Limitations Support Learning: A Design Based Study of Ethically Scaffolded Generative AI Tools for Long Form Argumentation and Metacognition

This design‑based research study implemented and refined an instructional model that used generative AI (GAI) tools within a nine‑week AP Capstone Seminar argumentation unit. The model combined teacher feedback and modeling, structured reflection tasks (including AI‑assisted reflection), AI assisted editing, AI-assisted literature searches, and counterargument support using restricted chatbots. During the nine week period, responsibility for feedback moved from teacher and AI supported scaffolding to student driven writing practice and revision. Data collection focus on four advanced high school students at a Florida public high school, including multiple drafts, GAI input and output logs, reflections, and pre‑ and post‑unit semi‑structured interviews. Data analysis showed that with these scaffolds in place, students improved their counterarguments, evidence integration, and essay structure. However, counterargument and rebuttal were not integrated with the main line of reasoning. As students became more self regulated and used more nuanced prompts with the GAI tools, metacognitive growth emerged. The tools supported this growth by prompting students to plan, monitor, and evaluate their writing choices, especially during revision and reflection tasks. AI-assisted reflections and editing tools gave students immediate feedback that encouraged intentional reconsideration of structure, evidence use, and argumentative clarity. In this model, limitations such as restricted tool access, scaffolded reflection, and gradually withdrawn support became productive constraints that helped students build independence and insight. Ethical awareness changed from naive acceptance to a more discerning evaluation of each tool’s benefits and limits. Over time, reliance on AI tools decreased, indicating both improved writing ability and strategic use of technology. Teaching students to question and direct AI output is as important as traditional writing instruction and improves their metacognition and ethical use of AI in high school writing classrooms.