Abstract:

This paper introduces adaptive exercises module in the open-source state-of-the-art APAS (Automated Programming Assessment System) Edgar. Adaptive exercises module is designed to facilitate student practice within existing course materials in a guided and adaptive manner, with minimal intervention from instructors. It leverages accumulated course questions which are categorized into three difficulty levels (easy, medium, hard). Students engage in a guided practice session, starting with easier questions and progressing based on their performance. This paper has a twofold contribution: we present this novel addition to our open-source APAS and address two key challenges: question difficulty categorization and adaptivity model governing student progression. To categorize questions efficiently, we utilize usage statistics from previous exams and employ three models of question difficulty: Naïve, 1-Parameter Logistic Model, and 2-Parameter Lord’s model from IRT (Item Response Theory). We present an algorithm for categorizing questions based on these models, where system suggests difficulty levels based on these models, with the instructor having the final say. Then we present the adaptivity model which defines the transition between difficulty levels based on parameters like the number of questions tracked and the number of correct answers required for progression (or regression). Proposed adaptivity model is flexible and can be modified or extended as needed. The system logs adaptive exercises activity and provides instructors with a dashboard for monitoring student progress and feedback. Overall, this module offers a structured yet flexible approach to student practice, with the final goal of enhancing learning outcomes.

Keywords:

e-learning, adaptive exercises, APAS, IRT, Edgar.