Effects of Causal Determinism on Causal Learning Trajectories
Abstract
Research on causal learning suggests that people are capable of learning nondeterministic causal relations, but might expectcausal relations to be deterministic in certain contexts. In two experiments, we demonstrated that peoples expectations ofcausal determinism are context-sensitive and can influence causal judgments in a sequential learning task. When the datawere deterministic (100% success) and participants expected the cause to be deterministic, their causal judgments wereat ceiling. When participants expectations were nondeterministic, causal ratings increased with accumulating positiveevidence. When the data were probabilistic (75% success), participants exhibited a high violation-of-expectation effectupon seeing the first failed event when they expected the causal relation to be deterministic, and much less so whentheir expectation was nondeterministic. We built a simple Bayesian model to explain participants violation-of-expectationeffect as a selection between two distinct hypotheses: that the causal relation in question is deterministic, and that it isnondeterministic.