Explanation-Guided Fairness Testing through Genetic Algorithm

TL;DR

This paper introduces ExpGA, a novel explanation-guided genetic algorithm for fairness testing that improves efficiency and effectiveness in detecting discriminatory samples across various models and datasets.

Contribution

ExpGA combines explanation methods with genetic algorithms to enhance fairness testing, overcoming limitations of low efficiency, low effectiveness, and model-specificity in prior approaches.

Findings

ExpGA outperforms four state-of-the-art methods in efficiency and effectiveness.

It generalizes well across different models and data types.

Experiments on real-world benchmarks validate its superior performance.

Abstract

The fairness characteristic is a critical attribute of trusted AI systems. A plethora of research has proposed diverse methods for individual fairness testing. However, they are suffering from three major limitations, i.e., low efficiency, low effectiveness, and model-specificity. This work proposes ExpGA, an explanationguided fairness testing approach through a genetic algorithm (GA). ExpGA employs the explanation results generated by interpretable methods to collect high-quality initial seeds, which are prone to derive discriminatory samples by slightly modifying feature values. ExpGA then adopts GA to search discriminatory sample candidates by optimizing a fitness value. Benefiting from this combination of explanation results and GA, ExpGA is both efficient and effective to detect discriminatory individuals. Moreover, ExpGA only requires prediction probabilities of the tested model, resulting in a better generalization capability to various models. Experiments on multiple real-world benchmarks, including tabular and text datasets, show that ExpGA presents higher efficiency and effectiveness than four state-of-the-art approaches.