Privilege Scores
This work addresses bias transformation in machine learning for fairness applications, offering a novel measurement and interpretation tool, though it is incremental in building on existing fairness methods.
The paper tackles the lack of explicit formulation for non-neutrality in fairness-aware machine learning by introducing privilege scores to measure privilege related to protected attributes, comparing real-world and fair-world predictions, and demonstrates applicability with insights into gender and racial privilege in mortgage and college admissions.
Bias-transforming methods of fairness-aware machine learning aim to correct a non-neutral status quo with respect to a protected attribute (PA). Current methods, however, lack an explicit formulation of what drives non-neutrality. We introduce privilege scores (PS) to measure PA-related privilege by comparing the model predictions in the real world with those in a fair world in which the influence of the PA is removed. At the individual level, PS can identify individuals who qualify for affirmative action; at the global level, PS can inform bias-transforming policies. After presenting estimation methods for PS, we propose privilege score contributions (PSCs), an interpretation method that attributes the origin of privilege to mediating features and direct effects. We provide confidence intervals for both PS and PSCs. Experiments on simulated and real-world data demonstrate the broad applicability of our methods and provide novel insights into gender and racial privilege in mortgage and college admissions applications.