Multinomial probabilistic values were introduced by one of us in reliability. Here we define them for all cooperative games and illustrate their behavior in practice by means of an application to the analysis of a political problem.Keywords: (TU) cooperative game, Shapley value, probabilistic value, binomial semivalue. Math. Subj. Class. (2000): 91A12.
IntroductionWeber's general model for assessing cooperative games [14] is based on probabilistic values.1 Every probabilistic value is defined by a set of weighting coefficients and allocates, to each player in each game of its domain, a convex sum of the marginal contributions of the player in the game. These allocations can be interpreted as a measure of players' bargaining relative strength. The most conspicuous member of this family (in fact, the inspiring one) is the Shapley value [13]. In the present paper we study a subfamily of probabilistic values that we call multinomial (probabilistic) values.2 Technically, their main characteristic is the systematic generation of the weighting coefficients in terms of a few parameters (one parameter per player).For more than a decade, our research group has been studying semivalues, a subfamily of probabilistic values introduced by Dubey et al. [9], characterized by anonymity and including the Shapley value as the only efficient member. In the analysis of certain cooperative problems we have successfully used binomial semivalues, a monoparametric subfamily defined by Puente [12]