Design and theoretical validation of a hybrid cognitive distractor generation model within a generative AI-based multi-stage assessment framework

When students get multiple-choice questions wrong, it s often not because they are careless but because they misunderstand key ideas. This study explores a better way to design assessments that not only measure learning but also help students understand their mistakes and improve.

At the heart of this research is a model called the Cognitive Distractor Generation Model (CDGM). It helps create incorrect answer choices ( distractors ) that are not random, but thoughtfully designed to reflect common student misconceptions. For example, if students often confuse square roots with division, the system includes a distractor that mirrors that mistake. This allows teachers and the system to diagnose what a student misunderstood based on the option they chose.

This model was built into a flexible, AI-powered testing system called GAMSA, which adjusts to each student s performance. When a student selects an incorrect answer, generative AI (like ChatGPT) instantly provides personalized feedback that explains why the answer is wrong and how to think differently. Students can then choose to try again immediately, applying what they ve learned.

Experts reviewed the system and confirmed its potential to help teachers better understand students' thinking. While the tool still needs to be tested with real students, this research lays a foundation for the future of assessment where tests don t just grade students, but guide them. It offers a vision of education where every wrong answer becomes a meaningful learning opportunity.