The ghosts in the computer: the role of agency and animacy attributions in "ghost controls"

Three studies evaluated the role of 4-year-old children's agency- and animacy-attributions when learning from a computerized ghost control (GC). In GCs, participants observe events occurring without an apparent agent, as if executed by a "ghost" or unobserved causal forces. Using a touch-screen, children in Experiment 1 responded to three pictures in a specific order under three learning conditions: (i) trial-and-error (Baseline), (ii) imitation and (iii) Ghost Control. Before testing in the GC, children were read one of three scripts that determined agency attributions. Post-test assessments confirmed that all children attributed agency to the computer and learned in all GCs. In Experiment 2, children were not trained on the computer prior to testing, and no scripts were used. Three different GCs, varying in number of agency cues, were used. Children failed to learn in these GCs, yet attributed agency and animacy to the computer. Experiment 3 evaluated whether children could learn from a human model in the absence of training under conditions where the information presented by the model and the computer was either consistent or inconsistent. Children evidenced learning in both of these conditions. Overall, learning in social conditions (Exp. 3) was significantly better than learning in GCs (Exp. 2). These results, together with other published research, suggest that children privilege social over non-social sources of information and are generally more adept at learning novel tasks from a human than from a computer or GC.