Citation

Low, Jing Hong and Zhang, Xiao (2026) Explainable Multi-View Modeling of AI-Driven Personalized Learning Adoption in TVET Systems. Journal of Logistics, Informatics and Service Science, 13 (4). pp. 125-135. ISSN 24092665

Text Vol.13.No.4.08.pdf - Published Version Restricted to Repository staff only Download (665kB)

Abstract

Artificial intelligence–driven personalized learning is increasingly embedded within Technical and Vocational Education and Training (TVET) systems as a data-enabled service innovation. However, large-scale institutional adoption remains uneven due to sociotechnical, governance, and organizational constraints. This study conceptualizes TVET personalization as a digital service system and investigates the critical determinants of adoption through an integrated explanatory–predictive framework. Building on UTAUT2 and organizational readiness perspectives, we model personalization quality, trust in AI, transparency, perceived privacy risk, instructor support, and facilitating conditions as key service-system drivers of behavioral intention and sustained use. Recent empirical studies on Generative AI adoption in higher education further validate the applicability of the UTAUT2 framework, confirming that facilitating conditions and performance expectancy are critical determinants for integrating educational technologies (Papadakis et al., 2025). We propose a Theory-Guided Multi-View Multi-Task (TG-MVMT) framework that fuses survey-based latent constructs with learning-platform behavioral traces and incorporates institution-aware regularization to enhance cross-campus robustness. The model is benchmarked against strong predictive baselines (logistic regression, random forest, gradient boosting, and TabNet) under nested cross-validation, and interpretability is achieved using SHAP-based attribution to translate predictive signals into actionable governance levers. Results indicate that combining perceptual factors with early behavioral indicators significantly improves adoption prediction and reveals a concentrated set of dominant determinants, with personalization quality and trust in AI consistently emerging as primary drivers. The findings contribute to service informatics research by demonstrating how explainable multi-view modeling can support AI governance, instructor-in-the-loop orchestration, and scalable deployment across heterogeneous TVET institutions.

Item Type:	Article
Uncontrolled Keywords:	TVET service systems, AI-driven
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science
Divisions:	Faculty of Management (FOM)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	04 May 2026 04:47
Last Modified:	04 May 2026 04:47
URII:	http://shdl.mmu.edu.my/id/eprint/15870

Downloads

Edit (login required)