Human Adaptation vs Machine Consistency. A Comparative Study of AI-Generated Feedback and Student-Teacher Provided Feedback in an Upper Secondary School Context.

Abstract

This study explores AI-provided feedback on written essays in the English as a foreign language (EFL) classroom, comparing it to student-teacher provided feedback. In light of increasing teacher workloads, growing class sizes and the time-consuming aspect of providing written corrective feedback (WCF). AI-tools like ChatGPT, with the ability of providing feedback in seconds, could provide a solution. As AI integration in the field of education expands, understanding its pedagogical implications becomes increasingly crucial. Despite growing interest, limited research has been done on integration of AI-generated feedback in a Swedish upper secondary school in a context. The aim of this research is to evaluate and compare the areas of focus and amount of feedback provided by ChatGPT and comparing it to student-teacher provided feedback. The data used for this study are essays written by 14 upper secondary students. Findings focused on four main eras, feedback on language, structure, content and what the grades given. A qualitative content analysis was conducted, using typologies from Sheen (2011) and Ferris (1997) to define and categorize feedback. The findings suggest that ChatGPT is a capable provider of feedback, delivering structured and consistent feedback. However, ChatGPT lacks adaptability and social and contextual awareness. The study suggests that ChatGPT can assist teachers with providing feedback on written texts, but preferably on non graded assignments.