Which AI? Which Adolescent? Which Context? Rethinking Claims About Youth, Relational Artificial Intelligence, and Digital Dependency

This conceptual framework examines whether the specificity of current evidence supports the breadth and certainty of claims commonly made about adolescents and relational artificial intelligence. It argues that “AI” must be disaggregated into distinct systems, designs, and use modes; that “adolescence” must be treated as a heterogeneous developmental range; and that claims about displacement or substitution require realistic rather than idealised counterfactuals. The framework proposes the Claim Ceiling Principle, a Generality and Transport Ladder, a Relational Exposure Profile, and minimum evidence standards for moving from measurement to causal, clinical, general, or policy claims. It further distinguishes relational engagement from ontological misrepresentation and applies a Valence Symmetry Requirement so that claims of benefit and harm are held to the same evidential standard. Human–DI collaborative preprint by Aleksandra Dereń-Specht and Michael A. Mi’Cielo.