A Fully Bayesian, Logistic Regression Tracking Algorithm for Mitigating Disparate Misclassification

TL;DR

This paper introduces a fully Bayesian logistic regression tracking algorithm designed to produce unbiased classification results across groups with different sensitive attributes, effectively addressing disparities in false prediction rates.

Contribution

It presents a novel Bayesian approach that dynamically estimates logistic regression parameters to mitigate bias in classification tasks involving sensitive variables.

Findings

Reduces false prediction rate disparities between groups.

Performs well on simulated and real datasets.

Offers a dynamic, unbiased classification method.

Abstract

We develop a fully Bayesian, logistic tracking algorithm with the purpose of providing classification results that are unbiased when applied uniformly to individuals with differing sensitive variable values. Here, we consider bias in the form of differences in false prediction rates between the different sensitive variable groups. Given that the method is fully Bayesian, it is well suited for situations where group parameters or logistic regression coefficients are dynamic quantities. We illustrate our method, in comparison to others, on both simulated datasets and the well-known ProPublica COMPAS dataset.