Fair Causal Feature Selection

TL;DR

This paper introduces FairCFS, a causal graph-based feature selection method that improves fairness in classification tasks by explicitly blocking sensitive information transmission, while maintaining competitive accuracy.

Contribution

The paper presents a novel FairCFS algorithm that constructs localized causal graphs to identify fair features, addressing limitations of existing methods in explaining causal relationships.

Findings

FairCFS achieves comparable accuracy to state-of-the-art methods.

FairCFS demonstrates superior fairness across multiple datasets.

Extensive experiments validate the effectiveness of FairCFS.

Abstract

Fair feature selection for classification decision tasks has recently garnered significant attention from researchers. However, existing fair feature selection algorithms fall short of providing a full explanation of the causal relationship between features and sensitive attributes, potentially impacting the accuracy of fair feature identification. To address this issue, we propose a Fair Causal Feature Selection algorithm, called FairCFS. Specifically, FairCFS constructs a localized causal graph that identifies the Markov blankets of class and sensitive variables, to block the transmission of sensitive information for selecting fair causal features. Extensive experiments on seven public real-world datasets validate that FairCFS has comparable accuracy compared to eight state-of-the-art feature selection algorithms, while presenting more superior fairness.