Fairness constraints and what they cost
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?
What prompted this
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.
Questions
- 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?
- 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?
- Which fairness definitions survive contact with operational data, where groups overlap and needs differ?
Data that could help
The instances and results of recent fairness-constrained models, and surveys of affected groups in a real deployment.
Next step
Re-solve ten published fairness-constrained models across the full range of the fairness parameter and plot the price of fairness for each.