Candy or Poison?

With AI's integration into healthcare, clinical decision systems are shifting the traditional doctor-patient dyad to an “(AI+doctor)-patient” triad. Current research mostly focuses on the human-machine interaction perspective, exploring the human-machine trust issue when patients directly use medical AI for self-service diagnosis, while there is a gap in the research on the impact of physician-assisted diagnosis with the help of AI on the trust of elderly patients from the human-human interaction dimension. This study addresses the gap through two experiments. The findings reveal that (1) the use of AI-assisted diagnosis by doctors significantly reduces the trust of elderly patients in their doctors, (2) patients' medical literacy plays a moderating role in the relationship between the application of AI and doctors' and patients' trust, (3) proactive AI disclosure fails to mitigate trust decline more effectively than passive notification, and (4) the patients' medical literacy further moderates the impact of the disclosure manner on doctor-patient trust.