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.
Large Language Models and User Trust: Consequence of Self-Referential Learning Loop and the Deskilling of Health Care Professionals
Citation. Choudhury, A., & Chaudhry, Z. (2024). Large Language Models and User Trust: Consequence of Self-Referential Learning Loop and the Deskilling of Health Care Professionals. Journal of Medical Internet Research, 26, e56764. https://doi.org/10.2196/56764
Summary. This viewpoint examines how clinician trust in large language models can shape reliance, verification, and professional skill retention in health care. It argues that productive use requires clinicians to offload routine work while retaining expert critical oversight, and it identifies feedback from AI-generated training data as a theoretical risk to model performance and data diversity.
Key claims
- The authors argue that undue clinician trust in LLM recommendations can foster automation bias, including confirmation and precautionary biases that weaken scrutiny of subsequent information.
- Repeated LLM reliance for critical tasks may erode clinicians’ diagnostic and critical-thinking skills, particularly for trainees, although the article presents this as a prospective risk rather than a measured effect.
- The proposed self-referential learning loop describes a theoretical pathway in which AI-generated content increasingly enters training data, reducing diversity, entrenching bias, and potentially degrading model performance.
- LLM integration in clinical work should preserve human validation, contextual expertise, transparency about model limits, and accountability for erroneous advice.
Evidence & limitations
- Peer-reviewed viewpoint article that synthesizes conceptual risks and prior literature rather than reporting a new empirical study or causal estimates.
- Its clinical framing may not generalize directly to other knowledge-work settings, and its deskilling and self-referential-loop claims are explicitly prospective.