Great Question! The Art and Science of Crafting High-Quality Multiple-Choice Questions

ChatGPT as a teaching tool: Preparing pathology residents for board examination with AI-generated digestive system pathology tests

OBJECTIVES

To evaluate the effectiveness of ChatGPT 4 in generating multiple-choice questions (MCQs) with explanations for pathology board examinations, specifically for digestive system pathology.

METHODS

The customized ChatGPT 4 model was developed for MCQ and explanation generation. Expert pathologists evaluated content accuracy and relevance. These MCQs were then administered to pathology residents, followed by an analysis focusing on question difficulty, accuracy, item discrimination, and internal consistency.

RESULTS

The customized ChatGPT 4 generated 80 MCQs covering various gastrointestinal and hepatobiliary topics. While the MCQs demonstrated moderate to high agreement in evaluation parameters such as content accuracy, clinical relevance, and overall quality, there were issues in cognitive level and distractor quality. The explanations were generally acceptable. Involving 9 residents with a median experience of 1 year, the average score was 57.4 (71.8%). Pairwise comparisons revealed a significant difference in performance between each year group (P < .01). The test analysis showed moderate difficulty, effective item discrimination (index = 0.15), and good internal consistency (Cronbach's α = 0.74).

CONCLUSIONS

ChatGPT 4 demonstrated significant potential as a supplementary educational tool in medical education, especially in generating MCQs with explanations similar to those seen in board examinations. While artificial intelligence-generated content was of high quality, it necessitated refinement and expert review.

Large Language Models as Tools to Generate Radiology Board-Style Multiple-Choice Questions

RATIONALE AND OBJECTIVES

To determine the potential of large language models (LLMs) to be used as tools by radiology educators to create radiology board-style multiple choice questions (MCQs), answers, and rationales.

METHODS

Two LLMs (Llama 2 and GPT-4) were used to develop 104 MCQs based on the American Board of Radiology exam blueprint. Two board-certified radiologists assessed each MCQ using a 10-point Likert scale across five criteria-clarity, relevance, suitability for a board exam based on level of difficulty, quality of distractors, and adequacy of rationale. For comparison, MCQs from prior American College of Radiology (ACR) Diagnostic Radiology In-Training (DXIT) exams were also assessed using these criteria, with radiologists blinded to the question source.

RESULTS

Mean scores (±standard deviation) for clarity, relevance, suitability, quality of distractors, and adequacy of rationale were 8.7 (±1.4), 9.2 (±1.3), 9.0 (±1.2), 8.4 (±1.9), and 7.2 (±2.2), respectively, for Llama 2; 9.9 (±0.4), 9.9 (±0.5), 9.9 (±0.4), 9.8 (±0.5), and 9.9 (±0.3), respectively, for GPT-4; and 9.9 (±0.3), 9.9 (±0.2), 9.9 (±0.2), 9.9 (±0.4), and 9.8 (±0.6), respectively, for ACR DXIT items (p < 0.001 for Llama 2 vs. ACR DXIT across all criteria; no statistically significant difference for GPT-4 vs. ACR DXIT). The accuracy of model-generated answers was 69% for Llama 2 and 100% for GPT-4.

CONCLUSION

A state-of-the art LLM such as GPT-4 may be used to develop radiology board-style MCQs and rationales to enhance exam preparation materials and expand exam banks, and may allow radiology educators to further use MCQs as teaching and learning tools.