The Agony of Choice: Optimal Policies for Value-based Decision Making

When faced with two equally high valued choices (e.g., a weekend in London or in Paris, all expenses covered), most people agonize and take an unusually long time to decide. This is quite counterintuitive: given that subjects are guaranteed a high reward regardless of their choice, they should decide quickly since there is little to lose. In addition, people take even longer if a third low reward choice (e.g. a weekend in Pyongyang) is added to the list of options. Why would a choice that is very unlikely to be picked, interfere with the other high valued choices? These puzzling behaviors have led to the notion that value-based decision making is highly suboptimal in humans. We revisited this issue and derived the optimal policy for value-based decision making. Surprisingly, the optimal strategy predicts long reaction times when confronted with two equally high valued choices as well as interferences from irrelevant choices. Moreover, this policy can be easily implemented in neural circuits using a form of nonlinearity known as normalization, which is known to exist in cortical areas implicated in decision making. While this work lays the foundations of a neural theory of optimal value-based decision making, I will argue that these models are nonetheless too limited to deal with complex decision making. I will end with speculations about what it will take to develop models of complex decision making.