Social World Models: Universal Structured Representations for Social Reasoning

ICLR 2026 Conference Submission13907 Authors

Keywords: social intelligence; theory of mind; world model

Abstract: Humans intuitively navigate social interactions by simulating unspoken dynamics and reasoning about others' perspectives, even with limited information. In contrast, AI systems struggle to automatically structure and reason about these implicit social contexts, largely due to traditional input representations (e.g., free text) being lossy, shaped by reporting biases, and often omitting crucial details. In this paper, we introduce a novel structured social world representation formalism (S3AP), designed to unlock social reasoning in AI systems. Following a POMDP-driven design, S3AP represents social interactions as structured tuples, such as state, observation, agent actions, and mental states, which can be automatically induced from free-form narratives or other inputs. To demonstrate the power of our representations, we first show S3AP can help LLMs better understand social narratives across five social reasoning tasks (e.g., +51% improvement on FANToM's theory-of-mind reasoning over OpenAI's o1), reaching new state-of-the-art (SOTA) performance. Then, we introduce an algorithm for social world models using S3AP, which enables AI agents to build models of their interlocutor and predict their next actions and mental states. Empirically, S3AP-enabled social world models yield up to +18% improvement on the SOTOPIA multi-turn social interaction benchmark. Our findings highlight the promise of S3AP as a powerful, general-purpose representation for social world states, enabling the development of more socially-aware systems that better navigate social interactions.

Primary Area: neurosymbolic & hybrid AI systems (physics-informed, logic & formal reasoning, etc.)

Submission Number: 13907