Too Relaxed to Be Fair
Michael Lohaus, Michael Perrot, Ulrike Von Luxburg
Abstract
We address the problem of classification under fairness constraints. Given a notion of fairness, the goal is to learn a classifier that is not discriminatory against a group of individuals. In the literature, this problem is often formulated as a constrained optimization problem and solved using relaxations of the fairness constraints. We show that many existing relaxations are unsatisfactory: even if a model satisfies the relaxed constraint, it can be surprisingly unfair. We propose a principled framework to solve this problem. This new approach uses a strongly convex formulation and comes with theoretical guarantees on the fairness of its solution. In practice, we show that this method gives promising results on real data.
BibTeX
@InProceedings{pmlr-v119-lohaus20a,
title = {Too Relaxed to Be Fair},
author = {Lohaus, Michael and Perrot, Michael and Luxburg, Ulrike Von},
booktitle = {Proceedings of the 37th International Conference on Machine Learning},
pages = {6360--6369},
year = {2020},
editor = {III, Hal Daumé and Singh, Aarti},
volume = {119},
series = {Proceedings of Machine Learning Research},
month = {13--18 Jul},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v119/lohaus20a/lohaus20a.pdf},
url = {https://proceedings.mlr.press/v119/lohaus20a.html},
abstract = {We address the problem of classification under fairness constraints. Given a notion of fairness, the goal is to learn a classifier that is not discriminatory against a group of individuals. In the literature, this problem is often formulated as a constrained optimization problem and solved using relaxations of the fairness constraints. We show that many existing relaxations are unsatisfactory: even if a model satisfies the relaxed constraint, it can be surprisingly unfair. We propose a principled framework to solve this problem. This new approach uses a strongly convex formulation and comes with theoretical guarantees on the fairness of its solution. In practice, we show that this method gives promising results on real data.}
}