A Maximal Correlation Approach to Imposing Fairness in Machine Learning

TL;DR

This paper introduces an information-theoretic maximal correlation framework to enforce fairness in machine learning, enabling efficient optimization and better tradeoffs between fairness and performance.

Contribution

It presents a novel maximal correlation approach for fairness constraints, leading to more computationally efficient algorithms that work with both discrete and continuous data.

Findings

Algorithms achieve smooth fairness-performance tradeoff curves.

Performance is competitive with state-of-the-art methods.

Applicable to both discrete and continuous datasets.

Abstract

As machine learning algorithms grow in popularity and diversify to many industries, ethical and legal concerns regarding their fairness have become increasingly relevant. We explore the problem of algorithmic fairness, taking an information-theoretic view. The maximal correlation framework is introduced for expressing fairness constraints and shown to be capable of being used to derive regularizers that enforce independence and separation-based fairness criteria, which admit optimization algorithms for both discrete and continuous variables which are more computationally efficient than existing algorithms. We show that these algorithms provide smooth performance-fairness tradeoff curves and perform competitively with state-of-the-art methods on both discrete datasets (COMPAS, Adult) and continuous datasets (Communities and Crimes).