Original title: Trust in AI Versus Human Clinicians: A Study of Diagnostic Judgements and Transparency
Ye Yang, Samuel Bennett · September 2026
Doctors face heavy workloads, and healthcare systems want to use artificial intelligence to help diagnose illnesses. However, people must trust these systems before clinics can adopt them. Researchers ran an experiment to test how people react to diagnoses from computers versus human doctors. They gave participants written scenarios where an artificial intelligence and a doctor either agreed on a diagnosis or disagreed with each other. The researchers then asked people about their trust, their preferences, and their worries about legal liabilityWho is legally responsible if something goes wrong. and transparencyThe ability to see and understand how a decision was made., which means knowing how the system makes decisions. The study found that people trust human doctors more than artificial intelligence. However, when the computer and the human doctor agree, trust in the final diagnosis goes up. Finally, people who worry more about transparency trust medical artificial intelligence less. Overall, trust depends on agreement with human experts and clear explanations of how the technology works.
You ask two friends for advice on fixing your bike. One friend explains step by step; the other just says 'tighten the bolt'. You trust the explaining friend more because you understand the reasoning. Even if both give the same advice, you're less confident in the quiet one. And if you're the kind of person who really needs to know why, you trust the quiet one even less. Here, the explaining friend is the human clinician, the quiet one is the AI, and the study finds that trust depends on both agreement and the ability to understand the logic.
As AI is considered for clinical use, patient trust becomes a critical barrier. This study finds that people trust human doctors more, and even when AI agrees with them, concerns about transparency still reduce trust. For AI to be accepted, developers and hospitals must prioritize explaining how AI reaches decisions and who is responsible for errors.
The abstract describes a vignette-based study, so findings may not generalize to real clinical encounters. Causal conclusions cannot be drawn because transparency concerns were measured, not manipulated. The sample and specific AI system are not detailed, limiting applicability.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Trust in AI Versus Human Clinicians: A Study of Diagnostic Judgements and Transparency — Ye Yang, Samuel Bennett — PsyArXiv — doi:10.31234/osf.io/4cbwe_v1
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.