From Learning Styles to Learning Signals: Dynamic Personalisation in AI-Mediated Education

As artificial intelligence transforms the way educational systems interpret students, personalise teaching and automate pedagogical decisions, the persistent use of learning styles as a basis for personalisation demands renewed critical review. Although learning styles remain attractive to teachers because they offer an accessible language for recognising learner diversity, their empirical fragility becomes more problematic when transferred to adaptive platforms, recommender systems and data-based learning environments. This scoping review examines the evidence supporting learning-style models as factors for personalisation and analyses how recent literature reconfigures this debate in the context of AI-mediated education. The findings suggest that the most defensible contribution of learning styles does not lie in their classificatory capacity, but in the pedagogical concern they express, namely avoiding uniform teaching. However, AI does not solve the problem by classifying students better. Its educational value lies in supporting dynamic, revisable, and teacher-mediated interpretations based on observable, contextual, and ethically used evidence about learning.