Spontaneous Theory of Mind for Artificial Intelligence
Nikolos Gurney, David V. Pynadath, Volkan Ustun
TL;DR
The paper interrogates the dominance of prompted Theory of Mind (ToM) research in AI and advocates a shift toward spontaneous ToM as a foundational aspect of robust Artificial Social Intelligence (ASI). It surveys historical and theoretical perspectives on ToM, contrasts spontaneous and prompted reasoning, and reviews empirical methods used to study ToM in humans and AI. It then proposes a principled framework for ASI development that emphasizes definable social intelligence skills, the impact of framing questions on answers, and the necessity of ground-truth benchmarks. The authors argue that combining prompt-responsive and spontaneous social reasoning capabilities will yield more generalizable, resilient ASI, with practical benefits for human–AI collaboration and safety.
Abstract
Existing approaches to Theory of Mind (ToM) in Artificial Intelligence (AI) overemphasize prompted, or cue-based, ToM, which may limit our collective ability to develop Artificial Social Intelligence (ASI). Drawing from research in computer science, cognitive science, and related disciplines, we contrast prompted ToM with what we call spontaneous ToM -- reasoning about others' mental states that is grounded in unintentional, possibly uncontrollable cognitive functions. We argue for a principled approach to studying and developing AI ToM and suggest that a robust, or general, ASI will respond to prompts \textit{and} spontaneously engage in social reasoning.