Magazine article AI Magazine

Embodied Communication in Humans and Machines

Magazine article AI Magazine

Embodied Communication in Humans and Machines

Article excerpt

Embodiment has become one of the most promising paradigms in cognitive science and a challenge to AI research. In an invitational conference hosted by the Center for Interdisciplinary Research (ZiF) at Bielefeld University in Germany, 24 highly acclaimed speakers from various disciplines presented their perspectives pertaining to conceptual issues of embodiment; the phylo--and ontogenesis of communication; bodily gestures; understanding and communicating intentions, emotions, and symbols; and the role of bodily action in language and speech. The talks were centered around the two aims of the research group, namely, (1) to obtain a more profound understanding of human communication and its evolution, and (2) to study machine communication both as a means of modeling human communicative abilities and of advancing the human-technology interface.

Gestures were a central topic of the conference. Primatologist Josep Call (Max Planck Institute of Evolutionary Anthropology, Leipzig, Ger many) reported on the rich and flexible gesture repertoire of non-human primates. Psychologist Bennett Bertenthal (University of Chicago, USA), and psycholinguists Sotaro Kita (University of Bristol, UK) and Susan Duncan (University of Chicago, USA) pointed out how gestures form an intimate connection with human speech. Moreover, gestures are responsive to the gestures of others, coupling speakers and listeners in a close "social loop" in which a real time processing of bodily signals seems at least as important as semantic analysis of words and sentences.

The question how the structure of bodily communication could be captured in rule systems was raised by communication psychologist Isabella Poggi (Universita Roma Tre, Rome, Italy) who presented ongoing research on mode-specific lexicons, such as "gestionaries," "gazeionaries" and "touchionaries," as an equivalent of dictionaries in spoken language. How these ideas could be integrated in human-machine interaction was addressed by computer scientist Catherine Pelachaud (Universite de Paris 8, Montreuil, France). She showed how Affective Presentation Markup Language (APML) markers are converted into facial signals to provide embodied conversational agents with multimodal synchronized behavior, including gestures, adaptation to context, and facial display.

As a glance at the future of human-computer interaction, two virtual creatures were around at the conference. Computer scientist Kristinn Thorisson (Reykjavik University, Iceland) demonstrated Gandalf, a virtual agent guide to our galaxy. Gandalf is able to rotate and zoom projected images of planets at the user's command (uttered verbally or by gesture) and answer questions about the planets. In Bielefeld University's virtual reality laboratory, the conference participants met the multi modal assembly expert (MAX). Developed by the Bielefeld AI group, Max can imitate human gestures and exhibit humanlike synthetic speech and coverbal gesture while constructing an airplane from a construction kit in cooperation with a human partner.

Imitation was described as a form of embodied communication by computer scientist Aude Billard (Ecole Polytechnique Lausanne, Switzerland). In her view, imitation is of great relevance in robot learning, namely, to find a way in which humans and robots could learn a common means of communication. Computer scientist Luc Steels (Free University, Brussels, Belgium and Sony Computer Science Laboratory, Paris) showed that comparably simple systems, linked to a surrounding world environment by a camera, are able not only to achieve agreement on the meaning of concepts but also to develop simple forms of syntax, without explicit rules of communication. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.