#1148: Meet Bip: Understanding Student–Peer Agent Interaction in an Embodied Exploratory Learning Environment

Peer agents have been shown to enhance learning in K–12 settings, yet the design of LLM-powered embodied peer agents for exploratory learning remains underexplored. This study examines how middle-school students interact with Bip, an embodied conversational agent in a Minecraft-based AI literacy game. Inspired by the Intelligent Novice paradigm, Bip observes players, provides hints, and scaffolds reflection during an AI literacy activity. Findings from nine students highlight the importance of spatial awareness, responsiveness, and bounded understanding in supporting exploratory learning. We discuss design implications for embodied agents that foster scientific inquiry.