Fair Feature Importance Scores via Feature Occlusion and Permutation

2026-02-09

AI Summary

This work addresses a critical gap in existing feature importance methods, which predominantly focus on predictive accuracy and fail to quantify the impact of features on model fairness. The paper introduces, for the first time, two model-agnostic measures of fair feature importance: one evaluates a feature’s contribution by assessing changes in fairness before and after its permutation, while the other enhances computational efficiency by masking specific features during training and leveraging minipatch learning to compare resulting fairness disparities. The proposed approach is interpretable, scalable, and computationally efficient, accurately identifying fairness-critical features across multiple prediction tasks. By offering a practical tool for fairness-aware model analysis, this method advances the development of responsible AI systems and fills a key void in fairness attribution research.

Technology Category

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

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

As machine learning models increasingly impact society, their opaque nature poses challenges to trust and accountability, particularly in fairness contexts. Understanding how individual features influence model outcomes is crucial for building interpretable and equitable models. While feature importance metrics for accuracy are well-established, methods for assessing feature contributions to fairness remain underexplored. We propose two model-agnostic approaches to measure fair feature importance. First, we propose to compare model fairness before and after permuting feature values. This simple intervention-based approach decouples a feature and model predictions to measure its contribution to training. Second, we evaluate the fairness of models trained with and without a given feature. This occlusion-based score enjoys dramatic computational simplification via minipatch learning. Our empirical results reflect the simplicity and effectiveness of our proposed metrics for multiple predictive tasks. Both methods offer simple, scalable, and interpretable solutions to quantify the influence of features on fairness, providing new tools for responsible machine learning development.

Problem

• fairness
• feature importance
• machine learning
• interpretability
• model accountability

Innovation

• fair feature importance
• feature occlusion
• feature permutation
• model-agnostic interpretability
• minipatch learning