Multiaccuracy: Black-Box Post-Processing for Fairness in Classification

TL;DR

This paper introduces MULTIACCURACY-BOOST, a black-box post-processing method that enhances fairness by ensuring accurate predictions across subgroups without requiring access to sensitive features.

Contribution

It develops a rigorous multiaccuracy framework and an efficient algorithm for improving fairness in classification models using limited labeled data and black-box access.

Findings

Improves accuracy for minority subgroups in diverse applications

Converges efficiently with guarantees on subgroup accuracy

Enhances fairness without explicit sensitive feature access

Abstract

Prediction systems are successfully deployed in applications ranging from disease diagnosis, to predicting credit worthiness, to image recognition. Even when the overall accuracy is high, these systems may exhibit systematic biases that harm specific subpopulations; such biases may arise inadvertently due to underrepresentation in the data used to train a machine-learning model, or as the result of intentional malicious discrimination. We develop a rigorous framework of *multiaccuracy* auditing and post-processing to ensure accurate predictions across *identifiable subgroups*. Our algorithm, MULTIACCURACY-BOOST, works in any setting where we have black-box access to a predictor and a relatively small set of labeled data for auditing; importantly, this black-box framework allows for improved fairness and accountability of predictions, even when the predictor is minimally transparent. We prove that MULTIACCURACY-BOOST converges efficiently and show that if the initial model is accurate on an identifiable subgroup, then the post-processed model will be also. We experimentally demonstrate the effectiveness of the approach to improve the accuracy among minority subgroups in diverse applications (image classification, finance, population health). Interestingly, MULTIACCURACY-BOOST can improve subpopulation accuracy (e.g. for "black women") even when the sensitive features (e.g. "race", "gender") are not given to the algorithm explicitly.