Infant discrimination of humanoid robots

Abstract

Recently, extremely humanlike robots called "androids" have been developed, some of which are already being used in the field of entertainment. In the context of psychological studies, androids are expected to be used in the future as fully controllable human stimuli to investigate human nature. In this study, we used an android to examine infant discrimination ability between human beings and non-human agents. Participants (N = 42 infants) were assigned to three groups based on their age, i.e., 6- to 8-month-olds, 9- to 11-month-olds, and 12- to 14-month-olds, and took part in a preferential looking paradigm. Of three types of agents involved in the paradigm-a human, an android modeled on the human, and a mechanical-looking robot made from the android-two at a time were presented side-by-side as they performed a grasping action. Infants' looking behavior was measured using an eye tracking system, and the amount of time spent focusing on each of three areas of interest (face, goal, and body) was analyzed. Results showed that all age groups predominantly looked at the robot and at the face area, and that infants aged over 9 months watched the goal area for longer than the body area. There was no difference in looking times and areas focused on between the human and the android. These findings suggest that 6- to 14-month-olds are unable to discriminate between the human and the android, although they can distinguish the mechanical robot from the human.

Keywords: android; eye tracking; humanoid robot; infant; preferential looking paradigm; uncanny valley.

FIGURE 1

Agents used as experimental stimuli. The android was designed to have the likeness of the human actor, and was identical in internal architecture to the robot. The original face of the robot was covered with a plastic mask to conceal its somewhat bizarre appearance, with naked eyeballs and gums.

FIGURE 2

Heat maps of mean gaze count across all trials of all participants, superimposed upon each agent after 7 × 7 pixel Gaussian smoothing was applied. Red represents an area that the greatest number of infants viewed. areas of interest (AOI) are depicted as white rectangles. The reason for the focused areas in the goal area of the android and the robot spreading vertically is probably due to the trajectories of the agents’ hands.

FIGURE 3

Proportions of total looking times at each AOI of each agent across the three age groups. Red solid lines, purple dotted lines, and blue thin lines represent the human, android, and robot agents, respectively. Circle, triangle, and square markers correspond to AOI of face, goal, and body, respectively. Error bars represent standard errors.