Lectures in Cooperative Game Theory

Lecture 1
Introduction and Denition of TU games

1.1 Introduction
The study of the theory of games gained popularity after the publication of the classical
book Theory of games and economic behaviour by von Neumann and Morgenstern
in 1944[6]. A game is a mathematical model of an interaction between rational agents.
The model rst contains the strategy space of any participants (i.e., the actions that are
available). From these different possible action, one can build the set of all possible
states of the interaction. Then, the model contains the rules of interaction. These rules
describe what happens when participants take some valid action, i.e., they determine
what state is reached when the participants take some actions. Each participant has
her own preferences between the different states, and each participant acts so as to
obtain the best possible outcome (this is what we call rational, and we will come back
to this later). The eld of strategic game studies conicting interactions and has been
made popular with the contributions of Nash. Cooperative games is another eld that
analyses cooperation between agents.
The basic framework of cooperative game was introduced by von Neumann and
Morgenstern in [6] and is called the characteristic function. A coalition is simply a
set of agents that interacts. The so called characteristic function provides a payoff to
each coalition. It is important to note that the payoff is given to a coalition, not to
individual agent. The natural question that arises is the following: if all the agents
want to cooperate, how should they share the payoff? Actually, if the agents also need
to decide whether they want to cooperate and with which other agents, we come up
with the following two key questions:
the selection problem: which coalitions are going to form?
the sharing problem: once the members have formed a coalition (i.e., they have
self-organized, interacted, and the coalition formed received a payoff), the prob-
lem is then how to distribute it to the different members of the coalition.
The goal of the course is to answer these two questions.
1
2 Lecture 1. Introduction and Denition of TU games
In some parts of the course, we will focus only on the sharing problem and we will
assume that all the participants intend to cooperate, forming one coalition containing
all the participants (we will call this coalition the grand coalition).
The solutions to the sharing problemare called solution concepts and they are based
on different interpretations of fairness. Unfortunately, there is no unique and accepted
solution concept. For example, one possible criterion is stability: participants should
not have an incentive to change (we will make this notion more formal later, but
for now, let us consider that no participant want to change of coalitions or to ask for a
different share of the payoff). As we will see, not all games can satisfy the most natural
criterion modeling stability. Consequently, many other solution concepts have been
proposed. A large part of the course will be about introducing the different solution
concepts and study their properties.
We will also study some interesting special classes of games (i.e., restrictions of
the model). A class of game may be interested because of its applications. For exam-
ple, the term coalition is often used in political science: parties may form alliances to
obtain more power. Consequently, we will study a class of games that models voting
situation. A classes of game can also be interested because of some special properties
(e.g., a solution with some properties is guaranteed to exist). The properties can also
be computational. However, with cooperative games, one needs to be careful when
dealing with computational issues as the input of the game is by nature exponential.
Indeed, one needs to reason about all possible coalitions, i.e., all possible subset of the
set of agents. Consequently, there are some interesting issues in representing the games
and computing a solution. There are also some interesting issues to use, in practice,
some solution concepts.
Finally, we will study different variations about modeling of cooperation. So far,
we have talked about sharing the value of a coalition. But in some cases, there is not
really a value to share: by forming a coalition, the members are in a specic state of
the world and experience a corresponding satisfaction (e.g. think about which group
of people to talk with at a party).
The course will mainly focus on the game theoretic aspect of cooperative games,
and we will also study AI related issues towards the end of the course. Here is a rough
outline of the course.
Solution concepts
The core
Games with coalition structure and the bargaining set
The nucleolus
The kernel
The Shapley value
Specic class of games
Voting games
Representation and complexity
different model
NTU games and hedonic games
misc
Coalition formation and related issues
1.2. TU games 3
There is no ofcial textbook for this course. There will be provide a lecture note
for each class as this one. Here are some sources I used to prepare the class:
The last three chapters of book A course in game theory by Osborne and Ru-
binstein [3] are devoted to cooperative games. I will use some of this material
for the lectures on the core, the bargaining set, the kernel, the nucleolus and the
Shapley value.
The book An introduction to the theory of cooperative gamesby Peleg and
Sudhlter [4] contains a rigorous and precise treatment of cooperative games. I
used this book for some precision, but it is a more advanced textbook.
For computational aspects and some advanced topics, you can read Computa-
tional Aspects of Cooperative Game Theory [1] by Chalkiadakis, Elkind and
Wooldridge.
Whenever appropriate, I will also refer to article from the literature.
1.2 TU games
The game theory community has extensively studied the coalition formation prob-
lem [2, 3]. The literature is divided into two main models, depending on whether
utility can be transferred between individuals. In a transferable utility game (or TU
game), it is assumed that agents can compare their utility and that a common scale of
utility exists. In this case, it is possible to dene a value for a coalition as the worth the
coalition can achieve through cooperation. The agents have to share the value of the
coalition, hence utility needs to be transferable. In a so-called non-transferable utility
game (or NTU game), inter-personal comparison of utility is not possible, and agents
have a preference over the different coalitions of which it is a member. In this section,
we introduce the TU games.
1.2.1 Denitions
In the following, we use a utility-based approach and we assume that everything has a
price: each agent has a utility function that is expressed in currency units. The use of
a common currency enables the agents to directly compare alternative outcomes, and
it also enables side payments. The denition of a TU game is simple: it involves a
set of players and a characteristic function (a map from sets of agents to real numbers)
which represents the value that a coalition can achieve. The characteristic function
is common knowledge and the value of a coalition depends only on the other players
present in its coalition.
Notations
We consider a set N of n agents. A coalition is a non-empty subset of N. The set N
is also known as the grand coalition. The set of all coalitions is 2
N
and its cardinality
is 2
n
. A coalition structure (CS) S = {C
1
, , C
m
} is a partition of N: each set C
i
is a
coalition with
m
i=1
C
i
= N and i = j C
i
C
j
= . We will denote S
C
the set of all
partitions of a set of agnets C N. The set of all CSs is then denoted as S
N
, its size
is of the order O(n
n
) and (n
n
2
) [5].
TU games
1.2.1. DEFINITION. A transferable utility game (TU game) is dened as a pair (N, v)
where N is the set of agents, and v : 2
N
R is a characteristic function.
The characteristic function (or valuation function) v : 2
N
R provides the worth or
utility of a coalition. Note that this denition assumes that the valuation of a coalition
C does not depend on the other coalitions present in the population.
Standard Assumptions: It is usually assumed that the value of the empty coalition
is 0 (i,.e. v() = 0; we will make that assumption throughout the class. Moreover,
it is often the case that the value of each coalition is non-negative (when agents make
prots) or else that the value of each coalition is non-posititive (when the members
share some costs). During this class, we will assume that for each coalition C, v(C)
0. However, most of the denitions and results can be easily adapted.
A rst example of a TU game is the majority game. Assume that the number of
agents n is odd and that the agents decide between two alternatives using a majority
vote. Also assume that no agent is indifferent, i.e., an agent always strictly prefers one
alternative over the other. We model this by assigning to a winning coalition the
value 1 and to the other ones the value 0, i.e.,
v(C) =
1 when |C| >

n
2
0 otherwise
Some types of TU games
We now describes some types of valuation functions. First, we introduce a notion that
will be useful on many occasion: the notion of marginal contribution. It represent the
contribution of an agent when it joins a coalition.
1.2.2. DEFINITION. The marginal contribution of agent i N for a coalition
C N \ {i} is mc
i
(C) = v(C {i}) v(C).
The maximal marginal contribution mc
max
i
= max
CN\{i}
mc
i
(C) can been seen
as a threat that an agent can use against a coalition: the agent can threatens to leave
its current coalition to join the coalition that produces mc
max
i
, arguing that it is able to
generate mc
max
i
utils. The minimal marginal contribution mc
min
i
= min
CN\{i}
mc
i
(C)
1.2. TU games 5
is a minimum acceptable payoff: if the agent joins any coalition, the coalition will
benet by at most mc
min
i
, hence agent i should get at least this amount.
Additive (or inessential): C
1
, C
2
N | C
1
C
2
= , v(C
1
C
2
) = v(C
1
) + v(C
2
).
When a TU game is additive, v(C) =
iC
v(i), i.e., the worth of each coalition
is the same whether its members cooperate or not: there is no gain in cooper-
ation or any synergies between coalitions, which explains the alternative name
(inessential) used for such games.
Monotone: C
1
C
2
N, v(C
1
) v(C
2
). For example, the valuation function of the
majority game is monotone: when more agents join a coalition, they cannot turn
the coalition from a winning to a losing one. Many games are monotone, how-
ever, we can imagine non-monotone games. For instance, the overhead caused
by costs of communication or the effort to cooperate may be such that adding
another agent may decrease the value of a coalition. Another example features
two agents that dislike each other: the productivity of a coalition may decrease
when both of them are members of the coalition.
Superadditive: C
1
, C
2
N | C
1
C
2
= , v(C
1
C
2
) v(C
1
) + v(C
2
), in other
words, any pair of coalitions is best off by merging into one. Many games are
super-additive. As we have assumed that the value of a coalition is positive, su-
peradditivity implies monotonicity (but the converse is not necessarily true). In
such games, social welfare is maximised by forming the grand coalition. Conse-
quently, the agents have incentives to form the grand coalition.
Subadditive: C
1
, C
2
N | C
1
C
2
= , v(C
1
C
2
) v(C
1
) +v(C
2
): the agents are
best off when they are on their own, i.e., cooperation not desirable.
Convex games: A valuation is convex if for all C T and i / T v(C {i}) v(C)
v(T {i}) v(T). So a valuation function is convex when the marginal con-
tribution of each player increases with the size of the coalition he joins. Convex
valuation functions are superadditive.
Unconstrained. The valuation function can be superadditive for some coalitions, and
subadditive for others: some coalitions should merge when others should remain
separated. This is the most difcult and interesting environment.
Solutions
The valuation function provides a value to a set of agents, not to individual agents. The
payoff distribution x = {x
1
, , x
n
} describes how the worth of the coalition is shared
between the agents, where x
i
is the payoff of agent i.
It will be useful to talk about the payoff obtained by the members of a coalition and
we will use the notation x(C) =
iC
x(i).
(N, v)
v : 2
N
R
(CS, x)
S S
N
, i.e.,
S = {C
1
, . . . , C
k
}, C
i
N, i = j C
i
C
j
=
x R
n
?
TU game
Payoff conguration
Figure 1.1: What is solving TU games?
We can nally formalize the solution of a TU game (N, v) by introducing the con-
cept of payoff conguration (PC). A payoff conguration (PC) is a pair (S, x) where
S S
N
is a CS and x is a payoff distribution. The CS answers the selection prob-
lem when the payoff distribution answers the sharing problem. This is illustrated in
Figure 1.1.
Let us now illustrate all these concepts with the following three-player TU game
described in Table 1.1. In this example, there are three agents named 1, 2 and 3. There
are 7 possible coalitions and the value of each coalition is given in the table. There
are 5 CSs which are the following: {{1}, {2}, {3}}, {{1, 2}, {3}}, {{1}, {2, 3}},
{{2}, {1, 3}}, {{1, 2, 3}}. This game is monotone and superadditive, but it is not
convex.
N = {1, 2, 3}
v({1}) = 0, v({2}) = 0, v({3}) = 0
v({1, 2}) = 90
v({1, 3}) = 80
v({2, 3}) = 70
v({1, 2, 3}) = 105
Table 1.1: An example of a TU game
What PC should be chosen? Should the agents form the grand coalition and share
equally the value? The choice of the coalition can be justied by arguing that it is
the coalition that generates the most utility for the society (the game is superadditive).
However, is an equal share justied? Agent 3 could propose to agent 1 to form {1, 3}
and to share equally the value of this coalition (hence, 40 for each agent). Actually,
agent 2 can make a better offer to agent 1 by proposing an equal share of 45 if they
form {1, 2}. Agent 3 could then propose to agent 1 to form {1, 3} and to let it get
46 (agent 3 would then have 34). Is there a PC that would be preferred by all agents
at the same time? In this course, you will learn different ways to solve this problem.
Unfortunately, as for many other games, we will see that there is not one unique best
solution.
1.2. TU games 7
1.2.2 Rationality concepts
In this section, we discuss some desirable properties that link the coalition values to
the agents individual payoff. In other words, these properties are constraints that one
would like to satisfy.
Feasible solution: First, one should not distribute more utility than is available. A
payoff x is feasible when
iN
x
i
v(N).
Anonymity: A solution is independent of the names of the agents. This is a pretty
mild solution that will always be satised.
Efciency: x(N) = v(N) the payoff distribution is an allocation of the whole worth
of the grand coalition to all the players. In other words, no utility is lost at the
level of the population. This is particularly relevant for superadditive games.
Individual rationality: An agent i will be a member of a coalition only when x
i

