2026 ASEE Annual Conference & Exposition

This paper reports on the design, classroom implementation, and evaluation of a game-based learning module aligned with relevant science and engineering curriculum standards for K12 students. Leveraging active learning principles, the module combines engaging game mechanics with a conversational LLM peer agent to provide on-demand hints and prompts that scaffold each student’s inquiry process during the play. To evaluate the module’s effectiveness, we conducted a mixed-methods study with three classroom conditions, employing multiple assessments: pre- and post-module concept tests measured learning gains; in-game behavioral analytics captured student engagement and inquiry behaviors; dialogue transcripts from student–LLM interactions were coded to analyze inquiry strategies; and student perceptions were gathered through surveys and focus groups.

Results indicate that students in the LLM-supported group demonstrated higher engagement and more successful inquiry-based problem solving than the other groups, highlighting the added value of interactive AI support over traditional instructional media. Student surveys and focus group feedback reflected a positive reception of the AI peer, with many reporting increased engagement and confidence during the learning process. Implementation insights highlight device and network constraints encountered when deploying a local LLM in the classroom, which informed iterative UI refinements and technical optimizations for smoother integration of the peer agent. These findings underscore the promise of AI-driven, active learning interventions for K–12 STEM education and illustrate the feasibility of integrating advanced AI tools in real classrooms under typical resource constraints.