Gauging Children’s Understanding of Artificially Intelligent Objects: A Presentation of “Counterfactuals”

Children aged 5 to 11 years and a comparison group of adults were presented with two instances where the behaviour of a computational object was contrary to what might normally be expected of such a device. In both instances findings are discussed with regard to children’s understanding of a computer program and resulting computational behaviour generally. In the first study, children viewed a ”lm featuring a number of robots either acting as traditionally programmed devices or, alternatively, with apparent intentionality. We examine to what extent, if at all, children were aware of this difference. Findings indicated that although the younger children mentioned other alleged differences between the robots, the issue of different loci of control was not a salient one. In the second study, children were encouraged to type two kinds of questions into a computer. The first kind (simple maths questions) required a general solution procedure commonly accessible to a computational object. The second (details of a biographical nature) did not. With respect to the first as well as the second kind of questions the computer was seen to provide apparently correct answers. Findings showed that although with increasing age children were better at articulating the difference between rote- and rule-generated solutions generally, this was not generally accompanied by an accurate assessment of the kinds of problems that could normally be expected to be solved by a computer.