Fairness without Demographics through Adversarially Reweighted Learning

TL;DR

This paper introduces Adversarially Reweighted Learning (ARL), a novel method to enhance fairness in machine learning models without requiring protected demographic data, by leveraging non-protected features and adversarial training.

Contribution

The paper proposes ARL, a new adversarial reweighting approach that improves fairness without protected features, outperforming existing methods across multiple datasets.

Findings

ARL improves Rawlsian Max-Min fairness.

ARL achieves higher AUC for worst-case protected groups.

ARL outperforms state-of-the-art fairness methods.

Abstract

Much of the previous machine learning (ML) fairness literature assumes that protected features such as race and sex are present in the dataset, and relies upon them to mitigate fairness concerns. However, in practice factors like privacy and regulation often preclude the collection of protected features, or their use for training or inference, severely limiting the applicability of traditional fairness research. Therefore we ask: How can we train an ML model to improve fairness when we do not even know the protected group memberships? In this work we address this problem by proposing Adversarially Reweighted Learning (ARL). In particular, we hypothesize that non-protected features and task labels are valuable for identifying fairness issues, and can be used to co-train an adversarial reweighting approach for improving fairness. Our results show that {ARL} improves Rawlsian Max-Min fairness, with notable AUC improvements for worst-case protected groups in multiple datasets, outperforming state-of-the-art alternatives.