This project will investigate human interaction with simulated agents in situations where humans expect empathic communication, such as healthcare. In domains such as this, the person interacts with computer-based interventions (CBI) as an education process evolves. Almost all psychosocial interventions currently available on the web are delivered merely via text. Research indicates that many users will lose interest and drop out, although completion is critical to achieving desired goals. Human-computer interaction literature suggests that interacting with simulated agents can increase the user's engagement, but that the user will expect social competence when interacting with them. This project will answer a set of research questions for the design of simulated characters with some empathic intelligence, to create a new modality for the delivery of CBIs. The specific health-related application area will facilitate development and evaluation of new techniques and design principles, that will have much wider applicability, potentially whenever people interact with computer-based systems through simulated agents. This research will advance the ability of computer scientists to create competent simulated agents that can adapt their verbal and non-verbal behavior to the user's affective states, and over time tailor their interaction to the specific user to produce the maximum positive impact in terms of users' engagement and achievement of their goals. The model of empathy and social competence developed will enable simulated characters to adapt in real time to a user's short-lived emotions over a single interactive session. Longer-term affective states will be modeled over long-term interaction via follow-up sessions with the same individual. Thus, development of rapport between a human and an artificial intelligence is a dynamic process over time, and the research is expected to discover new principles that might not be seen in exclusively short-term interactions. The project will provide: (1) a scheme to design tailored interaction by constructing a dynamic user-model with user's demographic information and fluctuating personal characteristics; (2) a model of empathic verbal communication built by combining motivational interviewing techniques with an ontology-based dialog system; (3) a computational model for the integration of verbal and non-verbal communication cues by adapting the character's facial expressions, vocal modulation, and kinesics to its verbal utterances in the context of the session. In addition, the research will engage scientific questions about diversity in communication style across human groups, facilitated by the fact that the project is housed at the region's principal minority-serving research university. The fact that the research will involve students in the development of systems that integrate multiple technologies will give them an excellent educational experience, gaining competence that will be valuable for a range of future careers.
