A system-theoretic model for cooperation, interaction and allocation (Q643011)

This paper presents a system-theoretic approach to cooperation, interaction and allocation. It simplifies, unifies and extends the results on classical cooperative games and their generalizations. The classical model of cooperative games suppose that arbitrary subsets of players can join to form feasible coalitions and create values in a given economic context. A fundamental model for the so-called probabilistic values, that includes, among others, the Shapley value, was developed for the classical model by \textit{R. J. Weber} [``Probabilistic values for games'', The Shapley value. Essays in honor of Lloyd S. Shapley, 101--119 (1989; Zbl 0707.90100)]. More general models assume restricted cooperation (see for example \textit{R. B. Myerson} [Math. Oper. Res. 2, 225--229 (1977; Zbl 0402.90106)], or \textit{J. M. Bilbao} [Cooperative Games on Combinatorial Structures. Theory and Decision Library. Series C: Game Theory, Mathematical Programming and Operations Research. 26. Dordrecht: Kluwer Academic Publishers (2000; Zbl 0983.91013)]). The current model is dynamic and based on the notion of states of cooperation that change under actions of agents. Careful distinction between ``local'' states of cooperation and general ``system'' states leads to a notion of entropy for arbitrary non-negative and efficient allocations and thus to a new information-theoretic criterion for fairness of allocation mechanisms. A cooperation system is a quadruple \(\Gamma=(N,V,A,\mathcal{A})\), where \(N\) is a finite set of agents, \(V\) a finite set of states of cooperation and \(A\) a finite set of feasible transitions \(x\rightarrow y\) between states. It is supposed that \(A\) is partitioned into pairwise disjoint blocks \(A_i\), indexed by the agents \(i\in N\), where \(A_i\) is the set of transitions governed by the agent \(i\). This partition is denoted by \(\mathcal{A}\) \(=\{A_i\, :\, i\in N\}\). In this way, \(G=(V,A)\) is a (directed) transition graph of \(\Gamma\) with a vertex set \(V\) and an arc set \(A\). It is assumed that there is one unique source \(s\in V\) and that the graph \(G\) is acyclic. If \(T\) is the collection of all sinks \(t\) of \(G\), \(\mathcal{P}\)\(_t\) will be the family of all paths from the source \(s\) to the sink \(t\) and \(\mathcal{P}\) \(=\bigcup_{t\in T}\) \(\mathcal{P}\)\(_t\). A trajectory \(P\in \mathcal{P}\) is called a cooperation instance and think of \(P\) as a pure state of the system \(\Gamma\). It is defined a mixed state as a convex combination \(\rho=\sum_{P\in \mathcal{P}}\pi_PP\), with \(\pi_P\geq 0\) and \(\sum_{P\in \mathcal{P}}\pi_P=1\). The states of cooperation are the vertices of the graph \(G\). Other important concepts introduced are the entropy of a system state \(\rho\), flows in the cooperation system \(\Gamma\), flow into or out of a state \(x\in V\), s-flow, potentials on \(\Gamma\) and marginal flow of a potential, discrete path integral of a flow and simple potentials. They show how the models of classical cooperative games and their generalizations fit into the framework of cooperation systems \(\Gamma\) and allocation mechanisms are defined, given an arbitrary cooperation system \(\Gamma\). Moreover, some axiomatic assumptions for allocation mechanisms are given. They associate with non-negative efficient allocation mechanisms a random walk on the transition graph \(G=(V,A)\). The so-called Shapley allocation mechanism is exhibited as arising from random walks with maximal entropy. The Shapley value (cf. [\textit{L. Shapley}, Contrib. Theory of Games, II, Ann. Math. Stud. No. 28, 307--317 (1953; Zbl 0050.14404)]) or the position value (cf. [\textit{P. Borm, G. Owen} and \textit{S. Tijs}, SIAM J. Discrete Math. 5, No. 3, 305--320 (1992; Zbl 0788.90087)]), for example, are special cases of such general Shapley allocation mechanism. For a large class of cooperative systems, a characterization of game symmetries in terms of the so-called \(\lambda\)-values is given; this generalizes to the present model a result of \textit{F. Carreras} and \textit{G. Owen}, [Int. J. Game Theory 26, No. 1, 1--10 (1997; Zbl 0871.90117)]. The \(\lambda\)-weighted Shapley value of \textit{E. Kalai} and \textit{D. Samet}, Int. J. Game Theory 16, 205--222 (1987; Zbl 0633.90100)] is a special \(\lambda\)-value. Interaction among agents in a cooperation system \(\Gamma\) is assessed and a systematic theory for the so-called interaction indices will be established. They finish the paper by proposing a concept for cores and Weber sets and showing that a Weber set of a game with the so-called selection structure always contains the core. This paper seems to offer a promising path of future research.