Direct vs. Reformulation Feedback through ChatGPT on EFL Learners’ Writings*1
2
조회
0
다운로드
0
인용
0
참고문헌
초록
This study investigates the short- and long-term effects of two types of written corrective feedback (WCF)—direct feedback and reformulation—delivered via ChatGPT on EFL learners’ writing performance. Twenty-seven university students participated in a three-stage writing process: an initial draft, a first revision immediately after feedback, and a delayed second revision two weeks later without additional support. Writing was evaluated across five dimensions: Content, Organization, Vocabulary, Language Use, and Mechanics. Descriptive and inferential analyses revealed that both feedback types led to short-term improvements, particularly in Content, Vocabulary, and overall writing scores. Direct WCF was effective in generating broad improvements across most areas, excluding Language Use. Reformulation feedback, in contrast, led to stronger gains in Content and Organization and continued to yield significant improvements in the delayed revision, especially in Content and Vocabulary. However, neither feedback type produced significant gains in grammatical accuracy (Language Use), indicating the need for supplementary grammar instruction. These findings suggest that reformulation feedback via ChatGPT is particularly effective in enhancing idea development and lexical variety, while direct WCF provides general but less targeted improvements. The study highlights the pedagogical potential of AI-mediated feedback and underscores the importance of differentiated instructional strategies to address persistent language accuracy issues.
Abstract
This study investigates the short- and long-term effects of two types of written corrective feedback (WCF)—direct feedback and reformulation—delivered via ChatGPT on EFL learners’ writing performance. Twenty-seven university students participated in a three-stage writing process: an initial draft, a first revision immediately after feedback, and a delayed second revision two weeks later without additional support. Writing was evaluated across five dimensions: Content, Organization, Vocabulary, Language Use, and Mechanics. Descriptive and inferential analyses revealed that both feedback types led to short-term improvements, particularly in Content, Vocabulary, and overall writing scores. Direct WCF was effective in generating broad improvements across most areas, excluding Language Use. Reformulation feedback, in contrast, led to stronger gains in Content and Organization and continued to yield significant improvements in the delayed revision, especially in Content and Vocabulary. However, neither feedback type produced significant gains in grammatical accuracy (Language Use), indicating the need for supplementary grammar instruction. These findings suggest that reformulation feedback via ChatGPT is particularly effective in enhancing idea development and lexical variety, while direct WCF provides general but less targeted improvements. The study highlights the pedagogical potential of AI-mediated feedback and underscores the importance of differentiated instructional strategies to address persistent language accuracy issues.