Designing AI-Supported Adaptive English Literacy Learning for Post-Disaster Education Recovery: A Simulation-Based Development Study

Post-disaster education is frequently disrupted not only by learning loss but also by unstable connectivity, limited devices, temporary learning spaces, and uneven access to teachers. This study aimed to design and conduct an initial simulation-based evaluation of an AI-supported adaptive English literacy learning model for primary-school education recovery. A Design and Development Research approach was employed. Needs were synthesized from secondary sources, policy documents, and recent research on education in emergencies, adaptive learning, and artificial intelligence in language education. The resulting dual-context model combines learner conditions (diagnostic performance, learning gap, pace, and progress) with disaster learning-environment conditions (connectivity, device availability, and learning access). A lightweight web prototype, Sahabat Belajar Adaptif, was developed to operationalize the model through learner profiling, diagnostic assessment, explainable rule-based recommendations, teacher review, adaptive learning content, formative assessment, and progress monitoring. Scenario-based simulation was organized across three recovery levels—Intensive, Guided, and Independent—and three access conditions—Connected, Limited Connectivity, and No Connectivity. The simulations showed that academic parameters consistently determined recovery level and content difficulty, while access parameters determined delivery mode through digital, semi-digital/offline, or manual learning pathways. The teacher-in-the-loop mechanism enabled recommendations to be accepted, modified, or overridden, preserving pedagogical control. Functional simulation also confirmed that the prototype could represent the full assessment–recommendation–learning–evaluation–readaptation cycle. The study contributes a context-sensitive model for adaptive English literacy learning in disrupted educational settings. Because the present evaluation relies on secondary data and synthetic scenarios rather than direct learner-effectiveness testing, future research should validate the model with teachers and primary-school learners in disaster-affected settings.