Lisetti, ChristineAmini, RezaYasavur, Ugan2018-01-082018-01-0820152015https://dl.gi.de/handle/20.500.12116/11455We discuss a large research project aimed at building socially expressive virtual health agents or assistants (VHA) that can deliver brief motivational interventions (BMI) for behavior change, in a communication style that individuals and patients not only accept, but also find emotionally supportive and socially appropriate. Because of their well-defined sequential structure, BMIs lend themselves well to automation, and are adaptable to address a variety of target behaviors, from obesity, to alcohol and drug use, to lack treatment adherence, among others. We discuss the advantages that VHAs provide for the delivery of health interventions. We describe components of our intelligent agent architecture that enables our virtual health agents to dialogue with users in realtime while delivering the appropriate intervention based on the patient’s specific needs at the time. We conclude by identifying open research challenges in developing virtual health agents.Behavior changeBrief interventionsComputational models of empathy and rapportEmbodied conversational agents (ECA)Hidden Markov models (HMM)Intelligent virtual health agentsMarkov decision process (MDP)Motivational interviewingSpoken dialog systems for HealthText/Journal Article1610-1987