Radiologists Exhibited Automation Bias in AI Study
A new study reveals that junior radiologists are more susceptible to errors when using AI-assisted bone age assessments.
Updated on Sept. 28, 2026 in Artificial Intelligence
Live Poll
Do you trust artificial intelligence outputs when you receive medical diagnoses from a doctor?
A randomized crossover study found that five out of six participating radiologists experienced significant performance differences when using accurate versus sham AI. The research highlights how automation bias can influence clinical decision-making during bone age assessment.
Why it matters
Understanding how clinicians interact with AI is essential to ensuring diagnostic accuracy. This research exposes the potential for automation bias to skew medical interpretations, particularly when AI outputs deviate significantly from reality.
The study utilized a sample of 200 radiographs per participant, with error rates increasing when AI discrepancies exceeded 6 months. Five participants showed significant shifts in accuracy with a p-value of 0.001, while the highest-performing radiologist remained steady.
The details
Researchers compared radiologist assessments using both accurate and randomized sham AI outputs to measure susceptibility to automation bias. The findings indicate that while senior radiologists frequently challenged AI, junior staff were more prone to accepting incorrect automated guidance.
Timeline
September 28, 2026: The research findings were published.
The Tech Race
This study underscores a critical shift in the evolution of medical diagnostics as AI transitions from a theoretical tool to a daily clinical assistant. It highlights the growing tension between relying on automated processing power and the necessity of human oversight in high-stakes fields.
This study suggests that patients may soon see tighter requirements for dual-review processes when AI is used in radiology clinics. It emphasizes that while AI can improve efficiency, its current limitations require clinicians to maintain rigorous, independent assessment workflows.
The takeaway
Clinicians must remain vigilant against trusting AI outputs without verification, especially when automated results differ by more than six months from their own clinical judgment. Future diagnostic workflows should prioritize strategies that mitigate automation bias to ensure consistent patient care.
Further reading
For broader context on how machine learning integrates into medical fields, visit the Artificial Intelligence section.
More information
View the complete results and methodology in the peer-reviewed research article.
Live Poll
Do you trust artificial intelligence outputs when you receive medical diagnoses from a doctor?
