Toward a comprehensive theory of emotion for biological and artificial agents

Abstract A new model of emotions that is applicable to both biological and artificial agents is proposed. The description includes theoretical foundations, internal representation, and the role of emotions in cognition. This model is based on definitions of emotions in valence and arousal space coupled with an adaptation of Maslow’s hierarchy and other ideas. The resulting architecture provides for a significantly more expressive range and organization of represented emotional experience compared to other models. Requirements for a satisfactory general computational theory of emotions are applied to the new theory and analyzed in terms of (i) neurological and psychological plausibility, (ii) range and complexity of human emotional experience, (iii) applicability to learning, memory, behavior, and decision-making, and (iv) consistency with well-accepted models and general facts about emotions. The model is implemented and studied through simulations of virtual agent-based systems. Presented results support the model’s applicability to perception, action selection, learning, and memory in virtual agents capable of human-like behavior. Paradigms and predictions allowing for further validation of the new model and emotion theories in general are discussed.