgets.</p> <h2>Next step</h2> <p>Use the staggered relaxation of registration rules across cities to compare migrants' family settlement and home purchases before and after.</p> https://scholar4ai.com/ideas/how-fragile-is-the-did-boom 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Difference-in-differences is the most-mentioned method on the equation ranking, and treatment effect heterogeneity is one of the fastest-rising keywords in economics. A first check finds that about three in ten recent DiD papers with searchable full text name a heterogeneity-robust estimator. How many staggered-adoption findings would survive one?

<h2>What prompted this</h2> <p>On the <a href="/rankings/equations">equation ranking</a>, <a href="/equations/difference-in-differences">difference-in-differences</a> is the most-mentioned method: 148 papers in the 30 days to 2026-10-23 (72 in economics journals, 53 in finance, 14 in management, 6 in accounting). Instrumental variables (112 papers), two-way fixed effects (69) and propensity score matching (60) follow. In economics, "treatment effect heterogeneity" had 14 papers in July–September 2026 against about 4 expected from its earlier share.</p> <p>Two-way fixed-effects DiD can be badly biased when a policy starts at different times for different units and its effect varies. Newer estimators fix this: Callaway and Sant'Anna, Sun and Abraham, de Chaisemartin and D'Haultfœuille, and Borusyak, Jaravel and Spiess, with the Goodman-Bacon decomposition as a diagnostic.</p> <h2>First check</h2> <p>OpenAlex has searchable full text for 55 of the 148 DiD papers. 17 of those 55 (31%) name at least one of these estimators or the decomposition: 11 of 20 in finance, 3 of 21 in economics, 2 of 9 in management and 1 of 4 in accounting. The samples are small and most papers have no searchable full text, so this is a first signal, not a finding.</p> <h2>Questions</h2> <ul> <li>Re-estimated with robust estimators, how often do recent staggered-adoption results change sign, size or significance?</li> <li>Which designs are most at risk: many adoption cohorts, long panels, or effects that build up over time?</li> <li>Do studies of Chinese pilot policies, a common setting in these journals, behave differently?</li> </ul> <h2>Data that could help</h2> <p>Replication packages where journals require them; for city pilot studies, the public city panels most of them use, such as the China City Statistical Yearbook.</p> <h2>Next step</h2> <p>Pick 20 recent staggered DiD papers with replication data and re-estimate each one with two robust estimators.</p> https://scholar4ai.com/ideas/housing-costs-and-the-fertility-decline 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Fertility is one of the fastest-growing keywords in the economics journals, as birth rates hit record lows in East Asia and elsewhere. Housing prices are a large topic too, but only 5 of 94 recent fertility papers carry a housing keyword. How much of the fertility decline do housing costs explain, and do housing policies raise births?

<h2>What prompted this</h2> <p>On the <a href="/categories/economics">economics dashboard</a>, "fertility" had 15 papers in July–September 2026 against about 8 expected from its earlier share, and "housing market" is growing as well (13 against 8). Since October 2025, 94 papers carry a fertility keyword (87 in economics journals). Only 5 of them also carry a housing keyword, while "housing prices" appears on 123 papers.</p> <h2>Questions</h2> <ul> <li>Do couples in places where housing costs rose faster have fewer children, or the same number later?</li> <li>Do owners and renters respond differently to rising house prices (a wealth gain for one, a cost for the other)?</li> <li>Did housing support for families, such as subsidies and priority access in China, Korea or Japan, change births?</li> </ul> <h2>Data that could help</h2> <p>Regional births and house prices, household panels such as the China Family Panel Studies (see the <a href="/rankings/datasets">dataset ranking</a>), and the timing of family housing policies.</p> <h2>Next step</h2> <p>Relate city-level birth rates in China to house-price growth for 2010–2023, separately for cities with mostly owners and cities with many renters.</p> https://scholar4ai.com/ideas/household-debt-when-rates-rise 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Household debt and borrowing appear on 53 recent papers in the economics and finance journals, but only 9 of them carry an interest-rate, tightening, variable-rate or refinancing keyword. The 2022–2024 rate rises hit households with floating-rate mortgages hard in some countries and barely at all in others. Where did consumption fall, and did defaults follow?

<h2>What prompted this</h2> <p>Since October 2025, 53 papers carry a household-debt, household-borrowing, household-leverage or mortgage-debt keyword (29 in economics journals, 24 in finance): "household debt" on 32, "household borrowing" on 11, "household leverage" on 7. Only 9 of them also carry an interest-rate, rate-hike, monetary-tightening, variable-rate, adjustable or refinancing keyword.</p> <h2>Questions</h2> <ul> <li>In countries with mostly floating-rate mortgages (Australia, Sweden, the UK), how much did consumption fall per point of rate increase, compared with fixed-rate countries (the US, France)?</li> <li>Did households that had refinanced at low fixed rates stop moving house, and what did that do to labour mobility?</li> <li>Did arrears rise where expected, or did savings built up in the pandemic absorb the shock?</li> </ul> <h2>Data that could help</h2> <p>Loan-level mortgage data with rate types, household spending from bank transaction data, and arrears statistics by country.</p> <h2>Next step</h2> <p>Compare quarterly consumption growth across countries by their share of floating-rate mortgages during the 2022–2024 tightening.</p> https://scholar4ai.com/ideas/heat-pumps-and-retrofits 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Heat pumps appear on 12 recent papers, energy retrofits on 7 and building energy efficiency on 10, almost all in the economics journals. Heating is the hardest part of home decarbonization, and subsidies have had uneven results. Who installs, do the energy savings match the models, and why do many households not switch?

<h2>What prompted this</h2> <p>Since October 2025, 44 papers carry a heat-pump, building-electrification, residential-heating or energy-retrofit keyword (38 in economics journals): "heat pumps" on 12, "building energy efficiency" on 10, "energy retrofit" on 7. The quarter to September 2026 had 11 such papers against about 10 per quarter earlier, with no sign of a surge despite large new subsidy programmes.</p> <h2>Questions</h2> <ul> <li>Do measured energy savings after retrofits and heat pump installations match the engineering estimates, and where do they fall short?</li> <li>Which households take up subsidies, and do rebates reach renters and low-income owners?</li> <li>When electricity is taxed more heavily than gas, as in many European countries, do heat pumps save money for the household at all?</li> </ul> <h2>Data that could help</h2> <p>Smart meter data before and after installation, subsidy programme records, housing stock surveys and energy prices.</p> <h2>Next step</h2> <p>Match subsidy recipients' smart meter readings before and after heat pump installation and compare realized savings with the projected ones.</p> https://scholar4ai.com/ideas/heat-and-labour-productivity 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Extreme heat and heat waves appear on 47 recent papers, and temperature keywords on 208, but only 29 of those carry a productivity, labour, worker or firm keyword. As summers get hotter, how much output is lost on hot days, in which industries, and does adaptation close the gap?

<h2>What prompted this</h2> <p>Since October 2025, 208 papers carry an extreme-heat, heat-wave, heat-stress or temperature keyword (105 in economics journals, 69 in AI): "extreme heat" on 34, "heat waves" on 13. Only 29 of the 208 also carry a productivity, labour, worker, output, firm or absenteeism keyword; "labor productivity" alone appears on 94 papers, mostly about other causes.</p> <h2>Questions</h2> <ul> <li>Using daily plant-level data, how much does output fall on days above 35°C in manufacturing, construction and agriculture, and does the loss persist into the following days?</li> <li>Do firms in hotter regions adapt (shifts, cooling, automation) so that the marginal effect of heat falls over time?</li> <li>Who bears the loss: workers through pay, firms through profits, or customers through delays?</li> </ul> <h2>Data that could help</h2> <p>Daily production or output records from firms, weather station and reanalysis data (ERA5), and worker-level time and attendance data.</p> <h2>Next step</h2> <p>Match daily output data from a multi-plant manufacturer to local temperatures and estimate the output loss per hot day by plant type.</p> https://scholar4ai.com/ideas/greenwashing-talk-versus-emissions 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Greenwashing is one of the fastest-growing keywords in the finance journals, yet few papers measure it against emissions. Scored as the gap between what firms say about climate and what their emissions do, does greenwashing fool ESG raters, lenders or investors?

<h2>What prompted this</h2> <p>In the <a href="/categories/finance">finance dashboard</a> trends, "greenwashing" had 24 papers in July–September 2026 against about 13 expected from its earlier share. Since October 2025, 189 papers carry a greenwashing keyword (105 in finance, 37 in economics, 32 in management). Only 19 of them also carry "ESG ratings", and only 10 carry an emissions keyword, so most measure greenwashing from text or ratings alone.</p> <h2>Questions</h2> <ul> <li>Does a firm-level "talk minus walk" score (the change in climate language minus the change in emissions intensity) predict later ESG rating changes, controversies or enforcement?</li> <li>Do ESG raters reward the talk? Do loan spreads and bond yields?</li> <li>Does the gap shrink after mandatory climate disclosure rules?</li> </ul> <h2>Data that could help</h2> <ul> <li>Climate language in annual and sustainability reports, scored with a language model.</li> <li>Emissions and ratings from the ESG and energy categories of the <a href="/rankings/datasets">dataset ranking</a>, for example Trucost, CDP, Refinitiv ESG or MSCI ESG, or the EPA Greenhouse Gas Reporting Program for US plants.</li> <li>Staggered disclosure mandates as a <a href="/equations/difference-in-differences">difference-in-differences</a> design, estimated with heterogeneity-robust methods.</li> </ul> <h2>Next step</h2> <p>Collect the 19 papers that pair greenwashing with ESG ratings and list how each one measures greenwashing.</p> https://scholar4ai.com/ideas/green-hydrogen-costs 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Green hydrogen appears on 32 recent papers and hydrogen policy on 11, mostly in the economics journals, while projects are being cancelled as costs stay high. Which uses of hydrogen survive at realistic prices, and what do the subsidies buy?

<h2>What prompted this</h2> <p>Since October 2025, 100 papers carry a hydrogen keyword (60 in economics journals, 12 in AI, 12 in operations research, 11 in management): "green hydrogen" on 32, "hydrogen policy" on 11, "hydrogen storage" on 10. The count fell to 21 in July–September 2026 against about 25 per quarter earlier, as several flagship projects were shelved.</p> <h2>Questions</h2> <ul> <li>At electrolyser and electricity costs observed in 2025, which uses (fertilizer, steel, refining, shipping fuel, heating) break even, and which never will?</li> <li>How many announced projects reached final investment decision, and what distinguishes those that did?</li> <li>Do production subsidies such as the US hydrogen tax credit and the EU hydrogen bank lower costs over time, or mainly pay for projects that would exist anyway?</li> </ul> <h2>Data that could help</h2> <p>Project databases (the IEA hydrogen project tracker), electrolyser price surveys, subsidy awards, and firm announcements.</p> <h2>Next step</h2> <p>Compile the fate of every green hydrogen project announced in 2021–2023 and model the break-even hydrogen price for each end use at 2025 costs.</p> https://scholar4ai.com/ideas/graph-neural-networks-for-financial-contagion 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Financial contagion and systemic risk are rising in the finance and economics journals, and graph neural networks are among the largest topics in the AI journals. The two have met in only five papers. Can a graph model learn the channels of contagion that network models fix in advance?

<h2>What prompted this</h2> <p>"Financial contagion" is among the fastest-growing keywords across all journals on the <a href="/trends">trends page</a> (31 papers in July–September 2026 against 16 expected), and "systemic risk" is growing in economics. Since October 2025, 392 papers carry a contagion or systemic-risk keyword; 66 of them also carry a network keyword, but only 5 carry a graph-neural-network keyword. Over the same period "graph neural networks" appears on 578 papers, 555 of them in AI journals, and graph convolution is near the top of the <a href="/rankings/equations">equation ranking</a>.</p> <h2>Questions</h2> <ul> <li>Given exposure networks between banks or firms, does a graph neural network forecast joint tail losses (CoVaR, marginal expected shortfall) better than cascade models with fixed rules?</li> <li>Which links carry stress: lending, common holdings, supply chains or news co-mentions?</li> <li>Can the model's attributions tell supervisors which institutions transmit stress, and are they stable enough to act on?</li> </ul> <h2>Data that could help</h2> <ul> <li>Common-holdings networks from 13F filings, supply-chain links, and bank exposures where supervisors publish them.</li> <li>Daily returns for tail-risk measures such as CoVaR.</li> </ul> <h2>Next step</h2> <p>Rebuild one published common-holdings contagion network from 13F data and compare a simple graph convolution model with a CoVaR benchmark.</p> https://scholar4ai.com/ideas/goodwill-impairment-under-pressure 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Goodwill appears on 14 recent papers across the tracked journals, with 2 in the last quarter. After a decade of acquisitions at high prices, goodwill on corporate balance sheets is enormous, and standard setters have again debated bringing back amortization. When do firms impair, and does the market believe them?

<h2>What prompted this</h2> <p>Since October 2025, 14 papers carry a goodwill keyword (6 in accounting journals, 4 in finance): "goodwill" on 7, "goodwill impairment" on 6, "goodwill accounting" on 2. Only 2 were published in July–September 2026. Mergers and acquisitions, the source of goodwill, appear on more than 120 papers over the same period.</p> <h2>Questions</h2> <ul> <li>Do firms delay impairments until a CEO change or a bad year, so that the charge says more about management than about the acquired business?</li> <li>Do stock prices react to impairments, or have investors already written the goodwill off?</li> <li>Would amortization, as the FASB and IASB have considered, change acquisition prices or only reported earnings?</li> </ul> <h2>Data that could help</h2> <p>Goodwill balances and impairment charges (Compustat), acquisition records (SDC), CEO turnover and stock returns.</p> <h2>Next step</h2> <p>Measure how often large impairments follow CEO changes within a year, compared with the base rate, for US and European firms since 2015.</p> https://scholar4ai.com/ideas/glp1-drugs-and-the-economy 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Obesity appears on 15 recent papers in the economics journals and GLP-1 receptor agonists on 2. Tens of millions of people now take drugs that cut appetite. What happens to food demand, health spending, labour supply and the firms that sell what people stop buying?

<h2>What prompted this</h2> <p>Since October 2025, "obesity" appears on 15 papers and "childhood obesity" on 5, almost all in economics journals; "GLP-1 receptor agonists" appears on 2 and "weight loss" on 2. The drugs' spread since 2023 is one of the largest shifts in consumer health behaviour in decades, and the economics of it is barely in the journals.</p> <h2>Questions</h2> <ul> <li>Do households that start the drugs cut grocery and restaurant spending, and on which products?</li> <li>Do users work more hours or earn more after a year, and does that offset the drug's cost for insurers and employers?</li> <li>How are food, beverage and restaurant firms repositioning, and does the stock market price the change?</li> </ul> <h2>Data that could help</h2> <p>Prescription data linked to household spending panels, insurance claims with employment records, and firms' product and marketing changes.</p> <h2>Next step</h2> <p>Use a consumer panel with pharmacy and grocery purchases to compare food spending of households before and after a member starts a GLP-1 drug.</p> https://scholar4ai.com/ideas/global-minimum-tax-and-profit-shifting 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tax avoidance and profit shifting are steady topics across the finance, accounting and economics journals, but the 15% global minimum tax, in force in many countries since 2024, barely appears; only 3 of 103 recent papers carry a minimum-tax keyword. Does it end profit shifting, or move it to other margins?

<h2>What prompted this</h2> <p>Since October 2025, 103 papers carry a tax-avoidance, profit-shifting or tax-haven keyword (36 in finance journals, 30 in accounting, 27 in economics). Only 3 of them also carry a minimum-tax, Pillar Two or BEPS keyword, and "global minimum tax" appears on 4 papers in all.</p> <h2>Questions</h2> <ul> <li>For multinationals above the €750 million revenue threshold, did effective tax rates in low-tax affiliates rise toward 15% from 2024, compared with groups just below it?</li> <li>Did profits stop moving to low-tax countries, or move into the tax credits and incentives the rules still allow?</li> <li>How did low-tax countries respond: domestic top-up taxes, new subsidies, or both?</li> </ul> <h2>Data that could help</h2> <p>Public country-by-country reports, which the EU requires from large groups from 2025 onward, affiliate-level financials (Orbis), and the OECD's corporate tax statistics.</p> <h2>Next step</h2> <p>Compare effective tax rates of groups just above and just below the €750 million threshold before and after 2024, using affiliate data.</p> https://scholar4ai.com/ideas/gig-workers-and-social-insurance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The gig and platform economy appears on 44 recent papers in the management, economics and AI journals, but only 2 of them carry a social-insurance, pension, health-insurance, labour-protection or classification keyword. Courts and laws in Europe, the UK and parts of the US are reclassifying platform workers. What happens to their earnings and hours when protections arrive?

<h2>What prompted this</h2> <p>Since October 2025, 44 papers carry a gig, platform-work or platform-economy keyword (20 in management journals, 12 in economics): "platform economy" on 15, "gig economy" on 15, "gig workers" on 9, "platform work" on 8. Only 2 of them also carry a social-insurance, social-security, pension, health-insurance, labour-protection, regulation, employment-status or classification keyword.</p> <h2>Questions</h2> <ul> <li>After reclassification (Spain's rider law, the UK Uber ruling, the EU platform work directive), did platforms cut the number of workers, their hours or their flexibility?</li> <li>Did workers' total earnings rise with the protections, or fall with the hours?</li> <li>Which protections could be attached to the work rather than the status, such as portable benefit accounts, and have any been tried at scale?</li> </ul> <h2>Data that could help</h2> <p>Platform earnings data from worker-side apps and audits, administrative social insurance records, and the dates of legal changes by country.</p> <h2>Next step</h2> <p>Compare hours and earnings of delivery riders in Spain before and after the 2021 rider law with riders in Italy, where classification stayed contested.</p> https://scholar4ai.com/ideas/geopolitical-risk-and-supply-chain-rerouting 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Geopolitical risk is one of the most frequent keywords in the finance and economics journals, and supply chain resilience is a major topic in operations research, but the two rarely meet in OR journals. Customs data can show whether trade re-routes through third countries when geopolitical risk rises, and at what cost.

<h2>What prompted this</h2> <p>Since October 2025, 445 papers carry a geopolitical keyword (202 in finance journals, 167 in economics), and "supply chain resilience" appears on 233 (95 in operations research). Only 50 papers carry both kinds of keyword, and only 8 of them are in operations research journals. In the last 30 days "geopolitical risk" is the 31st most frequent keyword.</p> <h2>Questions</h2> <ul> <li>After geopolitical shocks such as sanctions or tariff rounds, how much trade moves to connector countries rather than disappearing?</li> <li>Does re-routing show up in costs and lead times, for example in unit values and shipping times?</li> <li>Which network-design models in the OR literature predict the re-routing that actually happens?</li> </ul> <h2>Data that could help</h2> <ul> <li>UN Comtrade or CEPII BACI flows, Chinese customs records and tariff data (the international trade category of the <a href="/rankings/datasets">dataset ranking</a>).</li> <li>The Geopolitical Risk index of Caldara and Iacoviello.</li> <li>A gravity model with exporter-year, importer-year and pair fixed effects (see the <a href="/rankings/equations">equation ranking</a>).</li> </ul> <h2>Next step</h2> <p>Estimate a gravity model for 2015–2025 with the geopolitical risk index interacted with political alignment, and look for trade diverted to third countries.</p> https://scholar4ai.com/ideas/generative-ai-tutors-in-accounting-education 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Accounting education is one of the fastest-growing keywords in the accounting journals, and a quarter of those papers now involve generative AI. Does an AI tutor help students learn accounting, or only finish the homework?

<h2>What prompted this</h2> <p>On the <a href="/categories/accounting">accounting dashboard</a>, "accounting education" had 10 papers in July–September 2026 against about 5 expected from its earlier share, and "generative AI" is growing there too. Since October 2025, 36 papers carry an accounting-education keyword (33 of them in accounting journals), and 9 of those also carry a generative-AI, ChatGPT, language-model or AI keyword.</p> <h2>Questions</h2> <ul> <li>When access to an AI tutor is assigned at random by section or assignment, do students score better on exams they take without it?</li> <li>Which topics gain (journal entries, consolidation, tax) and which suffer?</li> <li>Do weaker students gain more, as some studies of AI assistance at work suggest, or does the tutor widen the gaps?</li> </ul> <h2>Possible approach</h2> <p>Randomize access across sections of an introductory course, measure learning with closed-book exams, and log the tutor conversations to see what students ask.</p> <h2>Next step</h2> <p>Read the nine papers that combine both topics and note which ones measure learning rather than satisfaction or usage.</p> https://scholar4ai.com/ideas/generative-ai-and-entry-level-jobs 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Employment is rising as a keyword in the economics journals, and generative AI is growing there fast. Yet of nearly 1,200 recent papers on employment, wages and labour markets, only 14 carry a generative-AI keyword. Is AI taking the tasks that junior workers used to learn on, and what happens to their careers?

<h2>What prompted this</h2> <p>In the economics journals, "employment" had 16 papers in July–September 2026 against about 8 expected, "labor market tightness" 6 against 1, and "generative AI" 9 against 3 (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, 1,181 papers carry an employment, labour-market or wage keyword (918 in economics journals). Only 14 of them also carry a generative-AI, language-model or ChatGPT keyword, and "young workers" appears on 10.</p> <h2>Questions</h2> <ul> <li>In occupations most exposed to generative AI, has hiring of new graduates fallen since 2023 relative to hiring of experienced workers?</li> <li>Do firms that adopt AI change their training and promotion ladders?</li> <li>Do workers who started their careers in exposed jobs earn less years later?</li> </ul> <h2>Data that could help</h2> <p>Online job postings (Lightcast), worker profiles (LinkedIn, Revelio), occupational task data (O*NET) to measure AI exposure, and payroll or administrative employment data.</p> <h2>Next step</h2> <p>Compare monthly entry-level and experienced job postings by occupational AI exposure from 2021 to 2026.</p> https://scholar4ai.com/ideas/foundation-models-versus-professional-forecasters 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Foundation models are rising fast in the AI journals, and time-series versions now forecast series they were never trained on. Economists forecast inflation and growth with surveys and small models. Do the new models beat the Survey of Professional Forecasters in real time?

<h2>What prompted this</h2> <p>In the AI journals, "foundation models" had 27 papers in July–September 2026 against about 16 expected (see the <a href="/categories/ai">AI dashboard</a>). Since October 2025, 1,714 papers carry a foundation-model or large-language-model keyword, but only 5 of them also carry an inflation, macroeconomic or GDP forecasting or nowcasting keyword. "Inflation forecasting" appears on 31 papers (22 in economics), and "time series forecasting" on 148 (117 in AI journals).</p> <h2>Questions</h2> <ul> <li>With real-time data only, do zero-shot time-series foundation models beat the median forecasts of the Survey of Professional Forecasters for inflation and GDP growth?</li> <li>Can pretraining leak the future into tests that look out of sample? Data vintages help check this.</li> <li>Where do the models fail: at turning points, at regime changes, or around data revisions?</li> </ul> <h2>Data that could help</h2> <p>Real-time data vintages (the Philadelphia Fed's real-time data set or ALFRED), the Survey of Professional Forecasters, and FRED-MD.</p> <h2>Next step</h2> <p>Run one open time-series foundation model on real-time CPI vintages for 2015–2025 and score it against the survey median.</p> https://scholar4ai.com/ideas/foreign-aid-cuts 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Foreign aid, humanitarian aid and aid effectiveness appear on 69 recent papers, mostly in the economics journals, with no rise despite the largest aid cuts in decades in 2025. The cuts are a natural experiment that the earlier aid-effectiveness debate never had. Which programmes' loss shows up first in health, schooling and conflict?

