#942: Tool, Tutor, or TBD? Examining Learners’ Perception Trajectories and Interaction Patterns through the Lens of Metacognitive Support

This study examines how students’ perceptions of StatPal, an AI conversational agent for statistics learning, evolved and shaped interaction patterns over the course of a college statistics class. Using survey data, open-ended reflections and conversational exchanges, we identified five distinct perception trajectories that began with students initially viewing the AI as either a Tool, a Tutor, or To-Be-Decided, and ended with two final orientations: Tool-oriented or Tutor-oriented. Tutor-oriented students reported receiving more metacognitive support in procedural and conditional knowledge. Using Epistemic-Network-Analysis, we found that perceptions of AI shaped both quantity and quality of learning interactions. Tutor-oriented students engaged with the AI in active exploration with cognitive activation and procedural inquiries, while tool-oriented students engaged in passive, surface-level inquiries relying primarily on declarative assistance. These findings connect learners’ evolving conceptualizations of AI with the depth of their metacognitive engagement, providing empirical guidance for designing educational AI that promotes deeper metacognitive engagement.