Doctors show reduced cancer detection skills after AI tool removal. (So don’t remove it?)

New research reveals that doctors using AI tools for colonoscopies became significantly worse at detecting precancerous growths when the technology was removed, marking the first evidence of “deskilling” in medical AI. The study, published in Lancet Gastroenterology and Hepatology, found that after just three months of AI assistance, physicians’ detection rates dropped from 28% to 22% when performing procedures without the technology.

What happened: Researchers at four Polish endoscopy centers gave doctors access to an AI tool that flagged suspicious growths during colonoscopies by drawing boxes around them in real time.

• After three months of using the AI assistance, the technology was removed to measure the doctors’ independent performance.
• Detection rates fell below their original baseline, dropping from about 28% to 22% of colonoscopies.
• This represents the first documented case of AI causing “deskilling” — the erosion of fundamental medical skills due to technology dependence.

The big picture: While AI tools have demonstrated clear benefits in medical diagnostics, including faster diagnoses and improved cancer detection, this study reveals a potential hidden cost of relying too heavily on automated assistance.

• Previous studies have consistently shown that AI-assisted colonoscopies significantly improve doctors’ detection rates of precancerous growths.
• The technology has been widely celebrated for enhancing physician performance across multiple medical specialties.

What they’re saying: Medical experts are noting the bidirectional nature of human-AI collaboration in healthcare settings.

• “This is a two-way process,” said Dr. Omer Ahmad, a gastroenterologist at University College Hospital London who published an editorial alongside the study. “We give A.I. inputs that affect its output, but it also seems to affect our behavior as well.”

Why this matters: The findings raise important questions about how medical professionals should integrate AI tools while maintaining their core diagnostic skills.

• Detection rates of precancerous growths serve as a widely accepted indicator of an endoscopist’s performance.
• The study suggests that healthcare systems may need to develop training protocols that prevent skill degradation while still leveraging AI’s benefits.