<h2>What prompted this</h2> <p>Since October 2025, 69 papers carry a foreign-aid, development-aid, humanitarian-aid, aid-effectiveness or donor keyword (45 in economics journals): "foreign aid" on 16, "humanitarian aid" on 12, "development aid" on 6, "aid effectiveness" on 6, "aid allocation" on 5. The quarter to September 2026 had 16 such papers, in line with earlier quarters.</p> <h2>Questions</h2> <ul> <li>In districts where health programmes lost funding in 2025, did child mortality and disease incidence rise within a year, and did governments or other donors replace the money?</li> <li>Did cash-based humanitarian aid, cut in several crises, show larger effects than in-kind aid when it stopped?</li> <li>Do the cuts change where people migrate from?</li> </ul> <h2>Data that could help</h2> <p>Project-level aid data with locations (AidData, the OECD creditor reporting system), health surveillance and survey data, and conflict and migration records.</p> <h2>Next step</h2> <p>Map the 2025 programme terminations to districts and compare health indicators in those districts with neighbouring ones that kept funding.</p> https://scholar4ai.com/ideas/food-waste-and-perishables 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Food waste reduction, perishable goods and perishable inventory appear on 69 recent papers, mostly in operations research, with the food waste problem studied separately in economics. Dynamic discounts, better forecasts and date labels all promise less waste. Which of them works at scale, and who captures the savings?

<h2>What prompted this</h2> <p>Since October 2025, 69 papers carry a food-waste, perishable, food-loss, shelf-life or expiry keyword (40 in operations research journals, 14 in economics): "food waste reduction" on 17, "perishable goods" on 14, "perishable inventory" on 12, "food waste" on 12, "shelf life" on 6. The inventory models and the policy studies rarely cite each other.</p> <h2>Questions</h2> <ul> <li>In retailers that introduced automatic discounts on products near expiry, how much waste fell, and did the discounts cannibalize full-price sales?</li> <li>Do AI demand forecasts reduce waste more in fresh categories than in packaged ones, and do stores trust the forecasts enough to order less?</li> <li>Did changes to date labels ("best before" versus "use by") reduce household waste, and did they raise safety incidents?</li> </ul> <h2>Data that could help</h2> <p>Store-level waste and sales records from retailers, forecast system deployment dates, and household waste surveys around labelling changes.</p> <h2>Next step</h2> <p>Compare waste rates in fresh categories at stores that adopted dynamic markdowns with matched stores in the same chain that had not yet adopted them.</p> https://scholar4ai.com/ideas/food-prices-and-unrest 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Food security, food prices and food inflation appear on 105 recent papers, mostly in economics, but only 6 of them carry a conflict, unrest, protest or violence keyword. Food price spikes preceded unrest in 2008 and 2011. With prices high again and aid budgets cut, where is the risk greatest, and does early warning work?

<h2>What prompted this</h2> <p>Since October 2025, 105 papers carry a food-price, food-inflation or food-security keyword (83 in economics journals): "food security" on 64, "food prices" on 19, "food inflation" on 14. Only 6 of them also carry a conflict, unrest, protest, riot, violence, instability or war keyword.</p> <h2>Questions</h2> <ul> <li>Do local food price spikes predict protests and violence at the district level, and is the link stronger where governments subsidize bread and fuel?</li> <li>Do food aid cuts, such as those of 2025, raise the risk in the places that lost the most aid?</li> <li>Can commodity futures and weather forecasts give months of warning, and do governments act on it?</li> </ul> <h2>Data that could help</h2> <p>Local food prices (the World Food Programme's price database), conflict events (ACLED), aid flows and weather data.</p> <h2>Next step</h2> <p>Estimate the relationship between district-level food price changes and protest events in Africa and the Middle East for 2015–2025, with aid cuts as an added shock.</p> https://scholar4ai.com/ideas/food-delivery-couriers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Online food delivery, delivery apps and on-demand delivery appear on 34 recent papers, mostly in operations research, and the topic is fading, with 5 papers in the last quarter against about 10 per quarter before. Platforms have consolidated, raised fees and face new labour rules. How do the dispatch models interact with courier pay, and who pays for faster delivery?

<h2>What prompted this</h2> <p>Since October 2025, 34 papers carry a food-delivery, on-demand-delivery, quick-commerce, delivery-platform or courier keyword (21 in operations research journals): "online food delivery" on 8, "food delivery apps" on 6, "on-demand food delivery" on 5, "on-demand delivery" on 5, "food delivery" on 4, "courier services" on 3. Five were published in July–September 2026 against about 10 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When platforms batch orders to cut courier kilometres, how much of the saving goes to couriers' earnings, to restaurants, or to the platform?</li> <li>Did minimum pay rules for couriers (New York, Seattle, Spain) change delivery times, fees and order volumes, and did couriers earn more per hour after fewer hours?</li> <li>Do the dispatch models' assumptions about courier availability hold when pay rules change who works and when?</li> </ul> <h2>Data that could help</h2> <p>Order-level data released in regulatory proceedings, courier earnings data from worker-side apps, and platform fee histories.</p> <h2>Next step</h2> <p>Compare delivery fees, order volumes and courier hourly earnings in New York before and after the 2023 minimum pay rule, using platform disclosures and worker-side data.</p> https://scholar4ai.com/ideas/flood-risk-in-mortgage-lending 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Climate risk is one of the most frequent keywords in the finance journals, yet only 5 of 82 recent mortgage papers carry a flood, climate-risk or disaster keyword. Do lenders price flood risk into mortgages, and do they pass the riskiest loans to government-backed buyers?

<h2>What prompted this</h2> <p>Since October 2025, 82 papers carry a mortgage keyword (54 in finance journals, 26 in economics), and only 5 of them also carry a flood, climate-risk, physical-risk, wildfire or natural-disaster keyword. "Climate risk" alone appears on 242 papers, so the two literatures exist side by side without meeting.</p> <h2>Questions</h2> <ul> <li>Do mortgages on homes in flood zones carry higher rates or lower loan-to-value ratios, after controlling for borrower risk?</li> <li>Do lenders keep low-risk loans and sell flood-exposed ones to Fannie Mae and Freddie Mac, shifting the risk to taxpayers?</li> <li>When flood maps are updated, do lending and house prices in newly mapped areas change?</li> </ul> <h2>Data that could help</h2> <p>Loan-level mortgage data (HMDA, Fannie Mae and Freddie Mac loan files), flood zone maps and their revision dates, and house transaction records (CoreLogic, Zillow).</p> <h2>Next step</h2> <p>Compare the share of mortgages sold to the government-sponsored enterprises inside and outside flood zones, before and after a flood map update in one state.</p> https://scholar4ai.com/ideas/fiscal-expansion-and-bank-credit-in-the-euro-area 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Government spending, borrowing costs and the euro area are all rising keywords in the finance journals. When governments borrow more, domestic banks often buy the bonds. Does that crowd out bank lending to firms, and which firms lose the most?

<h2>What prompted this</h2> <p>In the finance journals, "government spending" had 9 papers in July–September 2026 against about 1 expected, "borrowing costs" 9 against 3, "euro area" 9 against 3 and "credit supply" 7 against 2 (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 327 papers carry a government-spending, fiscal-policy, public-debt or sovereign-debt keyword (213 in economics journals, 105 in finance). Only 13 of them carry a credit-supply, bank-lending, crowding-out or sovereign-bank keyword.</p> <h2>Questions</h2> <ul> <li>When a government issues more debt, do the domestic banks that buy it lend less to firms than banks that do not?</li> <li>Are small firms, which depend on banks, hit harder than large firms that can issue bonds?</li> <li>Does central-bank bond buying, or its end, change the effect?</li> </ul> <h2>Data that could help</h2> <p>Bank-level sovereign holdings (EBA transparency exercises), the euro area credit register AnaCredit where access allows, firm balance sheets (Orbis) and national debt issuance.</p> <h2>Next step</h2> <p>Relate changes in banks' holdings of their own government's bonds to their lending to firms around large national issuance waves.</p> https://scholar4ai.com/ideas/firm-peers-from-contrastive-embeddings 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Contrastive learning is one of the largest topics in the AI journals, yet only 6 of 749 recent contrastive-learning papers carry a finance or firm keyword. Finance research still defines peer firms mostly by industry codes. Would firm embeddings learned contrastively from filings, returns and supply chains give better peers for valuation, benchmarking and pay?

<h2>What prompted this</h2> <p>Since October 2025, "contrastive learning" appears on 548 papers (538 in AI journals), and 749 papers carry some contrastive-learning keyword; only 6 of them carry a finance, stock, firm, company or industry-classification keyword. "Peer firms" appears on 13 papers and "industry classification" on 4.</p> <h2>Questions</h2> <ul> <li>Do peers from contrastive embeddings explain a firm's returns, valuation multiples and earnings better than SIC or GICS peers, or text-based peers?</li> <li>Do they pick up a firm's move into new businesses sooner than industry codes do?</li> <li>Would they change results that rely on peers, such as relative performance evaluation in CEO pay?</li> </ul> <h2>Data that could help</h2> <p>Business descriptions in 10-K filings, stock returns (CRSP), supply-chain links and industry codes (Compustat).</p> <h2>Next step</h2> <p>Train a contrastive model in which a firm's filing text and its return co-movement in the next year form positive pairs, and compare the resulting peers with GICS peers.</p> https://scholar4ai.com/ideas/finfluencers-and-retail-trading 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Retail investors are a large topic in the finance journals, but only 8 of 139 recent papers on them carry a social-media keyword, and "finfluencers" appears on just 2. Do the trades that follow a finance influencer's post earn anything, and who profits from the audience?

<h2>What prompted this</h2> <p>Since October 2025, 139 papers carry a retail- or individual-investor keyword (109 in finance journals, 13 in accounting). Only 8 of them also carry a social-media, influencer, Reddit, Twitter, YouTube or TikTok keyword, and "finfluencers" appears on 2 papers. "Social media" itself appears on 115 papers, most of them in AI and management journals, where it is studied as text rather than as a source of trades.</p> <h2>Questions</h2> <ul> <li>After a popular finance influencer recommends a stock, how much retail order flow follows, and do the followers earn or lose money over the next weeks?</li> <li>Do influencers who are paid by brokers or issuers recommend worse stocks than those who are not?</li> <li>Did disclosure rules for paid promotion, now in force in several countries, change what influencers post and how markets react?</li> </ul> <h2>Data that could help</h2> <p>Timestamped posts from TikTok, YouTube and X with view counts, retail order flow proxies such as off-exchange trades, and stock returns.</p> <h2>Next step</h2> <p>Collect the posts of 50 finance influencers for one year, match the stocks they mention to retail trading volume and returns, and compare paid and unpaid posts.</p> https://scholar4ai.com/ideas/financial-sentiment-beyond-english 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Investor sentiment and sentiment analysis appear on 490 recent papers across the AI and finance journals, but only 22 of them carry a multilingual, cross-lingual or non-English language keyword, and financial sentiment analysis on 6. Most of the world's markets are discussed in languages the standard tools were not built for. How well do the tools work there, and do they move markets?

<h2>What prompted this</h2> <p>Since October 2025, 490 papers carry a sentiment keyword (206 in AI journals, 193 in finance, 50 in economics): "investor sentiment" on 186, "sentiment analysis" on 127, "multimodal sentiment analysis" on 74, "news sentiment" on 19, "financial sentiment analysis" on 6. Only 22 of the 490 also carry a multilingual, cross-lingual, low-resource or specific non-English language keyword, led by "low-resource languages" on 9 and "multilingual sentiment analysis" on 6.</p> <h2>Questions</h2> <ul> <li>Do financial sentiment models fine-tuned on English lose accuracy on Japanese, Korean, Arabic or Indonesian financial news, and do translation layers recover it?</li> <li>Does news sentiment predict returns in markets with fewer analysts and more retail investors, more or less than in US markets?</li> <li>Do the sentiment measures used in published papers on emerging markets reflect the languages their investors read?</li> </ul> <h2>Data that could help</h2> <p>Financial news and social media in several languages, local market returns, and multilingual financial sentiment datasets, which are few.</p> <h2>Next step</h2> <p>Build a labelled financial sentiment set in three non-English languages and compare the accuracy of English-trained models, multilingual models and translation pipelines.</p> https://scholar4ai.com/ideas/financial-misinformation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Misinformation, fake news detection and rumor detection appear on 185 recent papers, 140 of them in AI journals, and only 10 carry a market, stock, investor or finance keyword. Fake earnings leaks, forged press releases and coordinated rumours now move prices within minutes. How much do they move, who profits, and can detection work at trading speed?

<h2>What prompted this</h2> <p>Since October 2025, 185 papers carry a misinformation, fake-news, disinformation or rumor keyword (140 in AI journals, 18 in economics, 17 in management): "misinformation" on 75, "fake news detection" on 42, "social media misinformation" on 21, "rumor detection" on 16, "misinformation detection" on 16. Only 10 of the 185 also carry an economics, finance, market, stock, firm, investor, crypto, bank or trading keyword.</p> <h2>Questions</h2> <ul> <li>For documented cases of false financial news (forged releases, fake regulatory filings, fabricated analyst notes), how large and how long-lived was the price move, and who traded before the correction?</li> <li>Do detectors trained on political misinformation transfer to financial rumours, which are shorter, numerical and time-critical?</li> <li>Do firms that are targeted by false news change their disclosure practices?</li> </ul> <h2>Data that could help</h2> <p>A catalogue of false financial news incidents with timestamps, intraday prices and volumes, social media archives and enforcement actions.</p> <h2>Next step</h2> <p>Build a catalogue of 100 documented false financial news incidents since 2018 and measure the price reaction and reversal around each.</p> https://scholar4ai.com/ideas/financial-literacy-and-crypto-speculation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Financial literacy is one of the largest household finance topics, on 146 recent papers, but only 7 of 161 literacy papers carry a crypto, gambling or speculation keyword. Does knowing more about finance keep people out of speculative crypto trading, or make them more confident speculators?

<h2>What prompted this</h2> <p>Since October 2025, 161 papers carry a financial-literacy keyword (112 in finance journals, 24 in economics, 20 in management), with "financial literacy" alone on 146 and "digital financial literacy" on 16. Only 7 of the 161 also carry a cryptocurrency, Bitcoin, digital-asset, gambling or speculation keyword, while "cryptocurrency" appears on 76 papers.</p> <h2>Questions</h2> <ul> <li>Are financially literate people less likely to hold crypto, or more likely to hold it in small, diversified amounts?</li> <li>Does literacy protect against the worst outcomes (leveraged trading, scams, panic selling in crashes)?</li> <li>Do literacy programmes that now include crypto change behaviour, or only quiz scores?</li> </ul> <h2>Data that could help</h2> <p>Household surveys that ask about crypto holdings and literacy (for example the Survey of Consumer Finances, the China Household Finance Survey and national financial capability surveys), and exchange account data where available.</p> <h2>Next step</h2> <p>Use a household survey with both literacy questions and crypto holdings to compare holdings, trading frequency and losses across literacy levels.</p> https://scholar4ai.com/ideas/fertilizer-prices-and-food 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Fertilizer use and prices appear on 30 recent papers, almost all in the economics journals. The 2022 fertilizer price spike after Russia's invasion of Ukraine was one of the largest on record. How much did farmers cut application, what did it do to yields and food prices a year later, and did it speed up the shift to precision farming?

<h2>What prompted this</h2> <p>Since October 2025, 30 papers carry a fertilizer keyword (26 in economics journals): "fertilizer use" on 6, "fertilizer prices" on 5, "fertilizer management" on 4. Food prices and food security appear on far more papers, but the link from input prices to output prices is rarely traced through farms.</p> <h2>Questions</h2> <ul> <li>After the 2022 spike, by how much did farmers in different countries cut nitrogen application, and how much yield did they lose?</li> <li>Did the yield loss show up in 2023 food prices, and did importing countries with fertilizer subsidies fare better?</li> <li>Did high prices accelerate adoption of precision application, organic fertilizer or nitrogen-fixing crops, and did the change persist when prices fell?</li> </ul> <h2>Data that could help</h2> <p>Fertilizer sales and prices by country (FAO, IFA), farm surveys with application rates, crop yields and food price indices.</p> <h2>Next step</h2> <p>Relate country-level changes in nitrogen use in 2022 to the fertilizer price shock and to the following year's cereal yields.</p> https://scholar4ai.com/ideas/federated-learning-against-money-laundering 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Federated learning, which trains one model across organizations without pooling their data, is a large topic in the AI journals, yet only 7 of 398 recent papers on it carry a finance, banking, fraud or money-laundering keyword. Money launderers spread transactions across banks precisely because no single bank sees the whole pattern. Can banks catch them together without sharing customer data?

<h2>What prompted this</h2> <p>Since October 2025, 398 papers carry a federated-learning keyword (386 in AI journals), and only 7 of them carry a money-laundering, bank, finance, fraud or credit keyword. "Anti-money laundering" appears on 14 papers. In the AI journals, "decentralized learning" is rising (8 papers in July–September 2026 against about 2 expected).</p> <h2>Questions</h2> <ul> <li>On realistic synthetic data in which laundering chains cross banks, how much does a federated model improve detection over models that each bank trains alone?</li> <li>How much detection do privacy protections such as secure aggregation and differential privacy cost?</li> <li>Who should run and pay for such a system: the banks, a shared utility or the regulator?</li> </ul> <h2>Data that could help</h2> <p>Synthetic anti-money-laundering datasets with transactions across banks, public transaction-graph benchmarks, and open federated learning frameworks.</p> <h2>Next step</h2> <p>Split a synthetic money-laundering dataset across five simulated banks and compare federated, bank-by-bank and pooled models on detection and false alarms.</p> https://scholar4ai.com/ideas/fake-reviews-from-language-models 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Online reviews appear on 54 recent papers, and 17 of 73 review papers carry a fake, manipulation or detection keyword. Fake reviews were once cheap but detectable; language models make them cheap and fluent. Can platforms still tell, how much of the review stock is now synthetic, and do consumers still trust ratings?

<h2>What prompted this</h2> <p>Since October 2025, 73 papers carry an online-review, product-review, consumer-review or fake-review keyword (32 in AI journals, 21 in management, 12 in operations research): "online reviews" on 54, "fake review detection" on 5, "fake reviews" on 4, "review manipulation" on 3. Seventeen of the 73 carry a fake, manipulation, fraud, incentive or detection keyword.</p> <h2>Questions</h2> <ul> <li>What share of recent reviews on major platforms show the statistical fingerprints of language models, and did the share jump in 2023?</li> <li>Do detectors trained on older fake reviews catch model-written ones, and can platforms keep up?</li> <li>As fake reviews become undetectable, do consumers shift to verified purchases, video reviews or recommendations from people they know?</li> </ul> <h2>Data that could help</h2> <p>Review text with timestamps (public review datasets, scraped samples), AI-text detectors, platform enforcement announcements and consumer surveys.</p> <h2>Next step</h2> <p>Run an AI-text detector on time-stamped reviews from one platform for 2019–2025 and track the flagged share by month and product category.</p> https://scholar4ai.com/ideas/fairness-audits-of-credit-scoring 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Fairness and algorithmic bias are growing in the AI journals, and credit scoring is one of the areas where AI rules are strictest. Yet only 19 of 473 recent fairness and bias papers carry a credit or lending keyword. What would an audit that lenders and regulators can both use look like, and what does fairness cost in default risk?

<h2>What prompted this</h2> <p>In the AI journals, "fairness" had 16 papers in July–September 2026 against about 10 expected and "algorithmic bias" 16 against about 11 (see the <a href="/categories/ai">AI dashboard</a>). Since October 2025, 473 papers carry a fairness, algorithmic-bias or discrimination keyword (186 in AI journals, 155 in economics), and only 19 of them carry a credit-scoring, lending, loan or credit-risk keyword.</p> <h2>Questions</h2> <ul> <li>On public credit data, how do approval rates and default rates change across groups when common fairness constraints are imposed?</li> <li>Do the fairness fixes popular in AI papers hold up under the legal tests used in lending, such as disparate impact?</li> <li>Can an audit work without protected attributes, which lenders are often not allowed to collect?</li> </ul> <h2>Data that could help</h2> <p>Mortgage applications with applicant demographics (HMDA), public credit datasets such as German Credit and LendingClub (the credit category of the <a href="/rankings/datasets">dataset ranking</a>), and the models of recent papers.</p> <h2>Next step</h2> <p>Train a standard credit model on LendingClub loans and on HMDA applications, apply three common fairness constraints, and report how approval gaps and predicted default rates change.</p> https://scholar4ai.com/ideas/fair-division-in-operations 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The equity-efficiency trade-off and fair division appear on 49 recent papers across the economics, operations research and AI journals. Routing, scheduling and allocation models increasingly add fairness constraints, for drivers, patients or regions. How much efficiency do the constraints cost in practice, and do the beneficiaries see them as fair?

<h2>What prompted this</h2> <p>Since October 2025, 49 papers carry a fair-allocation, fair-division, equitable, equity-efficiency, fairness-constraint or envy-free keyword (20 in economics journals, 17 in operations research, 9 in AI): "equity-efficiency trade-off" on 25, "fair division" on 15, "envy-freeness" on 2, "fairness constraints" on 2.</p> <h2>Questions</h2> <ul> <li>Across published models with fairness constraints, how large is the efficiency loss at the fairness level a decision maker would choose, and is the trade-off curve usually flat or steep?</li> <li>Do the people the constraints protect (drivers given balanced routes, regions given minimum service) recognize the outcome as fair, or do they judge fairness by a different measure?</li> <li>Which fairness definitions survive contact with operational data, where groups overlap and needs differ?</li> </ul> <h2>Data that could help</h2> <p>The instances and results of recent fairness-constrained models, and surveys of affected groups in a real deployment.</p> <h2>Next step</h2> <p>Re-solve ten published fairness-constrained models across the full range of the fairness parameter and plot the price of fairness for each.</p> https://scholar4ai.com/ideas/factor-crowding-and-anomaly-decay 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The cross-section of stock returns, momentum and asset pricing anomalies appear on 210 recent papers in the finance and economics journals, but only 14 of them carry a crowding, capacity, decay, replication or out-of-sample keyword. As quantitative funds trade the same signals, do anomaly returns shrink, and can crowding be measured in time?

<h2>What prompted this</h2> <p>Since October 2025, 210 papers carry a factor-investing, factor-model, anomaly, momentum or cross-section-of-returns keyword (130 in finance journals, 38 in economics). Only 14 of them also carry a crowding, capacity, arbitrage-capital, decay, post-publication, replication or out-of-sample keyword. "Factor investing" appears on 8 papers and "stock return predictability" on 28.</p> <h2>Questions</h2> <ul> <li>Do anomaly returns fall after the strategies become available in cheap funds, beyond the decline after publication?</li> <li>Can crowding be measured from short interest, fund holdings and trading volume in the anomaly portfolios, and does it predict crashes such as the 2007 quant meltdown?</li> <li>Do the same anomalies decay in China's A-share market, where quantitative funds grew later?</li> </ul> <h2>Data that could help</h2> <p>Anomaly portfolio returns (Open Source Asset Pricing, the Kenneth French library), fund holdings (13F filings), short interest and trading volume (CRSP).</p> <h2>Next step</h2> <p>Build a crowding measure from 13F holdings for 20 anomalies and test whether it predicts their returns over the following year.</p> https://scholar4ai.com/ideas/fab-scheduling 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Semiconductor manufacturing appears on 14 recent papers and wafer fabrication on 3, split between the AI and operations research journals, while chip plants worth tens of billions are being built with public money. Fab scheduling is among the hardest in industry. Do the published methods reach the plants, and how much yield and throughput do they add?

