Learning Adversarially Fair and Transferable Representations

TL;DR

This paper proposes adversarial representation learning to promote fairness in predictions, providing theoretical guarantees and demonstrating effective fair transfer learning with maintained utility across tasks.

Contribution

It introduces a novel adversarial approach to fair representation learning, linking fairness objectives to adversarial goals and validating transferability and utility in experiments.

Findings

Adversarial objectives are crucial for fairness guarantees.

Learned representations enable fair predictions on new tasks.

Empirical results show maintained utility in transfer learning.

Abstract

In this paper, we advocate for representation learning as the key to mitigating unfair prediction outcomes downstream. Motivated by a scenario where learned representations are used by third parties with unknown objectives, we propose and explore adversarial representation learning as a natural method of ensuring those parties act fairly. We connect group fairness (demographic parity, equalized odds, and equal opportunity) to different adversarial objectives. Through worst-case theoretical guarantees and experimental validation, we show that the choice of this objective is crucial to fair prediction. Furthermore, we present the first in-depth experimental demonstration of fair transfer learning and demonstrate empirically that our learned representations admit fair predictions on new tasks while maintaining utility, an essential goal of fair representation learning.