Intero-Expressive Robots: Externalizing Robot Internal Signals for Exploring Theory of Mind Attribution to Robots

We present the concept of Intero-Expressive Robots, proposing external representation of internal signals to make hidden states perceivable during interaction. In humans, external indications of interoceptive signals such as respiration or cardiovascular function support attribution of internal states to others (Theory of Mind) by providing information that cannot be inferred from behavior alone. Drawing on interoception as a design analogy, we discuss how exposing robotic variables—such as computational load or energy use—may support analogous forms of state attribution. We position intero-expression as complementary to robot transparency and legibility: externalizing internal dynamics that shape behavior and constrain performance. We describe preliminary feasibility work in a virtual environment, where a robot executes a task while a CPU-derived internal signal is logged. Identical actions produced different loads depending on task context, patterns not predictable from observation alone. This suggests that intero-expressive representations can be interaction-relevant cues.