ShadowAI: Fostering Children's AI Literacy and Critical Understanding of AI Reasoning

AI systems are increasingly shaping everyday life through data capture and reasoning processes that often remain difficult for children to grasp. This paper presents ShadowAI, a playful, metaphorical, and creative system designed to support children’s AI literacy in informal learning environments. It comprises the Shadow Capture Box and ShadowStories. Drawing on the metaphor of AI as a “ghost” that hides within the shadow of data, and on the benefits of creative making that makes AI concepts tangible, engaging, and personally meaningful. ShadowAI invites children to create play-doh artefacts whose shadows are captured and interpreted, generating images, narratives, and staged visualisations of AI reasoning. Rather than offering conceptual explanations of AI, the system uses ambiguity, family interaction, and discussion to help children critically engage with how AI observes, interprets, and reasons from partial input, while reflecting on data and agency.