About the Group
The Computational Emotion Group at the University of Southern California studies the processes underlying behaviour that people interpret as emotional:
- what is the information processing that underlies this behaviour;
- how is it physically and mentally manifested;
- what features of this behaviour drives an observers interpretation;
- how might these interpretations influence the behaviour of an observer and, most importantly,
- how might these phenomena be captured and exploited by computational models?
From an engineering perspective, we strive to exploit these findings to develop computational systems that communicate more effectively, particularly with respect to "virtual humans" (lifelike-autonomous agents that can participate in multi-model interaction with human user across a variety of educational applications).
From a scientific perspective, we strive to use computer generated characters as a tool to investigate how people interpret emotional behaviour and how these interpretations influence (directly or indirectly) memory and decision-making.
Major ongoing research efforts include the development of EMA, a computational model of appraisal theory that incorporates a detailed model of appraisal and coping and their influence on cognition, and the Virtual Human project, a long-term interdisciplinary research effort to develop an embodied conversational agent that utilizes EMA to model the cognitive and behavioural influences of emotion.
In addition to models of emotion, this project integrates advanced research in natural language processing (including individual research projects on speech recognition, natural language understanding, dialogue management, nonverbal communication, and animation).
Several smaller research efforts include: the development of machine learning techniques to characterize expressive behaviours, computation models of social attribution theory and social influence theory, and psychological studies that validate these models.