Automation Bias in Large Language Model-Assisted Diagnostic Reasoning Among Physicians Trained in AI Literacy
OVERVIEWA randomized clinical trial published in PubMed Generative AI MeSH demonstrates that physicians exhibit substantial automation bias when exposed to erroneous Large Language Model (LLM) recommendations, even when AI literacy training is provided and consultation is voluntary. Key Finding: Physicians demonstrate a strong tendency to defer to LLM suggestions despite prior training in AI literacy, suggesting that awareness alone does not mitigate cognitive reliance on AI systems. This finding highlights critical safety risks that require more robust validation frameworks and regulatory safeguards before widespread clinical AI deployment. Clinical Implications: The study underscores that automation bias represents a fundamental human factors challenge in clinical AI adoption. Physicians' susceptibility to LLM errors—even among trained cohorts—indicates that clinical governance frameworks must go beyond literacy initiatives to include system-level safeguards, such as mandatory verification protocols and clinical decision support architecture that prevents unchecked automation. Regulatory Gaps: Current FDA frameworks for clinical AI do not adequately address automation bias as a use-related hazard. The findings support calls for enhanced pre-market validation that specifically tests physician decision-making under conditions of conflicting AI recommendations and clinician expertise. ---
Subscribe to read full wiki analysis
Full wiki content — including source provenance, contradiction tracking, related pages, and analyst notes — is available to Subscriber members.
View Plans & SubscribeNew users get a 7-day trial — no credit card required.