A Student-Centered Approach Towards Implementing Large Language Models (LLMs) in Medical Education

Area of Science:

• Medical Education
• Artificial Intelligence
• Health Informatics

Background:

• Medical students increasingly use artificial intelligence (AI) generative text tools, commonly known as large language models (LLMs), often without formal instruction.
• Limited research exists on the self-reported experiences and educational expectations of medical students regarding LLM integration.

Purpose of the Study:

• To examine medical students' self-reported experiences with LLMs.
• To assess students' expectations for integrating LLMs into medical education.

Main Methods:

• A cross-sectional survey of 103 medical students across all training levels at two US institutions was conducted.
• Data analysis included descriptive statistics, non-parametric tests, and ordinal logistic regression to assess LLM engagement patterns.
• K-means clustering stratified students into low, moderate, and high LLM usage groups.

Main Results:

• High LLM usage correlated with greater knowledge, confidence in HIPAA compliance, and agreement on LLMs' educational benefits.
• LLMs were most valued for fact-finding, literature summarization, and differential diagnosis, but less so for flashcards.
• Students anticipated LLMs in documentation, administration, literature review, and patient education, with 86.4% endorsing formal training focused on critical thinking and ethics.

Conclusions:

• Medical students' perceptions and confidence in LLMs are linked to their usage levels.
• Students recognize LLMs' potential for administrative tasks but express concerns about learning impacts.
• Formal LLM education should prioritize critical thinking and ethical/legal considerations.