<![CDATA[Vocabulary learning in primary English classes is often constrained by heterogeneous learner readiness, where fixed game progression can under-serve both struggling and advanced students. This study aims to implement and technically validate a reproducible real-time adaptivity mechanism using Bayesian Knowledge Tracing (BKT) in a Unity-based Android vocabulary game. The research adopts a design-and-verification approach using the Game Development Life Cycle (GDLC), supported by requirements elicitation (classroom observation, teacher interview, and literature review). The adaptive engine applies BKT to update mastery after each quiz response and routes learners using a mastery-threshold policy, while event-level logs are stored locally and exportable for auditability. The main results demonstrate that adaptive mode activation, mastery updates, persistence of adaptive state, and mastery-gated progression function consistently in end-to-end black-box tests. Algorithm-level credibility is strengthened through white-box basis-path verification of the UpdateProbability() routine, ensuring independent execution paths for correct and incorrect responses are covered. This work contributes a deployable Unity/Android architecture for real-time BKT-driven adaptivity, accompanied by verification artifacts and reproducibility recommendations to support technical audit, replication, and subsequent controlled effectiveness studies.]]>
Copyrights © 2026
