As they take a crucial role in social decision makings, AI algorithms based on ML models should be not only accurate but also fair. Among many algorithms for fair AI, learning a prediction ML model by minimizing the empirical risk (e.g., cross-entropy) subject to a given fairness constraint has received much attention. To avoid computational difficulty, however, a given fairness constraint is replaced by a surrogate fairness constraint as the 0-1 loss is replaced by a convex surrogate loss for classification problems. In this paper, we investigate the validity of existing surrogate fairness constraints and propose a new surrogate fairness constraint called SLIDE, which is computationally feasible and asymptotically valid in the sense that the learned model satisfies the fairness constraint asymptotically and achieves a fast convergence rate. Numerical experiments confirm that the SLIDE works well for various benchmark datasets.
Keywords: Classification; Fairness AI; Learning theory; Machine learning; Supervised learning.
Copyright © 2022 Elsevier Ltd. All rights reserved.