Humanlike robots are designed to communicate with people using human verbal and nonverbal language. Social gaze cues play an important role in this communication. Although research in human-robot interaction has shown that people understand these gaze cues and interpret them as valid signals for human communication, whether they can serve as effective communicative mechanisms and lead to significant social and cognitive outcomes in human-robot interaction remains unknown. Furthermore, the theoretical and empirical foundations for how these mechanisms might be designed to work with the human communicative system have not been systematically explored.

The research questions that I seek to address in this area are as follows: What are the design variables for social gaze behavior? How do we design gaze behavior for humanlike robots? Can designed behaviors lead to positive, significant social and cognitive outcomes in human-robot interaction such as better learning, stronger affiliation, and active participation in conversations?

This dissertation seeks to find answers to these questions by exploring the design space to identify design variables for social gaze, adapting an approach based on modeling human behavior to designing robot behaviors, and evaluating the social and cognitive outcome of designed behaviors in three studies that focus on different functions of social gaze behavior using three robotic platforms, ASIMO, Robovie, and Geminoid. The first study focused on designing gaze behavior for communication of attention and found strong learning effects induced by a simple manipulation of how much ASIMO looks at an individual. The second study looked at how Robovie might use gaze cues to shape the participant roles of its conversational partners and found strong effects of gaze cues in behavioral and subjective measures of participation, attentiveness, liking, and feelings of groupness. The final study, which consisted of three experiments, explored how Robovie and Geminoid might use gaze cues to communicate mental states, and found task performance effects in a guessing game led by attributions of mental states.

This research contributes to the design of robotic systems with a theoretically and empirically grounded methodology for designing communicative mechanisms, human-robot interaction research with a better understanding of the social and cognitive outcomes of interacting with robots, and human communication research with new knowledge on and computational models of human gaze mechanisms.

249 pages

Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu