Choice under complete uncertainty when outcome spaces are state dependent (Q1025643)

The paper deals with the decision-making under uncertainty. It is motivated by the endeavour to limit the pesimism connected with the classical maximin procedure, realized due to Milnor's axioms. The approach to the problem, suggested in the paper, is based on the presumption that the results of a decision depend on the actual ``state of nature'', and that those states can be classified on ``good'' and ``bad'' ones. The authors suggest to choose the maximin decision for the class of good states, only, and the main results presented in the paper characterize the limits and consequences of such behavior.