On the Maximal Local Disparity of Fairness-Aware Classifiers

Fairness has become a crucial aspect in the development of trustworthy machine learning algorithms. Current fairness metrics to measure the violation of demographic parity have the following drawbacks: (i) theaverage differenceof model predictions on two groups cannot reflect theirdistribution disparity, and (ii) theoverallcalculation along all possible predictions conceals theextreme local disparityat or around certain predictions. In this work, we propose a novel fairness metric calledMaximalCumulative ratioDisparity along varyingPredictions' neighborhood (MCDP), for measuring the maximal local disparity of the fairness-aware classifiers. To accurately and efficiently calculate the MCDP, we develop a provably exact and an approximate calculation algorithm that greatly reduces the computational complexity with low estimation error. We further propose a bi-level optimization algorithm using a differentiable approximation of the MCDP for improving the algorithmic fairness. Extensive experiments on both tabular and image datasets validate that our fair training algorithm can achieve superior fairness-accuracy trade-offs.