Beyond World Models: The Causal Self Boundary | A Missing Layer Between Spatial Intelligence and Autonomous Agency

Abstract

Description World models are increasingly central to the development of artificial intelligence capable of representing space, time, physical dynamics, and the consequences of action. Yet the ability to model the world does not, by itself, explain how a system distinguishes its own actions and internal dynamics from changes originating in its environment. This conceptual hypothesis paper proposes the Causal Self Boundary (CSB) as a candidate architectural requirement for autonomous agency. The central proposition is that an autonomous system must maintain and update a distinction between self-caused and world-caused change, while preserving a sufficiently coherent representation of its own state, actions, and causal influence over time. Drawing on biological self-organization, theories of selfhood, sense-of-agency research, and existing approaches to world and self-modeling, the paper develops an account of agency as an endogenous property of a distributed system rather than the product of a central conscious controller. Within this framework, consciousness is considered a possible expressive and recursively influential interface through which aspects of the system's self/world distinction become available for reflection, coordination, and social interaction. Phenomenal consciousness is not proposed as a prerequisite for artificial autonomy. The paper situates the CSB hypothesis in relation to prior work, identifies its proposed contribution, and formulates testable predictions for biological and artificial systems. Its aim is to clarify the transition from world modeling to autonomous agency and to provide a framework for further theoretical and empirical investigation. Publication status: Conceptual hypothesis paper. Not peer reviewed. AI-assisted editing disclosure: The conceptual framework, hypotheses, arguments, and interpretations are those of the author. OpenAI ChatGPT was used to assist with language editing, structural refinement, literature discovery, and bibliographic cross-checking. The author takes full responsibility for the final content.