When people communicate face-to-face their interactions produce complex patterns of movement and meaning. Because of sheer complexity, a complete theory of such interaction will necessarily have to be in the form a runnable computer model. This presentation describes a computational approach to modeling realtime, multimodal turntaking. The model explains the tight-yet-flexible coordination between people engaged in face-to-face communication as the result of interactions between multiple goal-directed, cognitive processes. In this approach the patterns observed in turntaking are emergent, resulting from process interactions inside and between each participant. The approach builds on and extends prior work on YTTM - a computational model of multimodal turntaking addressing a large number of dialogue phenomena, including back-channel feedback, interruptions and speech content delivery.

Dr. Kristinn R. Þórisson is an Associate Professor in the Department of Computer Science at Reykjavík University and co-director of the Center for Analysis and Design of Intelligent Agents. He has been researching artificial intelligence for over 17 years, in academia and industry. At M.I.T. he pioneered technologies in the area of virtual, intelligent agents with integrated multi-modal perception and action control. At LEGO Digital he directed the creation of a large virtual world inhabited by interactive creatures. Kristinn has consulted on technology and business for British Telecom, Kaupthing Investment Bank, and NASA, among others, and taught A.I. courses at Columbia University and KTH. He is a co-founder of MINDMAKERS.ORG, an international consortium of developers in the field of interactive A.I. His present research focuses on construction of large A.I. systems with integrated perception, action and reasoning capabilities, and their application to robots. For more information see his home page (http://www.ru.is/faculty/thorisson) and CADIA (http://ailab.ru.is).