Humpty Dumpty and the craft of creating conversational agents

Internal Event
29. November 2017
CITEC building, 1.204

Conversational agents are created based on our observations and theories of human behaviour. In this talk, I would like to discuss the connection between two concerns that are troubling me lately. One concern has to do with instructing annotators that label multimodal corpora: how do we make sure that they agree on the meaning of the label. The second has to do with the variables in our computational models of conversational agents: how do we make sure that they are the relevant ones and that they do what we want them to do.

Dirk Heylen is Professor of Socially Intelligent Computing at the University of Twente (Enschede, The Netherlands) and president of the Association for the Advancement of Affective Computing (AAAC). His research interests cover both the machine analysis of human (conversational) behaviour and the generation of human-like (conversational) behaviour by virtual agents and robots. He is especially interested in the nonverbal and paraverbal aspects of dialogue and what these signals reveal about the mental state (cognitive, affective, social). He explores these topics both from a computational perspective and as basic research in the humanities. Other research interests that Dirk and CITEC have in common are: ambient intelligence, brain computer interfaces, touch in interaction, and virtual and augmented reality. See his website for projects, publications, and his more personal endeavours: