When an individual thinks about a problem, his decision to reason further may involve a tradeoff between cognitive costs and a notion of value. But it is not obvious that this is always the case, and the value of reasoning is not well-defined. This paper analyzes the primitive properties of the reasoning process that must hold for the decision to stop thinking to be represented by a cost-benefit analysis. We find that the properties that characterize the cost-benefit representation are weak and intuitive, suggesting that such a representation is justified for a large class of problems. We then provide additional properties that give more structure to the value of reasoning function, including ‘value of information’ and ‘maximum gain’ representations. We show how our model applies to a variety of settings, including contexts involving sequential heuristics in choice, response time, reasoning in games and research. Our model can also be used to understand economically relevant patterns of behavior for which the cost-benefit approach does not seem to hold. These include choking under pressure and (over)thinking aversion.
