Perceived Agency from Robot Gestures Reflecting Quantitative Meanings

Abstract

Humans perceive agency in artifacts that behave as if they can think. We believe that to realize artifacts that evoke the perception of agency, it's necessary for them to perform physical actions generated by a symbolic and quantitative conversion mechanism, thereby eliciting a sense of intelligence in humans. Thus, this study aims to investigate whether gestures generated by quantitative meanings extracted from the vector space of word-distributed representations enable the construction of artifacts that evoke the perception of agency. In this study, we conducted experiments using the robot gesture generation method proposed in our previous study. Using the Godspeed Questionnaire, commonly employed for subjective evaluation of robots in Human-Robot Interaction (HRI), we evaluated the intelligence of a robot with co-speech gesture. The results did not demonstrate the proposed method as a cause of agency perception. We aim to extend this method to develop a new artifact capable of eliciting agency perception and achieving natural interaction with humans.