Not yet human-reviewed
This source record was written automatically by the research pipeline and has not yet been reviewed by the maintainer or integrated into a topic page. Treat it as a provisional summary and verify claims against the original source.
Behavioral Impacts of AI Reliance in Diagnostics: Balancing Automation with Skill Retention
Citation. Sunday, O. (2025). Behavioral Impacts of AI Reliance in Diagnostics: Balancing Automation with Skill Retention. Epidemiology and Health Data Insights, 1(3), ehdi011. https://doi.org/10.63946/ehdi/16894
Summary. This narrative review examines how diagnostic AI can improve efficiency, accuracy, and confidence while increasing risks of automation bias, reduced situational awareness, and diagnostic deskilling. Drawing examples from radiology, pathology, laboratory medicine, clinical decision support, and aviation automation, it argues that AI should augment rather than replace clinicians’ judgment. It proposes recurring hands-on retraining, hybrid decision protocols, curriculum reform, explainable AI, and human-in-the-loop governance as safeguards for skill retention and patient safety.
Key claims
- The review argues that routine reliance on AI-generated diagnostic outputs can reduce clinicians’ independent reasoning, interpretive practice, situational awareness, and willingness to question automated recommendations.
- It identifies automation bias and complacency as mechanisms by which confidence in algorithmic outputs can displace critical evaluation, particularly for trainees and in high-stakes diagnostic settings.
- It proposes that hybrid decision-making, recurrent hands-on retraining, AI-integrated curricula, explainability, and retained human oversight can preserve diagnostic expertise while using AI support.
Evidence & limitations
- This is a narrative review and conceptual recommendations paper rather than a new empirical study; its claims draw on cited literature and examples across diagnostic domains and aviation automation.
- Its conclusions concern clinical diagnostics and may not transfer directly to other professional settings; it does not estimate effect sizes or test the proposed safeguards.
- The manuscript discloses use of ChatGPT-4 and Genmi for data synthesis, writing assistance, and formatting; the authors state they verified the factual information, interpretations, and final conclusions.