When is ai perceived as less biased than humans? Social unfairness perception, discrimination experience, and comparative bias judgments in gender-related decision-making

Abstract

As automated decision-making (ADM) becomes embedded in high-stakes domains, understanding how users perceive its fairness relative to human decision-making (HDM) is critical. The present study examined whether perceived social unfairness and personal experiences of gender discrimination moderate the perceived bias gap between HDM and ADM, and whether this gap predicts attitudes toward ADM. A web-based survey of 506 Korean adults assessed perceived gender bias in HDM and ADM across four domains (law, finance, recruitment, and education). Linear mixed models revealed that participants consistently perceived ADM as less gender-biased than HDM across all domains, with the gap largest in recruitment and education. This comparative bias advantage was moderated by perceived social unfairness but not by personal discrimination experience, and the magnitude of the advantage was significantly associated with more favorable attitudes toward ADM. Findings suggest that public preference for algorithmic decision-making is anchored in systemic evaluations of social fairness rather than in individual lived experience.