Back to the Drawing Board for Fair Representation Learning

arXiv:2405.18161v1 Announce Type: new
Abstract: The goal of Fair Representation Learning (FRL) is to mitigate biases in machine learning models by learning data representations that enable high accuracy on downstream tasks while minimizing discrimination based on sensitive attributes. The evaluation of FRL methods in many recent works primarily focuses on the tradeoff between downstream fairness and accuracy with respect to a single task that was used to approximate the utility of representations during training (proxy task). This incentivizes retaining only …

abstract accuracy arxiv attributes biases board cs.ai cs.lg data discrimination evaluation fair machine machine learning machine learning models representation representation learning tasks type while