Redesigning Writing Assessment: A Comparative Study of B2 learners’ writing with and without AI Assistance at a selected university.

Author : Bayan Bader

Abstract : The recent widespread of generative AI tools- in the educational context- has significantly affected L2 writing practices that raised several questions about the validity of current writing assessment. This paper addresses the need to redesign writing assessments by examining the impact of AI assistance on B2 students’ writing performance to explore the implications for redesigning L2 writing assessment. The researcher adopted a quasi comparative study design; it compared argumentative texts produced by B2 students at a selected university under two conditions: AI assisted writing and non AI assisted writing. Participants completed the same authentic writing task under the same time constraints. Texts were evaluated by using CEFR-aligned analytic frame work. It measured task fulfilment, organisation and coherence, linguistic range, accuracy and argumentative quality. The researcher collected reflective data to examine participants process writing and their strategic use of AI. Results indicated that AI assistance substantially enhanced accuracy and surface coherence. However, its impact on higher order thinking skills and dimensions such as task interpretation and argumentative reasoning is more limited. Therefore, this paper suggests a clear distinction between AI sensitive assessment criteria and non AI robust indicators or writing ability. In light of Messick’s validity framework, the study argues for an interpretive recalibration of writing assessment that distinguishes AI sensitive from AI robust criteria rather that excluding AI use from assessment practices.

Keywords : AI-Assisted L2 Writing and Implications for Assessment Validity

Conference Name : International Conference on Technology-Enhanced Teaching and Learning Innovation (ICTETLI-26)

Conference Place : Bristol, UK

Conference Date : 12th Mar 2026