De-biasing "bias" measurement

TL;DR

This paper identifies biases in common metrics used to measure model performance disparities across groups and introduces a simple, unbiased variance estimator to improve the accuracy of fairness assessments.

Contribution

The paper reveals statistical biases in existing disparity metrics and proposes a novel, easy-to-implement double-corrected variance estimator for unbiased measurement.

Findings

Existing disparity metrics are often biased estimators.

The proposed estimator provides unbiased variance estimates.

Applying the method changes significance conclusions in real data.

Abstract

When a model's performance differs across socially or culturally relevant groups--like race, gender, or the intersections of many such groups--it is often called "biased." While much of the work in algorithmic fairness over the last several years has focused on developing various definitions of model fairness (the absence of group-wise model performance disparities) and eliminating such "bias," much less work has gone into rigorously measuring it. In practice, it important to have high quality, human digestible measures of model performance disparities and associated uncertainty quantification about them that can serve as inputs into multi-faceted decision-making processes. In this paper, we show both mathematically and through simulation that many of the metrics used to measure group-wise model performance disparities are themselves statistically biased estimators of the underlying quantities they purport to represent. We argue that this can cause misleading conclusions about the relative group-wise model performance disparities along different dimensions, especially in cases where some sensitive variables consist of categories with few members. We propose the "double-corrected" variance estimator, which provides unbiased estimates and uncertainty quantification of the variance of model performance across groups. It is conceptually simple and easily implementable without statistical software package or numerical optimization. We demonstrate the utility of this approach through simulation and show on a real dataset that while statistically biased estimators of group-wise model performance disparities indicate statistically significant differences, when accounting for statistical bias in the estimator, the estimated between-group disparities are no longer statistically significant.