Two Simple Ways to Learn Individual Fairness Metrics from Data
This work addresses the barrier to adopting individual fairness in ML by providing practical ways to derive fair metrics, though it appears incremental as it builds on existing individual fairness concepts.
The paper tackles the problem of lacking widely accepted fair metrics for individual fairness in machine learning by presenting two simple methods to learn such metrics from data, showing empirically that fair training with these metrics improves fairness on three tasks susceptible to gender and racial biases.
Individual fairness is an intuitive definition of algorithmic fairness that addresses some of the drawbacks of group fairness. Despite its benefits, it depends on a task specific fair metric that encodes our intuition of what is fair and unfair for the ML task at hand, and the lack of a widely accepted fair metric for many ML tasks is the main barrier to broader adoption of individual fairness. In this paper, we present two simple ways to learn fair metrics from a variety of data types. We show empirically that fair training with the learned metrics leads to improved fairness on three machine learning tasks susceptible to gender and racial biases. We also provide theoretical guarantees on the statistical performance of both approaches.