Information Theory and Biased Beliefs

Beliefs can be incorrect or biased in many ways. I propose a unifying microfoundation for many well-documented biases using ideas from information theory. The common theme is that subjective beliefs are the solution to an optimization problem where one goal is accuracy, formalized by minimizing Kullback-Leibler divergence from the objective belief. Correct beliefs are a special case where accuracy is the only goal. Other goals or constraints produce biases such as motivated beliefs, partition dependence, anchoring, overconfidence, confirmation bias, base-rate neglect, and conservatism.