Censoring Representations with an Adversary

In practice, there are often constraints on the decisions that may be made for a decision problem, or in communicating data. One example of such a constraint is that a decision must not favour a particular group. Another is that data must not have identifying information. We address these two related issues by learning flexible representations that minimize the capability of an adversarial critic. This adversary is trying to predict the relevant sensitive variable from the representation, and so minimizing the performance of the adversary ensures there is little or no information in the representation about the sensitive variable. We demonstrate this in the specific contexts of making decisions free from discrimination and removing private information from images.