Child’s Perception of Robot’s Emotions: Effects of Platform, Context and Experience

Social robots may comfort and support children who have to cope with chronic diseases like diabetes. In social interactions, it is important to be able to express recognizable emotions. Studies show that the iCat robot, with its humanoid facial features, has this capability. In this paper we look if a Nao robot, without humanoid facial features, but with a body and colored eyes is also able to express recognizable emotions. We compare the recognition rates of the emotions between the Nao and the iCat. First a set of bodily expressions of the Nao for five basic emotions (angry, fear, happy, sad, surprise) was created and evaluated. With a signal detection task, the best recognizable bodily expression for each emotion was chosen for the final set. Then, fourteen children between 8 and 9 years old interacted both with the Nao and iCat to recognize the emotions within context, in a story-telling session, and without context. These interactions were repeated one week later to study the learning effect. For both robots, recognition rates for the expressions were relatively high (between 68 and 99 % accuracy). Only for the emotional state of sadness, the recognition was significantly higher for the iCat (95 %) than for the Nao (68 %). The emotions shown within context had higher recognition rates than those without context and during the second interaction the emotion recognition was also significantly higher than during the first session for both robots. To conclude: we succeeded to design a set of well-recognized dynamic emotional expressions for a robot platform, the Nao, without facial features. These expressions were better recognized when placed in a context, and when shown a week later. This set provides useful ingredients of social robot dialogs with children.