Robotic Faces: Exploring Dynamical Patterns of Social Interaction between Humans and Robots

Abstract

The purpose of this dissertation is two-fold: 1) to develop an empirically-based design for an interactive robotic face, and 2) to understand how dynamical aspects of social interaction may be leveraged to design better interactive technologies and/or further our understanding of social cognition.

Understanding the role that dynamics plays in social cognition is a challenging problem. This is particularly true in studying cognition via human-robot interaction, which entails both the natural social cognition of the human and the “artificial intelligence” of the robot. Clearly, humans who are interacting with other humans (or even other mammals such as dogs) are cognizant of the social nature of the interaction – their behavior in those cases differs from that when interacting with inanimate objects such as tools. Humans (and many other animals) have some awareness of “social”, some sense of other agents. However, it is not clear how or why.

Social interaction patterns vary across culture, context, and individual characteristics of the human interactor. These factors are subsumed into the larger interaction system, influencing the unfolding of the system over time (i.e. the dynamics). The overarching question is whether we can figure out how to utilize factors that influence the dynamics of the social interaction in order to imbue our interactive technologies (robots, clinical AI, decision support systems, etc.) with some "awareness of social", and potentially create more natural interaction paradigms for those technologies.

In this work, we explore the above questions across a range of studies, including lab-based experiments, field observations, and placing autonomous, interactive robotic faces in public spaces. We also discuss future work, how this research relates to making sense of what a robot "sees", creating data-driven models of robot social behavior, and development of robotic face personalities.