Choosing Between Causal Interpretations: An Experimental Study

Co-author: Heidi C. Thysen

Abstract
Good decisions require understanding the causal impact of actions on outcomes. Yet, data is often correlational and could stem from multiple causal mechanisms with divergent implications for choice. Our experiments comprehensively characterize choice under conflicting causal interpretations. Empirical behavior is heterogeneous and primarily reflects three types: following misspecified interpretations that make attractive promises, choosing cautiously, and assessing the fit between theory and data. We characterize the properties of ill-fitting interpretations that subjects struggle to recognize. Preferences for more complex models are twice as common as preferences consistent with Occam’s razor. Implications extend to the Causal Narratives and Model Persuasion literatures.