Perceptual evaluation of non-verbal cues from a humanoid robot for multi-party teleconferences

Humanoid robots that act as proxies for real people are useful for turn-taking in multi-party teleconferences due to their telepresence effect. However, it is still rather challenging to explore if the robot motions are effective or not as non-verbal cues. For this purpose, this paper presents a perceptual evaluation of robot motions in the context of a multi-party teleconference. Our main focus is on analyzing the relationship between the robot motions and the motivation to speak by testing the following two hypotheses. (H1) There are some kinds of robot motions that are more effective for expressing the motivation to speak than others. (H2) The intensity of robot motions is correlated with the perceived strength of the motivation to speak. In a preliminary experiment with five subjects, we confirmed that our hypotheses were reasonable and we formulated the robot motions that were used in the experiments. In each of 10 evaluation experiments, three participants attended a teleconference for about 30 minutes and one of them (called a subject) evaluated the robot motions in real time. The evaluation results from a total of 10 subjects in 10 experiments demonstrate that both hypotheses can be partly upheld.