Following Medical Advice of an AI or a Human Doctor? Experimental Evidence Based on Clinician-Patient Communication Pathway Model

Abstract

Medical large language models are being introduced to the public in collaboration with governments, medical institutions, and artificial intelligence (AI) researchers. However, a crucial question remains: Will patients follow the medical advice provided by AI doctors? The lack of user research makes it difficult to provide definitive answers. Based on the clinician-patient communication pathway model, this study conducted a factorial experiment with a 2 (medical provider, AI vs. human) × 2 (information support, low vs. high) × 2 (response latency, slow vs. fast) between-subjects design (n = 535). The results showed that participants exhibited significantly lower adherence to AI doctors’ advice than to human doctors. In addition, the interaction effect suggested that, under the slow-response latency condition, subjects perceived greater health benefits and patient-centeredness from human doctors, while the opposite was observed for AI doctors.