A large observational study from Poland suggests that routine use of AI for polyp detection may erode clinicians’ own diagnostic skills, raising concerns about patient outcomes. Published in The Lancet Gastroenterology & Hepatology, the research showed a 6% absolute drop in adenoma detection rates (ADR) when experienced endoscopists performed colonoscopies without AI after prolonged AI exposure.
Key Takeaways
Unexpected skill erosion
Even highly experienced endoscopists (each with >2000 colonoscopies performed) showed reduced ADR after integrating AI. The effect was most pronounced in centers with higher baseline ADRs and in younger female patients.
AI boosts, but also conditions
AI-assisted colonoscopies still showed higher detection rates (25.3%) compared to post-AI, non-assisted procedures (22.4%). Authors suggest clinicians may become conditioned to “rely” on AI, lowering vigilance when it isn’t present—much like drivers in self-driving cars.
A broader warning for medicine
Editorial commentary stresses that this phenomenon isn’t unique to GI. Overreliance on automation may dull human attention and reasoning, highlighting the need to integrate AI as a partner tool, not a substitute for clinical judgment.