Learning Optimal Individualized Decision Rules with Conditional Demographic Parity
This work is significant for researchers and practitioners developing individualized decision rules, particularly in sensitive societal applications, by offering a method to mitigate discriminatory outcomes for minority subgroups.
This paper addresses the problem of discriminatory effects in individualized decision rules (IDRs) by incorporating demographic parity (DP) and conditional demographic parity (CDP) constraints. The authors show that optimal IDRs under these constraints can be efficiently obtained by perturbing unconstrained optimal IDRs, and they derive convergence rates for policy value and fairness constraint terms.
Individualized decision rules (IDRs) have become increasingly prevalent in societal applications such as personalized marketing, healthcare, and public policy design. However, a critical ethical concern arises from the potential discriminatory effects of IDRs trained on biased data. These algorithms may disproportionately harm individuals from minority subgroups defined by sensitive attributes like gender, race, or language. To address this issue, we propose a novel framework that incorporates demographic parity (DP) and conditional demographic parity (CDP) constraints into the estimation of optimal IDRs. We show that the theoretically optimal IDRs under DP and CDP constraints can be obtained by applying perturbations to the unconstrained optimal IDRs, enabling a computationally efficient solution. Theoretically, we derive convergence rates for both policy value and the fairness constraint term. The effectiveness of our methods is illustrated through comprehensive simulation studies and an empirical application to the Oregon Health Insurance Experiment.