Exploring the Educational Applications of Large Language Models: A Systematic Review and Topic Analysis

In light of the accelerated growth of artificial intelligence (AI), large language models (LLMs) have become a central topic of interest in scientific research and practical applications across various fields. The present paper aims to perform a comprehensive systematic review of the scientific literature on LLMs in education published between 2023 and 2024, based on a dataset from the Web of Science, which includes 507 documents from 322 sources. The accelerated dynamics of research in this field are confirmed by the high annual growth rate of 369.66%. The study identifies the themes presented in the scientific literature by using thematic maps and analyzing the evolution of said thematic maps. In addition, Latent Dirichlet Allocation (LDA) and BERTopic are used to outline the research field more clearly. Due to LDA’s ability to discover high-level research topics using probabilistic discovery and BERTopic’s ability to capture deeper semantic patterns and the emergence of various topics by searching, this paper first identifies the main research topics in the extracted dataset, which are then discussed in the paper through a review of applications. As a result, a range of applications are discovered in areas related to teaching and learning, academic assessment, integrity, academic feedback, medical education, ethics, bias, regulation, and social challenges. The conclusions provide a roadmap for researchers, practitioners and stakeholders in highlighting the current situation of LLMs in educational practice, while opening the door for future explorations in this domain.