Human-human conversation is highly interactive\, on a very rapid \;time-scale\, with participants providing a range of overlapping and \;low-latency responses and signals\, as speech is occurring\, to \;facilitate a rapid and successful communication process. \; \;My research \;aims to advance our understanding of how such fluid turn-taking and \;interaction can be modeled and replicated computationally. \; \;I am also \;interested in the impact that specific system behaviors and \;interaction styles can have on user perceptions\, user disclosure of \;information to automated systems\, and overall user satisfaction. \;In this talk\, I highlight two recent empirical studies in which \;hundreds of human interlocutors have conversed with spoken dialogue \;systems and virtual human systems developed at the University of \;Southern California. \; \;Both studies compare human-human interaction to

human-agent interaction\, and provide insights into how improved \;turn-taking skills can facilitate more human-like interaction and \;improved performance in state-of-the-art dialogue systems. \;The first study focuses on a virtual human interviewer that conducts \;automated spoken interviews of users who may be suffering from \;psychological distress conditions such as depression and PTSD. \; \;The \;study illustrates how turn-taking and dialogue policy decisions can be \;customized to help build rapport and encourage disclosure of sensitive \;personal information by the system's human users.

The second study explores the extent to which fluid\, low-latency \;turn-taking can be achieved in an automated system designed to \;participate in a negotiation roleplay scenario with students of \;negotiation. \; \;This project relies on a combination of human-human \;negotiation data\, Wizard-of-Oz data\, and statistically trained \;language processing components in an effort to bring a more fluid and \;natural interaction style to an automated negotiation system.