Doctors' Cancer Detection Worsens After AI Reliance, Study Finds
The integration of artificial intelligence into professional fields has long been framed as a path to “upskilling” or “reskilling” — enhancing human capabilities or retraining for new roles. Yet, a recent study presents a troubling counter-narrative: the potential for “de-skilling,” where reliance on AI may actually diminish human proficiency. New research published this week reveals that doctors who routinely employed AI for cancer detection during colonoscopies experienced a significant decline in their ability to identify cancerous growths when the AI assistance was unavailable.
The study, featured in The Lancet Gastroenterology & Hepatology journal, was conducted by an international team of medical professionals and researchers from countries including Poland, Norway, Sweden, the United Kingdom, and Japan. Its primary objective was to investigate how continuous exposure to AI might influence doctors’ performance. To achieve this, researchers assessed how endoscopists — medical specialists performing colonoscopies — fared when AI was not in use, after they had become accustomed to frequent AI support.
The findings were concerning. Doctors who regularly used AI to assist in colonoscopy procedures demonstrated a detection rate approximately six percentage points lower when they were required to perform the task independently, without AI guidance. The study tracked doctors across four endoscopy centers in Poland, which were part of a trial program exploring AI’s application in colonoscopy for cancer prevention.
This research raises critical questions about the judicious use of AI in healthcare, particularly discerning when it genuinely aids and when it might inadvertently cause harm. It arrives amidst broader concerns regarding the reliability of medical AI models. Just recently, for instance, a major technology company’s healthcare AI model was reported to have potentially “hallucinated” a body part, further fueling discussions among medical professionals about the future trajectory of AI integration in the industry. As AI continues to permeate sensitive fields like medicine, ensuring that these tools augment rather than erode human expertise will be paramount.