Beyond Parity: Fairness Objectives for Collaborative Filtering

TL;DR

This paper introduces four new fairness metrics for collaborative-filtering recommender systems, addressing biases in historical data, and demonstrates their effectiveness in reducing unfairness through experiments.

Contribution

It proposes novel fairness metrics and optimization methods specifically designed to mitigate bias in collaborative filtering systems.

Findings

New metrics outperform baselines in measuring fairness

Fairness objectives reduce bias in recommendations

Effective on both synthetic and real datasets

Abstract

We study fairness in collaborative-filtering recommender systems, which are sensitive to discrimination that exists in historical data. Biased data can lead collaborative-filtering methods to make unfair predictions for users from minority groups. We identify the insufficiency of existing fairness metrics and propose four new metrics that address different forms of unfairness. These fairness metrics can be optimized by adding fairness terms to the learning objective. Experiments on synthetic and real data show that our new metrics can better measure fairness than the baseline, and that the fairness objectives effectively help reduce unfairness.