AI-assisted colonoscopies linked to clinician deskilling and reduced detection without AI
A Polish study of over 1,400 colonoscopies found that endoscopists who routinely used AI assistance to detect precancerous growths experienced a significant decline in detection ability when AI was not available, dropping 6 percentage points and representing about 20% worse performance than before AI was introduced. Researchers attributed this 'deskilling' effect to over-reliance on the technology reducing clinician motivation and cognitive engagement.
What happened
A study of over 1,400 non-AI-assisted colonoscopies at four endoscopy centres in Poland found that the adenoma detection rate (ADR) dropped from 28.4% in the three months before AI assistance was introduced to 22.4% in the three months after, representing a 6 percentage point decline.
The 19 experienced endoscopists in the study, each having performed more than 2,000 colonoscopies, experienced a roughly 20 per cent reduction in ability to detect tumours when AI assistance was removed compared to rates before AI was introduced.
Reported impact
The decline in ADR for non-AI-assisted colonoscopies suggests clinician deskilling from routine reliance on AI. The researchers attributed this to over-reliance leading to reduced motivation, focus, and responsibility in cognitive decision-making without AI assistance.
Endoscopists who became accustomed to AI assistance showed weakened visual search habits, altered gaze patterns, dulled pattern recognition, and reduced diagnostic confidence when performing procedures without the technology, potentially affecting patient outcomes by missing precancerous lesions.
Index analysis
While AI-assisted colonoscopy overall raised the ADR from 22.4% to 25.3%, the masking effect obscured the significant decline in performance during non-AI procedures, suggesting the apparent benefit of AI may partly reflect degradation in unassisted performance rather than genuine improvement.
The researchers compared this phenomenon to the 'Google Maps effect', where continuous reliance on decision support systems leads to atrophy of underlying skills and diminished ability to function when the tool is unavailable.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.