Unexpected Side Effects of AI Use Emerge Among Doctors
1. De-skilling in Colonoscopy Detection
A recent study published in The Lancet Gastroenterology & Hepatology found that doctors who regularly relied on AI-assisted colonoscopy tools experienced a notable drop in cancer detection rates when performing procedures without AI. Specifically, the adenoma detection rate declined from 28% to 22%, highlighting a concerning reduction in diagnostic proficiency due to over-reliance on AI .
2. Skill Decline Confirmed in Broader Study
Similarly, an observational study from Poland involving 1,400 patients confirmed that routine AI use in colonoscopies impaired physicians’ skills. Detection rates without AI dropped from 28.4% to 22.4%, while AI-assisted procedures maintained a 25.3% rate. Researchers urge caution and recommend deliberately incorporating non‑AI intervals to preserve doctors’ diagnostic capabilities.
3. Automation Bias in Pathology
In computational pathology, another study revealed that under time pressure, pathologists exhibited a 7% automation bias, where initial correct evaluations were overturned due to faulty AI advice. While AI integration improved overall performance, reliance on incorrect automated guidance led to decision errors.
Summary
Frequent use of AI tools in medical diagnostics may inadvertently weaken doctors’ independent judgment and observational skills—particularly in tasks like colonoscopy—while introducing cognitive vulnerabilities to misleading AI outputs. This underscores the necessity of balancing AI-assisted care with ongoing human expertise and robust safeguards.