Scaffolding young children’s AI use: opening pedagogical black boxes in early childhood classrooms

Abstract Given their real-time responsive capabilities, artificial intelligence (AI) technologies have been positioned as cognitive partners for young children. However, research and policy directives have tended toward technological determinism, which black-boxes how these technologies function in early childhood education (ECE). Applying Actor-Network Theory (ANT), this qualitative case study deconstructs pedagogical black boxes during a plant inquiry unit involving a teacher, four children (aged 5–6), and AI-integrated devices. The findings provide a typology of three pedagogical black boxes: operational (obscuring technical constraints), epistemic (obscuring knowledge construction), and relational (obscuring negotiations of agency and classroom authority). Analysis of the case illustrated how these black boxes opened through distance (children’s communication gaps), breakdown (AI errors), and innovation (negotiation over AI use), revealing hidden dynamics in AI integration, such as uncertainty, people at work, decisions, competition, and controversies. Teacher scaffolding emerged as a key mediator of child-AI interactions, translating children’s open-ended inquiries into AI-legible inputs, transforming AI errors into opportunities for evaluative judgement, and recalibrating children’s AI interactions into child-led scientific inquiry. Ultimately, the paper shows that the seamless integration of AI is neither realistic nor pedagogically effective, and exemplifies how emergent frictions can foster young learners’ AI literacy, evaluative judgement, and relational agency.