#324: “I Know It Because I've Seen It”: Epistemic Agency and Tool Selection in Student-LLM Mathematical Problem Solving

Though students are using large language models (LLMs), we do not yet know how students interact with LLMs as they solve problems or how they evaluate whether or not to use a LLM compared to other technological resources. This study addresses both of these gaps. We recruited ten undergraduates to participate in think aloud, mathematics task-based interviews. We analyzed students’ selections of technologies and the extent to which they built and verified their solutions through progressive inquiry. We found that students select LLMs more frequently than internet-based resources, and that progressive inquiry was related to students’ instrumental genesis (e.g., knowledge of tools and their affordances). We discuss implications for knowledge building with LLMs as they relate to human epistemic agency.