Uncovering Discrimination Clusters: Quantifying and Explaining Systematic Fairness Violations
Ranit Debnath Akash, Ashish Kumar, Verya Monjezi, Ashutosh Trivedi, Gang, Tan, Saeid Tizpaz-Niari
TL;DR
This work tackles the insufficiency of pairwise fairness checks by introducing discrimination clustering, a framework that identifies $k$-discriminant regions where protected attributes induce clustered, arbitrary outcomes in DNNs. HyFair combines formal verification (MILP/SMT) to certify $2$-fairness with randomized search (e.g., simulated annealing) to maximize $k$-discrimination and reveal clustered patterns, followed by an explanation method using decision trees to produce interpretable predicates and guardrails for mitigation. The approach yields a novel notion of $k$-discriminant and demonstrates that HyFair can outperform state-of-the-art methods in detecting, quantifying, and explaining fairness violations while enabling debiasing with relatively small accuracy trade-offs. This framework provides practical tools for diagnosing, explaining, and mitigating systematic biases in high-stakes, data-driven decision systems, with implications for governance and responsible deployment of neural-network-based classifiers. Specifically, it advances formal guarantees for fairness, scalable discovery of arbitrarily structured discrimination, and actionable mitigation via interpretable rules.
Abstract
Fairness in algorithmic decision-making is often framed in terms of individual fairness, which requires that similar individuals receive similar outcomes. A system violates individual fairness if there exists a pair of inputs differing only in protected attributes (such as race or gender) that lead to significantly different outcomes-for example, one favorable and the other unfavorable. While this notion highlights isolated instances of unfairness, it fails to capture broader patterns of systematic or clustered discrimination that may affect entire subgroups. We introduce and motivate the concept of discrimination clustering, a generalization of individual fairness violations. Rather than detecting single counterfactual disparities, we seek to uncover regions of the input space where small perturbations in protected features lead to k-significantly distinct clusters of outcomes. That is, for a given input, we identify a local neighborhood-differing only in protected attributes-whose members' outputs separate into many distinct clusters. These clusters reveal significant arbitrariness in treatment solely based on protected attributes that help expose patterns of algorithmic bias that elude pairwise fairness checks. We present HyFair, a hybrid technique that combines formal symbolic analysis (via SMT and MILP solvers) to certify individual fairness with randomized search to discover discriminatory clusters. This combination enables both formal guarantees-when no counterexamples exist-and the detection of severe violations that are computationally challenging for symbolic methods alone. Given a set of inputs exhibiting high k-unfairness, we introduce a novel explanation method to generate interpretable, decision-tree-style artifacts. Our experiments demonstrate that HyFair outperforms state-of-the-art fairness verification and local explanation methods.