Knowledge-aware Methods for Explainable Decision Support in Lifelong Learning

Scaling up digital education presents several critical challenges, including the management of large learner populations, the abundance of learning resources, and the difficulty of supporting informed learning decisions at scale. While digital platforms provide unprecedented access to education, learners often face overwhelming choices and limited guidance while navigation. The increasing availability of learning data and technological advancement creates significant opportunities for artificial intelligence (AI) to support lifelong learning. To make such intelligent support effective in real-world educational settings, careful planning, representation, reasoning are essential. This thesis addresses the design, implementation, and evaluation of explainable artificial intelligence methods for lifelong learning environments. The focus is on how AI systems can represent educational knowledge, reason over learning-related data, and communicate recommendations and explanations in a transparent and user-centered manner. The proposed contributions leverage learner-centered ontologies, educational knowledge graphs, path-based reasoning methods, conversational interfaces, and large language models to support explainable educational recommendation and knowledge exploration. Rather than prioritizing predictive accuracy alone, the thesis emphasizes explainability and learner trust as key requirements for AI-driven educational systems. Through a series of studies, this work shows how structured knowledge representations can enable explainability by construction, how conversational interfaces can facilitate intuitive interaction with educational knowledge graphs, and how path-based reasoning can support transparent recommendations. Furthermore, the thesis explores how large language models can transform structured, path-based explanations into user-friendly natural language narratives while preserving faithfulness to the underlying reasoning process. The findings provide methodological insights and design guidelines for combining structured reasoning and language-based explanation in educational AI systems. Overall, this thesis contributes to advancing explainable AI for lifelong learning by encouraging transparency, improving user acceptance, and strengthening the role of structured educational data as an asset for scalable and trustworthy learning support.