Medical study shows that doctors who rely on AI get worse at performing diagnoses without it

You wouldn’t trust a mechanic who struggled to identify a fuel injector, so why would you entrust your health to a doctor with similar difficulties? A new study conducted on doctors in Poland, and published by The Lancet, suggests that routine reliance on AI can diminish medical professionals’ skills when the crutch is removed.

The outcome of the study, which “assessed how endoscopists who regularly used AI performed colonoscopy when AI was not in use”, was not great. Those medical professionals who regularly relied on AI became less capable over time than those who didn’t.

AI capability

The study assessed the changes in adenoma detection rate, or ADR, during “standard, non-AI-assisted colonoscopy before and after AI exposure.” Assessments were performed between September 2021 and March 2022 and led to the “ADR of standard colonoscopy decreas[ing] significantly from 28·4% (226 of 795) before to 22·4% (145 of 648) after exposure to AI, corresponding with an absolute difference of –6·0%.”

A six percent downward change might not sound like much until you think about it. A martial artist who loses 6% of their acquired skills would be at a considerable disadvantage. The same goes for any task. Whether an accountant were to lose 6% of their numerical literacy or a person offloading a truck were to experience the same drop, the effects can be devastating. This is even more pronounced in a medical setting, where lives tend to be on the line.

It’s important to note that the study isn’t exhaustive. Extensive use of OpenAI’s biggest (and only) product “might reduce the ADR of standard non-AI assisted colonoscopy, suggesting a negative effect on endoscopist behaviour,” which doesn’t speak to its effect on the rest of the medical community. But, taken with Microsoft’s recent studies suggesting “cognitive decline” can be brought on by reliance on artificial intelligence, the current high-speed drive toward widespread adoption of the tech could surface more harmful effects than it claims to mitigate. Maybe don’t rely on ChatGPT for that university entrance essay, yeah?