October 18, 2025JAMA OphthalmologyOpen Access

Leveraging Large Language Models to Generate Multiple-Choice Questions for Ophthalmology Education

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Survey study evaluates large language models generating MCQs for ophthalmology, indicating potential to enhance training materials.

Quality of MCQs generated by GPT-4 was comparable to expert committee output, achieving median scores of 9.
LLM-generated questions showed moderate inter-rater reliability with an ICC of 0.63, indicating similar evaluations among graders.
Nearly 95% of generated MCQs had limited resemblance to existing questions, suggesting novelty in content creation.
Readability scores for questions from both sources were similar, reflecting equal clarity in comprehension.

Gholami et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark