What if? Counterfactual reasoning, pretense, and the role of possible worlds

Counterfactual thinking, where one envisions alternative possible events and their outcomes, is hypothesized to be one of the primary ways in which we reason about causal relationships (e.g., Pearl, 2000; Woodward, 2003). Recent computational and experimental work suggests that both adults and children may reason about causality in a manner consistent with probabilistic graphical models – coherent, complex representations of causal structure that allow distinctive kinds of inferences (e.g., Gopnik et al., 2004; Griffiths & Tenenbaum, 2009). In particular, the causal models approach supports and distinguishes two types of inferences, predictions, on the one hand, and interventions, including counterfactual interventions, on the other. In predictions, we take what we think is true now as a premise and then use the model to calculate what else will be true. In counterfactuals, we take some value of the model that we currently think is not true as a premise, and calculate what would follow if it were. Intuitively, childhood pretense bears a striking resemblance to counterfactual inference, but this relationship has not been widely explored. In general, pretend play seems paradoxical. Why should children spend so much time thinking about unreal worlds? Moreover, why would counterfactual inference itself be useful, since it is also about things that are not real? In this symposium we will explore the ways in which pretense and counterfactual thinking might be related (Buchsbaum, Walker & Gopnik; Rafetseder & Perner), the types of computations that might underly both kinds of thought (Lucas & Kemp; Chater) and the ways in which both might contribute to our causal understanding of the world, even without exposing us to new data (Chater; Danks).