Knowledge and Information Systems Взаимодействие человека и компьютера

A knowledge tracking model based on the potential knowledge states and consistency of learning benefits

Luo, Xianzeng, Wang, Ruxia, Meng, Yu
Springer Nature · 2026-11-14 · DOI 10.1007/s10115-026-02940-w

Аннотация

Knowledge tracking, which models learners’ evolving mastery of knowledge points, serves as a foundational technology for intelligent tutoring systems. To overcome the limitations of existing approaches—namely, their insufficient exploration of latent knowledge states, inconsistent handling of learning gains, and neglect of inter‑knowledge‑point correlations—we propose PKS‑CLB (Potential Knowledge States and Consistency of Learning Benefits), a knowledge tracking model that aligns potential knowledge states with learning gain consistency. The model employs a multi‑head attention mechanism to capture knowledge state improvements following both correct and incorrect responses and utilizes a knowledge graph to infer learners’ latent knowledge states. A bidirectional interaction module then produces a unified final knowledge state representation, which is used to predict performance on the next test question. Evaluated on a real‑world educational dataset, PKS‑CLB achieves superior predictive performance compared to state‑of‑the‑art baselines, demonstrating its effectiveness in advancing personalized learning support.

Читать у издателя →