Physicians vs. AI – effects of (perceived) expertise on trust and intention to comply with medical recommendations

We investigated the influence of artificial intelligence (AI) and expertise on trust in and intention to comply with medical recommendations. The goal was to examine effects beyond the “Gods in white” stereotype by comparing physicians and AI at different expertise levels. In avignette study (N = 243), participants were diagnosed with askin condition and given atreatment recommendation in achat by one of four diagnosis entities: AI vs. physician with high vs. low expertise. We measured perceived expertise, trust and intention to comply. AI was less trusted and complied with than physicians. Groups with the same advising entity were not trusted and complied with differently, with no effect of expertise manipulation. Yet correlational results show that perceived expertise predicts trust and intention to comply, for the latter moderated by entity nature. In an explorative grouping based on perceived expertise and advisor’s nature, trust and compliance differ between all groups except high-expertise AI and low-expertise physician (and both physicians for compliance). Results suggest that medical large language models (LLMs) perceived as competent are not less trusted and intended to comply with than a non-experienced physician. Future research needs to consider implicit attitudes on expertise of diagnosis entities explicitly.