Fairness metrics miss complexity in multi-class and multi-group data
Rethinking Fairness Metrics for Multi-Class and Multi-Group Assessment: Not Everything is Binary
Machine LearningArtificial Intelligence
Summary
Many fairness checks for AI assume groups and outcomes are just two choices, like yes or no. The authors find this is too simple and can hide unfairness when there are more groups or results. They create new, solid formulas to measure fairness that work directly with many groups and types of answers. Their work shows that the common 'one versus rest' method to simplify data can distort fairness measurements. This helps build fairer AI for complicated real-world choices.
What this means in practice
- •For machine learning engineers: Evaluate models on fairness metrics that handle multiple sensitive groups and classification outcomes directly, avoiding misleading binary simplifications.
- •For data ethics teams: Improve fairness audits by using metrics designed for complex, real-world data with multiple groups and labels, enhancing fairness assessments' accuracy.
Authors
Matteo Grazioso, Fabio Naccarella, Chiara Gallese, Marco S. Nobile
Abstract
Artificial Intelligence, and Machine Learning in general, is becoming pervasive across all domains and disciplines. By means of this data-driven approach, algorithms can learns complex task just by learning from samples. However, if the data-set is biased, they might lead to the development of unfair computer systems, that is, characterised by a prejudice towards specific individuals or groups, which might get an unmotivated output from the system. This circumstance lead to the development of multiple metrics designed to quantify the level of unfairness of an algorithm. However, a generalised assumption in the literature is that the sensitive features and the output labels are binary. When the data is not binary, it is forcefully binarized using a one-versus-the-rest simplification. In this work, we discuss the limitations of existing metrics and propose novel, mathematically sound, and generalized multi-group and multi-class formulations of the metrics. Ultimately, our results highlight the empirical assessment distortion caused by binarization strategies, which reveals the limitations of binary fairness metrics in the multi class setting.