Two voices, one draft: a diachronic comparison of human and AI feedback in EFL essay writing

Abstract

As intelligent technological tools increasingly permeate educational contexts, chatbots have emerged as prominent aids in language learning. This study investigates the comparative effectiveness of human- and AI-generated feedback on EFL learners’ essay writing, focusing on differences and similarities in feedback quality and impact. Specifically, the study examines how feedback from a human instructor and a generative AI (GenAI) chatbot, ChatGPT, varies across three dimensions: structure and organisation, content and ideas, and grammar. A diachronic perspective is adopted to analyse how each type of feedback contributes to student revision over time. Participants included twenty-six EFL college students enrolled in an English composition course. The study focused on the production of comparison essays through a three-stage process. In the first stage, students independently wrote their initial drafts at home, with access to dictionaries or online resources but without direct assistance. In the second stage, the drafts received written feedback from a human instructor. In the third stage, students revised their essays again based on feedback generated by ChatGPT. All drafts were subsequently evaluated by both human raters and the AI system using the aforementioned criteria. Findings show that AI-generated feedback in the third stage produced scores comparable to human feedback in the second stage, particularly in identifying grammatical issues and improving clarity. However, human feedback provided richer, more context-sensitive guidance, especially in enhancing content development and sentence complexity. In contrast, ChatGPT's responses tended to be more uniform and focused primarily on surface-level corrections. These results underscore the potential of integrating GenAI tools into EFL writing instruction while also highlighting their current limitations. The study contributes to a deeper understanding of how human and AI feedback complement each other, offering pedagogical insights into the thoughtful use of AI in academic writing contexts.

Ähnliche Arbeiten