CHATGPT-ASSISTED ACADEMIC WRITING: PERCEPTIONS AND ETHICAL CONCERNS AMONG ESL STUDENTS. A MIXED METHOD APPROACH

Abstract

Generative AI (GAI), including ChatGPT, has sparked a change in academic literacy for students in higher education and has placed the ESL students at the centre of the changes. Students learning English as a Second Language (ESL) play a pivotal role in the higher education academic literacy environment, and are being shaped by generative AI (GAI) tools like ChatGPT. What is so exciting about these tools is that they can help make writing easier as never before, but they raise a few issues that have ethical implications in terms of ownership, plagiarism, and cognitive offloading. The present mixed methods study is concerned with the perception and the moral dilemma of the academic writing with the help of ChatGPT among the diversified sample of 416 ESL learners of a big public university of Pakistan. The results of our qualitative study, which involved conducting semi structured focus group discussions, do not contradict the quantitative results from the survey in which students indicated that using ChatGPT to support them with a language skill is an equaliser, that is, "ease of use in relation to their performance" (C. Lee & Chew, 2023; Yan, 2023); at the same time, however, students do not feel confident of the boundaries of using ChatGPT as an assistive tool, and what constitutes academic transgression, which gives rise to “ethics anxiety” (Eaton, 2023; Perkins, 2023). A positive correlation (r = .68 p < .001) is statistically significant between the perception of a positive impact of AI on language acquisition and writing self-efficacy, while the negative correlation (r = –.61 p < .001) is statistically significant between the perception of a negative impact of AI on language acquisition and writing self-efficacy. They were analysed qualitatively and in themed forms and revealed four main ethical tensions: (1) Loss of voice of the author, (2) digital divide in algorithmic literacy, (3) Failure of current academic integrity frameworks, and (4) Potential intellectual deskilling. Finally, we propose a new “Dynamic AI Literacy Model” for teaching ESL that shifts from a punitive detection model to a constructive detection model, building on the students' culturally and linguistically diverse backgrounds with the knowledge of AI and engaging the students in developing explicit and contextual ethics guidelines.