When AI doctors lie about diagnosis: The effects of varying degrees of prosocial lies in patient–AI interactions

Instead of telling the whole truth, doctors sometimes resort to prosocial lies when diagnosing illness to protect patients from psychological harm. Recent advances in artificial intelligence (AI) have introduced AI doctors capable of telling prosocial lies in a medical setting. Accordingly, this study conducted a 3 (full truth vs. partial prosocial lie vs. full prosocial lie) × 2 (AI vs. human doctor) between-subjects experiment to examine how varying degrees of prosocial lying by different types of doctors are perceived by laypeople. The results showed that the full truth and partial prosocial lies elicited a similar level of acceptance, whereas full prosocial lies were the least acceptable. The effects of prosocial lying were mediated by autonomy violation and psychological benefits. Additionally, individuals preferred the truth from a human doctor rather than an AI doctor, but full prosocial lies were more acceptable from an AI doctor than from a human doctor.