RUME 28
2026
Alexandria, Virginia
Constructive or Procedural? Student Engagement With an AI Tutor in Multivariable Calculus
Page: 1199
Large language models (LLMs) such as ChatGPT are increasingly used in higher education, raising questions about how they influence student learning. While unstructured use often leads to shallow engagement, carefully designed AI tutors may foster deeper reasoning. This study examines student engagement with an AI tutor in a multivariable calculus course, drawing on the ICAP framework. The focal task asked students to use dot products and vector reasoning to estimate storm velocity from radar data. Preliminary results suggest that while the AI tutor frequently posed prompts aligned with constructive engagement, student responses were most often active, reflecting a “completionist” approach to problem solving. Constructive reasoning was difficult to sustain, and tutor follow-up was sometimes limited. These findings highlight both the promise and the complexity of designing AI tutors to elicit and maintain constructive mathematical engagement.