MOG 2007: Workshop on multimodal output generation: CTIT Proceedings of the Workshop on Multimodal Output Generation
I.F. van der Sluis, Mariët Theune, Ehud Reiter, Emiel Krahmer · Data Archiving and Networked Services (DANS) · 2007
Embodied conversational agents (ECAs)-both virtual and real-can exploit multiple output modalities, including speech, facial expressions, head motions and hand gestures.Understanding the nuances of human expression is a research goal in itself.But not all of those nuances may matter to people interacting with an ECA.So, what is the balance between the benefits and costs of increasing the realism of ECAs' multimodal output?This paper draws on experience emerging from a series of studies to argue as follows.First, people may prefer greater realism, but it remains an empirical question as to whether realism improves communication.Secondly, even if it does, there is no guarantee that more expressive communication means better task performance.Thirdly, this may be because ECAs in complex visual tasks attract only minimal, intermittent attention.Thus, in current and planned work on human-robot interaction, we are studying the dual impact of multimodal communicative output on what people do, and on what they look at.