Artificial agents that look and act like humans are a prominent research field in embodied communication and human-machine interaction. Researchers in the field of Affective Computing consider the emotional dimension important to let these agents successfully take part in social interaction.

This thesis presents the WASABI architecture ("WASABI Affect Simulation for Agents with Believable Interactivity"), a computational architecture for an emotionally believable agent that is based on established constructs and empirical evidence from the affective sciences. In contrast to most other approaches to modeling emotions, the idea of an emotion dynamics in three-dimensional affect space is central to the WASABI architecture, which naturally sustains mood congruency of emotions. Furthermore, the distinction between primary, onto-genetically earlier types of emotions and secondary, cognitively elaborated emotions is followed, both influencing the agent's bodily emotion dynamics. As a result, this architecture is not only more believable for interactants with an agent but also for theorists from other disciplines concerned about emotion.