L2 Teachers’ Perception of AI-based Assessment on Student Essay: The Case of CoGrader

Abstract

Abstract: Writing assessment in English as a foreign language (EFL) learning increasingly involves artificial intelligence (AI)-based technology. However, there are also issues regarding how well and how useful automated systems like CoGrader are in real classrooms. This study looks at how teachers view the assessment feedback of CoGrader on student essays. Three L2 teachers took part in this study by being interviewed in depth and being observed as they graded students’ essays using CoGrader. The data were then analyzed thematically using Brown's (2004) principles of language assessment, which include practicality, reliability, validity, authenticity, and learning impact (washback). The result shows that teachers appreciate CoGrader's speed in finding technical errors such as structure and language use implying that the principles of practicality and reliability are fulfilled in the assessment. However, teachers also view that the system cannot understand the personal context of students’ essays, which may result in their failing to make revisions in a meaningful way. These shortcomings indicate that the principles of validity, authenticity, and washback are not yet fulfilled when using CoGrader as an assessment tool. In practice, teachers still check, simplify, and change AI input before giving it to pupils. This underscores the critical role of teachers as a bridge between technological advancements and students' learning needs. The study emphasizes aspects of AI integration that are often overlooked and helps create a balanced writing assessment system that combines teachers' pedagogical knowledge with technological advancements, particularly in the learning process. 

Keywords: Generative AI, Feedback, Language Assessment, Teacher Perceptions