v({i}), i.e., to be part of a coalition, a player must be better off than when it is
on its own.
Group rationality: C N, x(C) v(C), i.e., the sum of the payoff of a coalition
should be at least the value of the coalition (there should not be any loss at the
level of a coalition).
Pareto optimal payoff distribution: It may be desirable to have a payoff distribution
where no agent can improve its payoff without lowering the payoff of another
agent. More formally, a payoff distribution x is Pareto optimal iff y R
n
| i
N | {y
i
> x
i
and j = i, y
j
x
j
}.
Reasonable from above: an agent should get at most its maximal threat, i.e., x
i

mc
max
i
.
Reasonable from below: the agent should get at least its minimum acceptable reward
x
i
mc
min
i
.
Some more notions will be helpful to discuss some solution concepts. The rst is
the notion of imputation, which is a payoff distribution with the minimal acceptable
constraints.
1.2.3. DEFINITION. An imputation is a payoff distribution that is efcient and individ-
ually rational for all agents.
An imputation is a solution candidate for a payoff distribution, and can also be used
to object a payoff distribution.
The second notion is the excess which can be seen as an amount of complaint or as
a potential strength depending on the view point.
1.2.4. DEFINITION. The excess related to a coalition C given a payoff distribution x is
e(C, x) = v(C) x(C).
When e(C, x) > 0, the excess can be seen as an amount of complaint for the current
members of C as some part of the value of the coalition is lost. When C is not actually
formed, some agent i C can also see the excess as a potential increase of its payoff if
C was to be formed. Some stability concepts (the kernel and the nucleolus, see below)
are based on the excess of coalitions. Another stability concept can also be dened in
terms of the excess.
Bibliography
[1] Georgios Chalkiadakis, Edith Elkind, and Michael Wooldridge. Computational
aspects of cooperative game theory. Morgan & Claypool, 2011.
[2] James P. Kahan and Amnon Rapoport. Theories of Coalition Formation. Lawrence
Erlbaum Associates, Publishers, 1984.
[3] Martin J. Osborne and Ariel Rubinstein. A Course in Game Theory. The MIT
Press, 1994.
[4] Bezalel Peleg and Peter Sudhlter. Introduction to the theory of cooperative coop-
erative games. Springer, 2nd edition, 2007.
[5] Tuomas W. Sandholm, Kate S. Larson, Martin Andersson, Onn Shehory, and Fer-
nando Tohm. Coalition structure generation with worst case guarantees. Articial
Intelligence, 111(12):209238, 1999.
[6] John von Neumann and Oskar Morgenstern. Theory of games and economic be-
havior. Princeton University Press, 2
nd
edition, 1947.
9
Lecture 2
The Core
Let us assume that we have a TU game (N, v) and that we want to form the grand
coalition. We model cooperation between all the agents in N and we focus on the
sharing problem: how to distribute the payoff v(N) to all agents. The idea for dening
one solution is to consider a payoff distribution in which no agent has an incentive to
change coalition to gain additional payoff. This is what is called stability.
The Core, which was rst introduced by Gillies [2], is the most attractive and natu-
ral way to dene stability. A payoff distribution is in the Core when no group of agents
has any incentive to form a different coalition. This is a strong condition for stability,
so strong that some games may have an empty core. In this lecture, we will rst intro-
duce the denition of the core and consider some graphical representations for games
with up to three players. Then, we will present some games that are guaranteed to have
a non-empty core. Finally, we will present a theorem that characterizes games with
non-empty core: the Bondareva-Shapley theorem. We will give some intuition about
the proof, relying on results from linear programming, and we will use this theorem to
show that market games have a non-empty core.
2.1 Denition and graphical representation for games
with up to three players
We consider a TU game (N, v). We assume that all the agents cooperate by forming
the grand coalition and that they receive a payoff distribution x. We want the grand
coalition to be stable, i.e., no agent should have an incentive to leave the grand coali-
tion. We will say that x is in the core of the game (N, v) when no group of agents has
an incentive to leave the grand coalition and form a separate coalition.
2.1.1. DEFINITION. [Core] A payoff distribution x R
n
is in the Core of a TU game
(N, v) iff x is an imputation that is group rational, i.e.,
Core(N, v) = x R
n
[
iN
x
i
= v(N) ( N x(() v(().
11
12 Lecture 2. The Core
A payoff distribution is in the Core when no group of agents has any interest in
rejecting it, i.e., no group of agents can gain by forming a different coalition. Note
that this condition has to be true for all subsets of N (group rationality). As a special
case, this ensures individual rationality. Another way to dene the Core is in terms of
excess:
2.1.2. DEFINITION. [Core] The Core of a TU game (N, v) is the set of payoff distri-
butions x R
n
, such that ( N, e((, x) 0.
In other words, a PC is in the Core when there exists no coalition that has a positive
excess. This denition is attractive as it shows that no coalition has any complaint:
each coalitions demand can be granted.
To be in the core, a payoff distribution must satisfy a set of 2
n
weak linear inequal-
ities: for each coalition ( N, we have v(() x((). The Core is therefore closed
and convex, and we can try to represent it geometrically.
Let us consider the following two-player game (1, 2, v) where v(1) = 5,
v(2) = 5, and v(1, 2) = 20. The core of the game is a segment dened as follows:
core(N, v) = (x
1
, x
2
) R
2
[ x
1
5, x
2
5, x
1
+ x
2
= 20 and is represented
in Figure 2.1. This example shows that, although the game is symmetric, most of the
payoffs in the core are not fair. Core allocations focus on stability only and they may
not be fair.
x
1
x
2
0 5 10 15 20
0
5
10
15
20
Figure 2.1: Example of a core allocation
It is possible to represent the core for game with three agents. For a game (1, 2, 3, v),
the efciency condition is v(1, 2, 3) = x
1
+ x
2
+ x
3
, which is a plane in a 3-
dimensional space. On this plane, we can draw the conditions for individual rationality
2.2. Games with non-empty core 13
and for group rationality. Each of these conditions partitions the space into two regions
separated by a line: one region is compatible with a core allocation, the other region
is not. The core is the intersection of all the compatible regions. Figure 2.2 represents
the core of a three-player game.
Core(N, v)
x
1
+x
2
= 4
x
1
+x
3
= 3
x
2
+x
3
= 5
(1, 6, 1) (1, 0, 7)
(7, 0, 1)
x
1
= 1
x
3
= 1 x
2
= 0
(
1
,
5
,
2
)
(
1
,
4
,
3
)
(
1
,
3
,
4
)
(
1
,
2
,
5
)
(
1
,
1
,
6
)
(2, 5, 1)
(3, 4, 1)
(4, 3, 1)
(5, 2, 1)
(6, 1, 1)
(2, 0, 6)
(3, 0, 5)
(4, 0, 4)
(5, 0, 3)
(6, 0, 2)
v(1) = 1 v(1, 2) = 4
v(2) = 0 v(1, 3) = 3
v(3) = 1 v(2, 3) = 5
v() = 0 v(1, 2, 3) = 8
Figure 2.2: Example of a three-player game: The core is the area in green
There are, however, multiple concerns associated with using the notion of the Core.
First and foremost, the Core can be empty: the conicts captured by the characteristic
function cannot satisfy all the players simultaneously. When the Core is empty, at least
one player is dissatised by the utility allocation and therefore blocks the coalition.
Let us consider the following example from [3]: v(A, B) = 90, v(A, C) = 80,
v(B, C) = 70, and v(N) = 120. In this case, the Core is the PC where the grand
coalition forms and the associated payoff distribution is (50, 40, 30). If v(N) is in-
creased, the size of the Core also increases. But if v(N) decreases, the Core becomes
empty.
Exercise: How can you modify the game in Figure 2.2 so that the core becomes empty?
2.2 Games with non-empty core
In the previous section, we saw that some games have an empty core. In this section,
we provide examples of some classes of games that are guaranteed to have a non-empty
core. In the following we will show that convex games and minimum cost spanning
tree games have a non empty core.
We start introducing an example that models bankruptcy: individuals have claims
in a resource, but the value of the resource is not sufcient to meet all of the claims
(e.g., a man leaves behind an estate worth less than the value of its debts). The problem
is then to share the value of the estate among all the claimants. The value of a coalition
( is dened as the amount of the estate which is not claimed by the complement of
(, in other words v(() is the amount of the estate that the coalition ( is guaranteed to
obtain.
2.2.1. DEFINITION. Bankruptcy game A Bankruptcy game (N, E, v) where N is the
set of claimants, E R
+
is the estate and c R
n
+
is the claim vector (i.e., c
i
is the
claim of the i
th
claimant. The valuation function v : 2
N
R is dened as follows.
For a coalition of claimants (, v(() = max
_
0, E
iN\C
c
i
_
.
First, we show that a bankruptcy game is convex.
2.2.2. THEOREM. Every bankruptcy game is convex.
Proof. Let (N, E, c) be a bankruptcy game. Let S T N, and i / T. We want to
show that
v(S i) v(S) v(T i) v(T),
or equivalently that
v(S i) +v(T) v(T i) +v(S).
For all C N, we note c(C) =
jC
c
j
, then we can write:
E
jN\C
c
j
= E
jN
c
j
+
jC
c
i
= E c(N) +c(C).
Let = E
jN
c
j
= E c(N). We have E
jN\C
c
j
= +c(C).
First, observe that (x, y) R
2
, max0, x + max0, y = max0, x, y, x +y.
v(S i) +v(T) = max
_
_
_
0, E
jN\(S{i})
c
j
_
_
_
+ max
_
_
_
0, E
jN\T
c
j
_
_
_
= max 0, +c(S) +c
i
+ max 0, +c(T)
= max 0, +c(S) +c
i
, +c(T), 2 +c(S) +c
i
+c(T)
v(T i) +v(S) = max
_
_
_
0, E
jN\(T{i})
c
j
_
_
_
+ max
_
_
_
0, E
jN\S
c
j
_
_
_
= max 0, +c(T) +c
i
+ max 0, +c(S)
= max 0, +c(T) +c
i
, +c(S), 2 +c(T) +c
i
+c(S)
Then, note that since S T, c(S) c(T). Then
max 0, +c(T) +c
i
, +c(S), 2 +c(T) +c
i
+c(S) =
max 0, +c(T) +c
i
, 2 +c(T) +c
i
+c(S).
2.2. Games with non-empty core 15
We also have:
+c(S) +c
i
+c(T) +c
i
.
+c(T) +c(T) +c
i
.
It follows that max 0, +c(S) +c
i
, +c(T), 2 +c(S) +c
i
+c(T)
max 0, +c(T) +c
i
, 2 +c(T) +c
i
+c(S)
which proves that v(S i) +v(T) v(T i) +v(S).
Now, we show an important property of convex games: they are guaranteed to have
a non-empty core. We dene a payoff distribution where each agent gets its marginal
contribution, given that the agents enter the grand coalition one at a time in a given
order, and we show that this payoff distribution is an imputation that is group rational.
2.2.3. THEOREM. A convex game has a non-empty core.
Proof. Let us assume a convex game (N, v). Let us dene a payoff vector x in the
following way: x
1
= v(1) and for all i 2, . . . , n, x
i
= v(1, 2, . . . , i)
v(1, 2, . . . , i 1). In other words, the payoff of the i
th
agent is its marginal contri-
bution to the coalition consisting of all previous agents in the order 1, 2, . . . , i 1.
Let us prove that the payoff vector is efcient by writing up and summing the payoff
of all agents:
x
1
= v(1)
x
2
= v(1, 2 v(1)
. . .
x
i
= v(1, 2, . . . , i) v(1, 2, . . . , i 1)
. . .
x
n
= v(1, 2, . . . , n) v(1, 2, . . . , n 1)
iN
x
n
= v(1, 2, . . . , n) = v(N)
By summing these n equalities, we obtain the efciency condition:
iN
x
n
= v(1, 2, . . . , n) = v(N).
Let us prove that the payoff vector is individually rational. By convexity, we have
v(i) v() v(1, 2, . . . , i) v(1, 2, . . . , i 1), hence v(i) x
i
.
Finally, let us prove that the payoff vector is group rational. Let C N, C =
a
1
, a
2
, . . . , a
k
and let us consider that a
1
< a
2
< . . . < a
k
. It is obvious that
a
1
, a
2
, . . . , a
k
1, 2, . . . , a
k
. Using the convexity assumption, we obtain the fol-
lowing:
v(a
1
) v() v(1, 2, . . . , a
1
) v(1, 2, . . . , a
1
1) = x
a1
v(a
1
, a
2
) v(a
1
) v(1, 2, . . . , a
2
) v(1, 2, . . . , a
2
1) = x
a2
. . .
v(a
1
, a
2
, . . . , a
l
) v(a
1
, a
2
, . . . , a
l1
) v(1, 2, . . . , a
l
) v(1, 2, . . . , a
l
1) = x
a
l
. . .
v(a
1
, a
2
, . . . , a
k
) v(a
1
, a
2
, . . . , a
k1
) v(1, 2, . . . , a
k
) v(1, 2, . . . , a
k
1) = x
a
k
v(C) = v(a
1
, a
2
, . . . , a
k
)

