#1025: Small Arguments, Big Ideas: Fostering Children’s Collaborative Reasoning with LLM Agents

Collaborative reasoning is at the core of critical thinking, problem solving, and effective communication, with growing attention being devoted to developing pedagogical and technological supports that enhance it. While many existing conversational agents (CAs) for education emphasize comprehension or factual recall during reading, the use of advanced reasoning behaviors, such as persuasion and inquiry, may further foster children’s engagement in collaborative reasoning. Based on recent advancements in large language models (LLMs), we created CAs that engage children in collaborative reasoning through argumentation during book-reading activities, by operationalizing three dialogue types from Walton’s framework of dialogue and argumentation: information-seeking, persuasion, and inquiry. We conducted a four-day exploratory study with 18 8-year-olds at a literacy-focused summer camp. Our findings revealed that while asking questions can support children’s collaborative reasoning, incorporating even small amounts of persuasion and inquiry elicits significantly more reasoning behaviors, especially explanations, clarifications, and reflections, from the children.