Abstract

In this study we address the lack of integrated decision-support frameworks for personalized learning pathway design in undergraduate general education, where learner diversity in prior knowledge, digital readiness, and learning goals continues to create persistent learning gaps. Although existing research has advanced intelligent tutoring systems and adaptive learning technologies, most studies have focused primarily on prediction accuracy or content adaptation rather than providing a comprehensive systems-level framework that integrates learner modeling, recommendation mechanisms, and explainable decision support. The objective of this research is to develop a conceptual framework for an Intelligent Personalized Learning Pathway Recommendation System positioned as an educational decision-support reference. We adopt a Design and Development approach and synthesizes relevant literature to construct a framework based on the Input–Process–Output–Feedback model. The framework integrates Deep Knowledge Tracing for temporal learner modelling, Retrieval-Augmented Generation combined with a Large Language Model for evidence-grounded recommendations, and Explainable Artificial Intelligence to enhance transparency and user trust. The framework was evaluated through expert validation involving 15 specialists using a five-point Likert-scale. The results indicate that the proposed framework is highly suitable (mean = 4.65, SD = 0.48), particularly in decision support and AI integration, suggesting strong conceptual coherence and feasibility for implementation in higher education.

This study contributes a unified systems-level reference that connects sequential learner knowledge tracking, explainable recommendations, and pathway design within a single architecture, extending prior research that has often treated these components separately. However, the study is limited to conceptual validation. In future studies, researchers should implement a prototype and conduct empirical evaluation with students to assess learning impact, engagement, usability, and recommendation reliability.

This work is licensed under a Creative Commons Attribution 4.0 License.