Human-AI Collaborative Feedback Enhances EFL Writing with Chain-of-Thought Prompting under SFL Guidance

• In Event: AER Roundtable Session Friday 1:45 pm - Gold 4
  In Roundtable Session: Using AI to Learn in Multi-lingual Contexts

Abstract

Despite the growing use of LLMs for writing feedback, a critical gap remains in understanding how human-AI collaborative feedback influences writing performance. We propose a feedback framework where students receive ChatGPT-generated feedback followed by teacher-led instruction, implemented in an eight-week study with 24 high school students. Prompts and rubrics are informed by SFL and chain-of-thought prompting. Results show significant gains in content, structure, and overall writing quality, while language accuracy remains stable. Content and structure follow distinct improvement patterns. Questionnaire (n=24) and interview (n=5) data indicate positive student perceptions, though effective use still requires teacher scaffolding and sustained practice. This study offers a replicable and theory-informed instructional model for future research on AI-supported writing feedback.

Authors

• Jialing Wu, East China Normal University

• Huirong Gao, East China Normal University