Evaluating The Efficacy of A Large Language Model in Scaffolding Research Report Writing for EFL Learners

This study examines how the Generative Pre-trained Transformer (GPT) architecture can assist English as a Foreign Language (EFL) learners in writing research reports. In a quasi-experimental design, 60 undergraduates were divided into two groups: a control group and an experimental group, where the latter utilized ChatGPT as a writing assistant. The intervention tested the model's performance in generating and improving text in five areas: organization and structure, research and evidence, analysis and critical thinking, writing quality and presentation, and content and relevance. Writing skills were measured with a 20-item achievement test before and after the intervention. Additionally, 15 participants took part in semi-structured interviews to provide qualitative feedback. The groups had similar skills before the intervention (p > 0.05). Afterward, the experimental group improved in all research report writing skills, particularly in language accuracy, organization, and content relevance. Participants praised ChatGPT for its user-friendly interface, time-saving features, and ability to improve grammar, vocabulary, and report structure. However, some raised concerns about plagiarism and reliance associated with the tool. These findings revealed that large language models (LLMs) can greatly improve the structure and language of student reports. Still, human oversight is needed to ensure critical thinking and factual accuracy.