Causal-Driven Neural Cognitive Diagnosis and Learning Attribution for Educational Data Analysis

Cognitive diagnosis is a key issue in the field of educational data mining. It primarily focuses on revealing students’ mastery of knowledge points. In recent years, neural network-based cognitive diagnosis has attracted increasing attention from researchers, with some studies also involving the modeling of relationships between knowledge points. However, current cognitive diagnosis methods often emphasize the accuracy of diagnostic results, making it difficult to identify key factors affecting students’ learning from a causal perspective.To address this issue, this paper proposes a diagnostic attribution process to tackle the problem from two aspects. First, we propose a Causal-Driven Neural Cognitive Diagnosis (CDNCD) model, to improve diagnostic results and model interpretability. Then, leveraging both the diagnostic results and causal graphs as prior inputs, we design an algorithm based on causal effects to precisely attribute students’ learning difficulties, identifying the prerequisite knowledge points that learners need to improve the most.These findings contribute to educational data mining by offering a causal-driven perspective for student assessment, facilitating more effective data-driven decision-making in education.