From Bit to Body: How AI Systems Can Develop a Sense of Body Ownership

Shuqin Ma

Abstract

This paper examines whether artificial intelligence systems can develop a sense of body ownership (SBO) by constructing virtual bodies through multimodal world models. The discussion begins by revisiting the concept of a “body” in SBO theory, framing it as a functional boundary of internal-external interaction, not confined to a purely physical structure. Building on this, the paper explores how theories of multimodal sensory integration can support the idea that virtual bodies may indeed foster a sense of body ownership. The idea of “world models” is then introduced as a framework for modeling environmental interactions, with an emphasis on how these models might achieve semantically centered multimodal sensory integration. This, in turn, enables AI systems to experience SBO over virtual bodies. Practical implementations, such as coupling these virtual bodies to external hardware or applications in virtual reality, are also considered. The paper concludes that granting SBO to AI systems can promote adaptability and decision-making quality while potentially increasing transparency, though it raises pressing questions about AI safety and the ethical design of embodied AI.