New Fairness Metrics for Recommendation that Embrace Differences

TL;DR

This paper introduces four novel fairness metrics for collaborative-filtering recommender systems to better detect and reduce biases against minority groups, improving fairness measurement and mitigation.

Contribution

The paper proposes new fairness metrics tailored for recommendation systems and demonstrates their effectiveness in measuring and reducing unfairness in experiments.

Findings

New metrics outperform baselines in fairness measurement

Fairness objectives reduce bias against minority groups

Metrics are effective on 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 against minority groups of users. 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.