A Study of a Single Robot Interacting with Groups of Children in a Rotation Game Scenario

We tested the hypothesis that children are more attentive to a robot if the robot appears to be interested in the children. In addition, we investigated if and how the quality and quantity of a child's attentive behaviour varies with the distance to the robot, reflecting the notion of "social spaces". Hereto, 16 groups of up to 10 children each were engaged in a play scenario in which they had to move closer to a robot over 6 successive rounds. The robot was endowed with a "camera eye" and an arm and hand. The camera could either be nonmoving ("static") or actively "searching" ("active searching"), giving the impression it was trying to select a child to focus on. Likewise, the arm and hand could either be fixed in a permanent pointing position ("permanent pointing") or actively rise to point selectively at a particular child when it stopped facing it ("selective pointing"). The results showed that: 1) The mean frequency of overall attentive behaviour by the children (including attention towards other children) was significantly higher when the robot was not selectively pointing at the children and independent of the state of the camera. 2) "Looking at" was the most frequently scored attentive activity for the children and was mainly targeted to the robot, but not correlated with any of their other attentive activities. 3) There was an interaction effect between the state of the camera and of the pointer: looking at the robot by the children occurred significantly more often when the camera and the arm were consistent in signalling apparent interest (i.e. camera "(active searching" and hand "selectively pointing" or camera "static" and hand "permanently pointing"). 4) There was no demonstrable effect of distance to the robot on the overall attentive behaviour of the children.