Probably Approximately Metric-Fair Learning

TL;DR

This paper introduces a relaxed, approximately fair learning framework that ensures fairness generalizes from training data to the population, and provides polynomial-time algorithms for linear and logistic models.

Contribution

It proposes a new approximate metric-fairness notion and demonstrates its generalization, along with efficient PACF learning algorithms for specific predictor classes.

Findings

Approximate metric-fairness generalizes from training to population.

Polynomial-time PACF algorithms are developed for linear and logistic predictors.

The approach balances fairness with predictive accuracy.

Abstract

The seminal work of Dwork {\em et al.} [ITCS 2012] introduced a metric-based notion of individual fairness. Given a task-specific similarity metric, their notion required that every pair of similar individuals should be treated similarly. In the context of machine learning, however, individual fairness does not generalize from a training set to the underlying population. We show that this can lead to computational intractability even for simple fair-learning tasks. With this motivation in mind, we introduce and study a relaxed notion of {\em approximate metric-fairness}: for a random pair of individuals sampled from the population, with all but a small probability of error, if they are similar then they should be treated similarly. We formalize the goal of achieving approximate metric-fairness simultaneously with best-possible accuracy as Probably Approximately Correct and Fair (PACF) Learning. We show that approximate metric-fairness {\em does} generalize, and leverage these generalization guarantees to construct polynomial-time PACF learning algorithms for the classes of linear and logistic predictors.