Generative AI in Intelligent Tutoring Systems: A Systematic Literature Review of Pedagogical Feedback, Conversational Interaction, and Learning Safety

Intelligent Tutoring Systems (ITSs) have advanced from rule-based learning environments to adaptive systems that use artificial intelligence techniques, modelling of learners and learning analytics. The wide range of systems and learning approaches makes it difficult to understand how learner data is mapped to adaptive instructional decision making. This paper reports a systematic literature review of pedagogical alignment, adaptivity, and analytics in Intelligent Tutoring Systems. This review synthesises relevant research in three inter-related dimensions of ITS design: pedagogical alignment (how tutoring strategies and learning objectives are aligned); adaptivity (how systems respond to learners' characteristics knowledge behaviour and performance); and learning analytics (how learners’ data are captured, analysed, and applied). The review mostly focuses on the integration of knowledge tracing, student modelling, educational data mining, learning analytics, and AI-based tutoring technologies in adaptive learning; it discusses the effectiveness of different adaptive learning processes; evaluation approaches; and technical and pedagogical issues related to adaptive learning (including data quality, data sparsity interpretability individualization, scalability, and model-to-pedagogy complementarity). The integration of the points of view presented in this study helps to provide a comprehensive view of the relationship between the data analyses adaptive decisions, and pedagogy used in intelligent tutoring systems. The review aims to lay a conceptual foundation for building adaptive ITSs that are more sensitive to individual learner needs, are informed by data, and are pedagogically consistent in their teaching strategies, and to suggest future research directions for the architecture of integrated adaptive ITSs in different educational settings.