If you can't see the above videos, please update your Flash Player here, and reload this page.
My research includes ethnographic fieldwork and exploratory laboratory studies to understand social aspects of robotic technology and create design variables. I use this understanding to create models of robot behavior and implement these models on different robotic platforms such as Honda’s ASIMO, ATR's Robovie, and Hiroshi Ishiguro's Geminoid. I then evaluate the effectiveness and appropriateness of these models in laboratory experiments by testing theories of human communication in human-robot interaction situations.
In my dissertation research, I investigate how social gaze behavior should be designed for humanoid robots through a series of studies. Each study focuses on a function of gaze behavior and follows a theoretically and empirically grounded design methodology.
Study I: Communication of Attention - Using data collected from a human storyteller and findings from human communication theory, I created a model of humanlike gaze behavior and implemented this model on ASIMO to perform humanike gaze behavior in a storytelling task in which it told stories to two people. I manipulated the amount of attention that the robot gave to the two people and found that people whom the robot gave more attention had better recall of ASIMO's story. Strong gender effects were observed. [Paper]
Study II: Establishing Participant Roles - Using data collected from different conversational structures and findings from theory on conversational participation, I designed gaze cues for Robovie to signal three kinds of participant roles: addressee, bystander, and overhearer. In a controlled experiment with human-robot triads, I found that people conformed to the participant role that the robot communicated to them 99% of the time. Participation in a conversation with the robot affected participants' attentiveness to the task and feelings of groupness. Being acknowledged by the robot led to mode liking of the robot. [Paper]
Study III: Communication of Intentionality - Using findings from literature on developmental functions of gaze and on cues for interpreting deception, I designed a gaze "leakage cue" for Robovie and Geminoid. In a controlled study, participants played a game of guessing with the two robots. I found that when the robot "leaked" information on "what it had in mind" by means of gaze, participants performed better in guessing what the robot had in mind. The effect was stronger with Geminoid. However, people were less likely to report identifying the gaze cue with Geminoid than Robovie. [Paper]
I conducted a long-term ethnographic study of how people interact with and perceive robots in a hospital organization. Through qualitative observations and semi-structured interviews, I collected data in two research sites over the period of a year. My analysis showed that people who worked in high-stress, emotinoally-involved jobs had negative experiences with the robot while workers who were in lower-stress positions thought that the robot improved their workflow and accepted the adoption of the robot to their workflow. [Paper]
Most research in robotics is built on the assumption that robots are co-operative agents. What happens if we put people in a competitive situation with a robot? In a controlled laboratory experiment, I investigated how competition and co-operation affected people's perceptions of Honda's ASIMO. I compared people’s perceptions of ASIMO in a scenario where the robot co-operated with a human partner and in one where it competed with the human partner. I found that people perceived the robot as more socially desirable and intelligent when they co-operated with the robot. Strong gender effects appeared in our analysis. [Short Paper | Full Paper]
I also have a continuing interest in ubiquitous computing, particularly in the development of intelligent technology embedded in our everyday environments such as wall, chair, etc.
The Abstract interfaces project explored the use of simple abstract attributes and motion in the design of social interfaces. This approach builds on previous work on perceptual causality, which suggests that simple displays in motion can evoke high-level social and emotional content. [Paper]
The SenseChair project explores how robotic technology, designed in forms as familiar as home appliances might be used to assist elders and those who provide care. SenseChair serves us as a research platform to understand how we can help elders stay independently, and offer them physical and emotional support. [Paper]
On the development end of this project, I am collaborating with Carlos Guestrin (MLD) and Andreas Krause (CSD) to achieve seating activity recognition with near-optimally placed minimum number of pressure sensors. [Paper]