Dialogue Systems for Health Communication:
Papers from the AAAI Fall Symposium
Timothy Bickmore, Chair
Since Eliza was developed in 1966, computer scientists and health researchers have attempted to build conversational systems that emulate interactions between health providers and patients. Although Eliza was developed only as a proof-ofconcept, more recent systems have been built with the intent to provide low-cost and widely accessible health care in limited treatment domains, and many of these systems have been proven effective in large-scale clinical trials. In addition to the unique challenges of developing health dialogue systems that are safe, scale to thousands of users, and can accommodate the complexity of multiple diseases or health behaviors, dozens of studies in the field of health communication indicate that psychosocial aspects of the provider-patient interaction—such as empathy, trust and liking —are crucial for maximizing outcomes and patient satisfaction, indicating that these should be addressed in automated systems as well. The goal of this symposium was to bring together researchers in AI—including computational linguistics, planning, user modeling and social agents—with researchers in health communication, public health and the medical sciences. The overall focus was the design, implementation and evaluation of effective health dialogue systems.