Long Term Fairness for Minority Groups via Performative Distributionally Robust Optimization

TL;DR

This paper extends performative prediction with a distributionally robust objective to address limitations of existing formal fairness criteria in machine learning, aiming to improve long-term fairness for minority groups.

Contribution

It introduces a novel approach combining performative prediction with distributionally robust optimization to enhance fairness for minority groups over the long term.

Findings

Addresses four key shortcomings of current fairness criteria

Proposes a new framework integrating robustness into performative prediction

Improves long-term fairness outcomes for minority groups

Abstract

Fairness researchers in machine learning (ML) have coalesced around several fairness criteria which provide formal definitions of what it means for an ML model to be fair. However, these criteria have some serious limitations. We identify four key shortcomings of these formal fairness criteria, and aim to help to address them by extending performative prediction to include a distributionally robust objective.