The two-model problem in rational decision making

Abstract

A model of a decision problem frames that problem in three dimensions: sample space, target probability and information structure. Each specific model imposes a specific rational decision. As a result, different models may impose different, even contradictory, rational decisions, creating choice ‘anomalies’ and ‘paradoxes’. So, decision making in real-life situations is different from decision making in an experiment. An experiment is a designed setting according to an experimenter’s model of the decision problem, while for a real-life situation it is not always obvious what the design is. A subject in an experiment may initially have a different model of the task than the experimenter and thus possibly make apparently irrational decisions from the experimenter’s model perspective. As a consequence a choice anomaly can be eliminated by learning what the experiment’s model is.