An Offline-First Adaptive Learning (OFALearn) Framework for Low-Bandwidth Educational Environments Using Lightweight AI and Dynamic Proficiency Modelling

Abstract

The increasing dependence of contemporary e-learning systems on continuous internet connectivity presents significant challenges for learners in low-bandwidth and resource-constrained environments. This study proposes OFALearn (Offline-First Adaptive Learning Framework), an intelligent educational platform that integrates adaptive learning, edge artificial intelligence, and local-first data management to support personalized learning without persistent network access. The framework employs a lightweight proficiency modeling approach that dynamically adapts learning content based on learner performance while enabling delayed synchronization when connectivity becomes available. To evaluate the effectiveness of the proposed framework, simulation experiments were conducted using a synthetic dataset comprising 10,000 learner interactions generated to reflect realistic educational usage patterns in low-bandwidth environments. Comparative evaluation was performed against Rule-Based Adaptive Learning (RBAL), Deep Knowledge Tracing (DKT), and Self-Attentive Knowledge Tracing (SAKT). Simulation results indicate that OFALearn achieved a learning gain improvement of 28.7%, adaptation accuracy of 82.9%, reduced data usage by 65.3%, and lowered system latency from 1200 ms to 410 ms when compared with conventional cloud-dependent learning systems. Although the findings are based on simulation experiments using synthetic data, they provide promising evidence for the feasibility of offline-first adaptive learning architectures in underserved educational contexts. Future work will focus on real-world deployment and validation with actual learners.