AI Unplugged: Exploring Pathways from Physical Simulation to Conceptualization of AI Reasoning Processes

Young adolescents interact with AI daily, but there is a lack of approachable, concrete opportunities for them to develop an understanding of AI reasoning processes. We leverage theories of embodied learning to design four unplugged activities— DataBites, Monster Mapping, Nature Net, and Kernel Quest —that foster young adolescents’ understanding of machine learning, convolutional neural networks, and semantic networks through physical manipulation and simulation of AI reasoning. We conducted an exploratory study with 20 middle-school-age participants to understand whether the activities led them to concretize AI reasoning processes by identifying with the AI and whether this improved their conceptualization of AI reasoning. Our results suggest that the activities engaged learners in physical simulation of AI reasoning processes and appeared to help them develop conceptual understanding by concretizing AI concepts and embodying the AI’s perspective. We reflect on implications for designing AI education interventions that emphasize physical manipulation and AI simulation to support AI literacy among young learners.