Learning Controllable Fair Representations

TL;DR

This paper introduces a novel information-theoretic framework for learning fair data representations that are highly expressive and controllably fair, enabling better trade-offs between fairness and utility.

Contribution

It proposes a new objective function for fair representation learning that allows explicit control over fairness levels and balances expressiveness and fairness using duality.

Findings

Achieves higher expressiveness with lower computational cost

Balances multiple fairness notions effectively

Outperforms existing methods in fairness-utility trade-offs

Abstract

Learning data representations that are transferable and are fair with respect to certain protected attributes is crucial to reducing unfair decisions while preserving the utility of the data. We propose an information-theoretically motivated objective for learning maximally expressive representations subject to fairness constraints. We demonstrate that a range of existing approaches optimize approximations to the Lagrangian dual of our objective. In contrast to these existing approaches, our objective allows the user to control the fairness of the representations by specifying limits on unfairness. Exploiting duality, we introduce a method that optimizes the model parameters as well as the expressiveness-fairness trade-off. Empirical evidence suggests that our proposed method can balance the trade-off between multiple notions of fairness and achieves higher expressiveness at a lower computational cost.