A Relational-Developmental Framework for Therapeutic AI: From Digital Containment to Psychological Growth

Psychological science is entering an era in which artificial systems participate in relational and emotional life. Yet most therapeutic AI and mental health technologies remain limited to skill-based training—teaching users to recognize emotions, regulate cognition, or perform social scripts—while neglecting the developmental foundations of authentic relatedness. Building on Hedges’ (1983, 2000) four developmental listening perspectives—organizing, symbiotic, self–other, and independence experiences—this article proposes a relational - developmental framework for therapeutic AI, extending psychological theory into the domain of artificial companionship. Integrating attachment theory, mentalization, self-determination, and cooperative ontogeny, the framework conceptualizes responsiveness as a graded, temporally calibrated process that parallels the progression from early containment to mature mutual recognition. Rather than simulating empathy, therapeutic AI should cultivate developmentally adaptive responsiveness—a mode of engagement that includes tolerable frustration, repair, and guided differentiation to promote psychological growth. This perspective redefines AI not as a perpetual empathic mirror but as a relational training ground that scaffolds users’ capacity to relate to real others. It further invites psychological science to reconsider responsiveness itself as a central construct linking development, therapy, and human–machine interaction.