대학 영어 글쓰기 평가에서 ChatGPT의 활용

• Published in: 언어학 연구, 0(74) pp. 203 - 222
• Main Author: 박혜경
• Format: Journal Article
• Language: Korean
• Published: 한국중원언어학회 01.01.2025
• Subjects: 언어학
• ISSN: 1975-8251; 2508-4259
• DOI: 10.17002/sil..74.202501.203

The purpose of this study was to examine the evaluation results of Korean college students’ writing assignments assessed by a human rater and ChatGPT (GPT-3.5 and GPT-4) to explore the feasibility of ChatGPT as an assistant tool in English writing evaluation. A total of 77 paragraph writing assignments were collected and evaluated using a five-point rubric. Quantitative analysis was conducted using one-way ANOVA, and qualitative analysis provided deeper insights into evaluation discrepancies. The findings revealed statistically significant differences among the scores of the human rater, GPT-3.5, and GPT-4, with GPT-3.5 demonstrating greater similarity to the human rater than GPT-4. Qualitative analysis highlighted that ChatGPT’s evaluations were occasionally simplistic, erroneous, or misaligned with the rubric criteria. Notably, ChatGPT tended to prioritize content evaluation over format assessment. These findings suggest that while ChatGPT can be effectively used as a supplementary tool in English writing evaluation, its current limitations in consistency and adherence to criteria indicate that it cannot fully replace human raters. KCI Citation Count: 0