Promoting Fairness through Hyperparameter Optimization

TL;DR

This paper introduces fairness-aware hyperparameter optimization methods that improve model fairness significantly with minimal impact on predictive performance, facilitating easier real-world adoption.

Contribution

It proposes and evaluates fairness-aware variants of popular hyperparameter optimization algorithms, demonstrating their effectiveness on real-world and benchmark datasets.

Findings

111% increase in mean fairness

6% decrease in performance

Feasible without extra training cost

Abstract

Considerable research effort has been guided towards algorithmic fairness but real-world adoption of bias reduction techniques is still scarce. Existing methods are either metric- or model-specific, require access to sensitive attributes at inference time, or carry high development or deployment costs. This work explores the unfairness that emerges when optimizing ML models solely for predictive performance, and how to mitigate it with a simple and easily deployed intervention: fairness-aware hyperparameter optimization (HO). We propose and evaluate fairness-aware variants of three popular HO algorithms: Fair Random Search, Fair TPE, and Fairband. We validate our approach on a real-world bank account opening fraud case-study, as well as on three datasets from the fairness literature. Results show that, without extra training cost, it is feasible to find models with 111% mean fairness increase and just 6% decrease in performance when compared with fairness-blind HO.