Towards Auditing Unsupervised Learning Algorithms and Human Processes For Fairness

TL;DR

This paper explores auditing the fairness of unsupervised learning algorithms across multiple groups using complex fairness definitions, extending beyond traditional binary protected status assessments.

Contribution

It introduces a framework for auditing multi-group fairness in unsupervised algorithms under complex fairness criteria, an area less studied in current research.

Findings

Developed methods for multi-group fairness auditing

Extended fairness definitions to complex, real-world scenarios

Provided insights into algorithmic bias across multiple demographic groups

Abstract

Existing work on fairness typically focuses on making known machine learning algorithms fairer. Fair variants of classification, clustering, outlier detection and other styles of algorithms exist. However, an understudied area is the topic of auditing an algorithm's output to determine fairness. Existing work has explored the two group classification problem for binary protected status variables using standard definitions of statistical parity. Here we build upon the area of auditing by exploring the multi-group setting under more complex definitions of fairness.