Design and Evaluation of Affective Expressions of a Zoomorphic Robot.

Social robots that are capable of showing affective expressions can improve human-robot interaction and users' experiences in many ways. This capability is important in many application areas. In this paper, we describe the design and evaluation of 11 affective expressions for Miro, an animal-like robot. Affective expressions were inspired by the animal and human behavior literature. The designs were evaluated through a video study on Mechanical Turk with 88 participants. Five of the expressions-happy, sad, excited, surprised, and tired-were correctly recognized by more than half of the participants. While fewer participants were able to recognize the other, more complex affective expressions, we observed a significant correlation between the recognition of the robot's displayed affective states and participants' understanding of human emotions. This suggested that the reduced accuracy in the recognition of the other affective expressions can be due to the general challenges involved in recognizing complex emotions.