Six-month-old infants' expectations for interactive-humanoid robots

Summary form only given. As technology advances, many human-like robots are being developed. These humanoid robots should be classified as inanimate objects; however, they share many properties with human beings. This raises the question of how infants classify them. Developmental psychology has addressed the issue of how infants come to characterize humans as agents having mental states that is indispensable foundation for sociality. Some studies suggest that infants attribute mental states only to humans. For instance, Legerstee et al. (2000) found that 6-month-old infants do expect people to communicate with people, not with objects. These results indicate that human cognition specializes in human in early infancy. Other studies have suggested, however, that infants attribute mental states to non-human objects that appear to be interactive with a person. For instance, Johnson et al. (1999) indicated that 12-month-old infants did gaze following to a non-human but interactive stuff. These results imply that interactivity between humans and objects is the key factor in mental attribution, however, interesting questions remain to be answered: do infants also have expectation for robots to communicate with person? In this study, we investigated whether 6-month-old infants expected an experimenter to talk to a humanoid robot "Robovie" [Ishiguro, et al., (2001) using infants' looking time as a measurement of violation-of-expectation. Violation-of-expectation method uses infants' property that they look longer at the event that they do not expect than at the event that they expect. During test trials, we show infants the stimulus in which an actor talks to the robot and another person. If infants regard robots as communicative existence like human, they will not be surprised and look at the robot as long as at the person. But if infants do not attribute communicational property to robots, they will look longer at the robot than at the person. To show infants how the robot behaved and interacted with people, we added a familiarization period prior to the test trials, which phase provided infants with prior knowledge about the robots. The stimuli in the familiarization of these conditions are as follows: 1) interactive robot condition: the robot behaved like a human, and the person and the robot interacted with each other; 2) non-active robot condition: the robot was stationary and the person was both active and talked to the robot; 3) active robot condition: the robot behaved like a human, and the person was stationary and silent. If the robots' appearance is dominant for expectation, the results of all condition are same. If robot' action is dominant, the results of the interactive robot condition and the active robot condition are same. And if human-robot interaction is dominant, the result of the interactive robot condition is only different. In the results, infants who had watched the interactive robot looked at the robot as long as at the person. However, infants who had previously observed other robots (non-active robot and active robot) looked longer at the robot than at the person. A previous study suggested that infants come to think of the same robot as intentional agents earlier than they can attribute simple geometric object as intentional agents (Kamewari, et al.). Therefore, it is thought that early infants have the cognitive base that specialized in appearance. Our findings, however, imply that early infants are also sensitive about external forms of human communications, such as "turn-taking" (Trevarthen, 1980), so they come to regard a non-human target as communicative by learning