As artificial intelligence (AI) systems become increasingly embedded in social contexts, understanding how individuals develop trust in AI relative to humans is critical. This study investigates the relative levels of trust in AI agents (embodied and disembodied) versus human social targets (intimate, intermediate, and distant groups), the psychological mechanisms underlying these trust patterns, and the potential cross-cultural differences between China and U.S. Moderated mediation models were tested to gain insights into how deception experience may affect trust via risk and trust propensity, with perceived honesty norms moderating the mediator-to-outcome pathways. Across both cultures, a consistent trust hierarchy emerged—AI was trusted less than close others but more than distant others. It is likely that, in China, embodied AI was evaluated through interpersonal trust schemas, while in the United States, AI was treated largely as a functional tool regardless of embodiment. Together, these findings clarify both the structure and the processes of AI trust, advancing theoretical debates over whether AI trust mirrors interpersonal trust and offering practical insights for designing culturally adaptive, trustworthy AI systems.
Zhang et al. (Thu,) studied this question.