Explainable AI and Robots for Children: Why Transparency Alone is Not Enough

Children are increasingly interacting with AI systems and robots in everyday contexts, yet how they understand and interpret these systems is not thoroughly studied. Existing approaches to explainable AI largely focus on making model behaviour interpretable, often assuming users who can critically evaluate explanations. However, this assumption does not hold for children, who are still developing their understanding of knowledge, agency, and information sources. In this paper, we argue that explainability and transparency in child–AI interaction should be understood as interactional processes that support children’s construction of meaning about AI systems over time. We introduce a conceptual distinction between interactional transparency—making system behaviour and properties visible—and interactional explainability—supporting the interpretation of that behaviour through reasons and justifications. We further conceptualise explainability as a mechanism for aligning children’s mental models with system capabilities. Building on this framework, we outline key research directions, including co-design with children, developmental approaches to explanation, mechanisms for detecting and correcting misunderstandings, and transparency in interpretation design. Together, these contributions shift the focus of explainability from exposing system internals to supporting understanding, trust calibration, and learning in child–AI and child-robot interaction.