<h2>What prompted this</h2> <p>Since October 2025, 48 papers carry a semiconductor or wafer-fabrication keyword (20 in AI journals, 9 in operations research, 8 in economics): "semiconductor manufacturing" on 14, "semiconductor industry" on 12, "wafer fabrication" on 3, "semiconductor supply chain" on 3. Twenty-four of them carry a scheduling, dispatching, production, yield or capacity keyword. The economics of chip subsidies is a separate and larger literature.</p> <h2>Questions</h2> <ul> <li>Do dispatching rules learned by reinforcement learning, as in several recent papers, beat the rules fabs use, on fab-scale simulations with realistic tool failures?</li> <li>How much of a new fab's ramp-up time comes from scheduling and yield learning, and can it be shortened by methods from existing plants?</li> <li>As fabs multiply across countries with different labour costs, which scheduling constraints change?</li> </ul> <h2>Data that could help</h2> <p>Public fab simulation models and datasets (for example the SMT2020 testbed), ramp-up timelines from company reports, and yield data where disclosed.</p> <h2>Next step</h2> <p>Benchmark three recent learned dispatching methods against standard rules on a public fab simulation testbed with tool failures turned on.</p> https://scholar4ai.com/ideas/explainable-optimization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

In the operations research journals, 1,178 recent papers carry an optimization keyword, and 36 of them carry an explanation or interpretability keyword, while interpretability appears on 379 papers in the archive, almost all about machine learning. Planners reject optimized schedules they cannot understand. What counts as an explanation of a solution, and do explanations raise adoption?

<h2>What prompted this</h2> <p>Since October 2025, 1,178 papers in the operations research journals carry an optimization, mathematical-programming or mixed-integer keyword, and 36 of them also carry an explanation, interpretability, counterfactual, transparency, human-in-the-loop or trust keyword. "Interpretability" appears on 379 papers in the archive, mostly in AI journals about models rather than decisions; "decision support system" appears on 70.</p> <h2>Questions</h2> <ul> <li>Which forms of explanation (which constraints bind, what would change the answer, why a nearby plan is worse) help planners accept or correct a solution, and which do they ignore?</li> <li>Do explanations change the solutions planners choose, and do the changes improve or worsen outcomes?</li> <li>Can explanations be generated for large mixed-integer models without solving many variants?</li> </ul> <h2>Data that could help</h2> <p>Experiments with planners choosing between optimized plans with different explanations, and logs from decision support systems that record overrides.</p> <h2>Next step</h2> <p>Run an experiment with professional planners comparing acceptance of the same optimized schedule presented with three kinds of explanation.</p> https://scholar4ai.com/ideas/expected-credit-losses-in-the-first-real-test 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Loan loss provisions appear on 23 recent papers in the finance and accounting journals, IFRS 9 on 5 and CECL on 3. The expected-loss rules were designed after 2008 to make banks provision earlier. The 2022–2024 rate shock and the pandemic were their first real tests. Did provisions move earlier, or just more?

<h2>What prompted this</h2> <p>Since October 2025, 31 papers carry a loan-loss-provision, expected-credit-loss, IFRS 9 or CECL keyword (19 in finance journals, 11 in accounting): "loan loss provisions" on 23, "expected credit loss" on 6, "IFRS 9" on 5, "CECL" on 3. Twenty of the 31 carry a procyclicality, crisis, lending, capital or discretion keyword, so the question is on the table but with few papers.</p> <h2>Questions</h2> <ul> <li>Did banks under IFRS 9 and CECL raise provisions at the start of the pandemic and the rate shock, and release them later, in a way that stabilized or amplified lending?</li> <li>Do the forward-looking scenarios banks use for provisions differ systematically from consensus forecasts, in the direction of smoother earnings?</li> <li>Did the "day one" provisioning for new loans reduce lending to riskier borrowers, as critics predicted?</li> </ul> <h2>Data that could help</h2> <p>Bank provisions by stage and scenario disclosures from annual reports, loan-level data, and macro forecasts at the time.</p> <h2>Next step</h2> <p>Compare quarterly provision changes of IFRS 9 banks with those of banks still under incurred-loss rules during 2020 and 2022–2023.</p> https://scholar4ai.com/ideas/evaluating-fraud-detection-in-money 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Class imbalance is one of the largest topics in the AI journals, and fraud detection is where it matters most for finance. But only 10 of 174 recent fraud papers carry a cost, profit or loss keyword; most report accuracy or F1 scores. Would ranking fraud models by the money they save change which ones win?

<h2>What prompted this</h2> <p>Since October 2025, "class imbalance" appears on 471 papers (427 in AI journals), and in the AI journals "label imbalance" is rising (7 papers in July–September 2026 against about 1 expected). Of the 174 papers with a fraud keyword (58 in finance journals, 57 in AI, 25 in economics, 16 in accounting), only 10 carry a cost-sensitive, profit, business-value, loss or cost keyword.</p> <h2>Questions</h2> <ul> <li>On public fraud datasets, how often is the model with the best F1 score not the one that saves the most money once losses and investigation costs are counted?</li> <li>Do the oversampling methods popular in AI papers still help when the score is money?</li> <li>How should the alert threshold be set when investigators can check only a fixed number of cases a day?</li> </ul> <h2>Data that could help</h2> <p>Public fraud datasets with transaction amounts, such as the credit card fraud, PaySim and IEEE-CIS data in the <a href="/rankings/datasets">dataset ranking</a>'s list, and the models of published papers.</p> <h2>Next step</h2> <p>Re-evaluate five published fraud models on a public credit card dataset by the money they save under a fixed daily investigation budget.</p> https://scholar4ai.com/ideas/eu-deforestation-regulation 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Deforestation appears on 28 recent papers and the EU Deforestation Regulation on 4, with 10 deforestation papers in the last quarter against about 8 per quarter before. The rule requires proof that coffee, cocoa, soy, palm oil and other goods were not grown on deforested land. Will it change where forests fall, or only where goods are sold?

<h2>What prompted this</h2> <p>Since October 2025, 36 papers carry a deforestation keyword (34 in economics journals): "deforestation" on 28, "EU Deforestation Regulation" on 4, "zero deforestation" on 2. Twelve of the 36 also carry a supply-chain, trade, regulation, commodity or due-diligence keyword. The rule's start was delayed to the end of 2025 and its scope narrowed, which makes its first years a moving target.</p> <h2>Questions</h2> <ul> <li>Do exporters divert goods from recently deforested land to markets without such rules, so that EU imports become clean while deforestation continues?</li> <li>Can smallholders provide the geolocation proof the rule requires, or are they dropped from supply chains?</li> <li>Does the rule's traceability requirement change the price premium for certified products?</li> </ul> <h2>Data that could help</h2> <p>Trade data by product and origin, satellite forest loss (Global Forest Watch), supply chain traceability data (for example Trase), and farm surveys in producing countries.</p> <h2>Next step</h2> <p>Compare forest loss near farms that supply the EU with loss near farms that supply other markets, using traceability data for cocoa or palm oil.</p> https://scholar4ai.com/ideas/esg-investing-under-political-backlash 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

ESG investing appears on 56 recent papers in the finance journals, but only 3 of 81 papers on responsible investing carry a backlash, political or polarization keyword. Several US states now restrict ESG in public funds, and some asset managers have stopped using the label. Did fund flows, fees and holdings change?

<h2>What prompted this</h2> <p>Since October 2025, 81 papers carry an ESG-investing, responsible-investing or sustainable-fund keyword (60 in finance journals); "ESG investing" alone appears on 56. Only 3 of the 81 also carry a backlash, anti-ESG, political, polarization or partisan keyword, although "political polarization" appears on 26 papers in the archive.</p> <h2>Questions</h2> <ul> <li>After a state restricted ESG considerations in its public funds, did the funds' holdings and returns change, or only the words in their reports?</li> <li>Did asset managers that dropped the ESG label change their voting and engagement, or only their marketing?</li> <li>Did investors in states with restrictions pull money from ESG funds, and did fees change?</li> </ul> <h2>Data that could help</h2> <p>Fund flows and holdings (Morningstar, CRSP Mutual Fund Database), state legislation dates, and fund voting records.</p> <h2>Next step</h2> <p>List the US state anti-ESG laws with their dates and compare flows into ESG-labelled funds held by investors in those states with flows elsewhere.</p> https://scholar4ai.com/ideas/employee-ownership 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Profit sharing and employee ownership appear on 24 recent papers spread thinly across the tracked journals. As founders of the baby-boom generation retire, employee buyouts and ownership trusts are one way to pass on firms. Do employee-owned firms survive longer, pay better, and invest differently?

<h2>What prompted this</h2> <p>Since October 2025, 24 papers carry an employee-ownership, profit-sharing, ESOP or worker-cooperative keyword, spread across economics (6), finance (6), operations research (4), accounting (3), management (3) and AI (2): "profit sharing" on 11, "employee ownership" on 10, "worker cooperatives" on 1.</p> <h2>Questions</h2> <ul> <li>When a retiring founder sells to employees rather than to a competitor or a fund, does the firm keep more jobs and plants in place over the next decade?</li> <li>Do employee-owned firms pay more, invest less, or take fewer risks, and do their workers hold too much of their wealth in one firm?</li> <li>Do the tax incentives for employee ownership trusts, as in the UK and proposed elsewhere, raise the number of such sales?</li> </ul> <h2>Data that could help</h2> <p>Ownership transition records, employee ownership registers, firm accounts (Orbis) and employment data.</p> <h2>Next step</h2> <p>Compare survival, employment and wages of UK firms sold to employee ownership trusts since 2014 with firms of similar size sold to trade buyers.</p> https://scholar4ai.com/ideas/electric-truck-routing-with-charging-queues 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Electric trucks are entering freight fleets, and routing them is a growing problem in operations research. But most models assume a charger is free when the truck arrives; only 5 of 66 recent papers on electric-vehicle routing and charging carry a queue, waiting or congestion keyword. How much do charging queues change routes, fleet size and costs?

<h2>What prompted this</h2> <p>Since October 2025, 66 papers carry an electric-vehicle-routing, electric-truck, charging-station or charging-scheduling keyword (42 in operations research journals); only 5 of them carry a queue, waiting-time or congestion keyword. Vehicle routing is one of the most frequent keywords of the year in the <a href="/categories/operations-research">operations research journals</a>.</p> <h2>Questions</h2> <ul> <li>When chargers are shared and queues form at busy times, how much longer and costlier do the best routes get?</li> <li>Is it cheaper to reserve charging slots, add battery capacity or build depot chargers?</li> <li>How do hourly electricity prices change when trucks should charge?</li> </ul> <h2>Possible approach</h2> <p>Add random queues at public chargers to a standard electric vehicle routing model with time windows, and solve it on benchmark instances with realistic numbers of chargers.</p> <h2>Next step</h2> <p>Extend an electric vehicle routing benchmark with charging queues drawn from observed station occupancy, and compare the plans made with and without queues.</p> https://scholar4ai.com/ideas/drone-inspection-of-infrastructure 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Industrial and infrastructure inspection appear on 139 recent papers, almost all in AI journals about detecting defects in images, and 25 of them involve drones or robots. Ageing bridges, pipelines and power lines need inspecting more often than inspectors can manage. How much do drone inspections cost and miss compared with people, and do they change maintenance decisions?

<h2>What prompted this</h2> <p>Since October 2025, 139 papers carry an inspection keyword (106 in AI journals, 12 in operations research, 11 in economics): "industrial inspection" on 36, "industrial visual inspection" on 11, "infrastructure inspection" on 6, "automated inspection" on 6. Twenty-five of them also carry a drone, UAV, robot, autonomous or computer-vision keyword; "UAV inspection" appears on 3 papers.</p> <h2>Questions</h2> <ul> <li>On the same structures, how do drone-based defect detections compare with human inspections in what they find and what they miss, and does the mix of misses matter for safety?</li> <li>Do agencies that adopt drone inspection inspect more often, and do they then repair earlier or defer repairs more?</li> <li>What does an inspection cost by drone and by crew once flight permissions, data review and false alarms are counted?</li> </ul> <h2>Data that could help</h2> <p>Inspection records with both methods from transport or utility agencies, maintenance decisions and costs, and defect ground truth from later repairs.</p> <h2>Next step</h2> <p>Compare drone and crew inspection findings for a set of bridges inspected both ways, and track which findings led to repairs.</p> https://scholar4ai.com/ideas/dollar-invoicing-and-de-dollarization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The US dollar, invoicing currency and de-dollarization appear on 67 recent papers in the finance and economics journals, and the yuan's use in trade settlement has risen since 2022. Yet only 23 of those papers connect the dollar to trade invoicing or the renminbi. Is de-dollarization happening in the currency of contracts, or only in the headlines?

<h2>What prompted this</h2> <p>Since October 2025, 67 papers carry a dollar, invoicing-currency, de-dollarization or currency-internationalization keyword (33 in finance journals, 32 in economics): "US dollar" on 21, "dollarization" on 6, "invoicing currency" on 5, "de-dollarization" on 4. Only 23 of the 67 also carry an invoicing, trade, pass-through, renminbi or reserve-currency keyword.</p> <h2>Questions</h2> <ul> <li>In countries that announced yuan settlement with China, what share of actual trade contracts switched currency, and did it change exchange rate pass-through to their import prices?</li> <li>Do exporters invoice in yuan when they can, or do they want dollars even for trade with China?</li> <li>Has the share of dollar invoicing in global trade changed measurably since 2022, outside of Russia?</li> </ul> <h2>Data that could help</h2> <p>Customs data with invoicing currency (available for the EU, some Asian countries and Russia), SWIFT currency shares, and central bank reserve composition (IMF COFER).</p> <h2>Next step</h2> <p>Compare invoicing currency shares in customs data for countries that signed yuan settlement agreements with those that did not, before and after the agreements.</p> https://scholar4ai.com/ideas/do-multi-criteria-rankings-agree 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Supplier selection is one of the fastest-growing keywords in the operations research journals, and much of it relies on multi-criteria methods such as AHP, TOPSIS and the best-worst method. These methods can rank the same options differently. How often would a paper's conclusion change with the method, and what should a robustness check look like?

<h2>What prompted this</h2> <p>In the operations research journals, "supplier selection" had 11 papers in July–September 2026 against about 4 expected (see the <a href="/categories/operations-research">operations research dashboard</a>). Since October 2025, 227 papers carry a multi-criteria keyword such as TOPSIS, AHP or the best-worst method (112 in AI journals, 74 in operations research). Only 15 of them carry a rank-reversal, sensitivity-analysis, robustness or comparative-analysis keyword. The formulas of these methods are on the <a href="/rankings/equations">equation ranking</a>'s list.</p> <h2>Questions</h2> <ul> <li>In published case studies, how often does the top choice change when the weighting method, the normalization or the aggregation changes?</li> <li>How sensitive are the rankings to small changes in expert judgments?</li> <li>Would reporting the share of method variants that agree on the winner be a useful standard?</li> </ul> <h2>Possible approach</h2> <p>Re-run the decision matrices reported in recent papers through a grid of method variants, and record how often the top choice and the full ranking agree.</p> <h2>Next step</h2> <p>Collect the decision matrices of 20 recent supplier-selection papers and re-rank each one with TOPSIS, VIKOR and the best-worst method.</p> https://scholar4ai.com/ideas/do-digital-twins-pay 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Digital twins appear on 87 recent papers, 64 of them in AI journals and 22 in operations research, almost all about building them. Digital twin adoption appears on 2 papers. Firms are buying twins of plants, ports and cities. Which of them changed a decision, and what did they cost relative to what they saved?

<h2>What prompted this</h2> <p>Since October 2025, 95 papers carry a digital-twin keyword (64 in AI journals, 22 in operations research, 7 in management): "digital twin" on 87, "human digital twin" on 3, "digital twin adoption" on 2. The quarter to September 2026 had 22 such papers against about 24 per quarter earlier. Almost all are about methods; the business case appears in almost none.</p> <h2>Questions</h2> <ul> <li>In firms that built digital twins of plants or logistics networks, which decisions were made differently because of the twin, and what were the savings?</li> <li>How much of the cost is data integration rather than modelling, and do the twins stay in use once the project team leaves?</li> <li>Do twins of public infrastructure (ports, cities) deliver benefits that justify public funding?</li> </ul> <h2>Data that could help</h2> <p>Case records from firms and public agencies with costs and decisions, vendor contract data, and surveys of plant managers.</p> <h2>Next step</h2> <p>Interview or survey managers at 30 firms that announced digital twin projects since 2021 about the decisions the twin changed and the project's cost and status.</p> https://scholar4ai.com/ideas/do-ai-explanations-agree 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Explainable AI is one of the largest topics in the AI journals and is spreading into management, finance and economics. Explanation fidelity (whether an explanation reflects what the model really does) is a fast-rising but still small keyword. When SHAP, LIME and attention maps disagree about the same prediction, which should a loan officer or a doctor believe?

<h2>What prompted this</h2> <p>Since October 2025, "explainable AI" appears on 419 papers (341 in AI journals, 28 in management, 20 in operations research, 15 each in finance and economics). "Explanation fidelity" appears on 11 papers and is rising in the AI journals (6 papers in July–September 2026 against about 1 expected).</p> <h2>Questions</h2> <ul> <li>How often do common explanation methods disagree about which features drive the same prediction?</li> <li>Which of them stay faithful when the model is changed in ways that should change the explanation?</li> <li>Do people decide better with explanations that are more faithful, or just with ones that look more confident?</li> </ul> <h2>Possible approach</h2> <p>Compare SHAP, LIME, integrated gradients and attention on standard tabular and text tasks, using tests that check fidelity by perturbing the model or its inputs.</p> <h2>Next step</h2> <p>Measure how often SHAP and LIME agree on the top features for one credit dataset and one medical dataset, across several model types.</p> https://scholar4ai.com/ideas/disclosure-for-machine-readers 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Information transparency and corporate disclosure are rising in the accounting journals, and language models now read filings for investors. Yet few disclosure papers mention language models or XBRL. Is disclosure written for machines changing what transparency means?

<h2>What prompted this</h2> <p>On the <a href="/categories/accounting">accounting dashboard</a>, "information transparency" (5 papers in July–September 2026 against about 1 expected) and "corporate disclosure" (7 against 3) are among the fastest-growing keywords. Since October 2025, 1,136 papers carry a disclosure or transparency keyword; only 43 of them also carry a language-model, generative-AI or XBRL keyword, and "XBRL" itself appears on 2.</p> <h2>Questions</h2> <ul> <li>Are filings that machines parse easily (structured tags, clean tables) priced faster after release?</li> <li>Earlier text-analysis work found that firms avoid the words that machine readers count as negative. Do language-model readers see through that?</li> <li>Does machine readability help small investors, or mostly the fastest traders?</li> </ul> <h2>Data that could help</h2> <ul> <li>SEC EDGAR filings with acceptance timestamps, inline XBRL data and intraday prices.</li> <li>The SEC's phase-in of inline XBRL (2019–2021) as a staggered change in machine readability.</li> </ul> <h2>Next step</h2> <p>Measure how fast prices react to 10-K releases before and after each group of filers moved to inline XBRL.</p> https://scholar4ai.com/ideas/digital-servitization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Servitization and digital servitization appear on 50 recent papers, mostly in management, with 15 in the last quarter against about 12 per quarter before. Manufacturers now sell uptime, outcomes and subscriptions instead of machines. Do the service contracts earn more than the products did, and what happens to the data the machines generate?

<h2>What prompted this</h2> <p>Since October 2025, 50 papers carry a servitization, product-service-system or as-a-service keyword (30 in management journals): "servitization" on 28, "digital servitization" on 10, "product-service systems" on 4. The quarter to September 2026 had 15 such papers against about 12 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Do manufacturers that shift to outcome-based contracts earn higher margins over the equipment's life, or do they take on risks they cannot price?</li> <li>Who owns and can use the operating data from connected equipment, and do customers pay for services with their data?</li> <li>Do service revenues make manufacturers' earnings steadier through downturns, as the strategy promises?</li> </ul> <h2>Data that could help</h2> <p>Segment reporting of manufacturers that disclose service revenues, contract terms from case firms, and equipment telemetry agreements.</p> <h2>Next step</h2> <p>Compare the margin and revenue volatility of manufacturers with high and low service revenue shares through the 2020 and 2022–2023 downturns.</p> https://scholar4ai.com/ideas/digital-public-infrastructure 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

E-government appears on 28 recent papers and digital identity on 4, while India's identity, payment and data-sharing systems have become a model that dozens of countries are copying. What have the systems changed for welfare delivery, tax collection and the firms built on top of them, and who is left out?

<h2>What prompted this</h2> <p>Since October 2025, 66 papers carry an e-government, digital-identity, digital-government or digital-ID keyword (23 in AI journals, 19 in finance, 13 in economics): "e-government" on 28, "digital identity" on 4. Most study adoption and satisfaction with online services rather than the economic effects of the underlying systems.</p> <h2>Questions</h2> <ul> <li>Did biometric identity and direct benefit transfers cut leakage in welfare programmes without excluding people who fail authentication?</li> <li>Did instant payment systems raise formal transactions and tax revenue, and which firms grew on top of them?</li> <li>When a country copies the model, which components matter, and what happens when the identity system goes down?</li> </ul> <h2>Data that could help</h2> <p>Programme administrative data (India's direct benefit transfer and UPI statistics), household surveys on exclusion, firm registrations and tax revenue.</p> <h2>Next step</h2> <p>Compare welfare payment leakage and exclusion rates before and after biometric authentication in Indian states that adopted it at different times.</p> https://scholar4ai.com/ideas/digital-land-registries 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Property rights appear on 42 recent papers and land tenure and land rights on 13, mostly in economics. Countries are digitizing land registries and using satellite imagery to map parcels. Does a digital title change credit, investment and disputes more than a paper one, and who loses when the map is wrong?

<h2>What prompted this</h2> <p>Since October 2025, 71 papers carry a land-titling, land-tenure, land-right, land-registration or property-rights keyword (56 in economics journals): "property rights" on 42, "land rights" on 7, "land tenure" on 6, "land registration" on 2. Few of them concern the digital registries and satellite-based mapping programmes now under way.</p> <h2>Questions</h2> <ul> <li>After a registry is digitized and parcels are mapped from imagery, do land transactions, mortgage lending and farm investment rise?</li> <li>Do disputes fall, or do mapping errors create new ones, especially for women and customary holders?</li> <li>Does digitization make land easier to tax, and do governments use it?</li> </ul> <h2>Data that could help</h2> <p>Registry transaction records before and after digitization, mortgage data, court records on land disputes and household surveys.</p> <h2>Next step</h2> <p>Use the district-by-district roll-out of a digital land registry in one country to compare land transactions and credit before and after.</p> https://scholar4ai.com/ideas/detecting-earnings-management-beyond-accruals 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Earnings management is a mainstay of the finance and accounting journals, still measured mostly with accrual regressions such as the modified Jones model. Only 4 of 156 recent papers on it carry a machine-learning or text keyword. Do learned models, including language models reading the filings, flag manipulation that accrual models miss?

