Abstract
Perceived algorithmic bias is defined as users' comprehensive subjective cognitive judgment, grounded in a cognitive attribution perspective, regarding whether an algorithmic system engages in differential treatment, whether its outcomes exhibit systematic deviations, and whether such differential treatment is justified. This study develops a theoretical model of perceived algorithmic bias and systematically elucidates the construct through three dimensions: perceived differential treatment, perceived outcome skew, and perceived lack of justification. The paper examines the mechanisms through which perceived algorithmic bias influences human-AI collaborative decision-making performance across three levels: cognitive processing, affect and morality, and human-AI interaction. It also analyzes the moderating roles of individual characteristics, algorithmic characteristics, and task characteristics. By systematically reviewing the conceptualizations of algorithmic bias and perceived algorithmic bias and clarifying their effects on and theoretical mechanisms underlying human-AI collaborative decision-making performance, this study offers theoretical implications for advancing research on perceived algorithmic bias. Future research may further develop a measurement scale for perceived algorithmic bias, enrich empirical investigations, test its underlying mechanisms, and explore its impact on human-AI collaborative decision-making performance across diverse application contexts.
Recommended Citation
Jia, Qiong and Zhang, Yuxin, "Theoretical Mechanisms and Framework of Perceived Algorithmic Bias Affecting Human-AI Collaborative Decision-Making Performance" (2026). ICEB 2026 Proceedings (Beijing, China). 115.
https://aisel.aisnet.org/iceb2026/115