Fair Risk Control: A Generalized Framework for Calibrating Multi-group Fairness Risks

TL;DR

This paper proposes a flexible post-processing framework based on multicalibration to ensure multi-group fairness in machine learning predictions across various applications.

Contribution

It introduces the $( extbf{s}, extbf{G}, extbf{α})$-GMC framework and algorithms for achieving multi-dimensional fairness guarantees in diverse machine learning tasks.

Findings

Effective fairness guarantees in image segmentation and language models

Applicable to multiple fairness concerns and scenarios

Numerical validation on various datasets

Abstract

This paper introduces a framework for post-processing machine learning models so that their predictions satisfy multi-group fairness guarantees. Based on the celebrated notion of multicalibration, we introduce $(\mathbf{s},\mathcal{G}, \alpha)-$GMC (Generalized Multi-Dimensional Multicalibration) for multi-dimensional mappings $\mathbf{s}$, constraint set $\mathcal{G}$, and a pre-specified threshold level $\alpha$. We propose associated algorithms to achieve this notion in general settings. This framework is then applied to diverse scenarios encompassing different fairness concerns, including false negative rate control in image segmentation, prediction set conditional uncertainty quantification in hierarchical classification, and de-biased text generation in language models. We conduct numerical studies on several datasets and tasks.