Medicine is not a 64-square board: human oversight requires independent judgment

Agreement: I Agree Body: Dear Editor, Morgan asks an important question about what doctors are for when artificial intelligence can already outperform clinicians in selected tasks. His chess analogy is powerful, but it also has limits. Chess has fixed rules, all the pieces are visible, and there is a clear objective. Clinical medicine rarely offers that certainty. Information may be incomplete, physiology can change quickly, several risks may compete, and the best course often depends on what matters to the individual patient. Morgan is right to ask whether adding a human can sometimes make a high-performing AI system worse. He also suggests that doctors will remain the emergency “manual override.” But there is another problem worth considering: seeing the AI answer first may shape the clinician’s own judgment, even when the system is working as intended. Being “in the loop” is not the same as providing meaningful oversight. In experimental work, inaccurate decision support has been shown to worsen clinicians’ diagnostic accuracy [1]. More recent evidence suggests that showing an AI diagnosis before the clinician has made an independent decision may produce stronger anchoring bias [2]. A clinician signing off an AI recommendation does not necessarily mean that an independent clinical judgment has taken place. This makes the way AI is introduced into clinical work important. For some high-risk decisions, it may be worth considering whether the clinician should form an initial assessment before seeing the AI recommendation. Systems should also make uncertainty clear and allow disagreements between the clinician and AI to be recognised and reviewed. The aim is not to force doctors to overrule AI when the technology performs better, but to make sure that human oversight means more than a final sign-off. AI can estimate probabilities, organise information, and support clinical decisions. But it does not remove the need for human judgment when decisions involve competing risks, values, or ethical concerns [3]. If doctors are expected to explain, supervise, or override AI-assisted decisions, they must still be able to reach and explain their own judgment rather than simply repeat the machine’s answer. The answer to Morgan’s question is therefore not that doctors must beat AI at every cognitive task. When human oversight is required, the clinician must still be able to think independently. Medicine is not won by finding one best move. It is practised by deciding what is best for this patient, at this moment, in circumstances that rarely fit neatly on a 64-square board. References 1. Gaube S, Suresh H, Raue M, et al. Do as AI say: susceptibility in deployment of clinical decision-aids. npj Digit Med 2021;4:31. doi:10.1038/s41746-021-00385-9. 2. Xu X, Hu H, Zhang H, et al. Divergent impacts of explainable AI for dermatological diagnosis on clinicians versus lay people. Nat Med 2026;32:3000–3009. doi:10.1038/s41591-026-04553-w. 3. Char DS, Shah NH, Magnus D. Implementing machine learning in health care—addressing ethical challenges. N Engl J Med 2018;378:981–983. doi:10.1056/NEJMp1714229. No competing Interests: Yes The following competing Interests: Electronic Publication Date: Tuesday, September 15, 2026 - 13:06 AI use: Yes I have used AI AI use details: AI was used for grammar and spelling correction only Highwire Comment Subject: Matt Morgan: After checkmate—if AI is better, what are doctors for? Workflow State: Released Full Title: Medicine is not a 64-square board: human oversight requires independent judgment Highwire Comment Response to: Matt Morgan: After checkmate—if AI is better, what are doctors for? Check this box if you would like your letter to appear anonymously:: Last Name: Modi First name and middle initial: Dr Girishkumar Email: girishmodi75@gmail.com Address: Al Ain, United Arab Emirates Occupation: Anesthesiology Affiliation: Tawam Hospital BMJ: Additional Article Info: Rapid response