WILL YOU NEVER LEARN? SELF DECEPTION AND BIASES IN INFORMATION PROCESSING

Preliminary Version This paper considers a repeated model of selective awareness and studies its implications for information processing. An individual receives a sequence of sig-nals. Each signal is informative about the state of the world.As in the static models of Benabou and Tirole (2002, 2004) and Benabou (2008a, 2008b), the indi-vidual can choose how to interpret each signal. The individuals behavior displays patterns consistent with observed biases in information processing. She displays a tendency to interpret information in ways that support original beliefs and attaches a disproportionately large weight to initial observations (Con\u85rmation Bias). She also updates beliefs in the right direction, but in insu ¢ cient amount compared to the update derived by Bayesrule (Conservatism Bias). Additionally, the individual misclassi\u85es information after a certain number of observations. As a consequence, learning is always incomplete. Moreover, the equilibrium of the model is discontinuous: arbitrarily small amounts of awareness selectivity may lead to large deviations from Bayesian decision making. When men wish to construct or support a theory, how they torture facts into their service