A New Value for Cooperative Games on Intersection-Closed Systems Journal: Mathematical Social Sciences
Martin Černý · 2025
Read A New Value for Cooperative Games on Intersection-Closed Systems Journal: Mathematical Social Sciences by Martin Černý (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-05169.txt.
Introduction
Martin Černý, writing alone, proposes a new rule for splitting payoffs in cooperative games where some coalition values are missing, in this 2025 arXiv paper. The uniform-dividend value spreads each family's surplus evenly among coalitions that look alike. His focus is on set systems closed under intersection, a structure common in such games.
On those systems the rule gives one definite answer, and it can be read as the average Shapley value across all completions of the game that keep surpluses nonnegative. Černý compares it with two older rules built for the same setting and gives fairness-style axioms supporting his. Tests suggest his rule and one rival usually land close together. Even off the closed systems, the value often stays unique.
Catalog record
| CREATOR | Martin Černý |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 51 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-05169
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A New Value for Cooperative Games on Intersection-Closed Systems Journal: Mathematical Social Sciences Martin Černý Abstract We introduce a new allocation rule, the uniform-dividend value (UD-value), for cooperative games whose characteristic function is incomplete. The UD-value assigns payoffs by distributing the total surplus of each family of indistinguishable coalitions uniformly among them. Our primary focus is on set systems that are intersection-closed , for which we show the UD-value is uniquely determined and can be interpreted as the expected Shapley value over all positive (i.e., nonnegative-surplus) extensions of the incomplete game. We compare the UD-value to two existing allocation rules for intersection-closed games: the R-value, defined as the Shapley value of a game that sets surplus of absent coalition values to zero, and the IC-value, tailored specifically for intersection-closed systems. We provide axiomatic characterizations of the UD-value motivated by characterizations of the IC-value and discuss further properties such as fairness and balanced contributions. Further, our experiments suggest that the UD-value and the R-value typically lie closer to each other than either does to the IC-value. Beyond intersection-closed systems, we find that while the UD-value is not always unique, a surprisingly large fraction of non-intersection-closed set systems still yield a unique UD-value, making it a practical choice in broader scenarios of incomplete cooperative games. JEL C71 , D81 1 Introduction In a cooperative game of n players, worth of cooperation of any subset of agents, a coalition S , is represented by a real value v(S) . The Shapley value \phi is a rule, which allocates the value of coalition of all agents, v(N) , among the agents in such a way that it reflects how the agents contribute to different coalitions. For each coalition, its surplus d_{v}(S) is computed and redistributed equally among its participants, i.e. each agent i receives \phi_{i}(v)=\sum_{S\subseteq N:\ i\in S}\frac{d_{v}(S)}{|S|}. When not all of the coalitions are possible to form, a cooperative game is enhanced with a list of feasible coalitions \mathcal{F} and at the same time, values of coalitions outside \mathcal{F} are not provided as they are, due to their nature, non-existent or meaningless. In literature, these games are referred to as games with restricted cooperation or cooperative games on set systems ( [ 1 , 2 ] ). Computing the Shapley value in this modified setting rises a problem; a surplus of a coalition depends on the values of all of its subcoalitons, which might not lie in \mathcal{F} . A way around this problem was introduced in [ 3 ] , where the surplus of unfeasible coalitions was set to zero. From the relation between surpluses of different coalitions, the rest of the surpluses can be uniquely determined and the Shapley value computed. Another model where some of the coalition values are not provided are incomplete cooperative games or partially defined cooperative games [ 4 , 5 ] . Formally, the model is equivalent to games on set systems; a list of known coalitions \mathcal{K} is provided and the values for coalitions outside \mathcal{K} are unknown. Although both models are formally equivalent, application of the R-value , introduced above, is not always suitable under incomplete information as can be seen on the following example. Let N=\{1,2,3\} be the set of players and the values being known for all coalitions but \{1,2\} : 1. v(\{1\})=v(\{2\})=v(\{3\})=1 , 2. v(\{1,3\})=v(\{2,3\})=3 , 3. v(\{1,2,3\})=7 . For all coalitions, except \{1,2\} and \{1,2,3\} , the surplus in uniquely defined. Under the R-value the surplus of \{1,2\} is assumed to be zero, which results in division (2.5,2.5,3) . If the game was modeling a situation where the collaboration of players 1 and 2 was infeasible and at the same time, player 3 can collaborate with either of the players, it seems reasonable to reward player 3 with a higher payoff as is done under the R-value. This division is less justified in situations, where collaboration of \{1,2\} is feasible but v(\{1,2\}) is unknown. Under the R-value, it is implicitly assumed that there is no surplus of coalition \{1,2\} . THis means that, under the assumption of non-negative surpluses, the R-value corresponds to the most pessimistic scenario. One might argue, following the pattern of the game, that v(\{1,2\})=3 as is the case for other coalitions of size 2. Without additional assumptions, all we know about the surplus of \{1,2\} is that it contributes to v(\{1,2,3\}) just as much as the surplus of \{1,2,3\} does. It is thus more fair to assume that both surpluses are equal. This idea leads to the definition of our uniform-dividend value (UD-value) . In the following section, we formally introduce this value and analyze when it is uniquely defined. Interestingly, uniqueness occurs when the set of known coalitions \mathcal{K} is intersection-closed. For such systems, Béal et al. [ 6 ] recently introduced a value, which we refer to as the IC-value . We also provide empirical analysis of uniqueness of the UD-value for set systems, which are not intersection-closed, however, our analysis restricts to intersection-closed set systems for the rest of this work. Section 3 is dedicated to characterizing the UD-value in terms of incomplete cooperative games. To illustrate, recall the example above. All the known surpluses are non-negative, and if we assume that the surpluses for \{1,2\} and \{1,2,3\} are also non-negative, then any assignment of values to these surpluses leads to what is called a positive extension of the incomplete game. The expected Shapley value over all these possible extensions is exactly the UD-value. In Section 3 , we formalize this intuition and highlight the correspondence between the UD-value and the average value introduced in [ 7 ] . We compare the UD-value with the IC-value and the R-value in Section 4 . We show a common framework for defining all these values and show how it can be used to analyse them in the scope of incomplete cooperative games. We further modify two known axiomatizations of the IC-value provided in [ 6 ] to characterize both the UD-value and the R-value and illustrate difference between the values by providing examples of incomplete games, which violate axioms of three known characterizations of the R-value. We conclude the section with experimental analysis of the difference of the three values and their distance from uniform allocation rule. 2 UD-value Recall a (complete) cooperative game (N,v) , where N is the player set and v\colon 2^{N}\to\mathbb{R} with v(\emptyset)=0 is the characteristic function. We call S\subseteq N a coalition and v(S)\in\mathbb{R} represents the value of coalition S . The Shapley value \phi(v)\in\mathbb{R}^{n} of a cooperative game (N,v) is defined as \phi_{i}(v)=\sum_{S\subseteq N,i\in S}\frac{d_{v}(S)}{|S|} , where d_{v}(S) is the surplus of coalition S in (N,v) and is defined recursively as d_{v}(S)=v(S)-\sum_{T\subsetneq S}d_{v}(T) with d_{v}(\{i\})=v(\{i\}) . When d_{v}(S)\geq 0 for every coalition S , we say (N,v) is positive and denote the set of all positive games on n players by \mathbb{P}^{n} . We write v instead of (N,v) when N is known from the context. For the sake of brevity, we write i instead of \{i\} , and by n,s,t , we denote the sizes of coalitions N , S , and T , respectively. Definition 1 . An incomplete cooperative game (N,\mathcal{K},v) is given by a set