<h2>What prompted this</h2> <p>Since October 2025, 156 papers carry an earnings-management or accrual keyword (88 in finance journals, 47 in accounting). Only 4 of them also carry a machine-learning, deep-learning, language-model, text or neural-network keyword. Accrual models such as the modified Jones model, on the <a href="/rankings/equations">equation ranking</a>'s list, remain the standard measure.</p> <h2>Questions</h2> <ul> <li>Against known cases (restatements, enforcement actions), do learned models built on raw financial statements and filing text beat accrual residuals?</li> <li>Do the measures disagree on which firms manage earnings, and do published results depend on the measure used?</li> <li>Can a language model explain why it flags a firm, in terms an auditor could check?</li> </ul> <h2>Data that could help</h2> <p>Restatements and enforcement actions (Audit Analytics, SEC releases), financial statements (Compustat or CSMAR) and filing text (SEC EDGAR).</p> <h2>Next step</h2> <p>Train a gradient-boosting model on raw accounting items to predict restatements, and compare its ranking of firms with modified-Jones residuals.</p> https://scholar4ai.com/ideas/depopulation-and-shrinking-regions-in-japan 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Population decline, shrinking cities and the aging society appear on 26 recent papers, but only 4 of them also concern Japan, the country furthest into the decline. Which local policies (consolidation of services, migration incentives, remote work) keep regions viable, and what does orderly decline cost compared with resisting it?

<h2>What prompted this</h2> <p>Since October 2025, 26 papers carry a depopulation, shrinking-city, population-decline, rural-decline or aging-society keyword (21 in economics journals): "population decline" on 10, "aging society" on 9, "shrinking cities" on 2. Only 4 of them also carry a Japan keyword, although "Japan" appears on 81 papers in the archive.</p> <h2>Questions</h2> <ul> <li>Do municipal mergers and the concentration of services in "compact city" cores lower costs per resident without accelerating out-migration?</li> <li>Did the hometown tax donation scheme and relocation subsidies bring people and money to shrinking towns, or shift them between towns?</li> <li>Which towns stabilized after remote work spread, and what did they have in common?</li> </ul> <h2>Data that could help</h2> <p>Municipal population, finance and service data (Japan's census and municipal finance statistics), the dates of mergers and compact city plans, and relocation subsidy records.</p> <h2>Next step</h2> <p>Compare per-resident public spending and net migration in Japanese municipalities that adopted compact city plans with similar municipalities that did not.</p> https://scholar4ai.com/ideas/demand-forecasting-with-llms 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Demand and sales forecasting appear on 66 recent papers in the operations research and AI journals, but only 9 of them carry a language-model, foundation-model or text keyword. Retailers now feed product descriptions, reviews, weather and news into forecasting systems. Does text improve forecasts for new products, where history is missing, and does it help where history is long?

<h2>What prompted this</h2> <p>Since October 2025, 66 papers carry a demand-forecasting, sales-forecasting or demand-prediction keyword (31 in operations research journals, 22 in AI): "demand forecasting" on 42, "sales forecasting" on 7, "retail demand forecasting" on 3. Only 9 of them also carry a language-model, foundation-model, transformer, deep-learning or text keyword.</p> <h2>Questions</h2> <ul> <li>For new products with no sales history, do embeddings of descriptions and images predict first-season demand better than analogue-based methods, and by how much?</li> <li>Does adding text (reviews, news, promotions) improve forecasts for established products, or add noise?</li> <li>Do the forecast gains translate into lower inventory and fewer stockouts once ordering rules use them?</li> </ul> <h2>Data that could help</h2> <p>Retail sales data with product descriptions and images (public datasets such as the M5 and fashion retail datasets), and inventory records from a retailer.</p> <h2>Next step</h2> <p>Compare new-product forecast errors from an embedding-based model and an analogue method on a fashion retail dataset with product images and text.</p> https://scholar4ai.com/ideas/delegating-to-ai-and-dishonesty 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Unethical behavior and moral disengagement are rising in the management journals, and AI tools now act on people's behalf. Recent experiments suggest people cheat more when a machine does the act for them. Does that happen at work, with sales targets, expense claims or performance reports?

<h2>What prompted this</h2> <p>In management, "unethical behavior" had 6 papers in July–September 2026 against about 2 expected, and "moral disengagement" 6 against 3 (see the <a href="/categories/management">management dashboard</a>). Since October 2025, 56 papers carry an unethical-behavior, dishonesty, moral-disengagement or cheating keyword (35 in management journals); only 6 of them carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>When employees can hand a task to an AI agent with a vague goal ("hit the target"), do they choose goals that lead to dishonest outcomes more often than when they act themselves?</li> <li>Does the effect shrink when the AI asks for explicit confirmation or keeps logs?</li> <li>Whom do peers and managers blame when an AI agent misreports?</li> </ul> <h2>Possible approach</h2> <p>Workplace experiments with AI assistants on incentivized tasks, and field data on reporting errors before and after AI tools were introduced.</p> <h2>Next step</h2> <p>Design an online experiment in which participants delegate a self-reported performance task to an AI assistant, with goals of varying vagueness.</p> https://scholar4ai.com/ideas/dei-under-backlash 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Diversity, equity and inclusion, inclusive leadership and affirmative action appear on 44 recent papers, with 17 in the last quarter against about 8 per quarter before, but only 5 carry a political, backlash or legal keyword. Firms in the US have cut or renamed diversity programmes after court rulings and political pressure. What changed in hiring and pay, and what only changed in name?

<h2>What prompted this</h2> <p>Since October 2025, 44 papers carry a diversity-and-inclusion, diversity-management, workplace-diversity, inclusive-leadership or affirmative-action keyword (23 in management journals): "diversity, equity, and inclusion" on 12, "inclusive leadership" on 8, "diversity and inclusion" on 8, "workplace diversity" on 5, "affirmative action" on 5, "diversity management" on 4. The quarter to September 2026 had 17 such papers against about 8 per quarter earlier. Only 5 of the 44 also carry a backlash, political, legal, rollback or ideology keyword.</p> <h2>Questions</h2> <ul> <li>At firms that announced cuts to diversity programmes, did the demographic mix of new hires and promotions change within a year, or did only the language in reports change?</li> <li>Did firms that kept programmes under new names lose or gain customers and applicants?</li> <li>Did the end of race-conscious college admissions change the pipeline that corporate programmes draw on?</li> </ul> <h2>Data that could help</h2> <p>Workforce demographic filings (EEO-1 reports), job postings, firms' diversity reports before and after 2024, and applicant data from hiring platforms.</p> <h2>Next step</h2> <p>Compare the demographics of new hires in EEO-1 filings for firms that publicly cut diversity programmes in 2024–2025 with matched firms that did not.</p> https://scholar4ai.com/ideas/defi-liquidation-cascades 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Decentralized finance is a growing topic in the finance journals, but only 11 of 54 recent papers on it carry a liquidation, leverage, collateral or stablecoin keyword. Lending protocols liquidate collateral automatically when prices fall. Do these liquidations feed on themselves, and could they reach traditional markets?

<h2>What prompted this</h2> <p>Since October 2025, "decentralized finance" appears on 54 papers (33 in finance journals), with 15 in July–September 2026 against about 12 per quarter over the nine months before. Only 11 of the 54 also carry a liquidation, leverage, collateral or stablecoin keyword. Financial contagion, a fast-rising topic on the <a href="/trends">trends page</a>, is studied almost entirely in traditional markets.</p> <h2>Questions</h2> <ul> <li>When a token's price falls, how much of the next fall is caused by forced liquidations on lending protocols, and how fast do they propagate across protocols?</li> <li>Do liquidation bots and oracle delays make cascades worse at times when exchanges are congested?</li> <li>As tokenized Treasury bills and stablecoins become collateral, could a cascade spill into short-term funding markets?</li> </ul> <h2>Data that could help</h2> <p>On-chain records of loans, collateral and liquidations from the main lending protocols, matched to exchange prices at high frequency.</p> <h2>Next step</h2> <p>Rebuild the liquidation chain of one major crypto sell-off from on-chain data and measure how much of the price drop happened during liquidation waves.</p> https://scholar4ai.com/ideas/deepfake-detection-in-the-wild 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Deepfake detection appears on 45 recent papers and AI-generated image detection on 13, almost all in AI journals, with 23 papers in the last quarter against about 21 per quarter before. Detectors score well on the datasets they were built with and fail on the next generator. How well do they work on real fraud, elections and markets, and who should run them?

<h2>What prompted this</h2> <p>Since October 2025, 86 papers carry a deepfake, face-forgery, synthetic-media or AI-generated-image keyword (78 in AI journals, 7 in management): "deepfake detection" on 45, "AI-generated image detection" on 13, "deepfakes" on 13, "synthetic media" on 7, "face forgery detection" on 6. The quarter to September 2026 had 23 such papers against about 21 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>On deepfakes collected from actual fraud cases, political campaigns and market rumours, rather than from research datasets, what is the detection rate of published detectors?</li> <li>How quickly does accuracy fall after a new generator is released, and can detectors be updated faster than generators?</li> <li>Do labels from platform detectors change how people react to a video, or do people believe what they expected to believe?</li> </ul> <h2>Data that could help</h2> <p>Verified deepfake collections from fact-checkers and fraud investigations, detector code from published papers, and platform labelling experiments.</p> <h2>Next step</h2> <p>Assemble 500 deepfakes verified by fact-checking organizations and measure the accuracy of five published detectors on them.</p> https://scholar4ai.com/ideas/deep-rl-for-portfolios-reproducibility 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Reinforcement learning appears on more than 700 recent papers, and 41 of them apply it to portfolios, trading or hedging, almost all in AI and operations research journals. The results are rarely reproduced and rarely compared with simple rules after costs. Which of the reported gains survive replication on new data?

<h2>What prompted this</h2> <p>Since October 2025, 738 papers carry a reinforcement-learning keyword (577 in AI journals, 120 in operations research): "reinforcement learning" on 255, "deep reinforcement learning" on 191, "multi-agent reinforcement learning" on 97. Forty-one of them also carry a portfolio, trading, stock, asset-allocation, hedging, market-making, finance or crypto keyword; "portfolio optimization" appears on 153 papers in the archive and "portfolio management" on 18.</p> <h2>Questions</h2> <ul> <li>When the published agents are retrained on the same data with different random seeds and then run on the two years after each paper's sample, how often do they beat equal weights and mean-variance after transaction costs?</li> <li>How much of the reported performance comes from look-ahead in data preparation and from tuning on the test period?</li> <li>Does the hedging use case, with a clearer objective, replicate better than the return-maximizing one?</li> </ul> <h2>Data that could help</h2> <p>The code and data of recent papers, daily prices for the post-sample period, and realistic transaction cost models.</p> <h2>Next step</h2> <p>Replicate ten published deep reinforcement learning portfolio papers with multiple seeds and evaluate them on the two years after each sample, after costs.</p> https://scholar4ai.com/ideas/data-leakage-in-benchmarks 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Data leakage appears on 16 recent papers and memorization on 3, almost all in AI journals, while thousands of papers report results on public benchmarks that language models may have seen in training. How much of reported progress is contamination, and can finance and economics benchmarks be protected?

<h2>What prompted this</h2> <p>Since October 2025, 29 papers carry a data-leakage, memorization, contamination or benchmark-leakage keyword (23 in AI journals): "data leakage" on 16, "memorization" on 3, "training data leakage" on 2. Benchmark results, by contrast, appear in a large share of the 17,000 AI-journal papers in the archive, and in the finance papers that test language models on public financial question sets.</p> <h2>Questions</h2> <ul> <li>For the benchmarks most used in the AI journals, how much do scores fall on freshly written items of the same kind, and does the fall differ between open and closed models?</li> <li>Have the financial and economic test sets used to evaluate language models (exam questions, sentiment sets) been published online before the models' training cutoffs?</li> <li>Can rolling benchmarks built from new filings, news and data releases give uncontaminated tests, and would journals require them?</li> </ul> <h2>Data that could help</h2> <p>The release dates of benchmark items and of model training cutoffs, fresh item sets written after the cutoffs, and the results tables of recent papers.</p> <h2>Next step</h2> <p>Rewrite 200 items from a widely used financial question benchmark with the same structure and new numbers, and compare model scores on the old and new items.</p> https://scholar4ai.com/ideas/data-centres-and-the-grid 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Data centres appear on just 16 recent papers across the tracked journals, though the count is rising, and electricity demand on 17. AI training and inference are driving the first sustained growth in electricity demand in rich countries for decades. Who pays for the new capacity, and do data centres crowd out electrification elsewhere?

<h2>What prompted this</h2> <p>Since October 2025, 16 papers carry a data-centre keyword (7 in AI journals, 4 in finance, 4 in economics), with 7 in July–September 2026 against about 3 per quarter earlier. "Electricity demand" appears on 17 papers. Most data-centre papers in the AI journals are about cooling and scheduling, not about the grid.</p> <h2>Questions</h2> <ul> <li>In regions with data-centre booms (Virginia, Ireland, Texas), how much have wholesale and retail electricity prices risen, and how much of the new capacity cost falls on households?</li> <li>Do data centres delay the connection of other new loads, such as factories and heat pumps, and the retirement of fossil plants?</li> <li>Do power purchase agreements with data centres bring new clean capacity, or claim capacity that would have been built anyway?</li> </ul> <h2>Data that could help</h2> <p>Grid connection queues, data-centre capacity by region, electricity prices and generation mix (ENTSO-E, US grid operators), and power purchase agreement records.</p> <h2>Next step</h2> <p>Relate changes in regional retail electricity prices since 2020 to data-centre capacity additions, controlling for fuel prices and other demand.</p> https://scholar4ai.com/ideas/cybersecurity-incident-disclosure 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Cybersecurity appears on 159 recent papers, most of them in AI journals about detection; only 22 carry a disclosure, reporting or investor keyword, and "cybersecurity disclosure" on 6. Since 2023, US firms must report material cyber incidents within four business days. What do they report, and what do markets learn?

<h2>What prompted this</h2> <p>Since October 2025, 159 papers carry a cybersecurity, cyber-risk, data-breach or cyberattack keyword (58 in AI journals, 47 in finance, 26 in management, 12 in accounting): "cybersecurity risks" on 39, "data breaches" on 10, "cybersecurity disclosure" on 6. Only 22 of the 159 also carry a disclosure, SEC, reporting, internal-control, investor or firm-value keyword.</p> <h2>Questions</h2> <ul> <li>Since the four-day rule, how many incidents are disclosed as material, how quickly, and what do firms say about the financial effect?</li> <li>Do stock prices react more to the first disclosure or to later updates, and do firms that disclose faster lose less value?</li> <li>Has the rule changed cyber insurance pricing or firms' investment in security?</li> </ul> <h2>Data that could help</h2> <p>Item 1.05 disclosures from 8-K filings since December 2023, breach notification records held by state attorneys general, stock returns and cyber insurance rates.</p> <h2>Next step</h2> <p>Collect all 8-K cyber incident disclosures since the rule took effect, code their timing and content, and measure the stock reactions.</p> https://scholar4ai.com/ideas/csrd-first-reports 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The Corporate Sustainability Reporting Directive appears on 9 recent papers and double materiality on 3, while sustainability reporting as a whole appears on 57. The first reports under the directive were published in 2025, and the rules were then scaled back. What did the first wave disclose, and what did the rollback remove?

<h2>What prompted this</h2> <p>Since October 2025, 12 papers carry a CSRD, European Sustainability Reporting Standards or double-materiality keyword (4 in finance journals, 4 in management, 3 in economics, 1 in accounting): "Corporate Sustainability Reporting Directive" on 9 and "double materiality" on 3. "Sustainability reporting" appears on 57 papers.</p> <h2>Questions</h2> <ul> <li>In the first reports, which topics did firms declare material, and does the double-materiality assessment differ across countries and auditors?</li> <li>Did firms that reported for the first time change their emissions targets or capital spending, or only their disclosure?</li> <li>After the 2025 simplification removed most mid-sized firms from the scope, did those firms stop reporting, or keep reporting voluntarily because their customers ask?</li> </ul> <h2>Data that could help</h2> <p>The first-wave reports (machine-readable from 2025), materiality assessments, firms' earlier voluntary reports, and the scope changes by firm size.</p> <h2>Next step</h2> <p>Code the material topics declared by 300 first-wave reporters and compare them across countries and sectors.</p> https://scholar4ai.com/ideas/crowdshipping 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Crowdshipping and crowdsourced delivery appear on 10 recent papers in the operations research journals, with 4 in the last quarter against about 2 per quarter before. The models route parcels through ordinary travellers who happen to be going the same way. Do enough people accept such tasks, and at what pay, for the routes to work?

<h2>What prompted this</h2> <p>Since October 2025, 10 papers carry a crowdshipping, crowdsourced-delivery or crowd-logistics keyword (9 in operations research journals): "crowdshipping" on 6, "crowdsourced delivery" on 4. Four of them were published in July–September 2026 against about 2 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In the platforms that actually run crowdshipping, what share of offered tasks are accepted, by whom, and at what compensation relative to professional couriers?</li> <li>Do the routing models' assumptions about the supply of occasional drivers match observed acceptance rates, and how much of the modelled savings survive when they do not?</li> <li>Does crowdshipping reduce vehicle kilometres, or add detours by drivers who were not going there anyway?</li> </ul> <h2>Data that could help</h2> <p>Task offer and acceptance logs from a crowdshipping platform, driver surveys, and GPS traces of accepted deliveries.</p> <h2>Next step</h2> <p>Obtain acceptance and detour data from one crowdshipping platform and recompute a published model's savings with the observed acceptance rates.</p> https://scholar4ai.com/ideas/critical-mineral-export-controls 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Critical minerals rose from almost nothing to one of the fastest-growing keywords across the tracked journals this year. Export controls on gallium, germanium, graphite and some rare earths since 2023 are natural experiments. Do they raise costs for clean-energy producers, speed up substitution and new supply, or both?

<h2>What prompted this</h2> <p>Across all journals, "critical minerals" had 9 papers in July–September 2026 against about 1 expected from its earlier share (see the <a href="/trends">trends page</a>). Since October 2025 it appears on 14 papers, "rare earth elements" on 12 and "lithium" on 11, most of them in economics and finance journals.</p> <h2>Questions</h2> <ul> <li>After an export control is announced, how do prices outside the controlling country, and the share prices of the firms that use the mineral, respond?</li> <li>Do controls speed up substitution (new materials and chemistries) and new mines elsewhere, and how fast?</li> <li>Which clean-energy supply chains are most exposed, by product and country?</li> </ul> <h2>Data that could help</h2> <p>Export-control announcement dates, trade flows by product (UN Comtrade, China Customs), mineral prices, firms' supplier links, and patents on substitute materials.</p> <h2>Next step</h2> <p>Build a list of critical-mineral export controls since 2023 and run an event study of the share prices of firms that use the minerals.</p> https://scholar4ai.com/ideas/counterfeits-and-luxury-resale 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Luxury brands, luxury consumption and counterfeit goods appear on 32 recent papers, mostly in management, with 5 in the last quarter against about 8 per quarter before. High-quality counterfeits and a booming resale market have blurred what authenticity means. Do counterfeits erode brand value, and does authentication technology restore it?

<h2>What prompted this</h2> <p>Since October 2025, 32 papers carry a counterfeit or luxury keyword (20 in management journals): "luxury brands" on 8, "counterfeit goods" on 4, "luxury consumption" on 3, "counterfeiting" on 2. Thirteen of them also carry a counterfeit, authentication, fake, resale or secondhand keyword. Five were published in July–September 2026 against about 8 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>Do consumers who buy high-quality counterfeits later buy the genuine product, or do they stop valuing the brand?</li> <li>Does authentication on resale platforms raise prices for genuine secondhand goods and reduce demand for new ones?</li> <li>Do brands' own resale programmes capture value from the secondhand market, or cannibalize new sales?</li> </ul> <h2>Data that could help</h2> <p>Resale platform listings and prices with authentication status, customs seizure records, consumer panel data and brand sales.</p> <h2>Next step</h2> <p>Compare resale prices of the same handbag models on platforms with and without authentication guarantees, and track them around counterfeit scandals.</p> https://scholar4ai.com/ideas/corporate-venture-capital-in-the-ai-cycle 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Corporate venture capital is a small but rising keyword, with 7 papers in July–September 2026 against about 3 per quarter before. Large technology firms now invest in AI startups that also buy their cloud and chips. Is this venture investing, customer financing, or a way to buy talent without a merger?

<h2>What prompted this</h2> <p>Since October 2025, "corporate venture capital" appears on 16 papers (8 in finance journals, 5 in economics, 3 in management), 7 of them in July–September 2026 against about 3 per quarter over the nine months before. "Open innovation" appears on 42 papers. Venture capital itself is one of the fastest-growing keywords in the finance journals (<a href="/trends">trends page</a>).</p> <h2>Questions</h2> <ul> <li>How much of corporate investment in AI startups comes back to the investor as cloud and chip revenue, and does that circular flow inflate valuations?</li> <li>Do startups backed by a corporate investor get acquired by it later at a discount, or by its rivals at a premium?</li> <li>Do "acqui-hire" deals, in which the team is hired and the company licensed, escape merger review that a purchase would face?</li> </ul> <h2>Data that could help</h2> <p>Deal data (PitchBook, Crunchbase) with investor identities, startups' cloud providers from job postings and technical documentation, and later exits.</p> <h2>Next step</h2> <p>List AI startups with corporate investors since 2022 and check what share of their reported spending goes to the investor's products.</p> https://scholar4ai.com/ideas/corporate-purpose-statements 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Stakeholder theory appears on 88 recent papers in the management and finance journals, but corporate purpose on 4 and stakeholder capitalism on 2. Hundreds of large firms signed statements putting stakeholders before shareholders in 2019. Several years on, did the signatories treat workers, suppliers and communities differently?

<h2>What prompted this</h2> <p>Since October 2025, 101 papers carry a stakeholder-theory, stakeholder-orientation, corporate-purpose or stakeholder-capitalism keyword (55 in management journals, 29 in finance): "stakeholder theory" on 88, "stakeholder orientation" on 6, "corporate purpose" on 4, "stakeholder capitalism" on 2. The theory is well cited; the statements that invoked it are rarely tested.</p> <h2>Questions</h2> <ul> <li>Did signatory firms differ from non-signatories in layoffs, supplier payment terms, environmental fines or executive pay in the years after 2019?</li> <li>Did they use the purpose language in shareholder disputes, and did courts or investors take it into account?</li> <li>Did the statements change which employees and customers chose the firm?</li> </ul> <h2>Data that could help</h2> <p>The list of signatories, firm-level employment, supplier payment data, enforcement actions and executive pay (Compustat, ExecuComp).</p> <h2>Next step</h2> <p>Compare layoff announcements and supplier payment days of Business Roundtable signatories with matched non-signatories for 2017–2025.</p> https://scholar4ai.com/ideas/corporate-misconduct-contagion 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Corporate misconduct, reputational risk and corporate scandals appear on 90 recent papers across the management, finance and accounting journals, mostly one firm at a time. Firms share directors, auditors, suppliers and bankers. Does misconduct travel along those links, and can network position predict who is next?

