MACHINE LEARNING IN MATHEMATICS AND COMPUTER SCIENCE EDUCATION: A SYSTEMATIC THEMATIC REVIEW OF DISCIPLINARY LEARNING SUPPORT

Source abstract

Machine learning (ML) is being increasingly used in mathematics and computer science education, but the educational value of machine learning is not theorized or assessed evenly. The current studies focus on predictive accuracy, automation, and personalization, which raises the question of whether such applications are relevant to supporting higher-order disciplinary learning. This gap is filled by the current study, which is a qualitative systematic review based on the PRISMA framework of open-access and Q1/Q2 peer-reviewed articles published between 2014 and 2025. Through reflexive thematic analysis, sixteen studies were synthesised in four dimensions of analysis, namely application types, evaluation practices, cognitive support, and pedagogical integration. The results show that there is a systematic predominance of predictive analytics and performance-based assessment, but little overt scaffolding in problem solving, abstraction, and conceptual knowledge. Notably, ML systems are frequently used as back-end analytics and not as a learning-focused instructional tool. This review is novel in that it is a discipline-specific, cognition-based synthesis, shifting the ML paradigm from outcome prediction to pedagogically grounded cognitive scaffolding. Such a point of view provides practical design and assessment guidelines to narrow the divide between ML innovation and disciplinary learning in the education of mathematics and computer science.