k
i=1
x
a
k
= x(C)
By summing these k inequalities, we obtain:
v(C) = v(a
1
, a
2
, . . . , a
k
)
k
i=1
x
a
k
= x(C), which is the group rationality con-
dition.
Consequently, if a game is convex, we know that we can guarantee a stable payoff
distribution. Moreover, we now know one easy way to compute one of these stable
payoffs.
Another example of games that have a non-empty core are the class of minimum
cost spanning tree game. This game features a set of houses that have to be connected
to a power plant. The houses can be directly linked to the power plant, or to another
house. Let N be the set of houses, and let P be the power plant. Let us dene N
=
N0. For (i, j) N
2
, i ,= j, the cost of connecting i and j by the edge e

ij
is c
i,j
. For
a coalition of houses ( N, let (() be a minimum cost spanning tree over the set of
edges (P. In other words, when the houses forma coalition (, they try to minimize
the cost of connected them to the power plant. Let (N, c) be the corresponding cost
game in which the cost of coalition ( N is dened as c(() =
(i,j)(C)
c
ij
.
1
2
3
4
5
0
2.2.4. THEOREM. Every minimum cost spanning tree game has a non-empty core.
Proof. Let us dene a cost distribution x and then we will show that x is in the core.
Let T = (N, E
N
) a minimum cost spanning tree for the graph
_
N
, c
{ij}N
2
_
. Let
i be a customer. Since T is a tree, there is a unique path (0, a
1
, . . . , a
k
, i) from 0 to i.
The cost paid by agent i is dened by x
i
= c
a
k
,i
.
This cost allocation is efcient by construction of x.
We need to show the cost allocation is group rational, i.e., for all coalition S, we
have x(S) v(S) (it is a cost, which explains the inequality).
Let S N and T
S
= (S 0, E
s
) be a minimum cost spanning tree of the graph
_
S 0, c
{ij}S{0}
_
. Let extand the tree T
S
to a graph T
+
S
= (N
, E
+
N
) by adding
2.3. Characterization of games with a non-empty core 17
the remaining customers N S, and for each customer i N S, we add the edge of
E
N
ending in i, i.e., we add the edge (a
k
, i). The graph T
+
S
has [S[ + [N S[ edges
an is connected. Hence, T
+
S
is a spanning tree. Now, we note that c(S) +x(N S) =
e
ij
E
+
N
c
ij

e
ij
E
N
= c(N) = x(N). The inequality is due to the fact that T
+
S
is
a spanning tree, and T is a minimum spanning tree. It follows that x(S) v(S).
2.3 Characterization of games with a non-empty core
We saw that the core may be empty, but that some classes of games have a non-empty
core. The next issue is whether we can characterize the games with non-empty core.
It turns out that the answer is yes, and the characterization has been found indepen-
dently by Bondareva (1963) and Shapley (1967), resulting in what is now known as
the BondarevaShapley theorem. This result connects results from linear program-
ming with the concept of the core. In the following, we will rst write the denition
of elements in the core as an optimization problem. Then, we will briey introduce
linear programming and we will use a result to charaterize the games with non-empty
core, which is the Bondareva-Shapley theorem. Finally, we will apply the Bondareva-
Shapley theorem to market games.
2.3.1 Expressing the core as an optimization problem
The main idea is to consider that the core can be expressed as a solution of a con-
straint linear optimization problem where the condition imposed by group rationality
are the constraints of the optimization problem and the objective function is the sum
of the payoffs of the agents. Let us consider a TU game (N, v), let x denote a payoff
distribution and let us consider the following optimization problem:
(LP)
_
min x(N)
subject to x(() v(() for all ( N, S ,=
The linear constraints are the constraints of group rationality: for each coalition (
N, we want x(() v((). Satisfying these constraints only is easy: one simply needs
to choose large enough values for each x
i
. If an element y R
n
satises all these
constraints (this is called a feasible solution), then y is group rational. The group
rationality assumption for the grand coalition guarantees that we have y(N) v(N).
For y to be in the core, it also needs to be efcient. This forces us to choose values that
are not too large for the y
i
. The idea is then to search for the elements that minimize
y(N) =
iN
y
i
.
When solving this optimization problem, two things may happen. Either the mini-
mum value found is v(N), or it is a value strictly greater.
In the rst case, the solutions x of the optimization problems are elements of the
core: they satisfy the constraints hence, they are group rational and since the
minimum is v(N), x is efcient as well.
In the second case, it is not possible to satisfy both group rationality and ef-
ciency, and the core of the game is empty.
there is no other cases as a solution of the optimization problem would satisfy
all the constraints, in particular the one for the grand coalition.
The optimization problem we wrote is called a linear program. It minimizes a
linear function of a vector x subject to a set of constraints where each constraint is an
inequality: a linear combination of x is larger than a constant. This problem is a well
established problem in optimization and in the following, we give a brief introduction
to such problems.
2.3.2 A very brief introduction to linear programming
The goal of this section is to briey introduce linear programming, which is a special
kind of optimization problems: the problem is about maximizing a linear function
subject to linear constraints. More formally, a linear program has the following form:
_
_
_
max c
T
x
subject to
_
Ax b,
x 0
where
x R
n
is a vector of n variables
c R
n
is the objective function
A is a mn matrix
b R
n
is a vector of size n
A and b represent the linear constraints. Let us look at a simple example:
_
_
_
maximize 8x
1
+ 10x
2
+ 5x
3
subject to
_
3x
1
+ 4x
2
+ 2x
3
7 (1)
x
1
+ x
2
+ x
3
2 (2)
In this example, we can recognize the different components A, B and C to be:
A =
_
3 4 2
1 1 1
_
b =
_
7
2
_
c =
_
_
8
10
5
_
_
.
We say that a solution is feasible when it satises the constraints. For our example,
we have:
0, 1, 1 is feasible, with objective function value 15.
1, 1, 0 is feasible, with objective function value 18, hence it is a better solution.
Next, we introduce the notion of the dual of a LP: it is another linear program
which goal is to nd an upper bound to the objective function of the original LP. Let us
rst look at our example and let us consider the following two linear transformations:
(1) 1 + (2) 6 9x
1
+ 10x
2
+ 8x
3
19
(1) 2 + (2) 2 8x
1
+ 10x
2
+ 6x
3
18
by taking linear combinations over the constraints, we are able to form a new constraint
that provides an upper bound for the objective function. The reason is that in the new
constraint we formed, the coefcients for x
1
, x
2
and x
3
are larger or equal to the ones of
the objective function, hence, it must be the case that the bound is an upper bound for
the objective function. Using the second new constraint, we observe that the solution
cannot be better than 18. But we already found one feasible solution with a value of
18, so we have solved the problem!
Hence, one idea of the dual is to nd a new constraint that is a linear combination
of all the constraints of the primal: y
T
A y
T
b (where y R
m
). This new constraint
must generate the lowest value as y
T
b will be the upper bound of a solution , and
the coefcient of y
T
A must be larger than the coefcients of the objective function,
i.e., y
T
A c
T
. Hence, the dual can be written in the following way:
Primal Dual
_
_
_
max c
T
x
subject to
_
Ax b,
x 0
_
_
_
min y
T
b
subject to
_
y
T
A c
T
,
y 0
The following theorems link the solution of the primal and the dual problems.
2.3.1. THEOREM (DUALITY THEOREM). When the primal and the dual are feasible,
they have optimal solutions with equal value of their objective function.
2.3.2. THEOREM (WEAK LP DUALITY). For a pair x, y of feasible solutions of the
primal LP and its dual LP, the objective functions are mutual bounds:
y
T
b c
T
x
If thereby c
T
x = y
T
b (equality holds), then these two solutions are optimal for both
LPs.
Proof. We have y
T
Ax y
T
b since Ax b, y 0, and y
T
Ax c
T
x since y
T
A c ,
x 0. It is immediate that equality of the objective functions implies optimality.
2.3.3 Linear programming and the core
Now, let us go back to the core. The linear programming problem that corresponds to
the core is:
(LP)
_
min x(N)
subject to x(() v(() for all ( N, S ,=
First, this formulation is not exactly the one what have just introduced since it is a
minimization and the constraints are of the form: a linear combination of x is greater
than a constant. It should not be difcult to get convinced that these two kinds of
optimization problems are symmetrical and have similar properties. In terms of the
conventional way to write the primal, we identify the following components:
the vector c R
n
is the vector 1, 1, . . . , 1.
the vector b R
2
n
contains the value of each coalition, i.e., we can index the
elements of b by using a coalition and the elements of b are b
C
= v(().
The matrix A has 2
n
rows (one for each coalition) and n columns (one for each
agent). The entries of A are either 0 or 1. Let us consider one coalition (, the
corresponding constraint for the core is
kC
x
k
v((). Let us say that the
value of coalition ( appears in row i of vector b, i.e. the constraint about ( is
expressed in the i
th
row of Ax b. Consequently, the i
th
row of A encodes
which agent are present in coalition (: the entry A(i, j) is 1 if j ( and 0
otherwise.
Now, we write the dual which maximises y
T
b over all vectors y R
2
n
+
.
_
_
_
max y
T
b
subject to
_
y
T
A c
T
,
y 0
Now, let us introduce some notations to help us write the matrix A.
2.3.3. DEFINITION. [Characteristic vector] Let ( N. The characteristic vector
C
R
N
of ( is the member of R
N
dened by
i
C
=
_
1 if i (
0 if i N (
The characteristic vector of a coalition simply encodes which agents are present in a
coalition. For example, for n = 4,
{2,4}
= 0, 1, 0, 1. This will be helpful to express
the rows of A.
2.3.4. DEFINITION. [Map] A map is a function 2
N
R
+
that gives a positive
weight to each coalition.
A map can be seen as a positive weight that is given to each coalition. Hence, the
solution y of the dual can be called a map.
2.3.5. DEFINITION. [Balanced map] A function : 2
N
R
+
is a balanced map
iff
CN
(()
C
=
N
. For convenience, we will write (() =
C
.
We provide an example in Table 2.1 for a three-player game.
C
is a scalar and
C
is
a vector of R
n
, so the condition features the equality between a sum over 2
n
vectors
of R
n
and
N
R
n
that is nothing but the vector of R
n
containing the value 1 for
each entry. This will be useful to write the constraints of the dual (we will give further
explanation in the following).
i 1 2 3
{1,2}
i
{1,2}
1
2
1
2
0
{1,3}
i
{1,3}
1
2
0
1
2
{2,3}
i
{2,3}
0
1
2
1
2
C
=
_
1
2
if [([ = 2
0 otherwise
Each of the column sums up to 1.
1
2
{1,2}
+
1
2
{1,3}
+
1
2
{2,3}
=
{1,2,3}
Table 2.1: Example of a balanced map for n = 3
One can interpret a balanced map as a percentage of time spent by each agent in
each possible coalition: for each agent i, the sum of the map for all coalitions contain-
ing agent i must sum up to one.
Given these notational tools, let us re-write the dual.
for the objective function: y
T
b is the dot product of the variable y with the
value of each coalition. If we use a coalition to index the entries of the vector y
or if we say we are using a map y , the objective function can be written as
CN
y
C
v(().
for the constraints, we have y
T
A c
T
. First c
T
is a vector composed of 1. It is
also the vector
N
as all the agents are present in N.
Then we have the dot product between y
T
and A: the result of this product is
a vector of size n. Let us consider the i
th
entry of the product: it is the dot
product between y
T
and the i
th
column of A (both vectors are of size 2
n
and we
can indexed them using coalition). We can write this as
CN
y
C
A((, i) and we
note that A((, i) = 1 if i ( and 0 otherwise. That is here that our notation
comes handy and we can write
CN
y
C
A((, i) =
CN
y
C
i
C
. Writing for the
entire vector, we nally have y
T
A =
CN
y
C
C
.
Finally, we have shown that the constraints become
CN
y
C
C

