Local Causal Discovery for Structural Evidence of Direct Discrimination

TL;DR

This paper introduces LD3, a local causal discovery method that efficiently identifies direct unfairness in decision systems by uncovering causal parents of outcomes, even with limited knowledge of the causal graph.

Contribution

LD3 provides a practical, efficient approach for detecting direct discrimination by identifying causal parents, handling latent confounders, and requiring fewer tests than global methods.

Findings

LD3 is more time-efficient than baseline methods.

LD3 produces more plausible results on real-world data.

LD3 effectively uncovers structural evidence of direct unfairness.

Abstract

Identifying the causal pathways of unfairness is a critical objective for improving policy design and algorithmic decision-making. Prior work in causal fairness analysis often requires knowledge of the causal graph, hindering practical applications in complex or low-knowledge domains. Moreover, global discovery methods that learn causal structure from data can display unstable performance on finite samples, preventing robust fairness conclusions. To mitigate these challenges, we introduce local discovery for direct discrimination (LD3): a method that uncovers structural evidence of direct unfairness by identifying the causal parents of an outcome variable. LD3 performs a linear number of conditional independence tests relative to variable set size, and allows for latent confounding under the sufficient condition that all parents of the outcome are observed. We show that LD3 returns a valid adjustment set (VAS) under a new graphical criterion for the weighted controlled direct effect, a qualitative indicator of direct discrimination. LD3 limits unnecessary adjustment, providing interpretable VAS for assessing unfairness. We use LD3 to analyze causal fairness in two complex decision systems: criminal recidivism prediction and liver transplant allocation. LD3 was more time-efficient and returned more plausible results on real-world data than baselines, which took 46$\times$ to 5870$\times$ longer to execute.