Article: Explainable knowledge tracking for industry-education integration based on a causal attention mechanism Journal: International Journal of Information and Communication Technology (IJICT) 2026 Vol.27 No.90 pp.49 - 70 Abstract: In industry-education integration scenarios, existing knowledge tracking models are largely based on statistical correlations derived from historical interactions, making it difficult to distinguish between skill deficiencies and random errors; moreover, their black-box nature undermines the credibility of decision-making. This paper proposes a causality-enhanced attention mechanism that models causal pathways between skill mastery, exercise interactions, and job requirements using structural causal graphs. By introducing counterfactual reasoning to generate representations of knowledge states after interventions, the model focuses on actual changes in skills rather than superficial patterns in interaction sequences. Experiments on the ASSISTments and industry-education integration platform datasets demonstrate that the model achieves area under the curves of 0.834 and 0.871, respectively, representing improvements of 3.1% and 2.7% over the best baseline. normalised discounted cumulative gain @5 also significantly outperforms comparison methods, and the visualisation of causal attention weights reveals the model's precise attribution of skill-related root causes. Inderscience Publishers - linking academia, business and industry through research

Title: Explainable knowledge tracking for industry-education integration based on a causal attention mechanism

Authors: Lai Zhou

Addresses: School of Information Engineering, Zhengzhou College of Finance and Economics, Zhengzhou, 450000, China

Abstract: In industry-education integration scenarios, existing knowledge tracking models are largely based on statistical correlations derived from historical interactions, making it difficult to distinguish between skill deficiencies and random errors; moreover, their black-box nature undermines the credibility of decision-making. This paper proposes a causality-enhanced attention mechanism that models causal pathways between skill mastery, exercise interactions, and job requirements using structural causal graphs. By introducing counterfactual reasoning to generate representations of knowledge states after interventions, the model focuses on actual changes in skills rather than superficial patterns in interaction sequences. Experiments on the ASSISTments and industry-education integration platform datasets demonstrate that the model achieves area under the curves of 0.834 and 0.871, respectively, representing improvements of 3.1% and 2.7% over the best baseline. normalised discounted cumulative gain @5 also significantly outperforms comparison methods, and the visualisation of causal attention weights reveals the model's precise attribution of skill-related root causes.

Keywords: knowledge tracking; causal inference; industry-academia integration; interpretability; attention mechanisms.

DOI: 10.1504/IJICT.2026.155732

International Journal of Information and Communication Technology, 2026 Vol.27 No.90, pp.49 - 70

Received: 03 Apr 2026
Accepted: 21 May 2026
Published online: 11 Aug 2026 *

Title: Explainable knowledge tracking for industry-education integration based on a causal attention mechanism

Keep up-to-date