N
.
With our notation, we can now write the dual of LP as:
(DLP)
_
_
_
max
CN
y
C
v(()
subject to
_
CN
y
C
C

N
and,
y
C
0 for all ( N, ( ,= .
Let us consider a game (N, v) with a non-empty core. This means that the dual is
feasible (there are payoff distributions that satisfy the constraints) and that the optimal
payoff has a value of v(N), i.e. it is efcient.
Note that the dual is also feasible. Since one can always dene a balanced map, we
are guaranteed that there exists some y R
2
n
+
such that
CN
y
C
C

N
).
Since v(N) is the minimum of the primal, by Theorem 2.3.2 it is an upper bound
of the dual and it follows that max
CN
y
C
v(() v(N). With this, we conclude that
if a game has a non-empty core we have max
CN
y
C
v(() v(N). To characterize
games with a non-empty core, we need to prove the converse. First, let us give a name
to our condition.
2.3.6. DEFINITION. [Balanced game] A game is balanced iff for each balanced map
we have
CN,C=
(()v(() v(N).
Let us consider that a game (N, v) is balanced, i.e., for each balanced map , we
have
CN

C
v(() v(N). We know that the dual is feasible (using any balanced
map). With the use of a balanced map, we reach the equality for the constraints (i.e.
each constraint is an inequality, but with the balanced map we reach an equality). Since
the coefcients are positive, we will not be able to improve the optimal value of the
dual. Hence, v(N) is the optimal value.
Now, let us go back to the primal. The vector 0, . . . , 0 is feasible, so the primal is
feasible. Using theorem 2.3.2, we know that v(N) is a lower bound for the primal, i.e
v(N) x(N). Applying group rationality to the grand coalition, we also know that
a solution must satisfy x(N) v(N). Consequently, v(N) is also the solution to the
primal. Hence, the core is non-empty.
We have thus proved a characterization of games with non-empty core. This results
was established independently by Bondareva (1963) and Shapley (1967).
2.3.7. THEOREM (BONDAREVA-SHAPLEY THEOREM). A TU game has a non-empty
core iff it is balanced.
This theorem completely characterizes the set of games with a non-empty core.
However, it is not always easy or computationally feasible to check that it is a balanced
game.
2.3.4 Application to market games
One example of coalitional games coming from the eld of economics is a market
game. This game models an environment where there is a given, xed quantity of a
set of continuous good. Initially, these goods are distributed among the players in an
arbitrary way. The quantity of each good is called the endowment of the good. Each
agent i has a valuation function that takes as input a vector describing its endowment
for each good and that output a utility for possessing these goods (the agents do not
perform any transformation, i.e., the goods are conserved as they are). To increase their
utility, the agents are free to trade goods. When the agents are forming a coalition, they
are trying to allocate the goods such that the social welfare of the coalition (i.e. the
sum of the utility of each member of the coalition) is maximized. We now provide the
formal denition.
A market is a quadruple (N, M, A, F) where
N is a set of traders
M is a set of m continuous good
A = (a
i
)
iN
is the initial endowment vector
F = (f
i
)
iN
is the valuation function vector, each f
i
is continuous and concave.
v(S) = max
_
iS
f
i
(x
i
)
x
i
R
m
+
,
iS
x
i
=
iS
a
i
_
we further assume that the f
i
are continuous and concave.
Let us assume that the players form the grand coalition: all the players are in the
market and try to maximize the sum of utility of the market. How should this utility be
shared amond the players? One way to answer this question is by using an allocation
that is in the core. One interesting property is that the core of such game is guaranteed
to be non-empty, and one way to prove it is to use the Bondareva-Shapley theorem.
2.3.8. THEOREM. Every Market Game is balanced.
Proof.
f : R
n
R is concave iff [0, 1], (x, y) R
n
, f(x + (1 )y)
f(x) + (1 )f(y). It follows from this denition that for f : R R, x R
n
,
R
n
+
such that
n
i=1
i
= 1, we have f(
n
i=1
i
x
i
)
n
i=1
i
f(x
i
).
Since the f
i
s are continuous,
iS
f
i
(x
i
) is a continuous mapping from
T=
_
(x
i
)
iS
[ i R
k
+
, x
i
R
k
+
,
iS
x
i
=
iS
a
i
_
to R. Moreover, T is compact
(it is closed and bounded). Thanks to the extreme value theorem from calculus, we
conclude that
iS
f
i
(x
i
) attains a maximum.
For a coalition S N, let x
S
= x
S
1
, . . . , x
S
n
be the endowment that achieves
the maximum value for the coalition S, i.e., v(S) =
iS
f
i
(x
S
i
). In other words, the
members of S have made some trades that have improved the value of the coalition S
up to its maximal value.
Let be a balanced map. Let y R
n
+
dened as follows: y
i
=
SC
i
S
x
S
i
where
(
i
is the set of coalitions that contains agent i.
First, note that y is a feasible payoff function.
iN
y
i
=
iN
SC
i
S
x
S
i
=
SN
iS
S
x
S
i
=
SN
iS
x
S
i
=
SN
iS
a
i
since x
S
i
was achieved by a sequence of trades within the members of S
=
iN
a
i
SC
i
S
=
iN
a
i
as is balanced,
_
i.e., the sum of the weights over all coalitions
of one agent sums up to 1
_
Then, by denition of v, we have v(N)
iN
f
i
(y
i
).
The f
i
are concave and since
SC
i
S
= 1, we have
f
i
(
SC
i
S
x
S
i
)
SC
i
S
f
i
(x
S
i
).
It follows:
v(N)
iN
f
i
(y
i
)
iN
f
i
(
SC
i
S
x
S
i
)
iN
SC
i
S
f
i
(x
S
i
)
SN
iS
f
i
(x
S
i
)
SN
S
v(S).
This inequality proves that the game is balanced.
2.4 Extension of the core
There are few extensions to the concept of the Core. As discussed above, one main is-
sue of the Core is that it can be empty. In particular, a member of a coalition may block
the formation so as to gain a very small payoff. When the cost of building a coalition is
considered, it can be argued that it is not worth blocking a coalition for a small utility
gain. The strong and weak -Core concepts model this possibility. The constraints
dening the strong (respectively the weak) -Core become T N, x(T) v(T) ,
(respectively T N, x(T) v(T) [T[ ). In the weak Core, the minimum amount
2.5. Games with Coalition Structure 25
of utility required to block a coalition is per player, whereas for the strong Core, it is a
xed amount. If one picks large enough, the strong or weak -core will exist. When
decreasing the value of , there will be a threshold
such that for <
the core
ceases to be non-empty. This special -core is then called the the least core.
2.5 Games with Coalition Structure
Thus far, we stated that the grand coalition is formed. With this denition, checking
whether the core is empty amounts to checking whether the grand coalition is stable.
In many studies in economics, the superadditivity of the valuation function is not ex-
plicitly stated, but it is implicitly assumed and hence, it makes sense to consider only
the grand coalition. But when the valuation function is not superadditive, agents may
have an incentive to form a different partition.
We recall that a coalition structure (CS) is a partition of the grand coalitions. If o
is a CS, then o = (
1
, . . . , (
m
where each (
i
is a coalition such that
m
i=1
(
i
= N and
i ,= j (
i
(
j
= .
Aumann and Drze discuss why the coalition formation process may generate a
CS that is not the grand coalition [1]. One reason they mention is that the valuation
may not be superadditive (and they provide some discussion about why it may be
the case). Another reason is that a CS may reect considerations that are excluded
from the formal description of the game by necessity (impossibility to measure or
communicate) or by choice [1]. For example, the afnities can be based on location,
or trust relations, etc.
2.5.1. DEFINITION. [Game with coalition structure] A game with coalition structure
is a triplet (N, v, S), where (N, v) is a TU game, and S is a particular CS. In addition,
transfer of utility is only permitted within (not between) the coalitions of S, i.e., (
S, x(() v(().
Another way to understand this denition is to consider that the problems of decid-
ing which coalition forms and how to share the coalitions payoff are decoupled: the
choice of the coalition is made rst and results in the CS. Only the payoff distribution
choice is left open. The agents are allowed to refer to the value of coalition with agents
oustide of their coalition (i.e., opportunities they would get outside of their coalition)
to negotiate a better payoff. Aumann and Drze use an example of researchers in game
theory that want to work in their own country, i.e., they want to belong to the coalition
of game theorists of their country. They can refer to offers from foreign countries in
order to negotiate their salaries. Note that the agents goal is not to change the CS, but
only to negotiate a better payoff for themselves.
First, we need to dene the set of possible payoffs: the payoff distributions such
that the sum of the payoff of the members of a coalition in the CS does not exceed the
value of that coalition. More formally:
2.5.2. DEFINITION. [Feasible payoff] Let (N, v, o) be a TU game with CS. The set of
feasible payoff distributions is X
(N,v,S)
= x R
n
[ ( ox(() v(().
A payoff distribution x is efcient with respect to a CS o when ( o,
iC
x
j
=
v((). A payoff distribution is an imputation when it is efcient (with respect to the
current CS) and individually rational (i.e., i N, x
i
v(i)). The set of all
imputations for a CS o is denoted by Jmp(o). We can now state the denition of the
core:
2.5.3. DEFINITION. [Core] The core of a game (N, v, o) is the set of all PCs (o, x)
such that x Jmp(o) and ( N,
iC
x
j
v((), i.e.,
core(N, v, o) = x R
n
[ (( o, x(() v(()) (( N, x(() v(()).
We now provide a theorem by Aumann and Drze which shows that the core satises
a desirable properties: if two agents can be substituted, then a core allocation must
provide them identical payoffs.
2.5.4. DEFINITION. [Substitutes] Let (N, v) be a game and (i, j) N
2
. Agents i and
j are substitutes iff ( N i, j, v(( i) = v(( j).
Since the agents have the same impact on all coalitions that do not include them, it
would be fair if they obtained the same payoff. For the core of a game in CS, this is
indeed the case.
2.5.5. THEOREM. Let (N, v, o) be a game with coalition structure, let i and j be sub-
stitutes, and let x core(N, v, o). If i and j belong to different members of o, then
x
i
= x
j
.
Proof. Let (i, j) N
2
be substitutes, ( o such that i ( and j / (. Let
x Core(N, v, o). Since i and j are substitutes, we have
v((( i) j) = v((( i) i) = v(().
Since x Core(N, v, o), we have ( N, x(() v((), we apply this to the
coalition (( i) j:
0 v((( i) j) x((( i) j) = v(() x(() + x
i
x
j
. Since ( o
and x Core(N, v, o), we have x(() = v((). We can then simplied the previous
expression and we obtain x
j
x
i
.
Since i and j play symmetric roles, we have also x
i
x
j
and nally, we obtain
x
i
= x
j
.
Aumann and Drze made a link from a game with CS to a special superadditive
game (N, v) called the superadditive cover [1].
