Causal Reasoning Under Ambiguity: An Illustration of Modeling Mixture Strategies

Causal reasoning with ambiguous observations requires subjects to estimate and evaluate the ambiguous observations. Detecting how people process ambiguous observations can be complicated by individual differences in causal reasoning. This paper proposes a hierarchical model that accounts for the uncertainty in both the distribution of the functional form selection and the distribution of the ambiguity treatment selection. The model provides an alternative to self-report measures for identifying subjects' strategic choices in reasoning about causal relationships under ambiguity. The posterior distribution of the causal estimates is determined by both the functional form and the ambiguity processing strategy adopted by the reasoner. Our model is tested in a simulation study where it demonstrates its ability to recover the strategies and functional forms adopted by simulated subjects across a range of hypothetical conditions. In addition, the model is applied to the results of an experimental study.