The DEEP project is concerned with the challenge of mapping the current situation in which a person finds himself to symbolic situational representations linked to individual emotions. The project develops and investigates a novel combination of a real-time interpretation of human communication signals with a real-time computer model for emotions. The combination uses an extended representation that integrates communicative emotions and internal emotions, possible emotion triggers and goals, possible emotion regulation strategies and associated sequences of social signals and their direction. Based on this, a Theory of Mind representation of user emotions with possible mental states and cognitive strategies will be created at runtime through the interpretation of social signals of the dialogue partner. For the first time, this approach enables real-time disambiguation of emotion triggers and targets. In addition, a real-time interpretation of social signals can open up possible emotion regulation strategies. The DEEP project is developing a software real-time model that describes on a symbolic level how human social communication signals can be linked to the current situational context and internal emotional states. The DEEP model is evaluated in dyadic dialogue situations between humans and social agents. Such models are essential for the research and application of the next generation of interactive social agents. The DEEP model is an extension of user models to include internal feelings and emotion regulation strategies. This enables deeper empathic adaptation to a dialogue partner.
Universität Augsburg Universität des Saarlandes Leopold-Franzens-Universität Innsbruck