The relational–epistemic stance: generative AI as a dynamic transitional object

Abstract Generative AI interaction produces divergent outcomes: genuine intellectual growth for some, convincing performance of competence for others. Multiple philosophical traditions—post-phenomenology, extended mind theory, and distributed cognition—have independently converged on this question, yet each reaches toward the developmental dimension without the theoretical resources to resolve it. This paper argues that the convergence reveals a structural gap: existing frameworks can describe how technologies mediate experience but lack the developmental apparatus to specify what determines whether a given encounter builds internal capacity or substitutes for it. Drawing on object relations theory—particularly Winnicott, Bion, and Bollas—the paper develops the Dynamic Transitional Object (DTO): a relational framework specifying the conditions under which AI interaction produces development rather than erosion. The framework identifies metabolization—the transformation of AI-generated material into owned thought—as the differentiating process, and epistemic seduction—the designed pull toward bypassing that work—as the structural force opposing it. What determines the outcome is not the technology but the relational–epistemic stance. When institutional evaluative frameworks cannot distinguish genuine capacity from its convincing performance, the developmental question ceases to be individual.