Abstract
The increasing use of generative large language models (LLMs) necessitates a fundamental reevaluation of traditional didactic lectures in medical education, particularly within psychiatry. The specialty's inherent diagnostic ambiguity, biopsychosocial complexity, and reliance on nuanced interpersonal skills demand an educational model that transcends mere information transfer, focusing instead on cultivating sophisticated clinical reasoning. This viewpoint argues for a shift from passive knowledge transmission to active, facilitated development of higher-order thinking, aligning with the Bloom taxonomy. We describe four core propositions: (1) shifting foundational knowledge acquisition to faculty-curated asynchronous artificial intelligence (AI)-assisted micromodules; (2) transforming synchronous time into "Ambiguity Seminars" for discussing nuanced cases, biopsychosocial formulation, and ethical dilemmas, leveraging faculty expertise in guiding reasoning; (3) integrating live LLM critical interaction drills to develop prompt engineering skills and critical appraisal of AI outputs; and (4) realigning assessment methods (eg, objective structured clinical examinations [OSCEs], reflective writing) to evaluate clinical reasoning and integrative skills rather than rote recall. Successful implementation requires comprehensive faculty development, explicit institutional investment, and a phased approach that addresses scalability across varying resource settings. This reimagined approach aims to cultivate clinical wisdom, equipping psychiatric trainees with adaptive reasoning frameworks essential for excellence in an AI-mediated future.