TrustLearn: An End-to-End Trustworthy Framework for Adaptive Microlearning in Net-Zero Energy Communities Using Blockchain-Enhanced Multi-Agent Federated Reinforcement Learning

Abstract

The optimization of microlearning dosage—encompassing duration, frequency, and difficulty—remains a critical challenge in personalized educational technology, particularly

within resource-constrained environments such as Net-Zero Energy Communities (NZECs).

This paper introduces TrustLearn, a novel end-to-end trustworthy framework that integrates

Blockchain technology, Multi-Agent Reinforcement Learning (MARL), and Federated Learning (FL) to maximize knowledge retention and learner motivation while ensuring privacy, security, and energy efficiency. Our approach leverages blockchain for immutable audit trails and secure traceability of learning interactions, FL for privacy-preserving model training across distributed mobile devices, and MARL for adaptive, personalized learning path optimization in decentralized settings. We address three critical research questions: (1) What microlearning dosage maximizes retention and motivation? (2) How can we ensure end-to-end trustworthiness in distributed learning environments? (3) What is the energy overhead and scalability of such systems in NZECs? Through extensive experimentation on real educational datasets (Duolingo SLAM, EdNet, and ASSISTments 2009-2010), we demonstrate that TrustLearn achieves 23.7% improvement in knowledge retention, 31.2% enhancement in learner engagement, and maintains privacy guarantees while incurring only 4.3% energy overhead. Our comprehensive evaluation includes adversarial attack resistance analysis, ablation studies, and scalability assessments up to 10,000 distributed learners. Statistical validation using Friedman tests and post-hoc Nemenyi analysis confirms the significance of our findings (p < 0.001). TrustLearn represents a paradigm shift toward sustainable, trustworthy, and adaptive educational systems for energy-conscious communities.