Discrepancies Between ChatGPT and Vietnamese EFL Teachers in Writing Assessment

Artificial intelligence, particularly ChatGPT, strongly influences education, especially in modern EFL classrooms. However, a significant gap still exists in Vietnam regarding how ChatGPT’s writing assessments differ from those of EFL teachers. To address this gap, this study directly compares ChatGPT’s writing assessments with those of Vietnamese EFL teachers to identify key discrepancies. Twenty experienced teachers working at English centers and universities in Ho Chi Minh City participated in this quantitative research. ChatGPT and human raters used an analytic rubric adapted from a standardized rubric to assess the written works of twenty students. A paired-sample t-test was then employed to compare the scores of human raters and ChatGPT in assessing twenty students’ essays across the rubric criteria. The findings revealed a statistically significant difference between ChatGPT and Vietnamese EFL teachers in writing evaluations (p = .034), with ChatGPT (M = 16.08, SD = 3.29) assigning a higher total score than human raters (M = 14.75, SD = 2.96). Particularly, ChatGPT tended to give higher scores than human raters across all criteria, including content, organization, vocabulary, and grammar, with the content criterion showing the greatest discrepancy.