Selection of the Best Policy under Fairness Constraints
Submitted to Operations Research, 2026
This paper formalizes the Selection of the Best with Fairness Constraints (SBFC) problem: identify the policy with the highest average performance among those that meet a minimum performance threshold in every pre-specified subpopulation. We establish an instance-specific lower bound on sample complexity and develop a Track-and-Stop with Constraints on Subpopulation (T-a-S-CS) algorithm that achieves it asymptotically. Numerical experiments and a case study using the International Stroke Trial demonstrate substantial efficiency gains over policy-level allocation baselines.
Authors: Tingyu Zhu, Yuhang Wu, Zeyu Zheng