Multi-Agent Graph-Enhanced Knowledge Tracing (MAGE-KT) is a sophisticated framework designed to improve the accuracy of Knowledge Tracing (KT) models, which aim to predict a student's future performance based on their learning history. At its core, MAGE-KT addresses the challenge of effectively representing complex relationships among students, questions, and knowledge concepts (KCs) within educational data. It works by constructing a multi-view heterogeneous graph that integrates both semantic relationships between KCs (extracted by a multi-agent system) and behavioral interactions between students and questions. This approach allows MAGE-KT to capture richer, more complementary signals than methods relying solely on interaction sequences. By retrieving compact, high-value subgraphs tailored to a specific student's history and fusing them via an Asymmetric Cross-attention Fusion Module, MAGE-KT enhances prediction fidelity while mitigating the computational cost and noise associated with full-graph encoding. This innovation is crucial for developing more precise and personalized adaptive learning systems, benefiting researchers in educational data mining, intelligent tutoring systems, and adaptive learning platforms.
MAGE-KT is a new method for predicting how well students will perform on future questions by better understanding their learning. It builds a special kind of graph that combines different types of information about students, questions, and knowledge topics. This helps it make more accurate predictions without getting bogged down by too much irrelevant data.
Multi-Agent Graph-Enhanced Knowledge Tracing
Was this definition helpful?