4
1

Feature Importance Disparities for Data Bias Investigations

Abstract

It is widely held that one cause of downstream bias in classifiers is bias present in the training data. Rectifying such biases may involve context-dependent interventions such as training separate models on subgroups, removing features with bias in the collection process, or even conducting real-world experiments to ascertain sources of bias. Despite the need for such data bias investigations, few automated methods exist to assist practitioners in these efforts. In this paper, we present one such method that given a dataset XX consisting of protected and unprotected features, outcomes yy, and a regressor hh that predicts yy given XX, outputs a tuple (fj,g)(f_j, g), with the following property: gg corresponds to a subset of the training dataset (X,y)(X, y), such that the jthj^{th} feature fjf_j has much larger (or smaller) influence in the subgroup gg, than on the dataset overall, which we call feature importance disparity (FID). We show across 44 datasets and 44 common feature importance methods of broad interest to the machine learning community that we can efficiently find subgroups with large FID values even over exponentially large subgroup classes and in practice these groups correspond to subgroups with potentially serious bias issues as measured by standard fairness metrics.

View on arXiv
Comments on this paper