The effects of AI-generated reformulation text as a form of feedback on EFL writing

Thesis

Exploring the ethical and effective use of generative AI tools in education, particularly in enhancing writing skills of second language (L2), is critical. Specifically, AI-generated reformulation texts can serve as an effective technique for written corrective feedback (WCF), helping to make students’ essays sound more native-like while preserving their original ideas. Despite its promise, the reformulation technique has been underutilized and under-researched in L2 writing due to its demanding nature. This study aims to investigate the impact of AI-generated reformulation texts on improving L2 writing through a three-stage process involving initial writing, revision with AI feedback, and a final rewrite. Sixty university students of English as Foreign Language (EFL) were divided into experimental and control groups, with the former receiving AI feedback and the latter self-correcting their work. Statistical analyses revealed that AI-generated feedback did not significantly improve EFL students’ argumentative writing scores compared to self-correction in the control group. However, the degree to which students accepted the AI-generated reformulations was a strong predictor of writing score improvements, particularly in organization, suggesting that adherence to AI feedback enhanced initial revisions. Additionally, participants in the experimental group noticed more language features than those in the control group, indicating that AI feedback improved learners’ language awareness. The study also contributes to the literature by examining students’ critical use of ChatGPT, focusing on their rejection of AI-generated changes. These rejections were influenced by internal, external, and AI-specific factors, highlighting learners’ critical thinking when interacting with AI tools like ChatGPT.

Authors: Guo, X

• DOI: 10.5287/ora-qqqnawdgq
• Type of award: MSc taught course
• Level of award: Masters
• Awarding institution: University of Oxford
• Language: English
• Keywords: academic writing; ALSLA2024; reformulation; EFL; quantitative research; stimulated recall; ALSLA; ChatGPT
• Subjects: English; Academic writing; Second language acquisition; Quantitative research; Applied linguistics