Adaptive LLM-based learning environments as a means of overcoming knowledge deficits under martial law

Abstract

This paper examines the profound educational disruptions precipitated by the full-scale military invasion of Ukraine and posits adaptive learning environments (ALEs) driven by large language models (LLMs) as a critical mechanism for mitigating consequent knowledge deficits. Highlighting the inadequacy of conventional pedagogical models amidst infrastructural collapse and continuous displacement, the study explores how LLMs facilitate real-time, personalised instructional scaffolding. By incorporating diagnostic assessment, adaptive remediation, guided practice, and teacher-led verification, these environments deliver essential compensatory adaptability for interrupted studies. Furthermore, the deployment of LLMs through edge computing and multimodal interfaces ensures pedagogical continuity even within resource-scarce and low-bandwidth conditions. Whilst acknowledging the inherent risks of model hallucination and over-reliance, the research argues that integrating retrieval-augmented generation (RAG) alongside rigorous human oversight can effectively safeguard academic integrity. Ultimately, LLM-based ALEs represent a resilient, trauma-informed pedagogical instrument capable of converting fragmented study patterns into manageable, continuous progression during prolonged wartime crises.