Artificial intelligence–driven learning style detection in interactive digital learning environments: a systematic review (2008–2025)

This systematic review examines how artificial intelligence (AI) and machine learning (ML) approaches have been applied to learning style detection in interactive digital learning environments between 2008 and 2025. Although the learning styles construct remains conceptually contested, it continues to inform adaptive and personalised learning systems. Following PRISMA 2020 guidelines, 1515 records were identified from Web of Science, Scopus, IEEE Xplore, and ERIC, of which 110 empirical studies met the inclusion criteria. The review analyses trends in learning style models, AI/ML approaches, algorithms, data sources, evaluation metrics, and educational contexts. Findings indicate a strong dominance of the Felder–Silverman and VARK models, widespread reliance on supervised classification techniques, and a persistent dependence on questionnaire-based data. While behavioural and multimodal data are increasingly used, their integration remains limited. The results also reveal methodological fragmentation, inconsistent evaluation practices, and a lack of evidence linking model performance to pedagogical impact. Despite growing technical sophistication, many studies prioritise predictive accuracy over educational validity. This review highlights key limitations in current research and outlines directions for more transparent, multimodal, and educationally meaningful AI-driven learner modelling in interactive learning environments.