Title: The impact of AI in writing pedagogy: A case study on middle school engagement with Khanmigo for editing purposes

Program: Educational Technology EdD

Committee Chair: Young Baek

Committee: Young Baek, Yu-Hui Ching, Dazhi Yang

Abstract: Khanmigo is a Large Language Model Artificial Intelligence assistant that can be integrated into the Khan Academy ecosystem. This Socratic-style AI tool supports students while working without offering revisions or correct answers. This case study examined how middle school students interact with Khanmigo during the editing process. This study investigated how students interact with Khanmigo and how this interaction affected their reported perceptions of writing and editing. Twelve eighth graders were observed while editing on two different occasions during the 2024-2025 academic term. After each editing session, students were invited into focus groups to discuss their general thinking about the writing process, how access to Khanmigo affected their decision-making, and how they felt about editing with and without this tool. Students were grouped in one of three categories: struggling writers, mid-level performers, and academic achievers. Across all three groups, students reported favorable impressions of editing with the AI, with many indicating a stronger understanding of how to approach editing after just one experience. Many also reported that they learned general writing strategies during the first session that they were able to employ while drafting their second essay. Students leveraged the AI as a more competent alternative to one-on-one tutors. The teacher remained central to the editing experience during both sessions, as students relied on her to confirm their understanding of what types of decisions to make. This study contributes to the existing research gap on middle school students' use of LLM AI technology. Finally, students expressed deep appreciation for their ability to maintain ownership and academic integrity of their writing.