Information and Signaling in Decentralized Decision Problems.

TLDR: In this article, it is shown that a decentralized many-person decision problem is one where each decision maker has different information and if one decision maker's information depends on what another decision maker had done, the information is called dynamic.

Abstract: : A decentralized many-person decision problem is one where each decision maker has different information. If one decision maker's information depends on what another decision maker has done, the information is called dynamic. In the past, problems involving dynamic information have been very difficult, if not impossible, to solve. Two specific examples which have been solved, one from economic theory and the other from classical information theory, will be investigated. It will be shown that they can be formulated as two-person decision problems with the type of dynamic information structure called signaling. The first example involves a model of the job market as a nonzero-sum game. New equilibrium solutions are found and properties of these solutions, such as stability, multiple solutions, and threshold effects of signaling cost and noise, are studied. The second example models the Shannon problem as a team theory problem. The concept of real-time information theory is introduced, where source and channel sequences are of a fixed length, and general results about real-time solutions are proved and demonstrated. (Author)