SoFaiR: Single Shot Fair Representation Learning

TL;DR

This paper introduces SoFaiR, a novel single-shot fair representation learning method that efficiently generates multiple fairness-information trade-offs from one model, based on rate-distortion theory, and explains the impact of fairness adjustments.

Contribution

Proposes SoFaiR, the first method to produce multiple fairness-information trade-offs with a single trained model, grounded in rate-distortion theory, and provides interpretability of fairness effects.

Findings

Achieves comparable fairness-information trade-offs to multi-shot methods

Reduces computational cost by generating multiple points with one model

Provides insights into how fairness adjustments affect information content

Abstract

To avoid discriminatory uses of their data, organizations can learn to map them into a representation that filters out information related to sensitive attributes. However, all existing methods in fair representation learning generate a fairness-information trade-off. To achieve different points on the fairness-information plane, one must train different models. In this paper, we first demonstrate that fairness-information trade-offs are fully characterized by rate-distortion trade-offs. Then, we use this key result and propose SoFaiR, a single shot fair representation learning method that generates with one trained model many points on the fairness-information plane. Besides its computational saving, our single-shot approach is, to the extent of our knowledge, the first fair representation learning method that explains what information is affected by changes in the fairness / distortion properties of the representation. Empirically, we find on three datasets that SoFaiR achieves similar fairness-information trade-offs as its multi-shot counterparts.