Comparisons of AI and Human Writing Feedback: Student Perceptions in a University EFL Workshop
2
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study examined the perceptions of university-level EFL students toward feedback on their persuasive writing from humans and artificial intelligence (AI). Post-feedback student writing was also analyzed to find any measurable differences attributable to the feedback sources. Students were not provided with the feedback sources in order to eliminate potential bias. The study was conducted through an asynchronous writing workshop with 32 students receiving feedback on a first draft of a writing assignment. They were then asked to produce a second draft and respond to a ten-item Likert questionnaire about perceived dimensions of clarity, motivation, and usefulness of the feedback. Second draft samples were analyzed for grammar and usage mistakes, readability, and general proficiency. Independent samples t-tests were performed, and no statistically significant differences were found between the groups on any measure, despite several medium to large effect sizes in four survey responses, three in favor of AI feedback. The results are discussed with regard to AI’s potential for integration into writing instruction. The importance of teaching skeptical attitudes towards AIgenerated material is underscored, and further research is encouraged to investigate effects beyond the short term.
Abstract
This study examined the perceptions of university-level EFL students toward feedback on their persuasive writing from humans and artificial intelligence (AI). Post-feedback student writing was also analyzed to find any measurable differences attributable to the feedback sources. Students were not provided with the feedback sources in order to eliminate potential bias. The study was conducted through an asynchronous writing workshop with 32 students receiving feedback on a first draft of a writing assignment. They were then asked to produce a second draft and respond to a ten-item Likert questionnaire about perceived dimensions of clarity, motivation, and usefulness of the feedback. Second draft samples were analyzed for grammar and usage mistakes, readability, and general proficiency. Independent samples t-tests were performed, and no statistically significant differences were found between the groups on any measure, despite several medium to large effect sizes in four survey responses, three in favor of AI feedback. The results are discussed with regard to AI’s potential for integration into writing instruction. The importance of teaching skeptical attitudes towards AIgenerated material is underscored, and further research is encouraged to investigate effects beyond the short term.