XAI-AGLearn: Toward Negotiable Personalization in Explainable Adaptive Gamified Learning Systems

Abstract

Adaptive learning systems, powered by AI, are proficient at personalizing educational material and remain opaque in their judgment; adding gamification increases engagement, but it is mostly implemented in a one-size-fits-all manner; and Explainable AI (XAI) continues to evolve, but has yet to be systematically integrated into adaptive gamified learning environments. These topics are frequently addressed in previous studies, but few frameworks simultaneously operationalize adaptivity, gamification, and explainability within one unified architecture. This paper proposes XAI-AGLearn, a four-layer model of design science system architecture, that is, learner modeling, an adaptive engine, a gamification engine, and an explainability interface. The framework produces five simple, concrete outputs: (1) explicit four-layer architecture data-flow specifications, (2) design principles for negotiable personalization, (3) operationalized explainability mechanisms (rule-based rationale, example-based explanation, and learner-facing natural language justification), (4) a mapping between learner state, adaptive game mechanics (badge type, challenge difficulty, quest narrative, reward cadence, leaderboard visibility), and explanation techniques, and (5) five negotiability mechanisms (user override, preference correction, explanation-based feedback loop, transparency toggle, and educator intervention point). The primary novelty positions explainability not as a post-decision add-on functionality but as an embedded interaction layer that turns opaque adaptation into negotiable adaptation. It allows personalization decisions to be made through the proposed framework, with traceable decision pathways and user-in-the-loop evolution, moving the paradigm from obscure to transparent.