Can Artificial Intelligence Personalize Learning? An Analysis of AI-Based Adaptive Learning Systems for Senior School Students

Abstract

Most classrooms still move at one speed. A teacher with thirty or forty students has to choose a pace that suits the middle of the class, which leaves weaker students behind and keeps stronger students waiting. AI-based adaptive learning systems promise to change this by watching how each student answers, estimating what that student already knows, and choosing the next task to match. This paper examines how far that promise holds for senior school students. It explains how adaptive systems work in simple terms, reviews findings from major meta-analyses and field experiments, including a randomised study of the Mindspark program in Delhi, and compares adaptive learning with the traditional one-size-fits-all approach. The evidence suggests that well-designed adaptive systems produce real and sometimes large learning gains, with reported effect sizes between about 0.2 and 0.8 standard deviations, and that students who are furthest behind often gain the most. However, the gains depend heavily on how the system is designed and how it is used. Systems that simply give answers can weaken learning, and results shrink when teachers are not involved. The paper concludes that AI can personalise practice, feedback and pacing to a significant extent, but it cannot replace the motivational, social and judgement-based work of a teacher. A blended model, in which AI handles adaptive practice and teachers use the data to guide instruction, is proposed for Indian senior schools.