Fair Bayesian Data Selection via Generalized Discrepancy Measures

TL;DR

This paper introduces a Bayesian data selection method that enhances fairness in machine learning by aligning group-specific distributions with a central distribution using flexible discrepancy measures, improving fairness and accuracy.

Contribution

It proposes a novel, scalable, data-centric fairness framework using generalized discrepancy measures for aligning distributions without explicit fairness constraints.

Findings

Outperforms existing fairness methods in accuracy and fairness metrics.

Supports flexible distribution alignment via Wasserstein, MMD, and $f$-divergence.

Provides theoretical guarantees for fairness improvements.

Abstract

Fairness concerns are increasingly critical as machine learning models are deployed in high-stakes applications. While existing fairness-aware methods typically intervene at the model level, they often suffer from high computational costs, limited scalability, and poor generalization. To address these challenges, we propose a Bayesian data selection framework that ensures fairness by aligning group-specific posterior distributions of model parameters and sample weights with a shared central distribution. Our framework supports flexible alignment via various distributional discrepancy measures, including Wasserstein distance, maximum mean discrepancy, and $f$-divergence, allowing geometry-aware control without imposing explicit fairness constraints. This data-centric approach mitigates group-specific biases in training data and improves fairness in downstream tasks, with theoretical guarantees. Experiments on benchmark datasets show that our method consistently outperforms existing data selection and model-based fairness methods in both fairness and accuracy.