One method that I have used is the Blicket Detector, which I helped to develop with Alison Gopnik . The detector is a devise that lights up and plays music when certain objects are placed upon it. You can watch an example of the blicket detector by clicking here . In these experiments, certain objects make the machine go when they are placed on the detector. Thus, the detector allows us to present children with novel, non-obvious causal property of objects, which we use to investigate several questions about cognitive development.
First, I am interested in the relationship between an object's causal properties and its category membership. We have shown that preschoolers will make category inferences based on the causal properties of objects, even when that information conflicts with the objects perceptual features. We have also shown that preschoolers understand that an object's causal properties will predict whether it has a particular internal property, especially when there is a relationship between those causal properties and the objects' labels.
Second, I am interested in questions about causal inference. I have investigated whether children can make deductive inferences about evidence they directly observe? Further, what kinds of inductive inferences do children make when they do not directly observe all the relevant information? In addition, I have examined whether children can use the prior probabilities of events occurring to make causal inferences. These experiments are based on a particular learning algorithm from the causal graphical model formalism, based on Bayesian inference, which was developed by Josh Tenenbaum and colleagues . Currently, I am examining whether the developmental implications of these inferential abilities in infants, toddlers, and preschoolers.