BIAS AMPLIFICATION IN AI RECRUITMENT SYSTEMS: A CAUSAL INFERENCE PERSPECTIVE

This paper explores the amplification of social bias in AI-powered recruitment platforms through the lens of causal inference. As automated hiring tools become more widespread, concerns have emerged regarding the reproduction—and even amplification—of gender, racial, and socioeconomic discrimination. We propose a framework that combines causal graphs, counterfactual analysis, and explainable AI (XAI) to measure and mitigate algorithmic bias in resume screening systems. Using real-world datasets and synthetic interventions, we demonstrate how seemingly neutral features can act as proxies for protected attributes. Our findings contribute to the growing body of research on fair AI, offering transparent methodologies for both academic evaluation and regulatory audit.