Large Language Models and Academic Writing in Practice: Exploring Participants' Utilization of Generative Pretrained Transformers during an AI-Assisted Course on Writing Research Papers

Esmaeel Ali Salimi ^1* Mahdi Hajinia ²

• ¹ Allameh Tabataba'i University, Tehran, Iran
• ² University of Ilam, Ilam, Ilam, Iran

This study investigates the integration of Generative Pre-trained Transformer (GPT) models in scholarly writing instruction, focusing on their impacts on researchers' academic writing in their second language (L2). Using a phenomenographic design, the research examines the practical implications of incorporating GPT-4 into real classroom settings, and tries to identify GPT's merits and challenges recorded by participants during the course of the study. The participants included 20 students from two universities, all with advanced English proficiency (IELTS > 7.5), and 5 instructors experienced in teaching academic writing held the sessions. Over five sessions, each designed to address a specific section of a research paper, students engaged in GPT-supported activities during class and home assignments. The findings demonstrate that GPT-4 enhances productivity and efficiency in academic writing while presenting challenges such as over-reliance and ethical concerns. The study highlights both the benefits and limitations of GPT, emphasizing its transformative potential in reshaping scholarly writing instruction. Practical recommendations include structured integration of GPT in educational settings to optimize its advantages while mitigating drawbacks. This research contributes empirical evidence on the impact of GPT on L2 scholarly writing, offering practical guidance for educators seeking to effectively integrate AI tools into their teaching practices. Future studies should explore the longitudinal effects of GPT on academic writing proficiency and investigate its integration in diverse educational contexts to broaden its applicability.