Exploring Learner Autonomy Through Chatbot Mode Selection: Interaction Patterns and Learning Outcomes in an Online Course

• In Event: Rethinking Learner Engagement in the Age of Intelligent Systems

Abstract

Despite the growing presence of chatbots, few studies have explored how students exercise autonomy when offered multiple interaction modes. This mixed-methods study investigates how learners engaged with a dual-mode chatbot, general (large language model, LLM-based) and reference-based (LLM and retrieval-augmented generation-based) modes, in a large-scale online course. Based on chat logs, learning management system logs, server logs, and post-course survey data (n = 121), the findings revealed that more than 40% of students were unaware of mode choices. Those aware favored the general mode for convenience and the reference-based mode for precision. Mode selection was significantly associated with interaction behaviors and perceived learning experiences. These findings highlight the importance of learner autonomy, awareness, and design integration in chatbot-mediated learning environments.

Author

• Songhee Han, Florida State University