Suchergebnisse

This book is concerned with situations in which several persons reach decisions independently and the final consequence depends, potentially, upon each of the decisions taken. Such situations may be described formally by an extensive form game: a mathematical object which specifies the order in which decisions are to be taken, the information available to the decision makers at each point in time, and the consequence that results for each possible combination of decisions. A necessary requirement for rational behavior in such games is that each decision maker should reach a decision that is optimal, given his preferences over his own decisions. This requirement is far from sufficient, however, since every decision maker should in addition base his preferences upon the conjecture that his opponents will act optimally as well. It is this principle that distinguishes noncooperative game theory from one-person decision theory. The main purpose of Rationality in Extensive Form Games is to discuss different formalizations of this principle in extensive form games, such as backward induction, Nash equilibrium, forward induction and rationalizability, under the assumption that the decision makers' preferences are given by subjective expected utility functions. The various formalizations, or rationality criteria, are illustrated by examples, and the relationships among the different criteria are explored

AbstractWe investigate the problem in which an agent has to find an object that moves between two locations according to a discrete Markov process (Pollock, Operat Res 18 (1970) 883–903). At every period, the agent has three options: searching left, searching right, and waiting. We assume that waiting is costless whereas searching is costly. Moreover, when the agent searches the location that contains the object, he finds it with probability 1 (i.e. there is no overlooking). Waiting can be useful because it could induce a more favorable probability distribution over the two locations next period. We find an essentially unique (nearly) optimal strategy, and prove that it is characterized by two thresholds (as conjectured by Weber, J Appl Probab 23 (1986) 708–717). We show, moreover, that it can never be optimal to search the location with the lower probability of containing the object. The latter result is far from obvious and is in clear contrast with the example in Ross (1983) for the model without waiting. © 2009 Wiley Periodicals, Inc. Naval Research Logistics 2009