Abstract
Artificial intelligence-driven adaptive learning systems are reshapingeducation through data-driven adaptation of learning experiences. Yet many ofthese systems lack transparency, offering limited insight into how decisionsare made. Most explainable AI (XAI) techniques focus on technical outputs butneglect user roles and comprehension. This paper proposes a hybrid frameworkthat integrates traditional XAI techniques with generative AI models and userpersonalisation to generate multimodal, personalised explanations tailored touser needs. We redefine explainability as a dynamic communication processtailored to user roles and learning goals. We outline the framework's design,key XAI limitations in education, and research directions on accuracy,fairness, and personalisation. Our aim is to move towards explainable AI thatenhances transparency while supporting user-centred experiences.