A Criterion of Model Decisiveness

Abstract

When faced with decision-relevant information, decision-makers are often exposed to a multiplicity of different models, or accounts of how information should be interpreted. This paper proposes a theory of model selection — an account of what models decision-makers find compelling, and ultimately adopt — based on the insight that individuals seek decisive models that provide clear guidance regarding the best course of action. The decisiveness criterion is characterized by a demand for extreme models, which generates inferential biases such as overprecision and confirmation bias, but predicts meaningful bounds on the extent of these biases. The dependence of the decisiveness criterion on the decision-maker’s objectives can produce documented patterns of preference reversals, rationalize seemingly contradictory patterns of inferential attribution errors, and generate novel predictions as to how belief polarization can arise along heterogeneity in decision-makers’ objectives. I discuss applications of the theory to financial decision-making, the provision of expert advice, and social learning through the exchange of models.

Keywords: Model selection, learning, polarization, confirmation bias

JEL Classification: D8,D9