Modeling Student Knowledge Components

• Magnus Culley, Computer Science, University of Delaware

• John Aromando, Computer Science, University of Delaware

Introductory Computer Science courses often have large student-to-instructor ratios, making it difficult to evaluate how assessments perform across students with different levels of understanding. One potential solution is to model student knowledge using artificial intelligence to simulate how students with varying mastery of key knowledge components would respond to exam questions. The goal of this research is to develop AI models that represent different levels of student understanding and use them to evaluate the difficulty and effectiveness of assessments before they are administered. Existing research has explored AI in education and student modeling, but there is limited work on using knowledge component-based models to simulate student performance on computer science exams. We are training and evaluating models using historical student work and measuring whether they exhibit realistic patterns of understanding across different knowledge components. This research will help determine whether AI-generated student models can become a practical tool for improving assessment design in introductory computer science courses.