POTENTIAL OF LARGE LANGUAGE MODEL ChatGPT [GENERATIVE PRE-TRAINED TRANSFORMER] IN CONSTRUCTING MULTIPLE CHOICE QUESTIONS ON PHARMACOLOGY OF DIABETES

Abstract

OBJECTIVE: The process of creating high-quality multiple-choice questions (MCQs) can be time-consuming and demanding. Therefore, the purpose of this text is to investigate whether ChatGPT has the ability to generate MCQs of satisfactory quality. The topic ‘pharmacology of diabetes’ was chosen.

METHODS: The ChatGPT, a large language model based on generative pre-trained transformer technology, has been utilized as an artificial intelligence tool to create various types of multiple-choice questions (MCQs). The answers generated by ChatGPT have been recorded for further analysis.

RESULTS: ChatGPT generates pharmacology MCQs covering cognitive and affective domains with correct answers. It creates MCQs of varying difficulty, corrects mistakes, and can frame negative type and case-based MCQs. It generates 3 or 4 options unless specified otherwise.

CONCLUSION: ChatGPT can generate high-quality MCQs quickly but has limitations such as a lack of medical expertise, context comprehension, answer verification, and visual aid production. Teachers should validate ChatGPT-generated MCQs for validity and reliability, ensuring alignment with the curriculum and providing context and visual aids for better comprehension and accuracy.