2.5. Games with Coalition Structure 27
2.5.6. DEFINITION. [Superadditive cover] The superadditive cover of (N, v) is the
game (N, v) dened by
_
_
v(() = max
PS
C
_
TP
v(T)
_
( N
v() = 0
In other words, v(() is the maximal value that can be generated by any partition of
(
1
. The superadditive cover is a superadditive game. The following theorem, from [1]
shows that a necessary condition for (N, v, o) to have a non empty core is that o is an
optimal CS.
2.5.7. THEOREM. Let (N, v, o) be a game with coalition structure. Then
a) Core(N, v, o) ,= iff Core(N, v) ,= v(N) =
CS
v(()
b) if Core(N, v, o) ,= , then Core(N, v, o) = Core(N, v)
Proof. Proof of part a)
Let x Core(N, v, o). We show that x Core(N, v) as well. Let ( N
and P
C
S
C
be a partition of (. By denition of the core, for every S N we
have x(S) v(S). The payoff of coalition ( is
x(() =
iC
x
i
=
SP
C
x(S)
SP
C
v(S),
which is valid for all partitions of (. Hence, x(() max
P
C
S
C
SP
C
v(S) = v(().
We have just proved ( N , x(() v((), and so x is group rational.
We now need to prove that v(N) =
CS
v(().
x(N) =
CS
v(() since x is in the core of (N, v, o) (efcient). Applying the
inequality above, we have x(N) =
CS
v(() v(N).
Applying the denition of the valuation function v, we have v(N)
CS
v(().
Consequently, v(N) =
CS
v(() and it follows that x is efcient for the game
(N, v)
Hence x Core(N, v).
1
Note that for the grand coalition, we have v(N) = max
PS
N
_
TP
v(T)
_
, i.e., v(N) is the maximum
value that can be produced by N. We call it the value of the optimal coalition structure. For some
application, on issue (that will be studied later) is to nd this value.
Lets assume x Core(N, v) and v(N) =
CS
v((). We need to prove that
x Core(N, v, o).
For every ( N, x(() v(() since x is in the core of Core(N, v). Then
x(() max
P
C
S
C
SP
C
v(S) v(() using ( as a partition of (, which proves x is
group rational.
x(N) = v(N) =
CS
v(() since x is efcient. It follows that ( o, we
must have x(() = v((), which proves x is feasible for the CS o, and that x is
efcient.
Hence, x Core(N, v, o).
proof of part b):
We have just proved that x Core(N, v) implies that x Core(N, v, o) and x
Core(N, v, o) implies that x Core(N, v). This proves that if Core(N, v, o) ,= ,
Core(N, v) = Core(N, v, o).
Bibliography
[1] Robert J. Aumann and Jacques H Drze. Cooperative games with coalition struc-
tures. International Journal of Game Theory, 3(4):217237, 1974.
[2] Donald B. Gillies. Some theorems on n-person games. PhD thesis, Department of
Mathematics, Princeton University, Princeton, N.J., 1953.
[3] James P. Kahan and Amnon Rapoport. Theories of Coalition Formation. Lawrence
Erlbaum Associates, Publishers, 1984.
29
Lecture 3
The bargaining set
The notion of core stability is maybe the most natural way to describe stability. How-
ever, some games have an empty core. If agents adopts the notion of the core, they
will be unable to reach an agreement about the payoff distribution. If they still want to
benet from the cooperation with other agents, they need to relax the stability require-
ments. In this lecture, we will see that the notion of bargaining set is one way to reach
an argument and to maintain some notion of stability (though of course, it is a weaker
version of stability).
The denition of the bargaining set is due to Davis and Maschler [1]. This notion is
about the stability of a given coalition structure (CS). The agents do not try to change
the nature of the CS, but simply to nd a way to distribute the value of the different
coalitions between the members of each coalition. Let us assume that a payoff distri-
bution is proposed. Some agents may form an objection against this payoff distribution
by pointing out a problem of that distribution and by offering a different payoff distri-
bution that eliminates this issue (or improves the situation). If all other agents agree
with this objection, the payoff distribution should change as proposed. However, some
other agents may form a counter-objection showing some shortcomings of the objec-
tion. The idea of stability in this context is to ensure that, for each possible objection,
there exists a counter-objection. When this is the case, there is no ground for chang-
ing the payoff distribution, which provides some stability. In the following, we will
describe the precise notion of objections and counter-objections.
3.1 Objections, counter-objections and the
prebargaining set
Let (N, v, S) be a game with coalition structure and x an imputation. For the bar-
gaining set, an objection from an agent i against a payoff distribution x is targeting a
particular agent j, in the hope of obtaining a payment from j. The goal of agent i is to
show that agent j gets too much payoff as there are some ways in which some agents
31
32 Lecture 3. The bargaining set
but j can benet. The objection can take the following (informal) form:
I get too little in the imputation x, and agent j gets too much! I can form a
coalition that excludes j in which some members benet and all members
are at least as well off as in x.
We recall that in a game with coalition structure (N, v, S), the agents do not try to
change the CS, but only obtain a better payoff. The set X
(N,v,S)
of feasible payoff
vectors for (N, v, S) is dened as X
(N,v,S)
= {x R
n
| C S,
iC
x
i
v(C)}. We
are now ready to formally dene an objection.
3.1.1. DEFINITION. [Objection] Let (N, v, S) be a game with coalition structure, x
X
(N,v,S)
, C S be a coalition, and i and j two distinct members of C ((i, j) C
2
,
i = j). An objection of i against j is a pair (P, y) where
P N is a coalition such that i P and j / P.
y R
p
where p is the size of P
y(P) v(P) (y is a feasible payoff distribution for the agents in P)
k P, y
k
x
k
and y
i
> x
i
(agent i strictly benets from y, and the other
members of P do not do worse in y than in x.)
An objection is a pair (P, y) that is announced by an agent i against a particular agent j
and a payoff distribution x. It can be understood as a potential threat to form coalition
P, which contains i but not j. If the agents in P really deviate, agent i will benet
(strictly), and the other agents in P are guaranteed not to be worse off (and may even
benet from the deviation). The goal is not to change the CS, but simply to update the
payoff distribution. In this case, agent i is calling for a transfer of utility from agent j
to agent i.
The agent that is targeted by the threat may try to show that she deserves the payoff
x
j
. To do so, her goal is to show that, if the threat was implemented, there is another
deviation that would ensure that j can still obtain x
j
and that no agent (except maybe
agent i) would be worse off. In that case, we say that the objection is ineffective. Agent
j can summarize her argument by saying:
I can form a coalition that excludes agent i in which all agents are at least
as well off as in x, and as well off as in the payoff proposed by i for those
who were offered to join i in the argument.
The formal denition of a counter-objection is the following.
3.1.2. DEFINITION. [Counter-objection] A counter-objection of agent j to the objec-
tion (P, y) of agent i is a pair (Q, z) where
Q N is a coalition such that j Q and i / Q.
3.1. Objections, counter-objections and the prebargaining set 33
z R
q
where q is the size of Q
z(Q) v(Q) (z is a feasible payoff distribution for the agents in Q)
k Q, z
k
x
k
(the members of Q get at least the value in x)
k Q P z
k
y
k
(the members of Q which are also members of P get at
least the value promised in the objection)
In a counter-objection, agent j must show that she can protect her payoff x
j
in spite
of the existing objection of i. Agents in the deviating coalition Q should improve their
payoff compared to x. For those who were also members of the deviating coalition
P with agent i, they should make sure that they obtain a better payoff than in y. In
this way, all agents in P and Q benet. Note that agent i is in P and not in Q, and
consequently, i may be worse off in this counter-objection.
When an objection has a counter-objection, no agent will be willing to follow agent
i and implement the threat. Hence, the agents do not have any incentives to change the
payoff distribution and the payoff is stable.
3.1.3. DEFINITION. [Stability] Let (N, v, S) a game with coalition structure. A vector
x X
(N,v,S)
is stable iff for each objection at x there is a counter-objection.
The denition of the pre-bargaining set is then simply the set of payoff distributions
that are stable. Not that they are stable for the specic denition of objections and
counter-objections we presented (we can, and we will, think about other ways to dene
objections and counter-objections).
3.1.4. DEFINITION. [Pre-bargaining set] The pre-bargaining set (preBS) is the set of
all stable members of X
(N,v,S)
.
We will explain later the presence of the prex pre in the denition. For now, we
need to wonder about the relationship with the core. The idea was to relax the stability
requirements of the core. The following lemma states that indeed, we have only relaxed
them:
3.1.5. LEMMA. Let (N, v, S) a game with coalition structure, we have
Core(N, v, S) preBS(N, v, S).
Proof. Let us assume that the core of (N, v, S) is non-empty and that x Core(N, v, S).
Given the payoff x, no agent i has any objection against any other agent j. Hence, there
are no objections, and the payoff is stable according to the pre-bargaining set.
3.2 An example
Let us now consider an example using a 7-player simple majority game, i.e., we con-
sider the game ({1, 2, 3, 4, 5, 6, 7}, v) dened as follows:
v(C) =
_
1 if |C| 4
0 otherwise
Let us consider x =
1
5
,
1
5
, . . . ,
1
5
. It is clear that x(N) = 1. Let us now prove
that x is in the pre-bargaining set of the game (N, v, {N}).
First note that objections within members of {2, 3, 4, 5, 6, 7} will have a counter-
objection by symmetry (i.e., for i, j {2, 3, 4, 5, 6, 7}, if i has an objection (P, y)
against j, j can use the counter-objection (Q, z) with Q = P \ {i} {j} and z
k
= y
k
for k P \ {i} and z
j
= y
i
).
Hence, we only have to consider two type of objections (P, y): the ones of 1 against
a member of {2, 3, 4, 5, 6, 7}, and the ones from a members of {2, 3, 4, 5, 6, 7} against
i. We are going to treat only the rst case, we leave the second as an exercise.
Let us consider an objection (P, y) of agent i against a member of {2, 3, 4, 5, 6, 7}.
Since the members {2, . . . , 7} play symmetric roles, we consider an objection of 1
against 7 using successively P = {1, 2, 3, 4, 5, 6}, P = {1, 2, 3, 4, 5}, P = {1, 2, 3, 4},
P = {1, 2, 3}, P = {1, 2} and P = {1}. For each case, we will look for a counter-
objection (Q, z) of player 7.
We consider that P = {1, 2, 3, 4, 5, 6}. We need to nd the payoff vector y R
6
so that (P, y) is an objection.
y = ,
1
5
+
2
,
1
5
+
3
, . . . ,
1
5
+
6
,
The conditions for (P, y) to be an objection are the following:
each agent is as well off as in x: >
1
5
,
i
0
y is feasible for coalition P:

