AI-DRIVEN ONLINE ADAPTIVE LEARNING SYSTEMS IN HIGHER EDUCATION: AN INTEGRATIVE REVIEW OF LEARNER MODELLING, ADAPTATION MECHANISMS AND EVIDENCE OF EFFECTIVENESS

The rapid expansion of online and blended learning in higher education has exposed a persistent limitation of conventional online courses: content is delivered to all students in the same sequence and at the same pace, regardless of differences in prior knowledge, learning progress and needs. Online adaptive learning systems, increasingly powered by artificial intelligence (AI), aim to overcome this limitation by continuously diagnosing the learner and personalising content, sequence, difficulty and feedback. Aim. This article reviews the theoretical foundations, learner-modelling techniques, adaptation mechanisms and empirical evidence on the effectiveness of AI-driven online adaptive learning systems, and derives implications for higher education systems undergoing digital transformation, with particular reference to Uzbekistan. Methods. An integrative literature review was conducted following the framework of Whittemore and Knafl. Seminal works and peer-reviewed publications indexed in Scopus, Web of Science, ERIC and Google Scholar, together with international policy documents, were analysed using reflexive thematic synthesis. Results. Six themes were identified: (1) pedagogical and psychological foundations; (2) learner modelling, from Bayesian knowledge tracing to deep learning; (3) adaptation targets and mechanisms; (4) evidence of effectiveness from meta-analyses; (5) the emerging role of generative AI; and (6) ethical, pedagogical and infrastructural challenges. Meta-analytic evidence shows moderate to large positive effects of intelligent and adaptive tutoring on learning outcomes, while gaps remain regarding teacher involvement, transparency and applicability to low-resource language contexts. Conclusion. Effective adaptive learning requires the integration of validated learner models, pedagogically grounded adaptation rules, interoperable learning platforms and a teacher-in-the-loop approach. A set of requirements for national adaptive learning systems is proposed.