#498: What AI “Sees” and What Facilitators Know: Supporting Collaboration in Constructionist Making Environments

As multimodal AI tools enter educational spaces, questions emerge about whether and how they might replace or complement educator expertise. Experimenting with AI-embedded tools in a middle-school constructionist making program, this study examines what AI and human facilitators attend to when encouraging collaboration between makers. We analyze AI-generated project descriptions and collaboration recommendations, alongside facilitator discussions of collaborations in debrief transcripts. We found that AI identified technical features and suggested direct feature-copying in student projects, while facilitators layered social dynamics, developmental trajectories, and pedagogical goals.. Our findings challenge replacement framings and offer AI design suggestions for leveraging pedagogical expertise. We argue that designing AI for constructionist environments requires strategic offloading: leveraging AI for tasks that extend facilitators’ observational reach while preserving facilitators’ judgment for decisions central to identity development and relationship building.