The Impact of a Social Robot Public Speaker on Audience Attention

Social robots acting as stand-ins for speakers or teachers would enable them to reach large audiences from anywhere in the world, increasing the options for distant learning. They would need to be endowed with effective public speaking skills though, in order to deliver their message, entertain, and maintain audience attention.In this paper, we report two user studies to understand the impact of a social robot public speaker on its audience and compare it to a skilled human speaker. The first study uses an in-house audience attention monitoring system based on computer vision and machine learning; the second study uses eye tracking technology. For both studies, we programmed the social robot Pepper to deliver a short speech in its robotic voice while mimicking the behaviour of a skilled human speaker (gestures and body movements).We found that, when the audience has a genuine interest in the speech, the robot manages to maintain audience attention level as well as the human speaker but fails to arouse interest as much. When the audience is not particularly interested in the speech, the human speaker is better at maintaining audience attention: the novelty of the robot does not compensate for the lack of interest in the speech content, and the robot's behaviour is found to be distracting. Finally, understanding of the speech content is significantly lower for the robot audience. It could be linked to lower audience attention levels, to the robot's lack of facial expressions and failure to convey enthusiasm, or to a feeling that the robot is not legitimate to speak about the topics touched on in the speeches.