<h2>What prompted this</h2> <p>Since October 2025, 90 papers carry a scandal, corporate-misconduct, corporate-social-irresponsibility, reputation-damage, crisis-communication or misconduct keyword (31 in management journals, 30 in finance, 11 in economics, 10 in accounting): "corporate misconduct" on 22, "reputational risk" on 21, "corporate social irresponsibility" on 9, "crisis communication" on 8, "corporate scandals" on 5.</p> <h2>Questions</h2> <ul> <li>Are firms that share a director or auditor with a firm caught in misconduct more likely to be caught later, and does the link run through practices or through enforcement attention?</li> <li>When a scandal hits one firm, do customers and investors punish its peers, and do the peers change behaviour?</li> <li>Can network measures improve the prediction of enforcement actions beyond firm characteristics?</li> </ul> <h2>Data that could help</h2> <p>Enforcement actions and restatements (Audit Analytics, SEC releases), director networks (BoardEx), auditor identities and supply chain links.</p> <h2>Next step</h2> <p>Build the director network of listed firms and test whether a board link to a sanctioned firm predicts later sanctions, controlling for industry and size.</p> https://scholar4ai.com/ideas/corporate-lobbying-on-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Lobbying, corporate political activity and campaign contributions appear on 49 recent papers, mostly about tax, trade and environmental policy, while AI regulation appears on 53 papers as a question of design. The EU AI Act and dozens of US state bills attracted record lobbying by technology firms. What did the money change in the texts?

<h2>What prompted this</h2> <p>Since October 2025, 49 papers carry a lobbying, corporate-political-activity or campaign-contribution keyword (23 in economics journals, 11 in management, 8 in accounting, 7 in finance): "lobbying" on 20, "corporate political activity" on 14, "campaign contributions" on 7, "corporate lobbying" on 6. "AI regulation" appears on 53 papers and "EU AI Act" on 25, in a separate literature on how the rules should look.</p> <h2>Questions</h2> <ul> <li>Which provisions of the EU AI Act changed between the Commission's draft and the final text, and do the changes line up with the positions filed by the largest lobbyists?</li> <li>In US states, do bills with the most technology-industry lobbying die, get delayed or get amended in predictable ways?</li> <li>Do firms that lobby on AI also publish "responsible AI" principles, and do the two say the same thing?</li> </ul> <h2>Data that could help</h2> <p>Lobbying registers (the EU Transparency Register, US federal and state filings), legislative texts across drafts, and firms' public policy statements.</p> <h2>Next step</h2> <p>Code the changes between the drafts and the final EU AI Act article by article and match them to the positions in lobby meeting records.</p> https://scholar4ai.com/ideas/corporate-bond-etfs-and-liquidity 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Corporate bonds appear on 54 recent papers in the finance and economics journals, but only 7 of them carry an ETF, fund or liquidity keyword. Exchange-traded funds now hold a large share of corporate bonds and trade far more than the bonds themselves. Do they add liquidity, or hide the lack of it until a stress?

<h2>What prompted this</h2> <p>Since October 2025, 54 papers carry a corporate-bond keyword (41 in finance journals); only 7 of them also carry an ETF, exchange-traded, liquidity or fund keyword. "Exchange-traded funds" appears on 29 papers, mostly about equity funds.</p> <h2>Questions</h2> <ul> <li>When bond ETFs trade at large discounts to their holdings, as in March 2020, do the discounts reflect the true price of the bonds or a dislocation in the ETF itself?</li> <li>Do bonds held heavily by ETFs become more liquid in normal times and less liquid in stress?</li> <li>Do the authorized participants that create and redeem ETF shares step back exactly when they are needed?</li> </ul> <h2>Data that could help</h2> <p>Bond transactions (TRACE), ETF holdings, prices and creation and redemption data, and dealer inventories.</p> <h2>Next step</h2> <p>Measure the price impact of trades in bonds with high and low ETF ownership across calm and stress periods using TRACE data.</p> https://scholar4ai.com/ideas/copyright-and-creators-in-the-ai-era 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Copyright protection, data ownership, content creators and the creative industries appear on 96 recent papers across the AI, management and economics journals, and 37 of them involve generative AI. Courts are deciding whether training on copyrighted work is fair use, and licensing deals are being signed. What are creators losing or gaining, and what would a licensing market look like?

<h2>What prompted this</h2> <p>Since October 2025, 96 papers carry a copyright, data-licensing, fair-use, data-ownership, creative-industries or content-creator keyword (49 in AI journals, 25 in management, 10 in economics): "copyright protection" on 19, "data ownership" on 11, "copyright" on 10, "cultural and creative industries" on 10, "creative industries" on 9, "content creators" on 8. Thirty-seven of the 96 also carry a generative-AI, language-model, diffusion-model or AI-generated keyword.</p> <h2>Questions</h2> <ul> <li>Since image and text generators became widespread, how have earnings of illustrators, stock photographers, translators and voice actors changed, and which work was lost first?</li> <li>What are the terms of the licensing deals between AI firms and publishers, and how is the money shared with the authors whose work is licensed?</li> <li>Would a collective licensing scheme, like those for music, be workable for training data, and what price would it set?</li> </ul> <h2>Data that could help</h2> <p>Freelance platform earnings by occupation, licensing deal announcements and terms, court filings, and royalty data from collecting societies.</p> <h2>Next step</h2> <p>Track earnings and job volumes of illustrators, translators and voice actors on freelance platforms from 2021 to 2025 relative to occupations less exposed to generators.</p> https://scholar4ai.com/ideas/coopetition-in-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Coopetition, cooperation between competitors, appears on 31 recent papers split between the management and operations research journals, with 7 in the last quarter, in line with earlier quarters. AI firms compete fiercely while sharing open models, benchmarks and safety research, and renting computing from each other. When does sharing help a rival more than oneself?

<h2>What prompted this</h2> <p>Since October 2025, "coopetition" appears on 31 papers, 14 in management journals and 14 in operations research, mostly on supply chains and platform ecosystems. Open-source models, shared benchmarks and cloud partnerships among AI firms are among the clearest current cases, and none of the 31 papers names them.</p> <h2>Questions</h2> <ul> <li>Do firms release open models to set standards, to recruit developers or to commoditize a rival's product, and do the stated and revealed motives match?</li> <li>When a firm both sells computing to a rival and competes with it in models, how does it price, and does the rival hedge by building its own?</li> <li>Does participation in shared safety and benchmark efforts correlate with market position, and does it change when a firm leads?</li> </ul> <h2>Data that could help</h2> <p>Model release histories and licences, cloud partnership announcements and contract terms, and benchmark and safety consortium memberships.</p> <h2>Next step</h2> <p>Build a timeline of open model releases and cloud partnerships among the main AI firms and relate release decisions to each firm's position in the market at the time.</p> https://scholar4ai.com/ideas/contracts-that-cut-scope-3-emissions 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Supply chain coordination is growing in the operations research journals, along with emissions trading and carbon pricing. For most firms, the Scope 3 emissions of their supply chains dwarf their own, yet "Scope 3 emissions" appears on only 9 recent papers. Which contracts get suppliers to decarbonize, and who pays?

<h2>What prompted this</h2> <p>In the operations research journals, "supply chain coordination" had 19 papers in July–September 2026 against 11 expected, "emissions trading" 7 against 3 and "carbon pricing" 5 against 3 (see the <a href="/categories/operations-research">operations research dashboard</a>). Since October 2025, 77 papers carry a supply-chain-coordination or supply-chain-contract keyword (58 in operations research journals); 16 of them carry an emission, carbon or Scope 3 keyword, and "Scope 3 emissions" appears on 9 papers in all.</p> <h2>Questions</h2> <ul> <li>Which contract forms (cost sharing, price premiums for low-carbon inputs, long-term purchase commitments) lead suppliers to invest in abatement, and at what cost to the buyer?</li> <li>How do carbon prices on the supplier's side, such as the EU ETS and the EU's border carbon adjustment, change the best contract?</li> <li>Can buyers verify supplier emissions well enough to write contracts on them?</li> </ul> <h2>Data that could help</h2> <p>Supplier-level emissions (the CDP supply chain programme), carbon prices, and purchasing contracts from case firms.</p> <h2>Next step</h2> <p>Model a buyer and a supplier that can invest in abatement under a carbon price, and compare cost sharing with a price premium in a two-stage Stackelberg game.</p> https://scholar4ai.com/ideas/congestion-pricing-in-new-york 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Congestion pricing, low emission zones and road tolls appear on just 10 recent papers, with one in the last quarter. New York's congestion charge began in January 2025, the first in the US, after years of delay. What did it do to traffic, transit, business and air quality, and who paid?

<h2>What prompted this</h2> <p>Since October 2025, 10 papers carry a congestion-pricing, congestion-charge, road-pricing, low-emission-zone or toll keyword (5 in economics journals, 4 in operations research): "congestion pricing" on 5, "low emission zones" on 4. Only one of them was published in July–September 2026. The evidence from London, Stockholm and Milan is old; New York's charge is a new, large and well-measured experiment.</p> <h2>Questions</h2> <ul> <li>How much did vehicle entries, travel times and bus speeds change in the charging zone, and did traffic rise just outside it?</li> <li>Did retail sales and restaurant visits in the zone fall, as opponents predicted, or rise with easier access?</li> <li>Who pays the charge by income and residence, and did transit ridership and the revenue reach the promised investments?</li> </ul> <h2>Data that could help</h2> <p>Traffic counts and travel times (the transit authority's public data, mapping services), card spending by location, transit ridership and air quality monitors.</p> <h2>Next step</h2> <p>Compare weekday vehicle entries, bus speeds and card spending in the zone with matched areas elsewhere in the city for the year before and after January 2025.</p> https://scholar4ai.com/ideas/concept-drift-in-credit-and-fraud-models 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Distribution shift, concept drift and non-stationary time series appear on 371 recent papers, almost all in AI journals, and only 27 of them concern credit, fraud, banking or markets. Credit and fraud models decay as borrowers, fraudsters and the economy change. How fast, how should it be detected, and when should a model be retrained rather than recalibrated?

<h2>What prompted this</h2> <p>Since October 2025, 371 papers carry a concept-drift, non-stationarity, distribution-shift, covariate-shift or dataset-shift keyword (347 in AI journals): "distribution shift" on 170, "concept drift" on 65, "non-stationary time series" on 42, "non-stationarity" on 22. Only 27 of the 371 also carry a credit, finance, fraud, bank, loan, stock, market, economics or trading keyword.</p> <h2>Questions</h2> <ul> <li>In credit scoring models deployed through the pandemic and the rate rises, how much did discrimination and calibration decay per year, and did the decay come from borrower mix or from the relationship between features and default?</li> <li>Which drift detectors raise the alarm early enough to matter and rarely enough to be trusted?</li> <li>When fraudsters adapt to a model, is the drift adversarial in a way that standard detectors miss?</li> </ul> <h2>Data that could help</h2> <p>Loan performance data over 2018–2025 (LendingClub, mortgage loan-level data), fraud transaction data with time stamps, and model monitoring logs from a lender.</p> <h2>Next step</h2> <p>Train a credit model on 2018–2019 loans and track its monthly discrimination and calibration through 2025, attributing the decay to feature drift and relationship drift.</p> https://scholar4ai.com/ideas/code-generation-and-software-work 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Code generation, program synthesis, vulnerability detection and software development appear on 75 recent papers, mostly in AI journals, and 35 of them touch productivity, quality or security. Coding assistants are the most widely adopted generative AI tool at work. What happened to developer output, code quality and the junior developer's job?

<h2>What prompted this</h2> <p>Since October 2025, 75 papers carry a code-generation, program-synthesis, coding-assistant, software-development, code-quality or vulnerability-detection keyword (60 in AI journals, 8 in management): "code generation" on 18, "program synthesis" on 6, "smart contract vulnerability detection" on 5, "software development" on 4, "code quality" on 3. Thirty-five of them also carry a productivity, developer, quality, security, bug or labour keyword, but almost none uses data on real teams.</p> <h2>Questions</h2> <ul> <li>In firms that adopted coding assistants, did output per developer rise, and did defect rates, security vulnerabilities and code churn rise with it?</li> <li>Did hiring of junior developers fall relative to senior ones, and did pay for review and architecture skills rise?</li> <li>Do open-source projects show more contributions and more low-quality pull requests since 2023?</li> </ul> <h2>Data that could help</h2> <p>Firm engineering metrics around assistant roll-outs, open-source repository histories, job postings for software roles by seniority, and developer surveys.</p> <h2>Next step</h2> <p>Measure pull request volume, acceptance rates and post-merge bug fixes in a large sample of open-source repositories before and after 2023.</p> https://scholar4ai.com/ideas/coal-phase-out-and-local-labour-markets 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

A "just transition" is now a common theme in the economics journals, and many coal plants and mines have announced closure dates. Closures known years in advance are a rare chance to see whether workers and towns can prepare. What happens to coal workers' earnings, and to the towns around them?

<h2>What prompted this</h2> <p>Since October 2025, "just transition" appears on 38 papers, 36 of them in economics journals, and "coal-fired power plants" on 12. Energy topics are rising on the <a href="/trends">trends page</a>, from energy markets to critical minerals.</p> <h2>Questions</h2> <ul> <li>When a coal plant or mine closes, how much do its workers' earnings fall, and for how long, compared with workers at plants that stay open?</li> <li>Do announced closure dates let workers and towns adjust earlier, softening the blow?</li> <li>Do transition funds, such as the EU's Just Transition Fund, reach the affected workers and towns, and do they help?</li> </ul> <h2>Data that could help</h2> <p>Plant closure dates (Global Coal Plant Tracker, national registers), linked employer-employee data or local labour statistics, and transition fund allocations.</p> <h2>Next step</h2> <p>Compile the announced and actual closure dates of coal plants in one country and compare employment in their counties with similar counties without closures.</p> https://scholar4ai.com/ideas/climate-stress-tests-and-lending 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Stress testing appears on 40 recent papers in the finance journals, and climate stress testing on 12. Central banks in Europe and Asia have run climate scenarios on their banks. Did banks that fared badly in the scenarios cut lending to exposed borrowers, or only write longer reports?

<h2>What prompted this</h2> <p>Since October 2025, 57 papers carry a stress-testing keyword (40 in finance journals), and 17 of them also carry a climate keyword; "climate stress testing" appears on 12 papers. Climate risk itself appears on 242 papers, so the stress-test exercises are a small corner of a large literature.</p> <h2>Questions</h2> <ul> <li>After the European Central Bank's 2022 climate stress test, did banks with larger projected losses reduce loans to carbon-intensive or flood-exposed borrowers?</li> <li>Did borrowers simply switch to banks that were not tested, or to bond markets?</li> <li>Do the scenarios' long horizons and lack of capital consequences mean that banks treat them as reporting exercises?</li> </ul> <h2>Data that could help</h2> <p>Published stress-test results by bank, loan-level data where available (AnaCredit, syndicated loans), and borrowers' emissions and locations.</p> <h2>Next step</h2> <p>Compare the syndicated lending of banks with high and low projected climate losses to carbon-intensive borrowers before and after the 2022 exercise.</p> https://scholar4ai.com/ideas/climate-migration 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Migration keywords appear on 283 recent papers, with 85 in the last quarter against about 61 per quarter before, but only 20 of them carry a climate, weather, drought, flood or disaster keyword, and "climate-induced migration" on 3. Do people move because of climate shocks, how far, and does the move help?

<h2>What prompted this</h2> <p>Since October 2025, 283 papers carry a migration, emigration, migrant, refugee or displaced keyword (235 in economics journals): "internal migration" on 27, "immigration" on 19, "migration" on 17, "rural-urban migration" on 15. The quarter to September 2026 had 85 such papers against about 61 per quarter earlier. Only 20 of the 283 also carry a climate, weather, drought, flood, temperature, sea-level, disaster or environmental keyword.</p> <h2>Questions</h2> <ul> <li>After droughts and floods, do people move within their country, mostly to nearby cities, rather than across borders, as earlier evidence suggests, and does that still hold for the largest recent shocks?</li> <li>Do the poorest households, who are most exposed, lack the means to move, so that shocks trap them instead?</li> <li>Do migrants who left after a shock earn more than those who stayed, and do they send money back that helps those who stayed adapt?</li> </ul> <h2>Data that could help</h2> <p>Census and survey data on migration, gridded weather and flood data, mobile phone mobility data, and remittance flows.</p> <h2>Next step</h2> <p>Match district-level migration flows from censuses to drought and flood exposure for several countries and compare the response of poor and non-poor households.</p> https://scholar4ai.com/ideas/climate-disclosure-mandates 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Climate risk disclosure appears on 19 recent papers, carbon disclosure on 10 and TCFD on 7; only 14 of 53 climate-disclosure papers carry a mandate, rule or regulation keyword. The US SEC rule was stayed, California's law survived, and the ISSB standards are being adopted by dozens of countries. Which regime changes what firms report, and what investors do with it?

<h2>What prompted this</h2> <p>Since October 2025, 53 papers carry a climate-disclosure, TCFD, carbon-disclosure or emission-disclosure keyword (27 in finance journals, 11 in accounting, 9 in economics): "climate risk disclosure" on 19, "carbon disclosure" on 10, "carbon emission disclosure" on 10, "TCFD" on 7. Only 14 of the 53 also carry a mandatory, rule, regulation, law, SEC or requirement keyword.</p> <h2>Questions</h2> <ul> <li>Do firms report different numbers under voluntary TCFD reporting, California's law and ISSB-based rules, for the same emissions?</li> <li>Did mandatory disclosure in the UK and New Zealand change investors' holdings of high-emitting firms, or lenders' terms?</li> <li>Do firms covered by California's law but not by any federal rule change their US operations or legal structure?</li> </ul> <h2>Data that could help</h2> <p>Disclosed emissions by regime (CDP, annual reports), adoption dates of mandates by country, institutional holdings and loan terms.</p> <h2>Next step</h2> <p>Compare the emissions reported by multinationals under the UK mandate with those reported voluntarily to CDP for the same years.</p> https://scholar4ai.com/ideas/city-logistics-and-cargo-bikes 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Urban logistics, urban freight transport and urban delivery appear on 46 recent papers, almost all in operations research, and the topic is fading, with 8 in the last quarter against about 13 per quarter before. Cities are restricting vans and funding cargo bikes and micro-hubs. Which combinations cut kilometres and cost, and do couriers and retailers adopt them without mandates?

<h2>What prompted this</h2> <p>Since October 2025, 46 papers carry an urban-logistics, urban-freight, urban-delivery, cargo-bike, micro-hub or consolidation keyword (41 in operations research journals): "urban logistics" on 29, "urban freight transport" on 8, "urban delivery" on 6, "cargo bikes" on 2, "urban consolidation centers" on 1. Eight were published in July–September 2026 against about 13 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In cities that restricted delivery vans in central zones, did couriers switch to cargo bikes and micro-hubs, and what happened to delivery cost and times?</li> <li>Do micro-hubs reduce total vehicle kilometres once the trips to stock them are counted?</li> <li>Which consolidation schemes survived without public subsidy, and what distinguished them from the many that closed?</li> </ul> <h2>Data that could help</h2> <p>Courier fleet and route data, city regulation dates, micro-hub operator records and delivery performance data.</p> <h2>Next step</h2> <p>Compare delivery costs and vehicle kilometres per parcel in central Paris or London before and after van restrictions, using courier data.</p> https://scholar4ai.com/ideas/circular-textiles 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The textile and apparel industries, fashion retail and fast fashion appear on dozens of recent papers across the AI, management and operations research journals, and 18 of them carry a circular, recycling, waste or resale keyword. The EU will require separate textile collection and producer responsibility. Can collection, sorting and recycling be made to pay, and does resale cut new production?

<h2>What prompted this</h2> <p>Since October 2025, 87 papers carry a textile, apparel, fashion, garment or clothing keyword (32 in AI journals, 30 in management, 11 in economics, 11 in operations research): "textile industry" on 12, "fashion retail" on 8, "apparel industry" on 7, "fast fashion" on 5. Eighteen of the 87 also carry a circular, recycling, waste, sustainability, secondhand, resale, rental or fast-fashion keyword.</p> <h2>Questions</h2> <ul> <li>At current prices for recycled fibres, which sorting and recycling technologies cover their costs, and how large must collection volumes be?</li> <li>Does secondhand resale reduce purchases of new clothing, or add to total consumption?</li> <li>Who pays under producer responsibility fees, and do fees change what fast fashion firms design?</li> </ul> <h2>Data that could help</h2> <p>Textile collection and sorting data from the countries that already collect separately, resale platform sales, apparel import volumes and producer responsibility fee schedules.</p> <h2>Next step</h2> <p>Build a cost model of a textile sorting and recycling chain from published plant data and compute the collection volume needed to break even at 2025 fibre prices.</p> https://scholar4ai.com/ideas/chinas-retirement-age-reform 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Retirement, retirement age and delayed retirement appear on 94 recent papers in the economics and finance journals, and 38 of them touch China, pension reform or labour supply. China's statutory retirement age began rising in 2025, from some of the lowest ages in the world. Who works longer, who cares for grandchildren less, and what happens to youth hiring?

<h2>What prompted this</h2> <p>Since October 2025, 94 papers carry a retirement keyword (54 in economics journals, 29 in finance): "retirement" on 19, "retirement age" on 12, "retirement planning" on 12, "delayed retirement" on 4. Thirty-eight of them also carry a China, labour-supply, employment, health or reform keyword, and "pension reform" appears on 10 of those. The 2025 reform raises the retirement age gradually over fifteen years, which makes it a long natural experiment.</p> <h2>Questions</h2> <ul> <li>Do workers affected by the first steps of the increase keep working, or leave the labour force without a pension?</li> <li>Does later retirement reduce grandparents' childcare and thereby lower their children's labour supply and fertility?</li> <li>Do firms hire fewer young workers when older workers stay, as the "lump of labour" argument claims, or not?</li> </ul> <h2>Data that could help</h2> <p>Household panels with retirement and childcare data (China Health and Retirement Longitudinal Study, China Family Panel Studies), firm employment records and the reform's cohort schedule.</p> <h2>Next step</h2> <p>Compare labour force participation and childcare hours of cohorts just above and just below the first retirement age steps in the 2025–2026 survey waves.</p> https://scholar4ai.com/ideas/china-local-debt-swap 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Local government finance is rising in the economics journals, and local government debt is a large topic in finance as China's property slump cut land revenue. In late 2024 China launched a large swap of local governments' hidden debt. Did it get suppliers paid and credit flowing, or just move risk to banks?

<h2>What prompted this</h2> <p>In economics, "local government finance" had 8 papers in July–September 2026 against 3 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, 77 papers carry a local-government-debt, land-finance or local-government-finance keyword (42 in finance journals, 31 in economics); 16 of them also carry a housing, real-estate, land or property keyword.</p> <h2>Questions</h2> <ul> <li>Did provinces with larger swap quotas pay overdue bills to firms sooner, and did those firms then invest or hire more?</li> <li>Did the swap lower borrowing costs of local government financing vehicles, and did the banks that bought the new bonds cut other lending?</li> <li>Does replacing hidden debt with official bonds change how local governments spend?</li> </ul> <h2>Data that could help</h2> <p>Swap quotas and bond issuance by province, bonds of local government financing vehicles (Wind, CSMAR), listed firms' receivables from government-linked customers, and bank balance sheets.</p> <h2>Next step</h2> <p>Collect provincial swap quotas for 2024–2026 and relate them to changes in listed firms' receivables from government-linked customers.</p> https://scholar4ai.com/ideas/chatbots-and-customer-satisfaction 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Customer satisfaction and customer engagement are among the fastest-growing keywords in the management journals, and firms keep replacing service agents with chatbots. The EU AI Act requires telling people when they are talking to an AI. Does disclosure lower satisfaction, and can good service recovery make up for it?

