Psychometrics-Guided Adaptive Quizzing: A Study Combining LLM Item Generation and Psychometric Calibration

Ashish K.C (Northwest Missouri State University), Rojina Raut (Northwest Missouri State University)

While personalized assessment is a gold standard in education, most university STEM courses still rely on one-size-fits-all quizzes. This creates a dual sided engagement problem: high ability students face boredom from lack of challenge, while struggling students face frustration from excessive difficulty. Furthermore, current AI-driven educational tools often prioritize content generation over psychometric rigor. There is a distinct lack of field-deployable frameworks that can generate quiz items while simultaneously calibrating their difficulty using empirical student data. This study seeks to bridge this gap by developing and testing a system that combines LLM-based item generation with longitudinal psychometric calibration to provide a truly adaptive learning experience.