New Model Improves Multi-Domain Knowledge Tracing

This research addresses a significant challenge in educational technology: accurately assessing student knowledge in multi-domain learning environments. Traditional knowledge tracing (KT) methods often focus on single domains, overlooking the complexities introduced by learning across multiple subjects simultaneously, specifically cognitive load and knowledge transfer. The proposed method, LT-MKT (incorporating cognitive Load and knowledge Transfer for Multi-domain Knowledge Tracing), tackles these factors. It first uses large language models (LLMs) to integrate textual information from questions and concepts, creating a Multi-domain Hierarchical Graph that bridges isolated domains. This graph helps represent the relationships between different knowledge areas. LT-MKT then explicitly models cross-domain features related to cognitive load in both temporal and knowledge dimensions. Additionally, a dedicated knowledge transfer module tracks how knowledge states in one domain influence others. By jointly considering these elements, the model achieves state-of-the-art performance in predicting students' future learning outcomes on real-world datasets.