Wu, S., Xu, C., Zhang, W. (2026) — arXiv preprint.
Analysis
This paper introduces PLKT (Probabilistic Logical Knowledge Tracing), which replaces deterministic vector embeddings with beta-distributed probabilistic embeddings, allowing explicit representation of uncertainty in each knowledge state.^[raw/papers/2605.09369.md]
The framework applies transparent logical operations over probabilistic states to construct auditable reasoning paths — showing educators which specific past actions led to a prediction. This bridges neural-symbolic-knowledge-tracing paradigms and supports intelligent-tutoring by providing explainable predictions that can be inspected and trusted.
Related Pages
- expert-cognition-dashboard — Shares goal of making learner models interpretable through cognition-level reasoning
- knowledge-tracing-irt — Probabilistic embeddings replacing deterministic KT vectors
- learning-analytics — Auditable reasoning paths for transparent learner modeling
- intelligent-tutoring — Beta-distributed knowledge states enabling confidence-aware tutoring
- student-experience — Explainable predictions linking historical behaviors to outcomes
- neural-symbolic-knowledge-tracing — Probabilistic logical reasoning bridges neural and symbolic KT
Citation
APA: Wu et al. (2026). Explainable Knowledge Tracing via Probabilistic Embeddings and Pattern-based Reasoning. arXiv:2605.09369. arXiv preprint.