<![CDATA[Background. Artificial intelligence is increasingly used to support EFL academic writing; however, most existing tools emphasize language correction, text generation, or general feedback rather than helping students examine the underlying structure of argumentative reasoning. In particular, EFL learners often experience difficulty connecting claims with relevant and sufficient evidence. Purpose. This study aimed to develop and evaluate ArguMap-AI, an intelligent claim–evidence mapping system designed to support argumentative writing by identifying argumentative components, visualizing claim–evidence relationships, and providing diagnostic feedback for revision. Methods. The study employed a design and development research approach supported by mixed-method evaluation. Participants were 21 fifth-semester students enrolled in the Writing Project course in the English Education Study Program at Universitas Muhammadiyah Enrekang during the 2025–2026 academic year. Data were collected through initial and revised argumentative drafts, an analytic writing rubric, ArguMap-AI-generated maps and feedback records, student perception questionnaires, and semi-structured interviews. Quantitative data were analyzed descriptively, while qualitative data were examined thematically. Results. The findings showed improvement across all assessed dimensions of argumentative writing, with the largest gains occurring in claim–evidence alignment, reasoning and explanation, and evidence relevance and sufficiency. The mean total writing score increased from 18.19 to 26.33. Students generally perceived the visual mapping and diagnostic feedback as useful for identifying unsupported claims and improving revision decisions. Challenges included occasional component misclassification, complex visual maps in longer texts, and the continuing need for lecturer guidance. Conclusion. ArguMap-AI shows potential as an intelligent reasoning scaffold for EFL academic writing by helping students visualize, evaluate, and revise argumentative relationships while preserving learner decision-making.]]>
Copyrights © 2026
