Fairness in Reinforcement Learning with Bisimulation Metrics

TL;DR

This paper introduces a novel reinforcement learning approach that uses bisimulation metrics to promote long-term fairness across groups, ensuring equitable treatment in sequential decision-making tasks.

Contribution

It establishes a connection between bisimulation metrics and group fairness, proposing a method to incorporate fairness into reward and observation modeling in RL.

Findings

Effective reduction of disparities in lending and college admission scenarios

Demonstrates improved fairness without sacrificing overall performance

Validates approach on standard fairness benchmarks

Abstract

Ensuring long-term fairness is crucial when developing automated decision making systems, specifically in dynamic and sequential environments. By maximizing their reward without consideration of fairness, AI agents can introduce disparities in their treatment of groups or individuals. In this paper, we establish the connection between bisimulation metrics and group fairness in reinforcement learning. We propose a novel approach that leverages bisimulation metrics to learn reward functions and observation dynamics, ensuring that learners treat groups fairly while reflecting the original problem. We demonstrate the effectiveness of our method in addressing disparities in sequential decision making problems through empirical evaluation on a standard fairness benchmark consisting of lending and college admission scenarios.