Strict With Humans, Lenient With Machines? Differential Effects of Artificial Intelligence Versus Human Unfairness on Trust

Existing research has produced mixed findings regarding how individuals respond to unfair decisions enacted by AI versus human decision-makers. A key limitation of prior research is that studies rarely distinguish among different forms of unfairness and have largely focused on moral reactions. Drawing on expectancy violation theory and mind perception theory, we conducted six vignette-based experiments (N = 1299) to examine how trust responses differ when distributive, procedural, and interactional unfairness are enacted by AI versus humans. Results show that (1) under distributive unfairness, participants reported significantly lower cognitive trust toward AI than toward humans; (2) under procedural and interactional unfairness, participants reported significantly higher cognitive trust toward AI than toward humans; and (3) across all types of unfairness, humans consistently evoked lower affective trust than AI. Mediation analyses further indicate that perceived ability and perceived benevolence serve as key explanatory mechanisms for cognitive and affective trust, respectively.