6
i=2
_
i
+
1
5
_
+ 1.
w.l.o.g 0
2

3

4

5

6
.
Then
6
i=2
_
1
5
+
i
_
+ =
5
5
+
6
i=2
i
+ = 1 +
6
i=2
i
+ 1.
Then
6
i=2
i
<
1
5
.
We need to nd a counter-objection for (P, y).
claim: we can choose Q = {2, 3, 4, 7} and z =
1
5
+
2
,
1
5
+
3
,
1
5
+
4
,
1
5
+
5
z(Q) =
1
5
+
2
+
1
5
+
3
+
1
5
+
4
+
1
5
+
5
=
4
5
+
5
i=2

i
1 since
5
i=2

i

6
i=2

i
<
1
5
so z is feasible.
3.2. An example 35
It is clear that i Q, z
i
x
i
and that i Q P, z
i
y
i
Hence, (Q, z) is a counter-objection.

Now, let us consider that P = {1, 2, 3, 4, 5}. The vector y = ,
1
5
+
2
,
1
5
+
3
,
1
5
+
4
,
1
5
+
5
is an objection when
>
1
5
,
i
0,
5
i=2
(
1
5
+
i
) + 1
This time, we have
5
i=2
(
1
5
+
i
) + =
4
5
+
5
i=2
i
+ 1
then
5
i=2
i
1
4
5
=
1
5
and nally
5
i=2
i

1
5
<
2
5
.
We need to nd a counter-objection to (P, y)
claim: we can choose Q = {2, 3, 6, 7}, z =
1
5
+
2
,
1
5
+
3
,
1
5
,
1
5
It is clear that i Q, z
i
x
i
and i P Q z
i
y
i
(for agent 2 and 3).
z(Q) =
1
5
+
2
+
1
5
+
3
+
1
5
+
1
5
=
4
5
+
2
+
3
. We have
2
+
3
<
1
5
, otherwise,
we would have
2
+
3

1
5
and since the
i
are ordered, we would then have
5
i=2
i

2
5
, which is not possible. Hence z(Q) 1 which proves z is feasible
Using similar arguments, we nd a counter-objection for each other objections (you

might want to ll in the details at home).
P = {1, 2, 3, 4}, y = ,
1
5
+
1
,
1
5
+
2
,
1
5
+
3
, >
1
5
,
i
0,
4
i=2

i
+
2
5

4
i=2

i

2
5
<
3
5
.
Q = {2, 5, 6, 7}, z =
1
5
+
2
,
1
5
,
1
5
,
1
5
since
2

1
5
|P| 3 P = {1, 2, 3}, v(P) = 0, y = ,
1
,
2
, >
1
5
,
i
0,
1
+
2

<
1
5
Q = {4, 5, 6, 7}, z =
1
5
,
1
5
,
1
5
,
1
5
will be a counter-objection (1 cannot provide
more than
1
5
to any other agent).
For each possible objection of 1, we found a counter-objection. Using similar
arguments, we can nd a counter-objection to any objection of player 7 against
player 1.
x preBS(N, v, S).
3.3 The bargaining set
In the example, agent 1 gets
1
5
when v(C) 0 for all coalition C N! This shows
that the pre-bargaining set may not be individually rational.
Let I(N, v, S) =
_
x X
(N,v,S)
| x
i
v({i})i N
_
be the set of individually
rational payoff vector in X
(N,v,S)
. Given most of the games, the set of individually
rational payoffs is non-empty. For some classes of games, we can show it is the case,
as shown by the following lemma.
3.3.1. LEMMA. If a game is weakly superadditive, I(N, v, S) = .
Proof. A game (N, v) is weakly superadditive when C N and i / C, we have that
v(C) + v({i}) v(C {i}). Let us consider that, for C S, each agent in C gets its
marginal contribution given an ordering of agents in C. Since for all C N, i / C
v(C {i}) v(C) v({i}), we know that x
i
v({i}). Hence, there exists a payoff
distribution in I(N, v, S).
Since we want a solution concept to be at least an imputation (i.e., efcient and
individually rational), we dene the bargaining set to be the set of payoff distributions
that are individually rational and in the pre-bargaining set.
3.3.2. DEFINITION. Bargaining set Let (N, v, S) a game in coalition structure. The
bargaining set (BS) is dened by
BS(N, v, S) = I(N, v, S) preBS(N, v, S).
Of course, this restriction does not have any negative impact on the relationship
between the core and the bargaining set.
3.3.3. LEMMA. We have Core(N, v, S) BS(N, v, S).
This lemma shows that we have relaxed the requirements of core stability. One
question is whether we have relaxed them enough to guarantee that this new stability
concept is guaranteed to be non-empty. The answer to this question is yes!
3.3.4. THEOREM. Let (N, v, S) a game with coalition structure. Assume that I(N, v, S) =
. Then the bargaining set BS(N, v, S) = .
It is possible to give a direct proof of this theorem (for example, se the proof in
Section 4.2 in Introduction to the Theory of Cooperative Games [2]. We will not
present this proof now, but we will prove this theorem differently in a coming lecture.
3.4. One issue with the bargaining set 37
3.4 One issue with the bargaining set
We relaxed the requirements of core stability to come up with a solution concept that
is guaranteed to be non-empty. In doing so, the stable payoff distributions may have
some issues, and we are going to consider one using the example of a weighted voting
game. We rst recall the denition.
3.4.1. DEFINITION. weighted voting games A game (N, w
iN
, q, v) is a weighted vot-
ing game when v satises unanimity, monotonicity and the valuation function is de-
ned as
v(S) =
_
_
_
1 when
iS
w
i
q
0 otherwise
We note such a game by (q : w
1
, . . . , w
n
)
We consider the 6-player weighted majority game (3:1,1,1,1,1,0). Agent 6 is a null
player since its weight is 0, in other words, its presence does not affect by any means
the decision taken by the other agents. Nevertheless the following payoff distribution
is in the bargaining set x =
1
7
, . . . ,
1
7
,
2
7
BS(N, v)! This may be quite surprising
as the null player receives the most payoff, but none of the other agents are able to
provide a objection that is not countered!
One of the desirable properties of a payoff was to be reasonable from above. We
recall that x is reasonable from above if i N x
i
mc
max
i
where mc
max
i
=
max
CN\{i}
v(C {i}) v(C). mc
max
i
is the strongest threat that an agent can use against
a coalition. It is desirable that no agent gets more than mc
max
i
, as it never con-
tributes more than mc
max
i
in any coalition i can join. The previous example shows
that the bargaining set is not reasonable from above: the dummy agent gets more than
max
CN\{6}
(v(C {6}) v(C)) = 0.
Proof.
This proof will be part of homework 2.
Bibliography
[1] Robert J. Aumann and M. Maschler. The bargaining set for cooperative games.
Advances in Game Theory (Annals of mathematics study), (52):217237, 1964.
[2] Bezalel Peleg and Peter Sudhlter. Introduction to the theory of cooperative coop-
erative games. Springer, 2nd edition, 2007.
39
Lecture 4
The nucleolus
The nucleolus is based on the notion of excess and has been introduced by Schmei-
dler [3]. The excess measures the amount of complaints of a coalition for a payoff
distribution. We already mentionned the excess and gave a denition of the core using
the excess. We now recall the denition.
4.0.2. DEFINITION. [Excess] Let (N, v) be a TU game, C N be a coalition, and x
be a payoff distribution over N. The excess e(C, x) of coalition C at x is the quantity
e(C, x) = v(C) x(C).
When a coalition has a positive excess, some utility is not provided to the coalitions
members, and the members complain about this. for a payoff distribution in the core,
there cannot be any complaint. The goal of the nucleolus is to reduce the amount of
complaint, and we are now going to see in what sense it is reduced.
4.1 Motivations and Denitions
Let us consider the game in Table 4.1 and we want to compare two payoff distributions
x and y. A priori, it is not clear which payoff should be preferred. To compare two
vectors of complaints, we can use the lexicographical order
1
.
4.1.1. DEFINITION. [Lexicographical ordering] Let (x, y) R
m
, x
lex
y. We say
that x is greater or equal to y in the lexicographical ordering, and we note x
lex
y,
when
_
x = y or
t, 1 t m such that i 1 i < t x
i
= y
i
and x
t
> y
t
For example, we have 1, 1, 0, 1, 2, 3, 3
lex
1, 0, 0, 0, 2, 3, 3. Let l be a
sequence of m reals. We denote by l
the reordering of l in decreasing order. In the ex-

ample, e(x) = 3, 3, 2, 1, 1, 1, 0 and then e(x)
= 1, 1, 0, 1, 2, 3, 3.
1
the order used for the names in a phonebook or words in a dictionary
41
42 Lecture 4. The nucleolus
Using the lexicographical ordering, we are now ready to compare the payoff distribu-
tions x and y and we note that y is better than x since e(x)
lex
e(y)
: there is a
smaller amount of complaints in y than in x given the lexicographical ordering.
N = {1, 2, 3},
v({i}) = 0 for i {1, 2, 3}
v({1, 2}) = 5, v({1, 3}) = 6, v({2, 3}) = 6
v(N) = 8
Let us consider two payoff vectors x = 3, 3, 2 and y = 2, 3, 3.
x = 3, 3, 2 y = 2, 3, 3
coalition C e(C, x)
{1} -3
{2} -3
{3} -2
{1, 2} -1
{1, 3} 1
{2, 3} 1
{1, 2, 3} 0
coalition C e(C, y)
{1} -2
{2} -3
{3} -3
{1, 2} 0
{1, 3} 1
{2, 3} 0
{1, 2, 3} 0
Table 4.1: A motivating example for the nucleolus
The rst entry of e(x)
is the maximum excess: the agents involved in the corre-

sponding coalition have the largest incentive to leave their current coalition and form
a new one. Put another way, the agents involved in that coalition have the most valid
complaint. If one selects the payoff distribution minimizing the most valid complaint,
there can be a large number of candidates. To rene the selection, among those pay-
off distribution with the smallest largest complaint, one can look at minimizing the
second largest complaint. A payoff distribution is in the nucleolus when it yields the
least problematic sequence of complaints according to the lexicographical ordering.
The nucleolus tries to minimise the possible complaints (or minimise the incentives to
create a new coalition) over all possible payoff distributions.
4.1.2. DEFINITION. Let Imp be the set of all imputations. The nucleolus Nu(N, v) is
the set
Nu(N, v) = {x Imp | y Imp e(y)
lex
e(x)
}.
Intuitively, this denition makes sense. It is another solution concept that focuses
on stability, and it relaxes the stability requirement of the core: the core requires no
complaint at all. The nucleolus may allow for some complaints, but tries to minimize
them.
4.2. Some properties of the nucleolus 43
4.2 Some properties of the nucleolus
We now provide some properties of the nucleolus. First, we consider the relationship
with the core of a game. The following theorem guarantees that the nucleolus of a
game is always included in the core. A payoff distribution in the core does not have
any complaint. If there are more than one payoff in the core, it is possible to use the
excess and the lexicographical ordering to rank the payoff according to the satisfaction
of the agents.
4.2.1. THEOREM. Let (N, v) be a TU game with a non-empty core. Then Nu(N, v)
core(N, v)
Proof. This will be an assignment of Homework 2.
if the core of a game in non-empty, one can use the nucleolus to discriminate between
different core members. Now, we turn to the important issue of the existence of payoff
distributions in the nucleolus. The following theorem guarantees that the nucleolus is
non-empty in most games:
4.2.2. THEOREM. Let (N, v) be a TU game and Imp is the set of imputations. If
Imp = , then the nucleolus Nu(N, v) is non-empty.
This property ensures that the agents will always nd an agreement if they use this
method, which is a great property. The assumption that the set of imputation is a very
mild assumption: if the game does not have any efcient and individually rational
payoff distributions, it is not such an interesting game. The following theorem shows
that in addition to always exist, the nucleolus is in fact unique.
4.2.3. THEOREM. The nucleolus has at most one element.
The proofs of both theorems are a bit involved, and are included in the next section.
The nucleolus is guaranteed to be non-empty and it is unique. These are two im-
portant property in favour of the nucleolus. Moreover, when the core is non-empty, the
nucleolus is in the core.
One drawback, however, is that the nucleolus is difcult to compute. It can be
computed using a sequence of linear programs of decreasing dimensions. The size of
each of these groups is, however, exponential. In some special cases, the nucleolus
can be computed in polynomial time [2, 1], but in the general case, computing the
nucleolus is not guaranteed to be polynomial. Only a few papers in the multiagent
systems community have used the nucleolus, e.g., [4].
4.3 Proofs of the main theorem
The results that the nucleolus is a unique payoff distribution is quite an important result.
We will also use this result to show that some other solution concepts are non-empty.
For this reason, it is worth stating one proof of this theorem, although it is quite a
technical result. To prove the theorem, one needs to use results from analysis. In the
following, we informally recall some denitions and theorems that will be used in the
proofs.
4.3.1 Elements of Analysis
Let E = R
m
and X E. ||.|| denote a distance in E, e.g., the euclidean distance.
We consider functions of the form u : N R
m
. Another viewpoint on u is an
innite sequence of elements indexed by natural numbers (u
0
, u
1
, . . . , u
k
, . . .) where
u
i
X. We recall some denitions:
convergent sequence: A sequence (u
t
) converges to l R
m
iff for all > 0,
T N s.t. t T, ||u
t
l|| .
extracted sequence: Let (u
t
) be an innite sequence and f : N N be a
monotonically increasing function. The sequence v is extracted from u iff
v = u f, i.e., v
t
= u
f(t)
.
closed set: a set X is closed if and only if it contains all of its limit points. In
other words, for all converging sequences (x
0
, x
1
. . .) of elements in X, the limit
of the sequence has to be in X as well.
For example, if X = (0, 1], (1,
1
2
,
1
3
,
1
4
, . . . ,
1
n
, . . .) is a converging sequence.
However, 0 is not in X, and hence, X is not closed.
One way to think about a closed set is by saying A closed set contains its bor-
ders.
bounded set: A subset X R
m
is bounded if it is contained in a ball of nite
radius, i.e. c R
m
and r R
+
s.t. x X ||x c|| r.
compact set: A subset X R
m
is a compact set iff from all sequences in X, we
can extract a convergent sequence in X.
A set is compact set of R
m
iff it is closed and bounded.
convex set: A set X is convex iff (x, y) X
2
, [0, 1], x +(1 )y X
(i.e. all points in a line from x to y is contained in X).
continuous function: Let X R
n
, f : R
n
R
m
.
f is continuous at x
0
X iff
4.3. Proofs of the main theorem 45
> 0 > 0 x X ||x x
0
|| < ||f(x) f(x
0
)|| < . In other
words, the function f does not contain any jump.
We now state some theorems. Let X R
n
.
Thm A
1
If f : R
n
R
m
is continuous and X E is a non-empty compact subset of
R
n
, then f(X) is a non-empty compact subset of R
m
.
Thm A
2
Extreme value theorem: Let X be a non-empty compact subset of R
n
, f : X
R a continuous function. Then f is bounded and it reaches its supremum.
Thm A
3
Let X be a non-empty compact subset of R
n
. f : X R is continuous iff for
every closed subset B R, the set f
1
(B) is compact.
4.3.2 Proofs
Let us assume that the following two theorems are valid. We will prove them later.
4.3.1. THEOREM. Assume we have a TU game (N, v), and consider its set Imp. If
Imp = , then set B = {e(x)
| x Imp} is a non-empty compact subset of R