<h2>What prompted this</h2> <p>In the management journals, "customer satisfaction" had 17 papers in July–September 2026 against about 6 expected, and "customer engagement" 13 against 8 (see the <a href="/categories/management">management dashboard</a>). Since October 2025, 90 papers carry a chatbot keyword (54 in AI journals, 29 in management), and only 7 of them also carry a satisfaction, service-quality, customer-service or service-recovery keyword.</p> <h2>Questions</h2> <ul> <li>Do customers rate the same service lower once they learn it came from a chatbot, and does the gap shrink as they get used to it?</li> <li>When a chatbot fails, does handing over to a human repair satisfaction better than an apology from the bot?</li> <li>Do disclosure rules change which firms deploy chatbots, and for which tasks?</li> </ul> <h2>Data that could help</h2> <p>Customer reviews that mention chatbots (Yelp, TripAdvisor, app stores), firms' chatbot launch dates, and field experiments with service providers.</p> <h2>Next step</h2> <p>Collect app-store reviews of banking apps that mention a chatbot, and compare ratings before and after each app began telling users that its assistant is an AI.</p> https://scholar4ai.com/ideas/ceo-pay-ratio-disclosure 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Executive compensation appears on 114 recent papers, yet only 14 of 144 pay papers carry a disclosure, pay-ratio, say-on-pay or transparency keyword. Firms in the US and UK now publish the ratio of CEO pay to median worker pay. Did it change CEO pay, worker pay or neither?

<h2>What prompted this</h2> <p>Since October 2025, 144 papers carry an executive-compensation, CEO-compensation or executive-pay keyword (73 in finance journals, 30 in accounting, 22 in management); "executive compensation" appears on 114 and "CEO compensation" on 33. Only 14 of the 144 also carry a pay-ratio, disclosure, say-on-pay or transparency keyword, and "executive compensation disclosure" on 2.</p> <h2>Questions</h2> <ul> <li>After pay ratio disclosure began, did firms with the highest ratios slow CEO pay growth, raise median worker pay, or outsource their lowest-paid jobs?</li> <li>Do investors vote against pay packages more often at high-ratio firms, and do customers or job applicants react?</li> <li>Do firms manage the ratio itself, for example by changing which employees count?</li> </ul> <h2>Data that could help</h2> <p>Pay ratio disclosures from proxy statements, ExecuComp, say-on-pay vote results and workforce data from annual reports.</p> <h2>Next step</h2> <p>Collect pay ratios for S&amp;P 1500 firms for 2018–2025 and test whether a firm's ratio in one year predicts changes in median worker pay in the next.</p> https://scholar4ai.com/ideas/central-banks-buying-gold 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Gold and safe-haven assets appear on 172 recent papers in the finance and economics journals, almost all about gold as a hedge for investors. Only 5 carry a central-bank, reserve or sanctions keyword. Central banks have been the largest buyers of gold since 2022. Is it insurance against sanctions, and does it move the price?

<h2>What prompted this</h2> <p>Since October 2025, 172 papers carry a gold or safe-haven keyword (99 in finance journals, 50 in economics): "gold" on 58, "safe-haven assets" on 49, "gold prices" on 24. Only 5 of the 172 also carry a central-bank, reserve, de-dollarization, sanctions or official keyword, and "gold reserves" appears on 2 papers.</p> <h2>Questions</h2> <ul> <li>Did central banks of countries at risk of sanctions buy more gold after the freezing of Russia's reserves in 2022, and did they move gold home?</li> <li>How much of the gold price rise since 2022 can be attributed to official buying rather than investor demand?</li> <li>Does gold in reserves reduce a country's borrowing costs or exchange rate volatility?</li> </ul> <h2>Data that could help</h2> <p>Official reserve data (IMF, World Gold Council), customs data on gold flows, gold prices and measures of geopolitical alignment.</p> <h2>Next step</h2> <p>Relate quarterly changes in each central bank's gold holdings since 2022 to its country's political distance from the US and its exposure to sanctions.</p> https://scholar4ai.com/ideas/central-bank-messages-for-households 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Central bank communication is one of the rising keywords across the tracked journals, but few of those papers look at households. Language models can now rewrite a policy statement for any reader. Would households who receive such messages update their inflation expectations?

<h2>What prompted this</h2> <p>"Central bank communication" grew across all journals (16 papers in July–September 2026 against about 9 expected) and in finance (10 against 4); see the <a href="/trends">trends page</a>. Since October 2025, 58 papers carry a central-bank-communication or forward-guidance keyword (32 in economics, 26 in finance). Only 13 of them also carry a household, consumer, survey, expectations or public keyword, while "inflation expectations" appears on 86 papers.</p> <h2>Questions</h2> <ul> <li>In a randomized survey, do households who read a plain-language or personalized version of a policy statement change their inflation expectations more than those who read the original?</li> <li>Does the effect last, and does it change spending plans?</li> <li>Do people trust a message less if they know a machine wrote it?</li> </ul> <h2>Data that could help</h2> <ul> <li>Household expectation surveys, such as the Michigan Surveys of Consumers or central banks' own consumer surveys, with an experiment embedded.</li> <li>Policy statements and the layered, plain-language versions that several central banks already publish.</li> </ul> <h2>Next step</h2> <p>Draft three versions of one recent policy statement (the original, a plain-language version and one personalized by a language model) and pilot them in an online survey with an expectations question.</p> https://scholar4ai.com/ideas/cbdc-and-bank-deposits 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Central bank digital currencies appear on 56 recent papers in the finance and economics journals, but only 14 of them carry a deposit, bank-lending, bank-run or financial-stability keyword. Pilots now run in China, Europe and elsewhere. If people can hold central bank money directly, how much would leave banks, and when?

<h2>What prompted this</h2> <p>Since October 2025, 56 papers carry a central-bank-digital-currency keyword (46 in finance journals); "CBDC adoption" appears on 11. Only 14 of the 56 also carry a deposit, disintermediation, bank-lending, bank-run or financial-stability keyword. "Bank deposits" appears on 8 papers.</p> <h2>Questions</h2> <ul> <li>In the pilots in China, Nigeria and the Bahamas, did adopters move balances out of bank deposits, and did banks respond with higher deposit rates?</li> <li>Would holding limits and no interest, as the digital euro design proposes, keep deposits in place during a bank stress?</li> <li>How does a digital currency compare with stablecoins and money market funds as a drain on deposits?</li> </ul> <h2>Data that could help</h2> <p>Pilot usage statistics published by central banks, bank-level deposit data, and survey evidence on who adopted the digital currency and why.</p> <h2>Next step</h2> <p>Compare deposit growth at banks in pilot cities with banks elsewhere during China's e-CNY roll-out.</p> https://scholar4ai.com/ideas/cbam-and-carbon-leakage 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Carbon leakage appears on 28 recent papers and the EU's carbon border adjustment on a handful, with the topic fading (5 papers in the last quarter against about 8 per quarter before) just as the mechanism starts charging in 2026. Which exporters pay, do they clean up or divert trade, and does leakage actually fall?

<h2>What prompted this</h2> <p>Since October 2025, 30 papers carry a carbon-leakage or border-carbon-adjustment keyword (24 in economics journals): "carbon leakage" on 28, "border carbon adjustment" on 4. Five were published in July–September 2026 against about 8 per quarter earlier. The keyword "CBAM" appears on 38 papers, but part of that count is an attention module in AI journals that shares the acronym.</p> <h2>Questions</h2> <ul> <li>For steel, cement, aluminium and fertilizer exporters to the EU, how much of the charge falls on them, and do they lower prices, cut emissions or sell elsewhere?</li> <li>Do exporters ship their cleanest production to Europe and dirtier output to other markets, so that measured leakage falls while global emissions do not?</li> <li>Do trading partners respond with their own carbon prices, as the mechanism was designed to encourage?</li> </ul> <h2>Data that could help</h2> <p>EU import data by product and origin with CBAM declarations, exporters' plant-level emissions, and carbon pricing changes in partner countries.</p> <h2>Next step</h2> <p>Track EU imports of the covered goods by origin country before and after January 2026 and compare them with those countries' exports to other destinations.</p> https://scholar4ai.com/ideas/causal-machine-learning-for-targeting 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Double machine learning, causal forests, conditional average treatment effects and policy learning appear on 50 recent papers, split between the AI and economics journals, with 14 in the last quarter against about 11 per quarter before. The methods promise to find who benefits most from a programme. Do the targeting rules they produce hold up in new samples, and are they fair?

<h2>What prompted this</h2> <p>Since October 2025, 50 papers carry a double-machine-learning, causal-forest, heterogeneous-treatment, uplift, conditional-average-treatment-effect or policy-learning keyword (19 in AI journals, 19 in economics, 8 in finance): "double machine learning" on 13, "causal forest" on 11, "conditional average treatment effect" on 8, "policy learning" on 5. The quarter to September 2026 had 14 such papers against about 11 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When targeting rules estimated from one experiment are applied to a new experiment of the same programme, how much of the predicted gain over uniform treatment is realized?</li> <li>Do the rules concentrate treatment on groups that are already better off, and how much is lost by adding fairness constraints?</li> <li>Are the heterogeneity estimates stable across estimators, or do causal forests and double machine learning disagree about who benefits?</li> </ul> <h2>Data that could help</h2> <p>Public experimental datasets with multiple waves or sites (job training, microfinance and cash transfer experiments), and replication code of the estimators.</p> <h2>Next step</h2> <p>Estimate targeting rules on one site of a multi-site experiment with two estimators and evaluate them on the other sites.</p> https://scholar4ai.com/ideas/cash-transfers-and-local-prices 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Cash transfers appear on 43 recent papers in the economics journals, but only 4 of them carry a price, inflation, multiplier or spillover keyword. If a programme gives cash to a large share of a village, shopkeepers may raise prices and non-recipients may gain or lose. How large are these effects, and when do they reverse the programme's purpose?

<h2>What prompted this</h2> <p>Since October 2025, 43 papers carry a cash-transfer keyword (42 in economics journals): "cash transfers" on 30, "conditional cash transfers" on 11, "unconditional cash transfers" on 4. Only 4 of them also carry a price, inflation, local-economy, multiplier, spillover or general-equilibrium keyword.</p> <h2>Questions</h2> <ul> <li>In villages where most households receive transfers, how much do food and rent prices rise, and for how long?</li> <li>Do non-recipients gain from the extra demand (as sellers and workers) or lose (as buyers)?</li> <li>Do mobile money transfers, which can be spent beyond the village, reduce the price effects?</li> </ul> <h2>Data that could help</h2> <p>Transfer programme rosters with saturation by village, local price surveys, and household surveys of recipients and non-recipients.</p> <h2>Next step</h2> <p>Use a programme with randomized saturation across villages to estimate price changes and the net effect on non-recipients.</p> https://scholar4ai.com/ideas/carbon-credit-quality 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Carbon credits and offsetting appear on 22 recent papers, with 9 in the last quarter against about 4 per quarter before, but only 3 carry a quality, additionality, integrity or greenwashing keyword. After investigations found that many forest credits avoided no emissions, the market shrank and new integrity labels appeared. Do buyers now pay for quality?

<h2>What prompted this</h2> <p>Since October 2025, 22 papers carry a carbon-credit, carbon-offset or voluntary-carbon keyword (14 in economics journals): "carbon credits" on 6, "carbon offsetting" on 5, "carbon credit trading" on 4. Nine of them were published in July–September 2026 against about 4 per quarter earlier. Only 3 of the 22 also carry a quality, additionality, integrity, rating or greenwashing keyword.</p> <h2>Questions</h2> <ul> <li>Do credits with integrity labels (for example the Core Carbon Principles) trade at a premium, and does the premium reflect measured additionality?</li> <li>Which firms buy the lowest-quality credits, and do their ESG ratings reward the purchase anyway?</li> <li>Can satellite data verify forest credits well enough to price them?</li> </ul> <h2>Data that could help</h2> <p>Registry data on credit issuance and retirement (Verra, Gold Standard), credit prices by project type, buyers' sustainability reports, and satellite forest cover.</p> <h2>Next step</h2> <p>Match retired credits in the public registries to the buyers' reported offset use and compare the project types chosen by firms with high and low ESG ratings.</p> https://scholar4ai.com/ideas/carbon-aware-computing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Edge and cloud computing appear on 151 recent papers, with 43 in the last quarter against about 34 per quarter before, and 46 of them carry an energy, carbon or green keyword, mostly about device energy. Carbon-aware scheduling, which moves computing jobs to times and places where the grid is clean, appears on one paper. How much carbon could it save, and who would pay for the delay?

<h2>What prompted this</h2> <p>Since October 2025, 151 papers carry a cloud-computing, edge-computing, data-centre, job-scheduling, task-offloading or serverless keyword (111 in AI journals, 14 in economics, 11 in finance, 9 in operations research): "edge computing" on 63, "cloud computing" on 22, "task offloading" on 14, "green data centers" on 4, "energy-efficient computing" on 11. The quarter to September 2026 had 43 such papers against about 34 per quarter earlier. Forty-six of the 151 carry an energy, carbon, renewable or green keyword, but "carbon-aware scheduling" appears on a single paper.</p> <h2>Questions</h2> <ul> <li>For training and batch inference jobs that can wait hours or move between regions, how much carbon can be cut at current grid carbon intensities, and at what cost in delay and transfer?</li> <li>Do cloud providers' carbon accounting methods (hourly matching versus annual) change which schedules look clean?</li> <li>Would a carbon price on computing move jobs enough to matter for the grid, and where?</li> </ul> <h2>Data that could help</h2> <p>Hourly grid carbon intensity by region (ENTSO-E, electricityMap), cloud job traces released by providers, and electricity prices.</p> <h2>Next step</h2> <p>Simulate shifting a public cloud job trace across regions and hours with 2025 grid carbon data and compute the carbon saved per hour of added delay.</p> https://scholar4ai.com/ideas/capital-flows-and-fed-tightening 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

International capital flows appear on 96 recent papers, and 24 of them carry a Federal Reserve, global financial cycle or spillover keyword. The 2022–2023 tightening was the fastest in forty years, yet the sudden stops predicted for emerging markets mostly did not happen. What was different, and will it last?

<h2>What prompted this</h2> <p>Since October 2025, 96 papers carry a capital-flow, portfolio-flow, sudden-stop, capital-flight or global-financial-cycle keyword (65 in finance journals, 28 in economics): "international capital flows" on 37, "capital flows" on 29, "global financial cycle" on 8, "sudden stops" on 6. Twenty-four of them also carry a Federal Reserve, US-monetary, dollar, spillover or monetary-policy keyword.</p> <h2>Questions</h2> <ul> <li>Which emerging markets avoided outflows in 2022–2023, and did early rate hikes, local-currency borrowing or reserves explain it?</li> <li>Did outflows show up in bond funds and ETFs rather than in bank lending, and does the channel matter for crises?</li> <li>Does the resilience hold for frontier markets that lost market access, such as those that defaulted?</li> </ul> <h2>Data that could help</h2> <p>Fund flow data (EPFR), balance of payments statistics, reserve data and policy rate changes by country.</p> <h2>Next step</h2> <p>Classify emerging markets by the timing of their rate hikes relative to the Fed and compare their portfolio flows during 2022–2023.</p> https://scholar4ai.com/ideas/call-centres-with-ai-agents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Customer service and call centres appear on 25 recent papers, and 10 of them carry a chatbot, automation or language-model keyword. Firms now route calls through voice agents that handle part of each conversation and hand the rest to people. How should queues, staffing and handovers be designed when the first responder is a model?

<h2>What prompted this</h2> <p>Since October 2025, 25 papers carry a call-centre, contact-centre or customer-service keyword (11 in management journals, 4 in operations research): "customer service" on 12, "customer service automation" on 2, "customer service employees" on 2, "call centers" on 2. Ten of them also carry a queue, staffing, chatbot, AI, automation, language-model or routing keyword. The queueing models that once dominated call-centre research have not yet absorbed AI agents.</p> <h2>Questions</h2> <ul> <li>When an AI agent resolves a share of calls and escalates the rest, how do waiting times and staffing needs change, and does the escalated traffic become harder per call?</li> <li>What is the optimal point to hand over to a person, and does the answer differ for customers who want to leave?</li> <li>Do AI-first systems lower costs more than they lower satisfaction and retention?</li> </ul> <h2>Data that could help</h2> <p>Call logs with routing, handover and resolution data from contact centres that deployed AI agents, and customer retention records.</p> <h2>Next step</h2> <p>Fit a queueing model with an AI first stage to call logs from one contact centre and test its staffing predictions against observed waiting times.</p> https://scholar4ai.com/ideas/calibrated-uncertainty-in-robust-optimization 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Distributionally robust optimization is one of the fastest-growing methods in operations research, and uncertainty estimation is rising in the AI journals, yet only two papers in the archive touch both. Can calibrated uncertainty from machine learning decide how much robustness an optimization model buys?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "distributionally robust optimization" is growing in the operations research journals (22 papers in July–September 2026 against about 15 expected) and "uncertainty estimation" in the AI journals (39 against 23). Since October 2025, 195 papers carry a robust-optimization keyword (158 in operations research), but only 2 of them also carry a conformal-prediction, uncertainty-quantification, uncertainty-estimation or calibration keyword. Uncertainty quantification (105 papers) and uncertainty estimation (117) stay almost entirely in the AI journals, while "demand uncertainty" is the second most frequent operations research keyword of the year.</p> <h2>Questions</h2> <ul> <li>If the size of the ambiguity set (say, a Wasserstein radius) comes from conformal prediction intervals of a demand forecast, does it beat a cross-validated radius out of sample?</li> <li>How much cost does too much robustness waste in inventory problems such as the newsvendor model?</li> <li>What happens under distribution shift, when the forecast's calibration breaks?</li> </ul> <h2>Data that could help</h2> <p>Public retail demand data such as the M5 competition data, and standard inventory and network-design instances.</p> <h2>Next step</h2> <p>Solve a newsvendor problem on M5 data with conformal intervals and with a cross-validated Wasserstein radius, and compare cost and service level.</p> https://scholar4ai.com/ideas/buybacks-under-the-excise-tax 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Share repurchases appear on 28 recent papers in the finance journals, and only one of all 35 buyback papers carries a tax keyword. The US has taxed buybacks at 1% since 2023, and a higher rate has been proposed. Did the tax move payouts to dividends, and did it change investment?

<h2>What prompted this</h2> <p>Since October 2025, 35 papers carry a share-repurchase or buyback keyword (29 in finance journals), with "share repurchases" on 28. Only one of them also carries a tax keyword. The literature on payout policy still mostly asks why firms buy back shares, not what a tax on it does.</p> <h2>Questions</h2> <ul> <li>After the 1% excise tax, did firms shift from buybacks to dividends, or did buybacks simply continue?</li> <li>Did firms with large buyback programmes invest or hire more, as supporters of the tax claimed?</li> <li>Do the responses predict what a 4% rate would do?</li> </ul> <h2>Data that could help</h2> <p>Quarterly repurchase and dividend data (Compustat), announcement dates of buyback programmes, and firms' investment and employment.</p> <h2>Next step</h2> <p>Compare buybacks and dividends of US firms before and after 2023 with those of Canadian firms, which faced a 2% tax from 2024, and with firms in countries without such a tax.</p> https://scholar4ai.com/ideas/buy-now-pay-later 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

"Buy now, pay later" appears on just 8 recent papers across the tracked journals, although it has become one of the fastest-growing forms of consumer credit and is now being brought into credit reporting. Who uses it, does it crowd out or add to other borrowing, and does it lead to missed payments elsewhere?

<h2>What prompted this</h2> <p>Since October 2025, "buy now pay later" appears on 8 papers in the archive, spread across finance, economics, management, operations research and AI journals, with 3 of them in July–September 2026. "Consumer credit" appears on 13. Credit scoring and household debt, by contrast, are well-established topics.</p> <h2>Questions</h2> <ul> <li>Do households that start using instalment plans at checkout spend more, and do they fall behind on other bills?</li> <li>Does access help people who are shut out of credit cards, or mainly people who already have them?</li> <li>Now that the main providers report to credit bureaus, do scores of frequent users rise or fall?</li> </ul> <h2>Data that could help</h2> <p>Transaction-level data from banks or account aggregators, credit bureau records that now include instalment plans, and the dates of regulatory changes in the UK, Australia and the US.</p> <h2>Next step</h2> <p>Use bank transaction data to identify new users of instalment plans and track their spending and late fees in the following six months.</p> https://scholar4ai.com/ideas/board-gender-quotas 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Board composition is one of the fastest-growing keywords across the tracked journals, and board gender diversity is a large topic in finance, accounting and management. But only 7 of 135 such papers carry a quota, mandate or regulation keyword, just as the EU's quota directive reaches its target date. Do quotas change how boards decide, or only who sits on them?

<h2>What prompted this</h2> <p>"Board composition" grew across all journals (13 papers in July–September 2026 against about 4 expected; see the <a href="/trends">trends page</a>). Since October 2025, 135 papers carry a board-gender-diversity, female-director or board-diversity keyword (51 in finance journals, 37 in accounting, 30 in management, 17 in economics). Only 7 of them also carry a quota, mandate, regulation, law or legislation keyword.</p> <h2>Questions</h2> <ul> <li>When a quota forces new appointments, do board decisions change (risk taking, payouts, CEO pay, sustainability), or do boards move decisions to committees that stay as they were?</li> <li>Do quota appointees differ from voluntary ones in experience and networks, and does that explain the results?</li> <li>Does the EU directive, with its June 2026 target, repeat the results of Norway's earlier quota, or do newer quotas work differently?</li> </ul> <h2>Data that could help</h2> <p>Director-level data (BoardEx), quota laws and their dates by country, and firm outcomes (Compustat Global, Orbis).</p> <h2>Next step</h2> <p>Code the quota laws and target dates across EU member states and count how many listed firms were below target the year before.</p> https://scholar4ai.com/ideas/big-data-tax-enforcement 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Tax compliance and enforcement appear on 118 recent papers, and 23 of them carry a digital, big-data or AI keyword, led by China's digitalized tax administration. Tax authorities now match invoices, bank data and platform records automatically. Which taxpayers change behaviour, and does automated enforcement crowd out voluntary compliance?

<h2>What prompted this</h2> <p>Since October 2025, 118 papers carry a tax-enforcement, tax-compliance, tax-administration or tax-audit keyword (53 in economics journals, 40 in finance, 15 in accounting): "tax compliance" on 49, "tax enforcement" on 34, "digitalization of tax administration" on 9, "Golden Tax Project" on 6. Twenty-three of the 118 also carry a digital, big-data, machine-learning or e-invoicing keyword.</p> <h2>Questions</h2> <ul> <li>When electronic invoicing becomes mandatory, how much does reported revenue rise, and does the effect fade as firms learn the system's blind spots?</li> <li>Does algorithmic audit selection target the right taxpayers, and does it create perceptions of unfairness that reduce voluntary compliance?</li> <li>Do firms respond to data-driven enforcement by moving to cash, to informality or across borders?</li> </ul> <h2>Data that could help</h2> <p>Firm-level tax data around the phase-in of electronic invoicing (China's Golden Tax Phase IV, e-invoicing mandates in Italy, India and Latin America), audit selection records where available.</p> <h2>Next step</h2> <p>Use the staggered roll-out of mandatory e-invoicing across regions in one country to measure changes in reported revenue and VAT.</p> https://scholar4ai.com/ideas/beneficial-ownership-registers-after-the-eu-ruling 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Corporate transparency is a rising keyword in the management journals, while beneficial ownership barely appears in the archive. In November 2022 the EU Court of Justice struck down general public access to the registers of companies' beneficial owners. Did the return of anonymity bring back shell companies and opaque property purchases?

