Can ChatGPT generate surgical multiple-choice questions comparable to those written by a surgeon?

Abstract

Background: This study aimed to determine whether surgical multiple-choice questions generated by ChatGPT are comparable to those written by human experts (surgeons). Methods: The study was conducted at a medical school and involved 112 fourth-year medical students. Based on five learning objectives in general surgery (colorectal, gastric, trauma, breast, thyroid), ChatGPT and surgeons generated five multiple-choice questions. No change was made to the ChatGPT-generated questions. The statistical properties of these questions, including correlations between two group of questions and correlations with total scores (item discrimination) in a general surgery clerkship exam, were reported. Results: There was a significant positive correlation between the ChatGPT-generated and human-written questions for one learning objective (colorectal). More importantly, only one ChatGPT-generated question (colorectal) achieved an acceptable discrimination level, while other four failed to achieve it. In contrast, human-written questions showed acceptable discrimination levels. Conclusion: While ChatGPT has the potential to generate multiple-choice questions comparable to human-written ones in specific contexts, the variability across surgical topics points to the need for human oversight and review before their use in exams. It is important to integrate artificial intelligence tools like ChatGPT with human expertise to enhance efficiency and quality.

Ähnliche Arbeiten