Empirical Risk Minimization under Fairness Constraints

TL;DR

This paper introduces a fairness-aware empirical risk minimization framework that enforces fairness constraints, ensuring classifiers are less biased by sensitive variables, with theoretical guarantees and practical effectiveness demonstrated through experiments.

Contribution

It proposes a novel fairness constraint integrated into empirical risk minimization, applicable to kernel and linear models, with theoretical bounds and empirical validation.

Findings

The method effectively enforces fairness constraints.

It achieves competitive performance against existing approaches.

The approach is adaptable to kernel and linear models.

Abstract

We address the problem of algorithmic fairness: ensuring that sensitive variables do not unfairly influence the outcome of a classifier. We present an approach based on empirical risk minimization, which incorporates a fairness constraint into the learning problem. It encourages the conditional risk of the learned classifier to be approximately constant with respect to the sensitive variable. We derive both risk and fairness bounds that support the statistical consistency of our approach. We specify our approach to kernel methods and observe that the fairness requirement implies an orthogonality constraint which can be easily added to these methods. We further observe that for linear models the constraint translates into a simple data preprocessing step. Experiments indicate that the method is empirically effective and performs favorably against state-of-the-art approaches.