Intelligent Tutoring Systems Using Large Language Models for Personalized and Context-Aware Learning

Abstract

Intelligent Tutoring Systems (ITS) have evolved from rule-based instructional environments toward adaptive learning platforms capable of modelling learner characteristics, providing individualized feedback, and dynamically adjusting instructional content. The emergence of Large Language Models (LLMs) introduces a significant opportunity to enhance ITS through natural-language interaction, contextual reasoning, formative assessment, personalized explanations, and learner-aware instructional adaptation. This paper examines the integration of LLMs into intelligent tutoring environments for delivering personalized and context-aware learning experiences. It conceptualizes an LLM-driven ITS architecture comprising learner modelling, contextual representation, pedagogical orchestration, content generation, feedback mechanisms, and continuous learner assessment. The study further examines the potential of LLMs to adapt explanations according to learner proficiency, learning history, misconceptions, interaction patterns, and task context. Particular attention is given to challenges involving hallucination, pedagogical reliability, bias, privacy, explainability, assessment validity, and over-reliance on generated responses. Existing research on intelligent tutoring, adaptive learning, conversational agents, and LLM-based education is synthesized to identify emerging research directions. The analysis indicates that LLM-enhanced ITS can substantially improve instructional flexibility and conversational personalization when combined with structured learner models, retrieval mechanisms, pedagogical constraints, and human oversight. The paper proposes a research-oriented framework for developing trustworthy, context-aware, and pedagogically aligned LLM-based tutoring systems capable of supporting diverse learners across educational settings.