Algorithm Aversion: Investigating the Role of Confidence in AI-Assisted Judgment

As artificial intelligence (AI) increasingly informs decision-making, understanding when and why individuals accept or reject algorithmic advice remains a critical issue. This study examines how individuals incorporate AI-generated advice into interpersonal judgments, an underexplored yet increasingly relevant domain for AI applications. Through an experimental study, we analyze the extent to which confidence in self-generated judgments moderates algorithm aversion. Specifically, we investigate whether individuals adjust their judgment of another person after receiving AI advice about that person's behavior and how their confidence shifts following the advice. Our findings suggest that algorithm aversion is strongly influenced by initial overconfidence rather than inherent distrust in AI. Furthermore, we find that receiving AI-generated advice can paradoxically decrease user confidence, highlighting potential risks of human vulnerability to AI influence in decision-making. These insights offer practical implications for organizations integrating AI into personnel evaluations, recruitment, and performance assessments, where interpersonal judgments play a crucial role.