Emotions play a key role in human-human communication. Because of our emotional understanding, we communicate empathic cues to others that act as foundations for building relationships and gaining trust. This is especially important in the healthcare domain, where empathy has been linked to improved doctor-patient relationships and stronger therapeutic bonds. However, as we move towards developing automated healthcare solutions, this rich channel of information can be lost.;To address this lack of emotional understanding by computational interfaces, researchers have developed techniques to detect and respond to the user's emotional state in real time. While research in emotion detection has achieved positive outcomes, the question of how automated systems could effectively use this information for empathic communication remains largely unanswered. Simple techniques such as mirroring back a user's emotions have been shown to be unsuccessful, and almost none of this work being done in the healthcare domain.;In this dissertation, I present a novel approach that allows computers to automatically adapt and respond to a user's emotional state. Informed by an analysis of empathic communication in the context of human-human counseling, I develop a theoretical framework that integrates empathy into automated healthcare systems. I demonstrate how this framework can be applied to develop an automated depression counseling system, and evaluate its efficacy in a series of evaluation studies. In a longitudinal evaluation study with 36 participants, I compare an affect-aware version of the system against a conventional counseling system. Results show that participants received improved therapeutic care and had higher levels of system engagement using the affect-aware version.
展开▼
