Models for pessimistic or optimistic decisions under different uncertain scenarios

Abstract

In many decision problems under uncertainty, agents are only able to provide a possibly incomplete preference relation on gambles, that can be “irrational” according to the classical expected utility paradigm. Furthermore, agents can find it easier to express their preference relation taking one or more specific scenarios as hypothesis. In order to handle these situations two betting scheme rationality conditions are introduced, which characterize those preference relations representable by a conditional Choquet expected value with respect to a conditional belief or plausibility function. Such conditions encode a pessimistic or an optimistic attitude towards uncertainty, respectively, and allow to take into account (possible) “null” scenarios by means of a weak ordering of “unexpectation”. Moreover, they are the basis of a completion technique for enlarging the set of preferences by preserving rationality.