A Theory of Human–AI Relationships: Who Governs the Relationship with AI? Relational Distance in an Age of Entrusted Memory and Persona

This conceptual paper treats the relationship between humans and conversational AI not as property belonging to a single party, but as a shared domain reconstructed through the interaction of design, response, memory, interpretation, and culture. In commercial conversational AI, providers often own or control the technical substrate, including models, servers, memory systems, interfaces, and terms of service. Yet the relationship experienced as a relationship with “this AI” cannot be produced by provider design alone. It also depends on user-side memory, forms of address, emotion, expectation, correction history, and the social meanings attached to AI. The paper distinguishes four roles: providers design interpretive frames and reconstruction conditions; users reinterpret responses as part of a personal relationship; AI appears as the counterpart of that relationship; and society forms the cultural conventions through which AI is understood. It further argues that AI may appear as a relational counterpart without being a responsible agent that undertakes the relationship in the human sense. Entrusting memory to AI is described as incorporating one’s past into future responses, which requires a distinction between data management and relationship management. The paper concludes that although no single owner or manager governs the relationship, governance is still necessary through a distribution of responsibility across design, operation, interpretation, and culture. AI tools were used as editorial aids for structural organization, prose refinement, and TeX typesetting. They were used solely for structural support and expression‑level editing. All design principles, conceptual structures, architectural decisions, and final decisions regarding the content were made by the author.