Int J of Soc Robotics DOI 10.1007/s12369-015-0282-2 Why Should We Imitate Robots? Effect of Back Imitation on Judgment of Imitative Skill Yasser Mohammad · Toyoaki Nishida Accepted: 13 January 2015 © Springer Science+Business Media Dordrecht 2015 Abstract How we perceive robots affects how we inter- act with them and vice versa. This leads us to hypothesize that imitating a robot (back imitation) would affect human’s perception of this robot. More specifically, we suggest that it would lead to the attribution to a higher imitative skill to the robot when it subsequently imitates the human. Given that one of the major challenges in learning from demonstra- tion (imitation) in robotics is the limited number of training examples that the demonstrator is usually willing to provide, it would be beneficial to design the interaction context in such a way to increase human’s subjective evaluation of the robot’s imitative skills and back-imitation may be a way to achieve that. Three studies were conducted—involving 78 subjects and 150 HRI sessions—to evaluate the effect of back imita- tion on human’s perception of the robot along several dimen- sions including imitation skill, motion human likeness, inter- action quality, humanness and likability. These studies show that people who imitated the robot for few minutes assigned it later higher imitative skill and motion human-likeness. More- over, back imitation was shown to lead to higher intention of future interaction. The paper reports the results of these stud- ies and discusses their implications for the design of imitation interactions. Keywords Learning through imitation · Back imitation · Perception of robots · Social factors in imitation Y. Mohammad (B ) · T. Nishida Assiut University, Assiut, Egypt e-mail: yasserfarouk@gmail.com; yasserm@aun.edu.eg T. Nishida e-mail: nishida@i.kyoto-u.ac.jp 1 Introduction and Research Question Imitation is an important factor in creating bonds between people as was shown in several social an psychological studies [7]. A famous example is the work of Chartrand and Bargh in which they experimentally showed that behavioral mimicry (“the chameleon effect”) has a significant effect on the interaction and increases empathy towards the interaction partner [4]. This work was followed up in HRI in several stud- ies. For example Riek et al. [21] used real-time head gesture mimicry to improve rapport between the human an a robot and Lee et al. [17] programmed a robotic penguin to nod in return to detected nods in order to consolidate the back- channel communication in a natural way. In a study with 3D agents rather than robots, Bailenson and Yee have shown that agents that imitated head movements of their interac- tion partners after a 4-s delay or utilized their speech style in presenting an argument were rated as more persuasive and received more positive traits ratings in general compared with non-imitating agents [3]. All of these studies have focused on the effect of robot’s imitative ability in human’s perception of this robot’s traits and convincingly argued for a positive effect. In this paper, we investigate the opposite direction of this connection: would human’s imitation of the robot affect how this robot is per- ceived? More specifically, our goal is to investigate the effect of this form of back -imitation on the attribution of motion human-likeness and imitative skill in general. As will be dis- cussed later, this may lead to higher acceptance of the robot according to either leading technology acceptance models. For the purposes of this study we define back imitation as the imitation of the learner by the teacher during, before or after the demonstration of a new task. One explanation of the chameleon effect in the human case can be found in the simulation theory of mind which argues 123