<h2>What prompted this</h2> <p>In the management journals, "corporate transparency" had 5 papers in July–September 2026 against about 2 expected, and in economics "transparency" grew too (9 against 4); see the <a href="/trends">trends page</a>. Since October 2025, "corporate transparency" appears on 50 papers (29 in finance journals), but "beneficial ownership" on just 2 and "anti-money laundering" on 14.</p> <h2>Questions</h2> <ul> <li>After several EU countries closed or restricted public access to their registers, did new companies with opaque ownership become more common?</li> <li>Did purchases of property through companies change in countries that closed their registers, compared with countries that kept them open?</li> <li>Do journalists, banks and tax authorities find fewer red flags without public access?</li> </ul> <h2>Data that could help</h2> <p>Company registers and ownership data (Orbis ownership, Open Ownership), property transactions with corporate buyers, and the dates on which each country restricted access.</p> <h2>Next step</h2> <p>List which EU countries closed, restricted or kept public access to their registers after the ruling, with dates, and count new company formations by legal form around those dates.</p> https://scholar4ai.com/ideas/benchmark-inertia-in-ai-journals 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The dataset ranking shows CIFAR and MNIST still among the most-mentioned image benchmarks in AI journal papers, while cross-dataset generalization and reproducibility are rising keywords. Do gains on these old benchmarks carry over to harder, shifted data?

<h2>What prompted this</h2> <p>On the <a href="/rankings/datasets">dataset ranking</a>, CIFAR-10/100 and MNIST are each named in the titles or abstracts of 7 AI journal papers in the 30 days to 2026-10-23, more than ImageNet (3). On the <a href="/categories/ai">AI dashboard</a>, "cross-dataset generalization" (24 papers in July–September 2026 against about 13 expected) and "reproducibility" (15 against 7) are growing. Since October 2025, 385 papers carry a benchmark keyword (287 in AI journals), and 28 of them also carry a reproducibility or generalization keyword.</p> <h2>Questions</h2> <ul> <li>For methods published with gains on CIFAR or MNIST, how much of the gain survives on newer or distribution-shifted datasets?</li> <li>How much of a reported gain comes from tuning effort rather than from the method?</li> <li>Should journals ask for at least one harder benchmark per paper?</li> </ul> <h2>Possible approach</h2> <p>Rerun released code from recent papers on a shifted benchmark, such as one of the domain-adaptation suites, with the same tuning budget for the method and its baselines.</p> <h2>Next step</h2> <p>Take ten recent AI-journal papers that report CIFAR gains and release their code, and rerun them on one distribution-shift benchmark.</p> https://scholar4ai.com/ideas/basic-income-pilots 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Universal basic income appears on 6 recent papers and unconditional cash transfers on 4, although large pilots in the US, Germany, Kenya and elsewhere have reported results in the last two years. What do the pilots agree on, where do they disagree, and what would scale change?

<h2>What prompted this</h2> <p>Since October 2025, 12 papers carry a basic-income, guaranteed-income or unconditional-cash keyword (11 in economics journals): "universal basic income" on 6, "unconditional cash transfers" on 4, "guaranteed income" on 2. The recent large pilots, with thousands of recipients and multi-year payments, have produced results that barely appear in these journals yet.</p> <h2>Questions</h2> <ul> <li>Across the pilots, how much did recipients reduce paid work, and did the hours go to care, education or job search?</li> <li>Did health, housing stability and children's outcomes improve by enough to offset the cost?</li> <li>What would differ at scale: prices, labour supply of non-recipients, and financing through taxes?</li> </ul> <h2>Data that could help</h2> <p>Published pilot data (the US OpenResearch study, the German Pilotprojekt Grundeinkommen, GiveDirectly's Kenya experiment), and administrative records where pilots were linked to them.</p> <h2>Next step</h2> <p>Tabulate the labour supply, income and well-being effects reported by each pilot on a common scale and compare them with the predictions of standard models.</p> https://scholar4ai.com/ideas/bank-runs-at-social-media-speed 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Deposit insurance and market discipline are rising topics after bank failures in which deposits left within hours. Yet only 3 recent papers on bank runs or deposits carry a social-media or digital-banking keyword. Does online talk now set the speed of runs, and should deposit insurance and liquidity rules change with it?

<h2>What prompted this</h2> <p>In the finance journals, "deposit insurance" had 7 papers in July–September 2026 against fewer than 2 expected from its earlier share, and "market discipline" is growing across all journals (12 against about 3); see the <a href="/trends">trends page</a>. Since October 2025, 37 papers carry a bank-run, deposit-insurance or bank-deposit keyword (23 in finance journals, 14 in economics). Only 3 of them carry a social-media, digital-banking or fintech keyword.</p> <h2>Questions</h2> <ul> <li>Do banks whose depositors talk more about them on social media lose deposits faster under stress, for the same share of uninsured deposits?</li> <li>Does online sentiment about a bank predict outflows days ahead, early enough for supervisors to act?</li> <li>Does app-based banking, where moving money takes minutes, make liquidity rules written for slower runs too lax?</li> </ul> <h2>Data that could help</h2> <ul> <li>Bank-level deposits from regulatory filings (Call Reports) and branch deposits (Summary of Deposits).</li> <li>Social media posts that mention banks, and each bank's share of deposits above the insurance limit.</li> <li>The 2023 US regional bank failures as a stress episode.</li> </ul> <h2>Next step</h2> <p>Build a weekly count of social media posts per US bank for 2022–2024 and relate it to deposit flows around March 2023.</p> https://scholar4ai.com/ideas/aviation-decarbonization-operations 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Airline, air traffic and aviation keywords appear on 142 recent papers, and the topic is fading, with 24 papers in the last quarter against about 38 per quarter before. Sustainable aviation fuel mandates began in Europe in 2025, and emissions trading now covers flights. Do the operations models and the climate policies meet, and what do the mandates cost per tonne?

<h2>What prompted this</h2> <p>Since October 2025, 142 papers carry an airline, aviation, flight-delay, air-traffic, air-transport, airport or aircraft keyword (56 in operations research journals, 39 in AI, 24 in economics): "airline industry" on 18, "air traffic management" on 12, "airline operations" on 6, "aviation emissions" on 5, "flight delays" on 5, "sustainable aviation fuel" on 4. Seventy-three of the 142 carry a pricing, scheduling, network, fuel, emission, delay or decarbonization keyword, but only a handful connect operations to the new mandates.</p> <h2>Questions</h2> <ul> <li>At the mandated blending shares for 2025–2030, what does sustainable aviation fuel cost per tonne of CO2 avoided compared with fleet renewal, routing changes and contrails avoidance?</li> <li>Do airlines respond to fuel mandates and emissions trading by changing networks, for example by refuelling outside the covered area?</li> <li>Do the scheduling and routing models that minimize fuel also minimize climate impact once contrails are included?</li> </ul> <h2>Data that could help</h2> <p>Flight-level data (ADS-B traces, schedules), fuel prices and mandate schedules, emissions trading data for aviation, and contrail models.</p> <h2>Next step</h2> <p>Estimate the cost per tonne avoided of the 2025 EU sustainable aviation fuel mandate from fuel price differentials and compare it with the cost of contrail avoidance on the same routes.</p> https://scholar4ai.com/ideas/autonomous-driving-liability 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Autonomous driving and autonomous vehicles appear on 314 recent papers, 270 of them in AI journals about perception and control, and only 29 carry a liability, insurance, regulation, acceptance or ethics keyword, with liability on 2. Driverless cars are now on public roads. When one crashes, who pays, and does the answer shape what gets deployed?

<h2>What prompted this</h2> <p>Since October 2025, 314 papers carry an autonomous-driving, autonomous-vehicle, self-driving or automated-vehicle keyword (270 in AI journals, 31 in operations research): "autonomous driving" on 152, "autonomous vehicles" on 64, "autonomous driving perception" on 45, "connected and automated vehicles" on 15, "autonomous driving safety" on 10. Only 29 of the 314 also carry a liability, insurance, accident, regulation, acceptance, trust, ethics, law or policy keyword, and "liability" appears on 2 papers.</p> <h2>Questions</h2> <ul> <li>In crashes involving driverless vehicles so far, how was fault assigned between the operator, the manufacturer and other road users, and how did insurers settle?</li> <li>Do strict manufacturer liability regimes slow deployment or improve safety, compared with regimes that treat the vehicle like a driver?</li> <li>How do crash rates per mile of driverless fleets compare with human drivers on the same roads, once reporting differences are corrected?</li> </ul> <h2>Data that could help</h2> <p>Crash reports filed with regulators (the US standing general order data, state DMV reports), court and settlement records, and fleet mileage disclosures.</p> <h2>Next step</h2> <p>Compile the crash reports of driverless fleets since 2021 with fault determinations and compare injury rates per mile with human-driven ride-hailing in the same cities.</p> https://scholar4ai.com/ideas/audit-market-concentration 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The Big 4 auditors appear on 32 recent papers in the finance and accounting journals, but only 9 of 38 papers on the audit market carry a concentration, competition or fee keyword. Audit fees have risen sharply, and regulators worry about four firms auditing almost every large company. Does concentration raise fees, lower quality or both?

<h2>What prompted this</h2> <p>Since October 2025, 38 papers carry a Big 4 or audit-market keyword (18 in finance journals, 13 in accounting); "Big 4 auditors" appears on 32. Only 9 of the 38 also carry a concentration, competition, fee, choice or market-share keyword, and "audit market concentration" on a single paper.</p> <h2>Questions</h2> <ul> <li>In local audit markets where one Big 4 office dominates, are fees higher and are restatements more or less common?</li> <li>When a Big 4 firm exits a market segment or a country, what happens to the fees and the quality of the clients it leaves?</li> <li>Do joint audits or shared audits, as tried in France and proposed in the UK, change concentration or only cost?</li> </ul> <h2>Data that could help</h2> <p>Audit fees and auditor identities (Audit Analytics), office-level market shares, restatements and inspection findings.</p> <h2>Next step</h2> <p>Compute office-level audit market concentration for US metropolitan areas and relate it to fee growth and restatement rates for 2015–2025.</p> https://scholar4ai.com/ideas/auction-design-for-procurement 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Auction design, first- and second-price auctions and bidding behaviour appear on 114 recent papers, mostly in economics, and the topic is fading, with 16 papers in the last quarter against about 32 per quarter before. Procurement auctions, which buy public goods and services worth trillions, appear on 2 papers. Which formats get governments the best price, and do bidders collude in the rest?

<h2>What prompted this</h2> <p>Since October 2025, 114 papers carry an auction keyword (81 in economics journals, 12 in finance, 10 in operations research): "auction design" on 15, "first-price auctions" on 12, "second-price auctions" on 9, "auctions" on 7, "auction revenue" on 6. Thirty-five of the 114 also carry a procurement, combinatorial, reverse-auction, supplier, bidding, public or government keyword, but "procurement auctions" appears on 2 and "reverse auctions" and "combinatorial auctions" on 4 each. Sixteen auction papers were published in July–September 2026 against about 32 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In public procurement data, do scoring auctions that weigh quality beat lowest-price auctions on final cost after change orders and delays?</li> <li>Can bid patterns in procurement data detect collusion, and do the detected cartels match prosecuted ones?</li> <li>Do combinatorial formats, used for spectrum and transport contracts, deliver the efficiency gains claimed, net of their complexity?</li> </ul> <h2>Data that could help</h2> <p>Procurement portal data with bids, awards and contract outcomes, spectrum auction records, and cartel prosecution records.</p> <h2>Next step</h2> <p>Compare final costs and delays of road contracts awarded by lowest price and by scoring rules in one country's procurement data.</p> https://scholar4ai.com/ideas/assurance-of-sustainability-reports 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Sustainability reporting appears on 57 recent papers, but assurance of those reports on far fewer: "ESG assurance" on 6, "external assurance" on 3, and the topic is rising. A global assurance standard took effect in 2026. Does assurance make reported emissions more accurate, and who should provide it?

<h2>What prompted this</h2> <p>Since October 2025, 34 papers carry an assurance keyword, 12 of which also carry a sustainability, ESG, CSR, non-financial, carbon or climate keyword: "ESG assurance" appears on 6 papers and "external assurance" on 3. The quarter to September 2026 had 12 assurance papers against about 7 per quarter earlier. "Sustainability reporting" appears on 57 papers.</p> <h2>Questions</h2> <ul> <li>Do firms whose emissions are assured restate them less often, and are their figures closer to independent estimates?</li> <li>Does it matter whether the assurer is an audit firm or an engineering consultancy, and whether the engagement is limited or reasonable?</li> <li>When assurance becomes mandatory under the EU's reporting directive, do the gaps between firms' claims and their emissions shrink?</li> </ul> <h2>Data that could help</h2> <p>Assurance statements attached to sustainability reports, reported emissions (CDP, Refinitiv ESG), independent emission estimates, and restatements.</p> <h2>Next step</h2> <p>Compare year-to-year revisions of reported Scope 1 emissions between firms with and without assurance for 2019–2025.</p> https://scholar4ai.com/ideas/annotation-labour-behind-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Annotation cost reduction and reinforcement learning from human feedback appear on 44 recent papers in the AI journals, among 71 papers on annotation and human feedback, but only 8 of them carry a worker, labour, wage, platform or ethics keyword. Millions of people label data and rate model outputs, often through platforms in low-income countries. What are they paid, and how does the pay shape the data?

<h2>What prompted this</h2> <p>Since October 2025, 71 papers carry an annotation-cost, human-annotation, annotator, crowdwork, human-feedback, RLHF or data-labelling keyword (69 in AI journals): "annotation cost reduction" on 27, "reinforcement learning from human feedback" on 17, "human-in-the-loop annotation" on 7, "expert annotation" on 5, "data annotation" on 6. Only 8 of the 71 also carry a worker, labour, wage, crowdwork, gig, pay, employment, economics, incentive, platform, fairness or ethics keyword; "ghost workers" and "precarious employment" appear on one and two papers.</p> <h2>Questions</h2> <ul> <li>What do annotators and raters earn per hour on the main platforms by country and task, and how has pay changed as demand for preference ratings grew?</li> <li>Does pay level change label quality and the preferences that are encoded in models, for example toward speed over care?</li> <li>Where is annotation work moving, and do the new locations' labour laws cover it?</li> </ul> <h2>Data that could help</h2> <p>Platform task postings and pay rates, worker surveys, and experiments that vary pay and measure label quality.</p> <h2>Next step</h2> <p>Scrape posted pay rates for annotation and rating tasks on the main platforms over a year and relate them to task type and worker location.</p> https://scholar4ai.com/ideas/analyst-coverage-when-ai-writes-summaries 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Analyst attention is rising in the finance journals, and AI tools now summarize filings and earnings calls for anyone. Only 13 of 246 recent papers on analysts carry an AI keyword. Do analysts drop the small firms that machines now summarize, or use the time saved to dig deeper?

<h2>What prompted this</h2> <p>In the finance journals, "analyst attention" had 12 papers in July–September 2026 against about 7 expected (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 246 papers carry an analyst keyword such as "analyst coverage" (96 papers) or "analyst attention" (46); 160 of them are in finance journals and 41 in accounting. Only 13 carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>After brokers adopt AI tools, does each analyst cover more firms, and which firms gain or lose coverage?</li> <li>Do forecasts become more accurate, or more alike, as analysts start from the same AI summaries?</li> <li>Do investors still react to analyst reports when an AI summary of the same filing is free?</li> </ul> <h2>Data that could help</h2> <p>Analyst forecasts and coverage (I/B/E/S), the dates brokers adopted AI tools (from announcements and job postings), report texts and stock price reactions.</p> <h2>Next step</h2> <p>Track forecast dispersion and coverage per analyst around brokers' announced AI rollouts between 2022 and 2025.</p> https://scholar4ai.com/ideas/ambulance-dispatch-and-emergency-care 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Emergency medical services, ambulance dispatch and emergency departments appear on 28 recent papers across the operations research, AI and economics journals, with 8 in the last quarter against about 6 per quarter before. Response times are rising in many countries as hospitals block ambulances at the door. Which dispatch and triage changes save lives, and which only move the queue?

<h2>What prompted this</h2> <p>Since October 2025, 28 papers carry an emergency-department, emergency-medical, ambulance or emergency-care keyword (14 in operations research journals, 7 in AI, 6 in economics): "emergency medical services" on 13, "ambulance dispatch" on 6, "emergency department visits" on 3, "emergency department" on 3, "emergency department triage" on 2. The quarter to September 2026 had 8 such papers against about 6 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When ambulances wait outside full emergency departments, how much of the lost response time comes from handover delays rather than from dispatch, and does the dispatch optimization literature model it?</li> <li>Do AI triage tools reduce waiting for the sickest patients, or shift waiting to those the tool misjudges?</li> <li>Does diverting low-acuity calls to telehealth or community paramedics shorten response times for the rest?</li> </ul> <h2>Data that could help</h2> <p>Ambulance dispatch and handover logs, emergency department arrival and waiting time records, and patient outcome data.</p> <h2>Next step</h2> <p>Decompose ambulance response times in one region into dispatch, travel and handover components for 2019–2025 and relate the handover component to hospital occupancy.</p> https://scholar4ai.com/ideas/algorithmic-management-and-wellbeing 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Algorithmic management and algorithmic control appear on 21 recent papers, mostly in management, and the topic is fading, with 2 papers in the last quarter against about 6 per quarter before. Yet software now sets schedules, rates and tasks for warehouse, delivery and call-centre workers. What does it do to effort, health and the manager's job?

<h2>What prompted this</h2> <p>Since October 2025, 21 papers carry an algorithmic-management, algorithmic-control, electronic-monitoring, workplace-surveillance or employee-monitoring keyword (17 in management journals): "algorithmic management" on 9, "algorithmic control" on 6. Only 2 of them were published in July–September 2026 against about 6 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>When algorithms set targets and monitor pace, do injuries, errors and turnover rise, and does output per worker rise by enough to pay for them?</li> <li>Do workers game the metrics, and does the algorithm then tighten, in a cycle?</li> <li>What do human managers do once scheduling and evaluation are automated, and do workers trust them more or less?</li> </ul> <h2>Data that could help</h2> <p>Worker-level productivity and injury records from warehouses and delivery platforms, scheduling system logs, and surveys of monitored workers.</p> <h2>Next step</h2> <p>Compare injury and turnover rates at warehouses before and after the roll-out of algorithmic pace management, using firm records or public injury filings.</p> https://scholar4ai.com/ideas/algorithmic-hiring-bias 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Recruitment, job matching and hiring discrimination appear on 109 recent papers across the economics, management and AI journals, with 32 in the last quarter against about 25 per quarter before; 32 of them carry a bias, fairness, algorithm or AI keyword. Screening tools now read millions of applications, and New York requires bias audits of them. What do the audits find, and do they change outcomes?

<h2>What prompted this</h2> <p>Since October 2025, 109 papers carry a hiring, recruitment, résumé, applicant, candidate-screening, job-matching or talent-acquisition keyword (54 in economics journals, 24 in management, 13 in finance, 8 in AI): "recruitment" on 8, "job matching" on 8, "hiring discrimination" on 6, "hiring decisions" on 6, "talent acquisition" on 5. Thirty-two of the 109 also carry a bias, fairness, discrimination, algorithm, language-model, AI, automated or machine-learning keyword. The quarter to September 2026 had 32 such papers against about 25 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In the bias audits published under New York's law, what disparities were found, and did employers change or drop the tools that showed them?</li> <li>When language models rank résumés, do they discriminate by name, gap years or address in correspondence tests, and does prompting the model to be fair fix it?</li> <li>Do algorithmic screens change who applies, as applicants learn to write for the machine?</li> </ul> <h2>Data that could help</h2> <p>Published bias audit reports, correspondence experiments with language-model screeners, applicant tracking data from employers, and job posting text over time.</p> <h2>Next step</h2> <p>Collect the bias audit reports published under the New York law and code the disparities found and the employers' subsequent tool use.</p> https://scholar4ai.com/ideas/algorithm-aversion-in-the-llm-era 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Trust in AI appears on 76 recent papers, algorithm aversion on 10 and algorithm appreciation on 3, across the AI and management journals, with 28 papers in the last quarter against about 25 per quarter before. People once rejected algorithmic advice after seeing it err. Language models that explain themselves fluently may get too much trust instead. Which way does it go, and for whom?

<h2>What prompted this</h2> <p>Since October 2025, 106 papers carry a trust-in-AI, algorithm-aversion, algorithm-appreciation or AI-advice keyword (52 in AI journals, 37 in management): "trust in AI" on 76, "algorithm aversion" on 10, "public trust in AI" on 4, "algorithm appreciation" on 3. The quarter to September 2026 had 28 such papers against about 25 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>In the same forecasting task, do people follow a language model's advice more than a statistical model's, and does the difference survive after both make the same error?</li> <li>Does a fluent explanation increase reliance even when it is wrong, and can calibration cues (stated uncertainty) correct that?</li> <li>Do professionals (doctors, loan officers, analysts) show aversion or over-reliance in their own domain, compared with laypeople?</li> </ul> <h2>Data that could help</h2> <p>Incentivized experiments with advice from different sources, and field data from professional settings where model recommendations and human decisions are both logged.</p> <h2>Next step</h2> <p>Run an experiment in which participants receive identical forecasts framed as coming from a statistical model, a language model with explanations, or a human expert, and measure reliance after errors.</p> https://scholar4ai.com/ideas/air-pollution-and-cognition 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Air pollution and PM2.5 appear on 213 recent papers, but only 33 of them carry a cognition, productivity, decision-making, education, mental-health or worker keyword, and "cognitive performance" on 5. Pollution affects brains as well as lungs. How much do bad-air days cost in exam scores, work errors and bad investment decisions?

<h2>What prompted this</h2> <p>Since October 2025, 213 papers carry an air-pollution, PM2.5, particulate or air-quality keyword (127 in economics journals, 54 in AI): "air pollution" on 85, "PM2.5" on 41, "air quality" on 24. Only 33 of them also carry a cognition, test-score, productivity, decision-making, education, mental-health, crime, worker, student or labour keyword, and "cognitive performance" appears on 5 papers.</p> <h2>Questions</h2> <ul> <li>On high-pollution days, do students score lower on high-stakes exams, do workers make more errors, and do traders and analysts make worse calls?</li> <li>Are the effects larger for tasks that need attention and memory than for routine tasks?</li> <li>Does indoor filtration undo the effect, and what would that imply for the value of clean-air policy?</li> </ul> <h2>Data that could help</h2> <p>Daily pollution readings, exam results with dates and locations, call-centre or factory error logs, and trading records matched to the trader's location.</p> <h2>Next step</h2> <p>Match the dates and locations of a national exam to daily PM2.5 readings and estimate the score loss per unit of pollution.</p> https://scholar4ai.com/ideas/ai-written-crowdfunding-appeals 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Charitable giving and crowdfunding appear on 168 recent papers, with 44 in the last quarter against about 38 per quarter before, but only 23 of them carry an AI, text, narrative or image keyword. Campaign success depends on the story. If fundraisers now write with language models, do donors give more, and do they give to the right causes?