2
|N|
4.3.2. THEOREM. Let A be a non-empty compact subset of R
m
.
{x A | y A x
lex
y} is non-empty.
We can use these theorems to prove that the nucleolus is non-empty.
Proof.
Let us take a TU game (N, v) and let us assume Imp = . From theorem 4.3.1,
we know that B = {e(x)
| x Imp} is a non-empty compact subset of R

2
|N|
.
Now let us apply the result of theorem 4.3.2 to B. We then have that
{e(x)
| (x Imp) (y Imp e(x)
lex
e(y)
)} is non-empty. From this, it

follows that: Nu(N, v) = {x Imp | y Imp e(y)
lex
e(x)
} = .
In the following, we need to prove both theorems 4.3.1 and 4.3.2. We start by the
rst one.
Proof. Let (N, v) be a TU game and consider its set Imp. Let us assume that Imp =
. We want to prove that B = {e(x)
| x Imp} is a non-empty compact subset of

R
2
|N|
.
First, let us prove that Imp is a non-empty compact subset of R
|N|
.
Imp non-empty by assumption.
To see that Imp is bounded, we need to show that for all i, x
i
is bounded by
some constant (independent of x). We have v({i}) x
i
by individual rationality
and x(N) = v(N) by efciency. Then x
i
+
n
j=1,j=i
v({j}) v(N), hence
x
i
v(N)
n
j=1,j=i
v({j}).
Imp is closed (this is trivial as the boundaries of Imp are members of Imp).
Imp non-empty, closed and bounded. By denition, it is a non-empty compact
subset of R
|N|
.
e()
is a continuous function and Imp is a non-empty and compact subset of R

2
|N|
.
Using thm A
1
, we can conclude that e(Imp)
= {e(x)
|x Imp} is a non-empty
compact subset of R
2
|N|
, which concludes the proof of theorem 4.3.1.
We now turn to the proof of theorem 4.3.2.
Proof. For a non-empty compact subset A of R
m
, we need to prove that the set
{x A | y A x
lex
y} is non-empty.
First, let
i
: R
m
R be the projection function such that
i
(x
1
, . . . , x
m
) = x
i
.
Then, let us dene the following sets:
_
A
0
= A
A
i+1
= argmin
xA
i
i+1
(x), i {0, 1, . . . , m 1}
Let us assume that we want to nd
the minimum of A according to the lexicographic order (say you want to nd the last
words in a text according to the order in a dictionary). You rst take the entire set,
then you select only the vectors that have the smallest rst entries (set A
1
). Then,
from this set, you select the vectors that have the smallest second entry (forming the
set A
2
) and you repeat the process until you reach m steps. At the end, you have
A
m
= {x A | y A x
lex
y}.
We want to prove by induction that each A
i
is non-empty compact subset of R
m
for i {1, . . . , m}. First, we need to show non-emptiness:
A
0
= A is non-empty compact of R
m
by hypothesis .
Let us assume that A
i
is a non-empty compact subset of R
m
and let us prove that
A
i+1
m
.
i+1
is a continuous function and A
i
m
.
Using the extreme value theorem A
2
, min
xA
i

i+1
(x) exists and it is reached in
A
i
, hence argmin
xA
i

i+1
(x) is non-empty.
Now, we need to show each A
i
is compact. We note by
1
i
: R R
m
the inverse
of
i
. Let R,
1
i
() is the set of all vectors x
1
, . . . , x
i1
, , x
i+1
, . . . , x
m
such
that x
j
R, j {1, . . . , m}, j = i. We can rewrite A
i+1
as:
A
i+1
=
1
i+1
_
min
xA
i
i+1
(x)
_
A
i
A
i+1
=
1
i+1
_
_
_
_
_
_
_
min
xA
i
i+1
(x)
. .
closed
_
_
_
_
_
_
_
. .
According to Thm A
3
, it is a compact subset of R
m
A
i
. .
is a compact subset of R
m
since
the intersection of two closed sets is closed and in R
m
,
and a closed subset of a compact subset of R
m
is a compact subset of R
m
Hence A
i+1
m
and the proof is complete.
For a TUgame (N, v) the nucleolus Nu(N, v) is non-empty when Imp = , which
is a great property as agents will always nd an agreement. But there is more! No we
need to prove that there is one agreement which is stable according to the nucleolus.
To prove the unicity of the nucleolus, we again need to prove two results.
4.3.3. THEOREM. Let A be a non-empty convex subset of R
m
. Then the set
{x A | y A x
lex
y
} has at most one element.

Proof.
Let A be a non-empty convex subset of R
m
, and
M
in
= {x A | y A x
lex
y
}. We now prove that |M

in
| 1.
Towards a contradiction, let us assume M
in
has at least two elements x and y,
x = y. By denition of M
in
, we must have x
= y
.
Let (0, 1) and be a permutation of {1, . . . , m} such that (x+(1 )y)
=
(x+(1)y) = (x)+(1)(y). Let us show by contradiction that (x) = x
and (y) = y
.
Let us assume that either (x) <
lex
x
or (y) <
lex
y
, it follows that
(x) + (1 )(y) <
lex
x
+ (1 )y
= x
.
Since A is convex, x+(1)y A. But this is a contradiction because by denition
of M
in
, x + (1 )y A cannot be strictly smaller than x
, y
in A. This proves
(x) = x
and (y) = y
.
Since x
= y
, we have (x) = (y), hence x = y. This contradicts the fact that

x = y. Hence, M
in
cannot have at least two elements, and |M
in
| 1.
4.3.4. THEOREM. Let (N, v) be a TU game such that Imp = .
(i) Imp is a non-empty and convex subset of R
|N|
(ii) {e(x) | x Imp} is a non-empty convex subset of R
2
|N|
Proof. Let (N, v) be a TUgame such that Imp = (in case Imp = , Imp is trivially
convex). Let (x, y) Imp
2
, [0, 1]. Let us prove Imp is convex by showing that
u = x + (1 )y Imp, i.e., that u is individually rational and efcient.
Individual rationality: Since x and y are individually rational, for all agents i,
u
i
= x
i
+ (1 )y
i
v({i}) + (1 )v({i}) = v({i}). Hence u is individually
rational.
Efciency: Since x and y are efcient, we have
iN
u
i
=
iN
x
i
+ (1 )y
i
=
iN
x
i
+ (1 )
iN
y
i
= v(N) + (1 )v(N).
Hence we have
iN
u
i
= v(N) and u is efcient.
Thus, u Imp.
Let (N, v) be a TU game and Imp its set of imputations. We need to show that the
set {e(z) | z Imp} is a non-empty convex subset of R
m
. Remember that e(z) is the
sequence of excesses of all coalitions for the payoff distribution z in a given order of the
coalitions (i.e., it is a vector of size 2
|N|
). Since Imp is non-empty, {e(z) | z Imp}
is trivially non non-empty. Then we just need to prove it is convex. Let (x, y) Imp
2
,
[0, 1], and C N. We consider the vector e(x) +(1 )e(y) and we look at the
entry corresponding to coalition C.
(e(x) + (1 )e(y))
C
= e(C, x) + (1 )e(C, y)
= (v(C) x(C)) + (1 )(v(C) y(C))
= v(C) (x(C) + (1 )y(C))
= v(C) ([x + (1 )y](C))
= e(x + (1 )y, C)
Since the previous equality is valid for all C N, both sequences are equal and we
can write
e(x) + (1 )e(y) = e(x + (1 )y).
Since Imp is convex, x + (1 )y Imp, it follows that
e(x + (1 )y) {e(z) | z Imp}. Hence, {e(z) | z Imp} is convex.
Now we nally are ready to prove that the nucleolus has at most one element.
Proof. Let (N, v) be a TU game, and Imp its set of imputations.
According to Theorem 4.3.4(ii), we have that {e(x) | x Imp} is a non-empty
convex subset of R
2
|N|
. Applying theorem4.3.3 with A = {e(x) | x Imp} we obtain
the following statement:
B = {e(x) | x Imp y Imp e(x)
lex
e(y)
} has at most one element.

