Analyse comparative entre la génération automatique de questionnaires à choix multiples par ChatGPT et le corps enseignant pour l’évaluation de l’apprentissage

Abstract

This study, conducted among university students, examines the use of ChatGPT-4 in the creation of multiple-choice questions (MCQs) by comparing AI-generated MCQs to those designed by instructors and analyzing the quality of distractors and the difficulty indices. The study reveals that AI-generated MCQs achieve high success rates but feature less effective distractors, limiting the possibility of assessing analytical thinking skills. This research highlights the importance of human intervention to ensure high-quality MCQs while acknowledging the potential of AI as a complementary tool.