MIFair: A Mutual-Information Framework for Intersectionality and Multiclass Fairness

2026-04-30

AI Summary

This work addresses the challenge of evaluating and mitigating bias in machine learning under intersecting and multi-category sensitive attributes by proposing a unified framework grounded in mutual information theory. The approach models statistical independence between predictions and sensitive attributes to construct a flexible template for fairness metrics and integrates a regularization-based training strategy to effectively reduce bias. It is the first to systematically apply mutual information theory to scenarios involving intersectional and multi-class fairness, thereby unifying classical fairness criteria—such as independence and separation—and establishing their theoretical equivalence. Empirical results demonstrate that the method significantly alleviates bias across complex subgroups on both real-world tabular and image datasets while maintaining strong predictive performance.

Abstract

Fairness in machine learning remains challenging due to its ethical complexity, the absence of a universal definition, and the need for context-specific bias metrics. Existing methods still struggle with intersectionality, multiclass settings, and limited flexibility and generality. To address these gaps, we introduce MIFair, a unified framework for bias assessment and mitigation based on mutual information. MIFair provides a flexible metric template and an in-processing mitigation method inspired by the Prejudice Remover, defining group fairness as statistical independence between prediction-derived variables and sensitive attributes. We further strengthen its information-theoretic foundation by establishing equivalences with widely used fairness notions such as independence and separation. MIFair naturally supports intersectionality, complex subgroup structures, and multiclass classification and employs regularization-based training to reduce bias according to the selected metric. Its key advantage is its versatility: it consolidates diverse fairness requirements into a single coherent framework, enabling consistent benchmarking and simplifying practical use. Experiments on real-world tabular and image datasets show that MIFair effectively reduces bias, including previously unaddressed multi-attribute scenarios, while maintaining strong predictive performance across the evaluated settings.

Problem

• fairness
• intersectionality
• multiclass classification
• bias mitigation
• machine learning

Innovation

• mutual information
• intersectionality
• multiclass fairness
• in-processing mitigation
• fairness regularization