Speaker
Description
In higher education, academic writing demands significant cognitive and linguistic effort from EFL students, making self-assessment a valuable pedagogical approach to foster self-regulated learning. With generative AI increasingly present in education, there is growing potential to leverage it beyond basic proofreading, yet its role in supporting the deeper cognitive and analytical dimensions of self-assessment remains largely underexplored. This qualitative study explores the aspects AI tools can support university students in assessing their own writing, and how students perceive the self-assessment with and without AI in Academic Writing. Grounded in Harris and Brown's integrated framework of self-regulated learning and formative assessment, the study employed rubric-based self-assessment documents and semi-structured interviews as primary instruments. Participants were divided into a non-AI group (n=8) and an AI group (n=9) using ChatGPT Plus. The findings revealed that while both groups could identify weaknesses in their writing, they differed in depth and accuracy. AI-supported students demonstrated stronger problem recognition in cognitively demanding criteria with more analytical justifications, whereas the non-AI group often only identified their performance level and failed to provide detailed evaluations. Both groups held positive but condition-dependent attitudes toward self-assessment, with shared challenges in rubric interpretation and personal bias. The AI group also demonstrated varying degrees of trust in AI feedback. The study concludes that AI enhances the analytical dimension of self-assessment but cannot replace teacher-guided evaluative calibration, offering practical implications for AI-integrated pedagogy in Vietnamese EFL higher education.
Keywords: self-assessment, academic writing, artificial intelligence, EFL, Vietnamese higher education
