A Causality-Based Interpretable Cognitive Diagnosis Model

Cognitive diagnosis model (abbr.CDM) aims to assess students’ cognitive processes during learning, enabling personalized support based on their needs. Nevertheless, deep learning-based CDMs are inherently opaque, posing challenges in providing psychological insights into the reasoning behind predicted outcomes. We address this by creating three interpretable parameters: skill mastery, exercise difficulty, and exercise discrimination. Inspired by Bayesian networks and neural networks, we use feature engineering for extraction of interpretable parameters and tree-enhanced naive Bayes classifiers for prediction. Our method balances interpretability and accuracy. Experimentally, we compare our approach to traditional and advanced models on four datasets, analyzing each feature’s impact. We conduct ablation studies on each feature to examine their contribution to student performance prediction. Thus, causality-based interpretable cognitive diagnosis model (CBICDM) has great potential for providing adaptive and personalized instructions with causal reasoning in real-world educational systems.