A Unified View of Group Fairness Tradeoffs Using Partial Information Decomposition

TL;DR

This paper offers an information-theoretic framework using partial information decomposition to analyze and understand the tradeoffs and overlaps among key group fairness notions in machine learning.

Contribution

It introduces a novel application of partial information decomposition to characterize the relationships and tradeoffs among statistical parity, equalized odds, and predictive parity.

Findings

Identifies regions where fairness measures overlap and conflict

Provides a theoretical basis for understanding fairness tradeoffs

Includes numerical simulations supporting the analysis

Abstract

This paper introduces a novel information-theoretic perspective on the relationship between prominent group fairness notions in machine learning, namely statistical parity, equalized odds, and predictive parity. It is well known that simultaneous satisfiability of these three fairness notions is usually impossible, motivating practitioners to resort to approximate fairness solutions rather than stringent satisfiability of these definitions. However, a comprehensive analysis of their interrelations, particularly when they are not exactly satisfied, remains largely unexplored. Our main contribution lies in elucidating an exact relationship between these three measures of (un)fairness by leveraging a body of work in information theory called partial information decomposition (PID). In this work, we leverage PID to identify the granular regions where these three measures of (un)fairness overlap and where they disagree with each other leading to potential tradeoffs. We also include numerical simulations to complement our results.