Reproduce “A Survey of Knowledge Tracing: Models, Variants, and Applications” on OpenResearch

About this paper

Shuanghong Shen, Qi Liu, Zhenya Huang, Yonghe Zheng, Minghao Yin, Minjuan Wang, Enhong Chen

A comprehensive survey systematically categorizes Knowledge Tracing (KT) models, from foundational Bayesian to advanced deep learning architectures, detailing their numerous variants that incorporate realistic learning factors and illustrating their diverse applications. This work also provides open-source libraries, EduData and EduKTM, to standardize research in the field.

Key insights

• KT models show a clear progression from interpretable probabilistic methods to high-performance deep learning architectures, often with a trade-off in explainability.
• Integrating realistic learning assumptions through model variants significantly enhances the practical utility of KT by more accurately reflecting complex learning processes.
• The applicability of KT extends beyond traditional educational assessment to diverse domains such as learning resource recommendation, adaptive learning, and user modeling in games or crowdsourcing.

Summary by alphaXiv.