Indirect information measure and dynamic learning
Weijie Zhong · RePEc: Research Papers in Economics · 2018
In this paper, I first showed that an indirect information measure is supported by expected learning cost minimization if and only if it satisfies: 1. monotonicity in Blackwell order, 2. sub-additivity in compound experiment and 3. linearity in mixing with no information. Then I studied a dynamic information acquisition problem with flexible design of information dynamics, costly waiting and costly information. When flow information measure satisfies the three conditions, dynamic problem can be solved in two steps: solving a static rational inattention problem, and implementing optimal learning dynamics. The optimal solution involves stationary Poisson direct signals: arrival of signal directly suggests optimal action, and non-arrival of signal provides no information.