Large Language Model-Based Academic Writing Support Framework for University English Learners
- Authors
-
-
Ivy Jones-Mensah
Author
-
- Keywords:
- large language models, academic writing, university english learners, automated writing evaluation, academic integrity, prompt literacy, higher education
- Abstract
-
The widespread adoption of large language models (LLMs) like ChatGPT in higher education has revolutionized the way university English learners compose, edit, and enhance their academic content. Previous studies report positive effects of LLM-generated feedback on grammar, coherence, and lexical range, but most of them target single tools without a pedagogical framework that specifies when, how, and why students should interact with the tools. This paper fills the void by proposing a “Large Language Model Based Academic Writing Support Framework (LAWSF)” tailored for the needs of university English learners, such as English as a Foreign Language (EFL) and English as a Second Language (ESL) students. The framework is based on a synthesis of 15 peer-reviewed studies related to generative artificial intelligence, automated writing evaluation, learner motivation, prompt engineering, and academic integrity and includes five intertwined layers: diagnostic profiling, scaffolded drafting, multi-dimensional feedback, integrity and reflection checkpoints, and instructor-mediated evaluation. In contrast to traditional automated writing assessment tools that focus primarily on error correction, LAWSF places the LLM in a pedagogical cycle that is engaged by human teachers and students, where students are encouraged to be active agents in the process while avoiding overreliance and thereby promoting the development of prompt literacy and self-regulation. A suggested implementation and evaluation protocol is outlined, including instruments for assessing gains in writing quality, student engagement and perceived ownership of the writing throughout the semester-long intervention. Implications for curriculum design, teacher professional development, institutional AI-use policy are discussed as well as limitations due to model variability, linguistic bias and inequities in access. The paper ends by calling for a structured, integrity-oriented construct to transform the potential of LLMs into ongoing academic writing learning for university English learners.
- Downloads
- Published
- 2025-03-31
- Issue
- Vol. 14 No. 1 (2025)
- Section
- Articles