AI-Enhanced Exam Generator Program: A Case Study in Live University Exam Settings

Abstract

Creating exams is time-consuming for educators, and despite existing tools, no solution has been universally adopted. This study evaluates EGAL+, a hybrid artificial intelligence and metaheuristics-based exam generation tool, in real university exam settings. Students were randomly assigned to traditional or EGAL+-generated exams. Student performance and exam quality were assessed using objective metrics, and qualitative feedback from teachers and students. Results show that EGAL+ significantly reduces exam preparation time without harming student performance, while improving exam quality through better alignment with teachers’ preferences, greater question diversity, and more consistent difficulty. These findings indicate that EGAL+ reduces teacher workload while maintaining or enhancing exam quality, with no observed drawbacks.