Fair Bayesian Data Selection via Generalized Discrepancy Measures

2025-11-10

AI Summary

Existing fairness interventions for machine learning models in high-stakes settings predominantly operate at the model level, suffering from high computational overhead and poor generalization. This paper proposes a Bayesian data selection framework that, for the first time, incorporates generalized distributional discrepancy measures—such as Wasserstein distance, maximum mean discrepancy (MMD), and f-divergences—into Bayesian data weighting. By aligning the parameter-weight posterior distributions across demographic groups to a shared central distribution, the method mitigates bias at the data source without imposing explicit fairness constraints. It is theoretically grounded and geometrically aware. Extensive experiments on multiple benchmark datasets demonstrate that our approach significantly outperforms state-of-the-art data-level and model-level fairness methods in both fairness metrics—including equal opportunity difference and statistical parity—and predictive accuracy.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Philosophy and Ethics of AI: Bias, Fairness & Equity
• Computer Vision: Bias, Fairness & Privacy

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Fairness, privacy, and diversity in economic environments
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

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.

Problem

• Addresses fairness concerns in machine learning model deployment
• Mitigates group-specific biases in training data selection
• Ensures fairness without explicit constraints using Bayesian framework

Innovation

• Bayesian data selection framework ensures fairness
• Aligns group-specific posteriors with central distribution
• Uses flexible discrepancy measures for geometry-aware control