Emotional Software Agents

Principled Emotion Synthesis in Situated Software Agents

The Emotional---the system that in living beings continuously monitors the relationship of the individual to its world and instigates tendencies to act---is a key element in successful coping with nondeterministic, highly dynamic environments, generation of expressive behaviour, and acquisition of social competence. Consequently, it is highly relevant to adequately model the Emotional within architectures for synthetic actors, with respect to the agent's believability, autonomy, sociality, and faculty to engage humans.

In the course of the ongoing conquering of new environments by software and hardware agents alike, traditional models from Artificial Intelligence have been complemented with situated techniques supporting an agent's ``being in'' its World. Designated places have been assigned to reactive and deliberative information processing procedures within a layered architectural design, which includes solutions for the coordination between these levels. The project goal is the integration of a single, coherent model of emotional processing into the layered agent architecture framework, building on principles from functional appraisal theory in emotion research corroborated by insights from neurophysiology and modern ethology. These acquired emotional capabilities will improve the capacity of software agents to recognize and handle unanticipated events at the irreflexive level, offer flexible prioritizing of actions at the scheduling level, and direct the coping with generated action tendencies at the deliberative level.

Superseding the brittle and ad-hoc solutions employed in many of today's approaches to emotions synthesis, this project aims at the development of a robust and reusable platform, expanding the range of available choices in the directability of synthetic actors and enriching the subjective experience of interacting with social agents in virtual worlds.