A Feasibility and Implementation Integrity Study of the Community Builder (CoBi): An AI-based Collaboration Support System in K-12 Classrooms

Abstract We present the “Community Builder” or CoBi—a classroom-wide AI system to help students develop their collaboration skills. CoBi uses speech recognition and language understanding to create non-evaluative and privacy-preserving visualizations of uplifting small group student discourse. The visualizations are intended to serve as resources for students to reflect on their classroom collaborations with guidance from their teacher. We conducted an interview study with three middle school teachers and 12 students to probe attitudes about CoBi and seek feedback on initial design concepts. Both groups appreciated the intention behind CoBi and expressed a preference for more qualitative over quantitative types of visualizations of their classroom collaborations; students also raised the importance of privacy and teachers expressed concerns about classroom integration. Prototypes were tested in two classroom studies—a technical feasibility and initial implementation study and an implementation integrity study—with 61 students across six middle school classrooms in the Western US. Findings indicated that students and teachers valued an AI system to potentially improve their collaboration skills, but CoBi also requires significant professional learning for high integrity use. Teachers pursued a range of strategies for integrating CoBi into their pedagogies and activating students’ contemplations and reflections around collaboration as a skill. We discuss transferable design principles and recommendations for future AI learning systems.