Reducing Dependence on AI-Generated Text through Contextual Writing Assessment

Abstract

Although generative artificial intelligence (GAI) increasingly impacts academic writing and integrity in higher education, empirical evidence on assessment strategies to reduce students' reliance on AI-generated texts remains limited. This study aimed to examine the effectiveness of the Contextual Writing Assessment in reducing AI-generated text dependence and improving writing performance, writing originality, academic integrity, and learner agency. A quasi-experimental mixed-methods design with an embedded explanatory approach was employed in an English course at the Faculty of Engineering at the University of Muhammadiyah Tangerang. The participants comprised 150 second-semester students from six classes and four lecturers. Three classes were assigned to the experimental group and received Contextual Writing Assessment, while three classes served as the control group and received the conventional writing assessment. Data were collected through writing tests, AI-dependence questionnaires, academic integrity and learner agency questionnaires, originality scores, AI-use logs, reflective journals, draft analyses, observations, and interviews.  The results demonstrated the effectiveness of the intervention, as the experimental group outperformed the control group in reducing AI-generated text dependence and exhibited substantial improvements in writing performance, originality, academic integrity, and learner agency. Qualitative evidence indicated more transparent AI use, meaningful revisions, stronger reflections, and reduced copy-paste behavior. The study concludes that Contextual Writing Assessment offers an ethical, process-based, and AI-resilient approach to writing assessment.