M$^2$FGB: A Min-Max Gradient Boosting Framework for Subgroup Fairness

2025-04-16

AI Summary

This work addresses subgroup unfairness in supervised learning arising from protected attributes such as gender and race. We propose a novel gradient boosting framework that, for the first time, incorporates a min-max fairness objective directly into each boosting iteration. The resulting primal-dual boosting algorithm jointly optimizes overall predictive accuracy and worst-subgroup fairness, with provable convergence guarantees. Our framework unifies multiple subgroup fairness definitions—including equal opportunity—via fairness-aware regularization terms that enforce fairness constraints. Theoretical analysis establishes convergence under standard assumptions. Empirical evaluation across diverse binary classification and regression benchmarks demonstrates substantial improvements in worst-subgroup fairness (e.g., a 37% reduction in equal opportunity difference) while preserving predictive accuracy—no statistically significant degradation is observed.

Technology Category

• Machine Learning: Ethics, Bias, and Fairness
• Computer Vision: Bias, Fairness & Privacy
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

In recent years, fairness in machine learning has emerged as a critical concern to ensure that developed and deployed predictive models do not have disadvantageous predictions for marginalized groups. It is essential to mitigate discrimination against individuals based on protected attributes such as gender and race. In this work, we consider applying subgroup justice concepts to gradient-boosting machines designed for supervised learning problems. Our approach expanded gradient-boosting methodologies to explore a broader range of objective functions, which combines conventional losses such as the ones from classification and regression and a min-max fairness term. We study relevant theoretical properties of the solution of the min-max optimization problem. The optimization process explored the primal-dual problems at each boosting round. This generic framework can be adapted to diverse fairness concepts. The proposed min-max primal-dual gradient boosting algorithm was theoretically shown to converge under mild conditions and empirically shown to be a powerful and flexible approach to address binary and subgroup fairness.

Problem

• Ensuring fairness in machine learning predictions for marginalized groups
• Mitigating discrimination based on protected attributes like gender and race
• Developing a min-max gradient boosting framework for subgroup fairness

Innovation

• Min-max fairness term in gradient boosting
• Primal-dual optimization per boosting round
• Flexible framework for diverse fairness concepts