When the Algorithm Teaches — Promise and Peril of AI in Physician Learning

Journal of the American Medical Association (05/21/26) McMahon, Graham T.

A new generation of clinician-facing, large language model tools is transforming physician learning by delivering instant, patient-specific guidance that feels authoritative but is not always complete, nuanced or correct, writes Graham T. McMahon, MD, MMSc, president and CEO of the Accreditation Council for Continuing Medical Education. This, in turn, creates risks of overreliance, automation bias, and erosion of clinical reasoning. Although these systems promise more timely, embedded, and adaptive education that could reduce errors and standardize evidence-based care, their fluent explanations can produce an illusion of understanding, encourage uncritical acceptance, and contribute to deskilling as clinicians rely on artificial intelligence (AI) rather than engaging in deliberate reasoning or shared professional dialogue. McMahon warns that commercial influence, opaque system design, and solitary AI-mediated learning could undermine trust and professionalism, emphasizing that AI should augment — not replace — clinical judgment. To preserve expertise, clinicians will need explicit education about system limitations, active evaluation of outputs, and continued engagement in reflective, social learning, while developers and accrediting bodies ensure transparency, independence, and designs that support rather than supplant clinician judgment. "Artificial intelligence should extend, not replace, physician learning," McMahon concludes. "Maintaining clinical expertise will require a continued commitment to learning that cannot be delegated."

Read More - May Require Paid Subscription