XKT: Toward Explainable Knowledge Tracing Model With Cognitive Learning Theories for Questions of Multiple Knowledge Concepts

Abstract

Deep learning (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DL</i>) based knowledge tracing (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">KT</i>) models have challenges for uninterpretable prediction and parameter representation in educational applications, though they achieved remarkable outcomes in predicting the exercise performance of students. This paper proposes a novel knowledge tracing model of high precision and interpretability (named <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">XKT</i>) for questions with multiple knowledge concepts based on cognitive learning theories and multidimensional item response theory (<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MIRT</i>). The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">XKT</i> consists of three differentiable network components: multi-feature embedding, cognition processing network, and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MIRT</i>-based neural predictor, which aim to provide an explainable prediction of student exercise performance. Specifically, in <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">XKT</i>, multi-feature embedding learns the rich semantic representation (e.g., knowledge distribution information) to enhance knowledge tracing using a cognition processing network. The cognition processing network performs selective perception, ability memory processing, and long-term knowledge memory processing to ensure the explainable factor representation for the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MIRT</i>-based neural predictor. Lastly, the <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MIRT</i>-based neural predictor employs psychometric parameters to interpret student exercise predictions better. Extensive experiments on four real-world datasets show that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">XKT</i> outperforms existing <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">KT</i> methods in predicting future learner responses. Moreover, ablation studies further show that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">XKT</i> offers good interpretability of student performance predictions with multiple knowledge concepts, indicating excellent potential in real-world educational applications.