Decisional agreement, perceived value similarity, and trust in AI systems across medical, investment, and driving contexts

Trust is central to appropriate reliance on artificial intelligence (AI) in consequential decisions, but it remains unclear how users evaluate AI and human agents when decisions involve value-laden trade-offs. We examined whether decisional agreement, stated favorable outcome probability, and subjective perceived risk are associated with trust across three contexts. Across three vignette-based studies ( N = 1,609), participants evaluated AI or human agents in medical decision-making (Study 1, laypersons), investment decisions (Study 2, laypersons), and driving-related moral dilemmas (Study 3, professional taxi drivers). Each study used a 2 × 2 × 2 between-subjects design crossing agent type, decisional agreement (agent agreed vs. disagreed with the participant’s initial choice), and favorable outcome probability (20% vs. 80%). Participants then rated trust, perceived value similarity, and perceived risk. Decisional agreement was associated with higher perceived value similarity and greater trust in both AI and human agents across all three studies. The 20% favorable-outcome condition increased perceived risk relative to the 80% condition, but favorable outcome probability did not significantly affect trust in any study. Perceived risk showed negative concurrent associations with trust in some models, but these associations were generally modest, varied across studies and trust-scoring specifications. Lay participants trusted human agents more than AI agents in medical and investment contexts, whereas professional taxi drivers showed no reliable agent-type difference in driving. Exploratory analyses of the interaction between favorable outcome probability and decisional agreement did not provide reliable evidence of moderation, although the Study 2 result remained inconclusive. In brief value-laden encounters with AI and human agents, trust judgments were more consistently associated with agreement-based value cues than with favorable outcome probability. The findings suggest that users may treat decisional agreement as an accessible cue to perceived value similarity, whereas perceived risk reflects a separate, context-dependent appraisal rather than a direct proxy for stated outcome probability.