Computational Models of Cognitive and Affective Theory of Mind

Theory of Mind is described as the capability to attribute mental states to oneself and others, and it can be essential for robots to favor more collaborative, adaptive, and emotionally appropriate behaviors when they are deployed in human-centered environments. In this work, we survey existing methodologies to introduce ToM in Human-Robot Interaction, focusing on two main formalizations: Cognitive ToM, focusing on reasoning about beliefs and intentions in a more task oriented way, and Affective ToM, which requires the agent to recognize and adapt to others' emotional states. While both approaches have advanced robot adaptability and user engagement, they are often developed in isolation. For this reason, we will discuss the need for integrated models that combine cognitive and affective reasoning, and outline future directions for more socially intelligent and emotionally aware robotic systems.