Abstract
Education systems worldwide face persistent challenges in delivering personalized, curriculum-aligned instruction at scale. While individualized tutoring is effective, it is costly and difficult to deploy equitably, and existing digital learning platforms primarily emphasize content delivery rather than adaptive instructional decision-making. Recent advances in artificial intelligence have enabled new forms of educational support; however, generic conversational AI systems lack the pedagogical control, curriculum alignment, and governance required for formal education contexts.
This study presents SmartTutor AI, a curriculum-aligned adaptive tutoring system designed as a computational artifact for use in formal education settings. The system integrates learner modeling, knowledge tracing, adaptive difficulty selection, and closed-loop assessment within a modular architecture, treating generative AI as a constrained instructional component rather than an open-ended tutor. Instructional decisions are governed by explicit pedagogical logic and structured curriculum representations, enabling personalized learning progression while maintaining safety, reliability, and exam alignment.
As a computational research contribution, this work reports the design and specification of the SmartTutor AI system and introduces a structured evaluation framework as a research output. The proposed framework defines metrics and instruments for assessing instructional effectiveness, adaptive quality, assessment fidelity, teacher workload support, and governance readiness in future empirical studies.
By emphasizing deployability, ethical alignment, and system-level design, this study positions SmartTutor AI as a practical pathway for translating applied artificial intelligence into scalable educational impact, with relevance for national education systems.