Context-Aware Personalized Education Using Artificial Intelligence

Intelligent educational systems increasingly support personalised learning, yet the integration of narrative-based instruction with mastery-driven adaptation remains underexplored. This paper presents TaleTutor AI, an adaptive AI-driven framework that uses pedagogically structured narrative progression as its primary instructional mechanism. A controller-driven multi-agent pipeline generates educational narratives from ScienceQA lessons following a Noticing--Testing--Using sequence grounded in established learning theory. Chapter-level formative assessment updates a lightweight mastery estimate, which determines subsequent instructional support from extensive scaffolding to increasingly independent learning. Evaluation across seven model configurations on twenty ScienceQA lessons assessed pedagogical compliance, factual consistency, semantic alignment, and readability. The selected GPT-4o outline-first configuration demonstrated strong phase progression, plan adherence, and scaffolding differentiation. Following style-guided simplification, phase progression remained at 4.96, while scaffolding differentiation remained high at 4.90 and 4.95 for the Testing and Using chapters, respectively. Qualitative analysis further showed structurally distinct mastery-conditioned narratives and phase-aligned assessment questions. These findings demonstrate the integration of narrative generation, formative assessment, and mastery-conditioned scaffolding within a unified adaptive tutoring pipeline and provide a foundation for future learner-centred evaluation.