Is AI ruining our skills? Early results are in — and they’re not good
Studies show that medical specialists’ ability to detect lesions in endoscopy drops significantly when AI assistance is removed, and similar trials are underway in software engineering. Experts warn that over-reliance on AI can impair hard-won skills and call for research into ways to preserve human expertise.
- Polish endoscopy specialists' adenoma-detection rate dropped from 28.4% to 22.4% when AI assistance was switched off after they'd grown used to it.
- 70% of US nurses and 77% of physicians already worry AI reliance is eroding their clinical skills.
- Anthropic's trial with 52 software engineers found those who habitually used AI took longer and made more errors solving problems without it.
- Researchers like Yuichi Mori say there's no clear fix yet and want "deskilling" made a top research priority.
Seventy per cent of US nurses and 77 percent of physicians told researchers they worry that relying on AI could weaken their clinical skills. That anxiety isn’t just theoretical. In Poland, a team of endoscopy specialists—all with at least 2,000 colonoscopies under their belts—used an AI tool that highlights adenomas in real time. Before the system arrived, those doctors spotted at least one precancerous lesion in 28.4 percent of procedures. After adding AI assistance, their adenoma‐detection rate fell to 22.4 percent on days when the tool was switched off. The drop suggests experts quickly lose some of their sharpness when they start offloading tasks to algorithms.
Robert Wachter of UCSF, whose book examines AI’s role in health care, points out that over time clinicians may grow “less motivated, less focused, and less responsible” without their digital aide. Yuichi Mori at the University of Oslo, a co-author on the colonoscopy paper, says no clear fixes exist yet—and he wants deskilling to become a top research priority over the next decade.
In computer science, Anthropic ran a small randomized trial with 52 software engineers tackling a basic coding challenge. Everyone could look up documentation online, but half the group also received prompts to lean on an AI assistant. Early results hint that coders who habitually summoned AI help took longer to solve problems independently and made more errors when the assistant wasn’t at hand.
Researchers are sounding the alarm: as AI tools become ubiquitous, experts in medicine, programming and beyond could slowly, almost imperceptibly, lose the edge they built through years of practice. At stake isn’t just efficiency, but the very depth of human expertise that machines are designed to support.