Fairness-Aware Grouping for Continuous Sensitive Variables: Application for Debiasing Face Analysis with respect to Skin Tone

TL;DR

This paper introduces a fairness-aware grouping method for continuous sensitive attributes like skin tone, enabling nuanced discrimination analysis and debiasing in face analysis models, with stable results across datasets.

Contribution

It proposes a novel grouping approach based on discrimination levels for continuous attributes, improving fairness analysis and debiasing in face analysis applications.

Findings

Uncovers nuanced discrimination patterns in face datasets.

Demonstrates robustness of the method across datasets.

Achieves fairness improvements with minimal accuracy loss.

Abstract

Within a legal framework, fairness in datasets and models is typically assessed by dividing observations into predefined groups and then computing fairness measures (e.g., Disparate Impact or Equality of Odds with respect to gender). However, when sensitive attributes such as skin color are continuous, dividing into default groups may overlook or obscure the discrimination experienced by certain minority subpopulations. To address this limitation, we propose a fairness-based grouping approach for continuous (possibly multidimensional) sensitive attributes. By grouping data according to observed levels of discrimination, our method identifies the partition that maximizes a novel criterion based on inter-group variance in discrimination, thereby isolating the most critical subgroups. We validate the proposed approach using multiple synthetic datasets and demonstrate its robustness under changing population distributions - revealing how discrimination is manifested within the space of sensitive attributes. Furthermore, we examine a specialized setting of monotonic fairness for the case of skin color. Our empirical results on both CelebA and FFHQ, leveraging the skin tone as predicted by an industrial proprietary algorithm, show that the proposed segmentation uncovers more nuanced patterns of discrimination than previously reported, and that these findings remain stable across datasets for a given model. Finally, we leverage our grouping model for debiasing purpose, aiming at predicting fair scores with group-by-group post-processing. The results demonstrate that our approach improves fairness while having minimal impact on accuracy, thus confirming our partition method and opening the door for industrial deployment.