<![CDATA[Within educational technology, the adoption of Large Language Models (LLMs) and Generative AI (GenAI) has accelerated notably, yielding applications spanning intelligent tutoring systems, adaptive assessment pipelines, and automated formative feedback. Yet the fusion of these generative capabilities with Cognitive Digital Twin (CDT) architectures has proceeded unevenly, leaving theoretical and practical gaps unaddressed. This study reports a systematic literature review conducted in accordance with PRISMA 2020 principles. After screening 1,400 records from ScienceDirect and IEEE Xplore, 111 Core and Supporting studies were included in the qualitative synthesis. A narrative thematic synthesis was then carried out by mapping the included studies to five research questions and five thematic clusters. The five thematic clusters comprise CDT and learner modeling, LLM and GenAI tutoring, adaptive assessment and feedback, ethics and explainability, and learning analytics. The analysis reveals that automated feedback and assessment constitute the most densely evidenced application domains, whereas learner modeling and CDT remain underdeveloped as sustained personalization mechanisms. The paper’s primary contribution is a five-layer LLM-driven CDT framework comprising a Learner Data Layer, a Learner Digital Twin Layer, an LLM Intelligence Layer, an Adaptive Learning Service Layer, and a cross-cutting Trust, Ethics, and Governance Layer. Taken together, these layers constitute a learner-aware, adaptive, explainable, and human-centered architecture intended to ground and direct future empirical inquiry.]]>
Copyrights © 2026
