Dissertation Research - Robots Without Faces: Non-Verbal Social Human-Robot Interaction

My dissertation asks: Can the use of non-facial and non-verbal affective expression provide a mechanism for naturalistic social interaction between a functional, appearance-constrained robot and the human with which it is interacting? My work expands the fundamental theory of affect for robots and will permit companies to develop non-anthropomorphic robots that are more affective and less cognitively dissonant simply by modifying the software. To date, I have developed design guidelines for roboticists to utilize to determine the most appropriate method of non-facial and non-verbal affective expression (body movement, posture, orientation, illuminated color, and sound) to use based on the distance between the person and the robot with which he or she is interacting. These guidelines apply to both non-anthropomorphic appearance-constrained and anthropomorphic robots.

In the Fall of 2008, I completed the first known comprehensive human-robot interaction study on non-facial and non-verbal affective expression in appearance-constrained and anthropomorphic robots using 128 human participants (79 females and 49 males) interacting with two actual rescue robots (an iRobot Packbot Scout and an Inuktun Extreme-VGTV - See Figure 1) in conditions duplicating a building collapse (See Figure 2). This was the first time an HRI study used significant sample sizes with psychophysiological measures, as HRI studies reported in the literature to date have had less than 35 participants undermining statistical significance. The study hypothesis was: Victims will be calmer and view the experience more positively when interacting with a robot that expresses affective responses through body movement, postures, orientation, and light illuminated color.

In the study, participants were placed in a confined space box in the dark in a simulated disaster site and exposed to both robots (See Figure 2). The robots were programmed to act in either a standard, impersonal mode or in an emotive mode and modified to carry IR devices for operating and recording in the dark. Analyses used psychological self-assessments, psychophysiological measures (EKG, heart rate, inter-beat interval, respiration - thoracic and abdominal, blood pulse volume, and skin conductance), participant interviews, and video recording from four different perspectives (face view, robot view, overhead view, and participant view).

Correlation studies will be conducted to determine any relationships between participants’ self-assessments and their physiological responses to the robots. A doubly multivariate analysis of variance will also be conducted to determine the overall significance of participants’ self-assessments evaluating participants’ valence and arousal responses. Preliminary results will be printed in upcoming publications.