#910: Quantitative Reasoning During Student-LLM Interactions: A Comparative Case Study

Interpreting statistical information in ways that maintain humanistic perspectives requires strong quantitative reasoning skills. Yet it remains unclear whether students are using LLMs in ways that exercise or short circuit reasoning processes that strengthen such skills. To address this gap, we ask: Does using an LLM during problem solving influence how students reason about contextual statistics tasks? We observed and think aloud, task-based interviews with 6 non-STEM majors who were given access to ChatGPT and analyzed moment-to-moment reasoning to identify how LLM interactions shaped their attention to problem quantities. Findings show unexpected behavioral patterns according to students’ prior experiences with LLMs. We discuss implications for quantitative reasoning and related skills.