Penalizing Unfairness in Binary Classification

TL;DR

This paper introduces a new method to reduce unfairness in binary classifiers by balancing false positive and false negative rates across two populations, demonstrated through empirical evaluation on real-world datasets.

Contribution

It proposes a novel fairness mitigation approach specifically targeting false positive and false negative rate parity in binary classification.

Findings

Achieves comparable false positive and false negative rates across populations

Maintains high classification accuracy while reducing unfairness

Effective on datasets from criminal risk, credit, lending, and college admissions

Abstract

We present a new approach for mitigating unfairness in learned classifiers. In particular, we focus on binary classification tasks over individuals from two populations, where, as our criterion for fairness, we wish to achieve similar false positive rates in both populations, and similar false negative rates in both populations. As a proof of concept, we implement our approach and empirically evaluate its ability to achieve both fairness and accuracy, using datasets from the fields of criminal risk assessment, credit, lending, and college admissions.