<h2>What prompted this</h2> <p>Since October 2025, 168 papers carry a crowdfunding, charitable-giving, donation, philanthropy or fundraising keyword (69 in economics journals, 50 in management, 32 in finance): "charitable giving" on 45, "crowdfunding" on 44, "equity crowdfunding" on 18, "crowdfunding campaign success" on 14. The quarter to September 2026 had 44 such papers against about 38 per quarter earlier. Only 23 of the 168 also carry an AI, text, image, narrative or language keyword.</p> <h2>Questions</h2> <ul> <li>Since 2023, has the language of crowdfunding appeals converged on model-like text, and did success rates change for campaigns that use it?</li> <li>Do donors give more to polished appeals, and does that shift money from causes with less-skilled writers?</li> <li>Do platforms that offer built-in writing assistants see more campaigns and lower average pledges?</li> </ul> <h2>Data that could help</h2> <p>Campaign text and outcomes from crowdfunding platforms with timestamps, AI-text detectors, and platform feature release dates.</p> <h2>Next step</h2> <p>Score the text of campaigns on one platform for 2021–2025 with an AI-text detector and relate the score to funding success before and after 2023.</p> https://scholar4ai.com/ideas/ai-weather-forecasts-and-their-economic-value 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Weather forecasting, precipitation forecasting and typhoons appear on 45 recent papers, mostly in AI journals, and only 9 carry an economic, energy, agricultural, insurance or market keyword. Machine learning models now beat the physics-based systems on many measures at a fraction of the cost. Who captures the value, and does it reach the farmers, utilities and insurers who need it?

<h2>What prompted this</h2> <p>Since October 2025, 45 papers carry a weather-forecasting, weather-prediction, precipitation-forecasting, numerical-weather or typhoon keyword (35 in AI journals, 7 in economics): "weather forecasting" on 21, "precipitation forecasting" on 8, "typhoons" on 5, "numerical weather prediction" on 3. Only 9 of the 45 also carry an economics, finance, energy, agriculture, insurance, commodity, value, market, cost or decision keyword.</p> <h2>Questions</h2> <ul> <li>For which decisions (electricity trading, crop planting, flood warnings, airline operations) does the accuracy gain of machine learning forecasts translate into money, and how much?</li> <li>If a few firms own the best forecasts, do they sell them, trade on them, or both, and does public forecasting lose funding?</li> <li>Do the new models forecast the extremes that matter for insurance and disaster response as well as they forecast the averages?</li> </ul> <h2>Data that could help</h2> <p>Public forecast archives from both physics-based and machine learning systems, decision records from utilities and farms, and market prices sensitive to weather.</p> <h2>Next step</h2> <p>Compare the value of physics-based and machine learning forecasts for day-ahead electricity trading in one market over a year, using realized prices.</p> https://scholar4ai.com/ideas/ai-washing-and-crash-risk 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Artificial intelligence and stock price crash risk are both rising in the finance journals, and "AI washing" has appeared as a keyword. When the AI story in disclosures runs ahead of actual AI investment, does the stock carry more crash risk?

<h2>What prompted this</h2> <p>In the finance journals, "artificial intelligence" had 31 papers in July–September 2026 against about 17 expected from its earlier share, and "stock price crash risk" had 27 against 17 (see the <a href="/categories/finance">finance dashboard</a>). Since October 2025, 119 papers carry a crash-risk keyword, and 10 of them also carry an AI keyword. The keyword "AI washing" appears on 11 papers (8 in finance journals, 3 in management).</p> <h2>Questions</h2> <ul> <li>Can an AI-washing score be built from AI talk in filings and calls relative to measurable AI inputs, such as AI job postings, AI patents and computing spending?</li> <li>Is a high score followed by more crash risk (negative return skewness, down-to-up volatility), as theories of hoarded bad news predict?</li> <li>Did the first regulatory actions against AI washing change how firms talk about AI?</li> </ul> <h2>Data that could help</h2> <ul> <li>Filing and call text (the corporate filings category of the <a href="/rankings/datasets">dataset ranking</a>), online job postings, patents and daily returns.</li> <li>Crash-risk measures from firm-specific weekly returns.</li> </ul> <h2>Next step</h2> <p>Count AI mentions in a sample of 10-K filings from 2020 to 2025 and compare them, firm by firm, with AI job postings.</p> https://scholar4ai.com/ideas/ai-shopping-assistants 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Purchase intention, purchase decisions and consumer choice appear on 178 recent papers, mostly in management and economics, but only 14 of them carry a generative-AI, conversational-AI or recommendation keyword. Shoppers are starting to let AI assistants search, compare and buy. Who gets recommended, and what happens to brands and prices when a model makes the choice?

<h2>What prompted this</h2> <p>Since October 2025, 178 papers carry a purchase-intention, purchase-decision, consumer-choice or online-shopping keyword (88 in management journals, 48 in economics, 24 in operations research): "purchase intention" on 48, "purchase decision" on 38, "consumer choice" on 37, "online shopping" on 19. Only 14 of them also carry a generative-AI, ChatGPT, language-model, AI-assistant, conversational or recommendation keyword, while "conversational AI" appears on 20 papers in the archive.</p> <h2>Questions</h2> <ul> <li>When an assistant picks products, how concentrated are its choices across brands compared with human shoppers, and do the choices favour the assistant's owner or advertisers?</li> <li>Do consumers who delegate to assistants pay less, buy more, or buy different things?</li> <li>How do sellers optimize for assistants rather than for human eyes, and does that change product descriptions and prices?</li> </ul> <h2>Data that could help</h2> <p>Logs of assistant-mediated shopping sessions, experiments comparing assistant and human choices for the same shopping tasks, and retailer price and description data over time.</p> <h2>Next step</h2> <p>Ask several shopping assistants to buy the same 100 everyday products and compare brand concentration and prices with what human panels chose.</p> https://scholar4ai.com/ideas/ai-patents 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Patents appear on 377 recent papers, and patent quality on 24, but only 32 of the patent papers carry an AI or software keyword, and "AI patents" on 14. AI patent filings have exploded, especially in China, while much AI progress is published as open code. What do AI patents protect, and do they predict anything?

<h2>What prompted this</h2> <p>Since October 2025, 377 papers carry a patent keyword (148 in economics journals, 101 in management, 98 in finance): "patent citations" on 33, "patents" on 31, "patent applications" on 29, "patent quality" on 24. Only 32 of the 377 also carry an AI, machine-learning, generative or software keyword, and "AI patents" appears on 14 papers.</p> <h2>Questions</h2> <ul> <li>Do AI patents get cited, licensed and litigated like other patents, or are they filed for signalling and subsidies?</li> <li>Do firms that patent AI release more or fewer open models and papers, and do the two forms of disclosure substitute for each other?</li> <li>Do AI patent counts predict firms' AI revenue or productivity any better than AI job postings do?</li> </ul> <h2>Data that could help</h2> <p>Patent records classified by technology (PATSTAT, PatentsView, CNIPA), citations and litigation, open-source releases and job postings.</p> <h2>Next step</h2> <p>Compare forward citations, renewals and litigation rates of AI patents with those of other software patents from the same firms and years.</p> https://scholar4ai.com/ideas/ai-liability-and-insurance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

AI regulation, trustworthy AI, responsible AI and the EU AI Act appear on 229 recent papers, mostly in AI and management journals, but only 12 of them carry an insurance, liability, compensation or damages keyword, and AI liability on 2. Rules now assign duties; they say little about who pays for harm. Can AI harms be insured, and would insurance discipline deployment?

<h2>What prompted this</h2> <p>Since October 2025, 229 papers carry an AI-regulation, AI-safety, EU AI Act, AI-alignment, responsible-AI or trustworthy-AI keyword (157 in AI journals, 53 in management): "AI regulation" on 53, "trustworthy AI" on 51, "responsible AI" on 46, "AI safety" on 28, "EU AI Act" on 25. Only 12 of the 229 also carry an insurance, liability, tort, compensation, legal-responsibility or damages keyword, and "AI liability" appears on 2 papers.</p> <h2>Questions</h2> <ul> <li>For harms already litigated (wrong credit decisions, biased hiring tools, chatbot misinformation, autonomous vehicle crashes), who paid, and how long did it take?</li> <li>Could insurers price AI liability given correlated failures across customers of the same model, and would premiums vary with testing and monitoring practices?</li> <li>Does the withdrawal of the EU's AI liability directive leave victims with workable claims under existing law?</li> </ul> <h2>Data that could help</h2> <p>Court records and settlements involving AI systems, insurance policy wordings and exclusions, and the texts of liability rules across jurisdictions.</p> <h2>Next step</h2> <p>Compile a database of litigated AI harms since 2018 with the outcome, the party that paid and the time to resolution.</p> https://scholar4ai.com/ideas/ai-in-peer-review 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Peer review, research productivity and academic writing appear on 56 recent papers across the economics and AI journals, with 17 in the last quarter against about 12 per quarter before, and 17 of them involve language models. Journals report AI-written reviews and AI-written submissions. How much of each is there, and does it change what gets published?

<h2>What prompted this</h2> <p>Since October 2025, 56 papers carry a peer-review, scientific-writing, research-productivity, academic-writing or scientific-discovery keyword (22 in economics journals, 20 in AI): "peer review" on 19, "research productivity" on 16, "academic writing" on 8, "scientific discovery" on 3. Seventeen of them also carry a language-model, generative-AI or AI keyword. The quarter to September 2026 had 17 such papers against about 12 per quarter earlier.</p> <h2>Questions</h2> <ul> <li>What share of referee reports at economics and finance journals since 2023 show the statistical signs of language-model drafting, and do such reports reach different recommendations?</li> <li>Do submissions written with AI assistance get accepted at different rates, and has submission volume risen faster than acceptance?</li> <li>Would AI screening for errors and missing references improve reviews, and would authors and editors accept it?</li> </ul> <h2>Data that could help</h2> <p>Referee report and submission archives from journals that permit research access, AI-text detectors, and editor surveys.</p> <h2>Next step</h2> <p>Run an AI-text detector on anonymized referee reports from a journal for 2021–2025 and relate flagged reports to their recommendations and length.</p> https://scholar4ai.com/ideas/ai-governance-meets-corporate-governance 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

The keyword "AI governance" is growing fastest across the tracked journals, while corporate governance keeps rising in economics and management. Is board-level oversight of AI turning into a research stream of its own?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "AI governance" gained more papers than any other keyword over the last three months, mostly in AI and management journals. At the same time "corporate governance" is climbing in economics. The two literatures rarely cite each other's mechanisms.</p> <h2>Questions</h2> <ul> <li>Do firms that put AI oversight on the board agenda take different risks, or get different audit outcomes?</li> <li>Which corporate-governance tools (independent directors, committees, disclosure rules) carry over to AI systems, and which break?</li> <li>Is AI-governance research led by different institutions and countries than corporate-governance research?</li> </ul> <h2>Data that could help</h2> <ul> <li>Papers that carry both keywords in <a href="/data/papers.json">papers.json</a>, to map the overlap.</li> <li>Annual-report disclosures of AI committees, hand-collected or text-mined.</li> </ul> <h2>Next step</h2> <p>Read the ten most recent papers that carry both keywords and note which governance mechanisms they test.</p> https://scholar4ai.com/ideas/ai-generated-advertising 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Advertising appears on 15 recent papers, online advertising on 9 and AI-generated advertising on 3, in the management and economics journals, while AI-generated content appears on 33. Brands now generate images, copy and whole campaigns with models. Does disclosure of AI use change how consumers respond, and does generated advertising work?

<h2>What prompted this</h2> <p>Since October 2025, 99 papers carry an advertising keyword (40 in management journals, 25 in economics), including "advertising" on 15, "online advertising" on 9, "advertising expenditure" on 10 and "AI-generated advertising" on 3. Only 14 of the 99 also carry an AI-generated, generative, synthetic, AI or algorithm keyword, while "AI-generated content" appears on 33 papers, mostly in AI journals about detection.</p> <h2>Questions</h2> <ul> <li>In field experiments, do AI-generated advertisements get the same click and conversion rates as agency-made ones, and does the gap depend on product type?</li> <li>When platforms label AI-generated advertisements, as some now do, does the label reduce trust and purchases, and for which brands?</li> <li>Do cheaper creative costs lead firms to run more variants and more total advertising, and does that change competition for attention?</li> </ul> <h2>Data that could help</h2> <p>Advertising platform experiments with creative-level performance, labelling policy dates, and advertising spending data by firm.</p> <h2>Next step</h2> <p>Run a field experiment on an advertising platform comparing labelled and unlabelled AI-generated creatives with human-made ones for the same products.</p> https://scholar4ai.com/ideas/ai-and-the-offshoring-of-services 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Outsourcing, offshoring and digital trade appear on 104 recent papers, but only one of them carries an AI, remote-work or online-labour keyword. If language models make call centres, coding and back-office work cheaper to automate than to offshore, the countries that built service exports on cheap labour face a new competitor. Is service offshoring already slowing?

<h2>What prompted this</h2> <p>Since October 2025, 104 papers carry an outsourcing, offshoring, digital-trade or trade-in-services keyword (44 in economics journals, 23 in finance, 22 in operations research): "outsourcing" on 29, "digital trade" on 18, "offshoring" on 13, "trade in services" on 12. Only one of the 104 also carries an AI, remote-work, telework, freelancing or online-labour keyword.</p> <h2>Questions</h2> <ul> <li>Have service exports from India and the Philippines slowed in the tasks most exposed to language models (customer support, transcription, basic coding) relative to other tasks?</li> <li>Do firms replace offshore contracts with AI, or do offshore providers adopt AI and cut prices?</li> <li>Does the shift change who in the exporting countries loses, by skill and gender?</li> </ul> <h2>Data that could help</h2> <p>Service export statistics by category, online freelancing platform data on jobs posted and prices, and firms' outsourcing contracts and job postings.</p> <h2>Next step</h2> <p>Compare the growth of online freelance work in AI-exposed and non-exposed task categories since 2022 using platform data.</p> https://scholar4ai.com/ideas/ai-and-data-on-the-balance-sheet 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Firms spend heavily on data, software and AI models, but much of that spending is expensed and barely shows on balance sheets. Intangible assets are a steady finance topic, yet only one of 57 recent papers on them carries an AI, data or software keyword. How large is the missing AI capital, and does it distort book-based measures?

<h2>What prompted this</h2> <p>Since October 2025, 57 papers carry an intangible-assets, intangible-capital or intangible-investment keyword (30 in finance journals, 11 in economics, 10 in management). Only one of them also carries an AI, data-asset, software or digital-asset keyword, while AI keeps rising in almost every category on the <a href="/trends">trends page</a>.</p> <h2>Questions</h2> <ul> <li>How much do listed firms spend on AI and data each year, and how much would a capitalized "AI capital" add to their assets?</li> <li>Does adding it change book-to-market sorts, the value premium or measured returns on capital?</li> <li>Where the rules allow capitalization, are firms that capitalize more of it valued differently?</li> </ul> <h2>Data that could help</h2> <ul> <li>R&amp;D, software and overhead expenses (Compustat), and AI job postings as a proxy for AI labour spending.</li> <li>China's accounting rules for enterprise data resources, in force since 2024, which let some firms report data as assets.</li> </ul> <h2>Next step</h2> <p>Estimate AI labour spending of listed firms from job postings and salaries, capitalize it with a perpetual-inventory method, and recompute book-to-market.</p> https://scholar4ai.com/ideas/ai-agents-and-the-boundaries-of-the-firm 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Transaction costs are back as a growing keyword in the management journals, while agentic AI and language-model agents rise in AI. Coase argued that firms exist because using markets is costly. If AI agents can search, negotiate and monitor contracts cheaply, do firms buy more and make less?

<h2>What prompted this</h2> <p>In the management journals, "transaction costs" had 6 papers in July–September 2026 against about 1 expected (see the <a href="/categories/management">management dashboard</a>); in the AI journals, "agentic AI" (11 against 5) and "LLM agents" (14 against 8) are growing. Since October 2025, 170 papers carry a transaction-cost keyword (65 in finance journals, 49 in economics, 37 in management), and only 7 of them carry an AI keyword. "Agentic AI" appears on 41 papers, 27 in AI journals and 11 in management.</p> <h2>Questions</h2> <ul> <li>Which transaction costs do AI agents lower most (search, bargaining or monitoring), and which do they raise (verification, security)?</li> <li>Do firms that adopt AI tools for procurement and contracting outsource more and integrate less?</li> <li>Does the change favour small firms, which can now buy what they could not organize, or large firms with more data?</li> </ul> <h2>Data that could help</h2> <p>Firm-level purchases of goods and services (procurement records, firm-level input data), AI adoption measures (job postings, software use) and supplier counts.</p> <h2>Next step</h2> <p>Track the number and type of suppliers per firm before and after adoption of AI procurement software, for firms whose adoption dates vendors have published.</p> https://scholar4ai.com/ideas/ageing-populations-and-asset-prices 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Overlapping-generations models are a rising tool in the economics journals, and population aging is a large topic, but few papers connect aging to asset prices. Of 239 recent papers with an aging or demographic keyword, 17 carry a keyword about stocks, interest rates or house prices. As Japan, Korea, China and Europe age, will returns and interest rates fall?

<h2>What prompted this</h2> <p>In economics, "overlapping generations model" had 14 papers in July–September 2026 against about 9 expected (see the <a href="/categories/economics">economics dashboard</a>). Since October 2025, "population aging" appears on 114 papers (61 in economics journals, 35 in finance). Of the 239 papers with a population-aging or demographic keyword, only 17 also carry a keyword about stocks, asset pricing, interest rates, the equity premium or house prices.</p> <h2>Questions</h2> <ul> <li>Do countries whose working-age share falls faster see lower real interest rates and equity valuations, at the same growth rate?</li> <li>Do retirees sell their assets as life-cycle models predict, or hold on to them, as many Japanese households do?</li> <li>How much do pension reforms that raise saving or the retirement age offset the effect?</li> </ul> <h2>Data that could help</h2> <p>Population projections by age (UN), real interest rates and equity valuations by country, and household portfolio surveys such as the Survey of Consumer Finances, SHARE or the China Household Finance Survey.</p> <h2>Next step</h2> <p>Regress long-run real interest rates and price-to-earnings ratios on the ratio of middle-aged to old people across OECD countries for 1970–2025.</p> https://scholar4ai.com/ideas/accounting-for-crypto-at-fair-value 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Cryptocurrency and digital assets appear on 476 recent papers, but only 6 of them carry an accounting, fair-value, reporting, disclosure or audit keyword. Since 2025, US firms report crypto holdings at fair value through earnings, and some firms now hold Bitcoin as their main asset. What does fair value accounting do to earnings, and does it change who holds crypto?

<h2>What prompted this</h2> <p>Since October 2025, 476 papers carry a crypto, digital-asset or Bitcoin keyword (299 in finance journals, 81 in economics, 65 in AI); "digital assets" appears on 32. Only 6 of the 476 also carry an accounting, fair-value, financial-reporting, disclosure or audit keyword, and "fair value accounting" appears on 4 papers.</p> <h2>Questions</h2> <ul> <li>After the switch from impairment-only to fair value, did firms' earnings become dominated by crypto swings, and did analysts start excluding them?</li> <li>Did the rule change encourage more firms to hold crypto, as treasury-strategy firms claim, or did it mostly affect the few that already did?</li> <li>How do auditors verify ownership and valuation of crypto held through custodians, and how often do they qualify opinions?</li> </ul> <h2>Data that could help</h2> <p>Financial statements of crypto-holding firms, their footnotes and audit reports, and earnings call transcripts.</p> <h2>Next step</h2> <p>Identify every US listed firm reporting crypto assets since 2023 and decompose their earnings volatility into crypto and non-crypto components.</p> https://scholar4ai.com/ideas/absorptive-capacity-and-ai 2026-10-23T00:00:00+09:00 2026-10-23T00:00:00+09:00

Absorptive capacity tops the list of growing keywords across the tracked journals, while AI adoption keeps rising. Do firms need prior know-how to profit from AI, so that AI widens the gap between leaders and laggards?

<h2>What prompted this</h2> <p>On the <a href="/trends">trends page</a>, "absorptive capacity" tops the growing keywords across all journals: 44 papers in July–September 2026 against 22 expected from its earlier share, with growth in both management (21 against 9) and economics (17 against 7). In economics, "artificial intelligence" (26 against 17) and "total factor productivity" (33 against 21) are growing too. Since October 2025, 123 papers carry an absorptive-capacity keyword, but only 11 of them also carry an AI keyword.</p> <h2>Questions</h2> <ul> <li>Do firms with more R&amp;D, data infrastructure or skilled staff get larger productivity gains from adopting AI?</li> <li>Is the binding constraint people (hiring and training) or complementary assets (data and processes)?</li> <li>Can subsidies or training programmes narrow the gap?</li> </ul> <h2>Data that could help</h2> <ul> <li>AI adoption from online job postings, and total factor productivity estimated from firm accounts (see the <a href="/rankings/equations">equation ranking</a>).</li> <li>R&amp;D spending and patents as measures of absorptive capacity before adoption.</li> </ul> <h2>Next step</h2> <p>Merge AI job postings with firm-level productivity for listed firms and test whether the gain from AI hiring rises with R&amp;D intensity before adoption.</p> https://scholar4ai.com/ideas/cost-curve-of-drone-delivery 2026-10-22T00:00:00+09:00 2026-10-22T00:00:00+09:00

Drone delivery is one of the fastest-rising topics in the operations research journals. Most papers optimise routes; few ask who pays, and how drones compare with vans as delivery density changes.

<h2>What prompted this</h2> <p>"Drone delivery" climbs among the <a href="/categories/operations-research">operations research keywords</a> this year, next to "supply chain coordination" and "price competition". The economics journals barely mention it.</p> <h2>Questions</h2> <ul> <li>At what delivery density does a drone fleet beat vans on cost per parcel, once batteries, staff and regulation are priced in?</li> <li>Who captures the savings: platforms, retailers or customers?</li> <li>Do the routing models in the literature assume demand that real cities actually have?</li> </ul> <h2>Possible approach</h2> <p>Combine published routing models with public data on parcel volumes per district, and compare cost curves under a few regulatory scenarios.</p> https://scholar4ai.com/ideas/ai-trading-and-price-discovery 2026-10-21T00:00:00+09:00 2026-10-21T00:00:00+09:00

In the finance journals both "artificial intelligence" and "price discovery" gained papers this year, but rarely in the same paper. Do markets with more AI-assisted trading take in news faster, or just louder?

<h2>What prompted this</h2> <p>The <a href="/categories/finance">finance dashboard</a> shows "artificial intelligence" and "price discovery" among the keywords growing fastest over the year. The papers that study AI in finance mostly look at prediction; the price-discovery papers mostly look at market structure.</p> <h2>Questions</h2> <ul> <li>Around earnings announcements, do stocks with more AI-driven trading reach their new price faster?</li> <li>Does that speed come with more overshooting and reversal?</li> <li>Did anything change after large language models became widely used for reading news?</li> </ul> <h2>Data that could help</h2> <ul> <li>Intraday prices around scheduled announcements.</li> <li>A proxy for AI-assisted trading, for example fund disclosures or order-flow patterns.</li> </ul>