METACOGNITION-BASED ACADEMIC WRITINGWITH GENERATIVE-AI SCAFFOLDING:STUDENT AND INSTRUCTOR NEEDS

Abstract

This study examines metacognition-oriented academic writing needs and the supportive role of generative AI among cross-disciplinary master’s students. A convergent mixed-methods design combined an online survey (n = 120; 20 Likert items), semi-structured interviews (12 students, 8 faculty), and document analysis of syllabi, rubrics, and anonymized student texts to compare intended scaffolding in curricula with scaffolding experienced in practice. Results show that metacognitive awareness exceeded the scale midpoint, with the highest scores in Linguistics, while difficulties in constructing coherent arguments were more evident in Management and Law. Attitudes toward generative AI were generally positive in these professional programs, though ethical policies and reflective opportunities varied across courses. Thematic and document analyses indicate that paragraph-level argument quality improves when rubric criteria are accompanied by worked examples of claim–evidence–warrant patterns and when feedback cycles require explicit articulation of reasoning links and transitions. Pilot reliability (Cronbach’s α < .70 on some subscales) suggests several items require refinement; quantitative data were therefore interpreted descriptively and triangulated with qualitative and documentary findings. Recommendations include embedding brief, rubric-aligned reflective checkpoints, using worked-example and self-explanation exercises to stabilize claim–evidence–warrant connections, and providing explicit ethical guidance that positions generative AI as a reasoning companion rather than a substitute for reflection.