Behavioral Social Neuroscience Seminar

In economics, we typically assume that people make choices "as if" they were maximizing a utility function. We account for inconsistent or noisy choices using a logistic (softmax) choice function and related Quantal Response Equilibira. This approach to behavior treats the utility function and the choice function separately and is agnostic about what generates noise in the choices. In psychology and neuroscience, there is a large literature on modeling perceptual choices based on noisy information. Drift-diffusion models of these decisions have proven to be very successful because they provide a precise quantitative explanation for the relationship between choices and reaction times for different levels of task difficulty (i.e. amounts of noise). Here I show that this class of models can be successfully applied to various domains of economic decision making including purchasing behavior, temporal discounting, and social preferences. The basic idea behind this approach is that people don't know which option to choose but must instead consider how good each option is for them in their current state. This process takes time, and people only make a decision once they are confident enough which option is better for them. The model correctly predicts, counterintuitively, that people will take more time to make choices when the two options are very similar in value, even though the benefit of making the "correct" decision is minimal. Better than being able to merely fit the model to the data, I show that the model and parameters are robust across different tasks and datasets, allowing us to accurately predict choice and reaction time profiles in novel settings. Finally, I show that measuring reaction times can dramatically improve our ability to predict individual subjects' choices. From this I argue that economics could greatly benefit by adopting a more "as is" approach to modeling decision making.