Despite the growing discussions surrounding the use of artificial intelligence (AI) judges, empirical research examining how the social attributes of AI influence the perceived reliability of judicial decisions remains limited. This study investigates...
Despite the growing discussions surrounding the use of artificial intelligence (AI) judges, empirical research examining how the social attributes of AI influence the perceived reliability of judicial decisions remains limited. This study investigates whether the perceived social attributes(warmth, competence, discomfort) of an AI judge affect the perceived reliability of its outcomes depending on the AI’s appearance and sentencing outcome. A total of 220 university students were randomly assigned to one of four conditions in a 2 (AI appearance: human-like vs. machine-like) × 2 (sentencing outcome: beneficial vs. detrimental to the defendant) between-subjects design. Participants first evaluated the social attributes of the AI presented to them and then read a scenario corresponding to their assigned condition, after which they evaluated trust, acceptance, reuse intention, and perceived need for revision regarding the AI judge. The results showed that, first, when a human-like AI judge rendered a outcome detrimental to the defendant, higher perceived competence increased trust in the AI judge. Second, when a human-like AI judge rendered a beneficial outcome, higher perceived competence increased acceptance of the AI judge. Third, when the AI judge delivered a beneficial outcome, perceived competence increased reuse intention for the human-like AI judge, whereas perceived warmth increased reuse intention for the machine-like AI judge. Finally, when the human-like AI judge rendered a detrimental outcome, higher perceived warmth increased the perceived need for revision. Overall, competence played a key role for human-like AI judges, whereas warmth played a key role for machine-like AI judges in shaping perceived reliability of AI judicial outcomes. These findings suggest that the effects of perceived social attributes on the perceived reliability of AI judges vary depending on the AI judge’s appearance and the sentencing outcome.