This arXiv paper presents Maike, an AI chatbot built to support secondary-school students' argumentative writing. Its question is how AI writing support can connect what students reflect on with how they revise. The authors start from the position that helping students write arguments with AI requires linking their reflection to their revision process.
Maike is a Socratic chatbot. It diagnoses the arguments in a student's essay and asks targeted questions based on that diagnosis, while a student model tracks the learner across 2 modes, Think and Write. It supports argumentative reasoning without generating essay text for the student. To test it, the authors ran a classroom study with 65 secondary-school students, who used either Maike or a control chatbot. The control shared Maike's interface and language model, and it offered the same Think and Write modes.
About 40% of students improved their essay scores, and the abstract doesn't report a score advantage for either group. Differences showed up in how students worked. Maike users engaged more in idea-level dialogue, 56% compared with 44% for the control group, and they moved back and forth between discussion and editing. Across the study, transitions from ideas discussed in dialogue to actual revisions were positively associated with score gains.
From these results the authors propose a design target for AI writing support, which is coordinating reflection with revision. They describe 2 contributions, an architecture for supporting argumentative reasoning without producing essay text and an approach that links interaction patterns to writing processes and outcomes. They argue that as AI enters classrooms, understanding these connections is needed both to judge its educational value and to design for it.