How do humans want causes to combine their effects? The role of analytically-defined causal invariance for...

The present paper reports two experiments (N = 232, 254) addressing the question: How do reasoners reconcile the desire to have useable (i.e., invariant) causal knowledge - knowledge that holds true when applied in new circumstances/contexts - with the reality that causes often interact with other causes present in the context? The experiments test two views of how reasoners learn and generalize potentially complex causal knowledge. Previous work has focused on reasoners' ability to learn rules (functions) describing how pre-defined candidate causes combine, potentially interactively, to produce an outcome in a domain. This empirical-function-learning view predicts that participants would generalize an acquired combination rule based on similarity to stimuli they experienced in the domain. An alternative causal-invariance view goes beyond empirical learning: it allows for the possibility that one's current representation may not yield useable causal knowledge. This view posits that the human causal-induction process incorporates invariant knowledge as an aspiration, entailing that observed deviation from causal invariance when the knowledge is applied serves as a signal for a need to revise causal knowledge: Only invariance across contexts with potentially new causal factors justifies generalization across them. The invariance view therefore predicts that reasoners would revise their representation so that they have whole causes - potentially consisting of interacting components - that do not interact with each other, even when in their relevant experience all (pre-defined) causes interact. Across both experiments, our results favor the causal-invariance view: Participants generalize their empirically learned function (which may involve interactions) to new stimuli, but switch to the analytic causal-invariance function for both old and new stimuli at the level of the whole cause, indicating that how humans want causes to combine their effects shapes the knowledge they induce.