Optimal Myopic Information Acquisition (with Annie Liang and Vasilis Syrgkanis) 

Abstract. A decision-maker repeatedly samples from many flexibly correlated information sources and takes actions. His payoff depends on the actions taken as well as an unknown state. Myopic information acquisition corresponds to (history-independent) acquisition of signals that maximize next-period precision of beliefs. We identify a canonical class of informational environments (normal-linear), in which the optimal dynamic information acquisition rule is myopic from period 1 when agents sample suf- ficiently many signals each period. For all sample sizes, the optimal rule is generically eventually myopic. These results demonstrate the possibility of robust and “simple” optimal information acquisition, and simplify the analysis of dynamic information ac- quisition in an environment that is commonly used in economics.