Modification of the Banzhaf value for games with a coalition structure (Q1614139)

What is a reasonable solution to a conflict situation? A natural way to answer this question is to formulate natural conditions on the desired solution, such as Pareto optimality, symmetry, etc., and to look for solutions that satisfy these properties. It is known that if we list all seemingly reasonable conditions, then it is not possible to satisfy all of them; however, for many classes of games, we can select subsets of conditions that uniquely determine a solution. For example, for \(n\)-person cooperative games with transferable utility, depending on which conditions we select, we get either the Shapley value or the Banzhaf value. In the general \(n\)-person cooperative game, in principle, all coalitions are possible, and all coalitions have to be considered to describe, e.g., the Shapley value. In many real-life conflict situations, however, there is a prior partition of the set of players, and coalitions can only form between these partitions; for example, in a parliament, political parties can be viewed as such prior coalitions. In such situations, it is reasonable to consider a new symmetry property: symmetry under coalition structure; in other words, if two parties provide a similar effect on whether the resulting coalition is winning, then these parties should get the same overall value. There are several possible approaches to this situation. In particular, if we use the standard Banzhaf approach, then we do not get the new symmetry property. Another approach is to consider this situation as a quotient game, in which we only consider prior coalitions as players; then, after assigning values to the prior coalitions, we divide these values between the participants of each prior coalition. The authors show that the resulting solution is the only one that satisfies several reasonable conditions including the new party-based symmetry condition.