B is the image of the nucleolus under the function e. We need to make sure that an
e(x) corresponds to at most one element in Imp. This is true since for (x, y) Imp
2
,
we have x = y e(x) = e(y).
Hence Nu(N, v) = {x | x Imp y Imp e(x)
lex
e(y)
} has at most
one element!
Bibliography
[1] Xiaotie Deng, Qizhi Fang, and Xiaoxun Sun. Finding nucleolus of ow game.
In SODA 06: Proceedings of the seventeenth annual ACM-SIAM symposium on
Discrete algorithm, pages 124131, New York, NY, USA, 2006. ACM Press.
[2] Jeroen Kuipers, Ulrich Faigle, and Walter Kern. On the computation of the nucle-
olus of a cooperative game. International Journal of Game Theory, 30(1):7998,
2001.
[3] D. Schmeidler. The nucleolus of a characteristic function game. SIAM Journal of
applied mathematics, 17, 1969.
[4] Makoto Yokoo, Vincent Conitzer, Tuomas Sandholm, Naoki Ohta, and Atsushi
Iwasaki. A compact representation scheme for coalitional games in open anony-
mous environments. In Proceedings of the Twenty First National Conference on
Articial Intelligence, pages . AAAI Press AAAI Press / The MIT Press, 2006.
51
Lecture 5
The Kernel
The Kernel is another stability concept that weakens the stability requirements of the
core. It was rst introduced by Davis and Maschler [3]. The denition of the kernel is
based on the excess of a coalition. For the nucleolus, a positive excess was interpreted
as an amount of complaint as by forming a coalition with positive excess, some payoff
was lost. In the kernel, a positive excess is interpreted as a measure of threat: in the
current payoff distribution, if some agents deviate by forming coalition with positive
excess, they are able to increase their payoff by redistributing the excess between them.
When any two agents in a coalition have similar threatening powers, the kernel consid-
ers that the payoff is stable. In the following, we will see two denitions of the kernel
and we will see that it is guaranteed to be non-empty.
5.1 Denition of the Kernel
We recall that the excess related to coalition ( for a payoff distribution x is dened as
e((, x) = v(() x((). We saw that a positive excess can be interpreted as an amount
of complaint for a coalition. We can also interpret the excess as a potential to generate
more utility. Let us consider that the agents are forming a CS o = (
1
, . . . , (
k
, and
let consider that the excess of a coalition ( / o is positive. Agent i ( can view the
positive excess as a measure of his strength: if she leaves its current coalition in o and
forms coalition ( N, she has the power to generate some surplus e((, x). When
two agents want to compare their strength, they can compare the maximum excess of
a coalition that contains them and excludes the other agent, and the kernel is based on
this idea.
5.1.1. DEFINITION. [Maximum surplus] For a TU game (N, v), the maximum surplus
s
k,l
(x) of agent k over agent l with respect to a payoff distribution x is
s
k,l
(x) = max
CN | kC, l / C
e((, x).
53
54 Lecture 5. The Kernel
For two agents k and l, the maximum surplus s
k,l
of agent k over agent l with re-
spect to x is the maximum excess from a coalition that includes k but does exclude l.
This maximum surplus can be used by agent k to show its strength over agent l: as-
suming it is positive and that agent k can claim all of it, she can argue that she will
be better off without agent l; hence she should be compensated with more utility for
staying in the current coalition. When any two agents in a coalition have the same
maximum surplus (except for a special case), the agents are said to be in equilibrium.
A payoff distribution is in the Kernel when all agents are in equilibrium. The formal
denitions follow:
5.1.2. DEFINITION. [kernel] Let (N, v, o) be a TU game with coalition structure. The
kernel is the set of imputations x Jmp(o) such that for every coalition ( o, if
(k, l) (
2
, k ,= l, then we have either s
kl
(x) s
lk
(x) or x
k
= v(k).
First, note that the denition of the kernel is for a particular coalition structure. The
agents do not try to change the structure, but they argue about the payoff distribution.
Another observation is that the kernel is a subset of the set of imputations(we could de-
ne a pre-kernel and show this solution does not satisfy individual rationality). Finally,
the stability condition must be satised by any two agents that are in the same coali-
tion. This condition may appear to surprising at rst as one would expect an equality
between the maximum surpluses of the two agents. The condition s
kl
(x) < s
lk
(x) calls
for a transfer of utility from k to l unless it is prevented by individual rationality, i.e.,
by the fact that x
k
= v(k). Hence, it is possible that agent l has a strictly greater
maximum excess than k when the payoff of k is her minimum payoff v(k).
5.2 Another denition
As for the bargaining set and the nucleolus, the kernel can be dened using objec-
tions and counter objections. For the kernel, objections and counter-objections are
exchanged between two members of the same coalition in o. Objections and counter-
objections take the form of coalitions only (unlike for the the bargaining set and the nu-
cleolus for which a payoff distribution was part of an objection and counter-objection).
Let us consider a game with CS N, v, o and let us consider a coalition ( o
such that both agents k and l are in (.
Objection: A coalition P N is an objection of k against l to x iff k P, l / P and
x
l
> v(l).
P is a coalition that contains k, excludes l and which sacrices too
much (or gains too little).
An objection of agent k against agent l is simply a coalition P that has some
issues according to agent k. The only constraint is that agent l is not a member of
5.3. Properties of the kernel 55
that coalition. A counter-objection of agent l is then a coalition Q that excludes
agent k which has a greater (or equal) excess than e(P, x).
Counter-objection: A coalition Q N is a counter-objection to the objection P of k
against l at x iff l Q, k / Q and e(Q, x) e(P, x).
ks demand is not justied: Q is a coalition that contains l and ex-
cludes k and that sacrices even more (or gains even less).
Note that if the inequality is strict, we can view Q as a new objection against agent
l. Remember that the set of feasible payoff vectors for (N, v, o) is X
(N,v,S)
= x
R
n
[ for every ( o : x(() v((). We are now ready to dene the kernel.
5.2.1. DEFINITION. [Kernel] Let (N, v, o) be a TU game in coalition structure. The
kernel is the set of imputations x X
(N,v,S)
s.t. for any coalition ( o, for each
objection P of an agent k ( over any other member l ( to x, there is a counter-
objection of l to P.
This denition using objections and counter-objections is very intuitive compared to
the rst denition that was presented. It should be clear that both denitions coincide.
5.3 Properties of the kernel
The Kernel and the nucleolus are linked: the following result shows that the nucleolus
is included in the kernel. As a consequence, this guarantees that the Kernel is non-
empty. The second part of the result is that the kernel is included in the bargaining set.
As a consequence, the nucleolus is also included in the bargaining set and the bargain-
ing set is non-empty, a result we did not prove during the lecture on the bargaining set.
A nal observation is that the kernel can be seen as a renement of the bargaining set.
5.3.1. THEOREM. Let (N, v, o) a game with coalition structure, and let Jmp ,= .
Then we have:
(i) Nu(N, v, o) K(N, v, o)
(ii) K(N, v, o) BS(N, v, o)
Proof.
Let us start by proving (i).
Let x / K(N, v, o), we want to show that x / Nu(N, v, o).
Since x / K(N, v, o), there exists a coalition ( CS and two members k and l of
coalition ( such that s
lk
(x) > s
kl
(x) and x
k
> v(k). Let y be a payoff distribution
corresponding to a transfer of utility > 0 from k to l:
y
i
=
x
i
if i ,= k and i ,= l
x
k
if i = k
x
l
+ if i = l
Since x
k
> v(k) and s
lk
(x) > s
kl
(x), we can choose > 0 small enough such that
x
k
> v(k)
s
lk
(y) > s
kl
(y)
We want to show that e(y)
lex
e(x)
. Note that for any coalition S N such

that e(S, x) ,= e(S, y) we have either:
k S and l / S (e(S, x) > e(S, y) since e(S, y) = e(S, x) + > e(S, x))
k / S and l S (e(S, x) < e(S, y) since e(S, y) = e(S, x) < e(S, x))
Let B
1
(x), . . . , B
M
(x) a partition of the set of all coalitions such that
(S, T) B
i
(x) iff e(S, x) = e(T, x). We denote by e
i
(x) the common value of
the excess in B
i
(x), i.e. e
i
(x) = e(S, x) for all S B
i
(x).
e
1
(x) > e
2
(x) > . . . > e
M
(x)
In other words, e(x)
= e
1
(x), . . . , e
1
(x)

|B
1
(x)|times
, . . . , e
M
(x), . . . , e
M
(x)

|B
M
(x)|times
.
Let i
be the minimal value of i 1, . . . , M such that there is a coalition (

B
i
(x) with e((, x) ,= e((, y). For all i < i
, we have B
i
(x) = B
i
(y) and e
i
(x) =
e
i
(y). Since s
lk
(x) > s
kl
(x), B
i
contains
at least one coalition S that contains l but not k, for such coalition, we must have
e(S, x) > e(S, y)
no coalition that contains k but not l.
If B
i
(x) contains either
coalitions that contain both k and l
or coalitions that do not contain both k and l
Then, for any such coalitions S, we have e(S, x) = e(S, y), and it follows that B
i
(y)
B
i
(x).
Otherwise, we have e
i
(y) < e
i
(x).
In both cases, we have e(y) is lexicographically less than e(x), and hence y is not
in the nucleolus of the game (N, v, o).
We now turn to proving (ii).
Let (N, v, o) a TU game with coalition structure. Let x K(N, v, o). We want to
prove that x BS(N, v, o). To do so, we need to show that for any objection (P, y)
from any player i against any player j at x, there is a counter objection (Q, z) to (P, y).
For the bargaining set, An objection of i against j is a pair (P, y) where
5.3. Properties of the kernel 57
P N is a coalition such that i P and j / P.
y R
p
where p is the size of P
y(P) v(P) (y is a feasible payoff for members of P)
k P, y
k
x
k
and y
i
> x
i
An counter-objection to (P, y) is a pair (Q, z) where
Q N is a coalition such that j Q and i / Q.
z R
q
where q is the size of Q
z(Q) v(Q) (z is a feasible payoff for members of Q)
k Q, z
k
x
k
k Q P z
k
y
k
Let (P, y) be an objection of player i against player j to x. i P, j / P, y(P)
v(P) and y(P) > x(P). We choose y(P) = v(P).
x
j
= v(j): Then (j, v(j)) is a counter objection to (P, y).
x
j
> v(j): Since x K(N, v, o) we have s
ji
(x) s
ij
(x) v(P) x(P)
y(P) x(P) since i P, j / P.
Let Q N such that j Q, i / Q and s
ji
(x) = v(Q) x(Q).
We have v(Q) x(Q) y(P) x(P). Then, we have
v(Q) y(P) + x(Q) x(P)
y(P Q) + y(P Q) + x(Q P) x(P Q)
> y(P Q) + x(Q P) since i P Q, y(P Q) > x(P Q)
Let us dene z as follows
x
k
if k Q P
y
k
if k Q P
(Q, z) is a counter-objection to (P, y).
Finally x BS(N, v, o).
These properties allow us to conclude that when the set of imputation is non-empty,
both the kernel and the bargaining set are non-empty.
5.3.2. THEOREM. Let N, v, o a game with coalition structure such that the set of
imputations Jmp ,= . The kernel K(N, v, o) and the bargaining set BS(N, v, o) are
non-empty.
5.4 Computational Issues
One method for computing the Kernel is the Stearns method [7]. The idea is to build
a sequence of side-payments between agents to decrease the difference of surpluses
between the agents. At each step of the sequence, the agents with the largest maxi-
mum surplus difference exchange utility so as to decrease their surplus: the agent with
smaller surplus makes a payment to an agent with higher surplus so as to decrease their
surplus difference. After each side-payment, the maximum surplus over all agents de-
creases. In the limit, the process converges to an element in the Kernel. Computing an
element in the Kernel may require an innite number of steps as the side payments can
become arbitrarily small, and the use of the -Kernel can alleviate this issue. A crite-
ria to terminate Stearns method is proposed in [6], and we present the corresponding
algorithm in Algorithm 1.
Algorithm 1: Transfer scheme to converge to a -Kernel-stable payoff distribu-
tion for the CS S
compute--Kernel(, S)
repeat
for each coalition ( S do
for each member i ( do
for each member j (, j ,= i, do // compute the surplus
for two members of a coalition in S
s
ij
max
RC|(iR, j / R)
v(R) x(R)
max
(i,j)N
2 [s
ij
s
ji
[;
(i
, j
) argmax
(i,j)N
2 s
ij
s
ji
;
if

x
j
v(j) <

2
then // payment should be individually

rational
d x
j
v(j
);
else
d

2
;
x
i
x
i
+ d;
x
j
x
j
d;
until

v(S)
;
Computing a Kernel distribution is of exponential complexity. In Algorithm 1,
computing the surpluses is expensive, as we need to search through all coalitions
that contains a particular agent and does not contain another agent. Note that when
a side-payment is performed, it is necessary to recompute the maximum surpluses.
The derivation of the complexity of the Stearns method to compute a payoff in the
-Kernel can be found in [4, 6], and the complexity for one side-payment is O(n 2
n
).
Of course, the number of side-payments depends on the precision and on the initial
5.5. Fuzzy Kernel 59
payoff distribution. They derive an upper bound for the number of iterations: con-
verging to an element of the -Kernel requires nlog
2
(

0
v(S)
), where
0
is the maximum
surplus difference in the initial payoff distribution. To derive a polynomial algorithm,
the number of coalitions must be bounded. The solution used in [4, 6] is to only con-
sider coalitions whose size is bounded in the interval K
1
, K
2
. The complexity of the
truncated algorithm is O(n
2
n
coalitions
) where n
coalitions
is the number of coalitions
with a size between K
1
and K
2
, which is a polynomial of order K
2
.
5.5 Fuzzy Kernel
In order to take into account the uncertainty in the knowledge of the utility function, a
fuzzy version of stability concept can be used. Blankenburg et al. consider a coalition
to be Kernel-stable with a degree of certainty [2]. This work also presents a side-
payment scheme and shows that the complexity is similar to the crisp Kernel, and the
idea from [4] can be used for ensuring a polynomial coalition formation algorithm.
This approach assumes a linear relationship of the membership and coalition values.
Fuzzy coalitions can also allow agents to be members of multiple coalitions at
the same time, with possibly different degrees of involvement [1]. It can be mutually
benecial for an agent to be in two different coalitions. It may be benecial for the
agent to be in both coalition. In addition, the two coalitions may need the competence
of the same agent, though the coalitions do not have any incentive to merge as they may
not have anything to do with each other. This solution may allow to form coalitions
that involve only agents that need to work together. In the previous example, without
the possibility of being member of multiple coalitions, the two coalitions should merge
to benet from the agent participation, and agents that would not need to be in the same
coalition are forced to be in the same coalition. In [1], the degree of involvement of an
agent in different coalition is a function of the risk involved in being in that coalition.
The risk is quantied using nancial risk measures. This work presents a denition
of the Kernel based on partial membership in coalitions and introduces a coalition
formation protocol that runs in polynomial time.
Bibliography
[1] Bastian Blankenburg, Minghua He, Matthias Klusch, and Nicholas R. Jennings.
Risk-bounded formation of fuzzy coalitions among service agents. In Proceedings
of 10th International Workshop on Cooperative Information Agents, 2006.
[2] Bastian Blankenburg, Matthias Klusch, and Onn Shehory. Fuzzy kernel-stable
coalitions between rational agents. In Proceedings of the second international joint
conference on Autonomous agents and multiagent systems (AAMAS-03). ACM
Press, 2003.
[3] M. Davis and M. Maschler. The kernel of a cooperative game. Naval Research
Logistics Quarterly, 12, 1965.
[4] Matthias Klusch and Onn Shehory. A polynomial kernel-oriented coalition algo-
rithm for rational information agents. In Proceedings of the Second International
Conference on Multi-Agent Systems, pages 157 164. AAAI Press, December
1996.
Press, 1994.
[6] Onn Shehory and Sarit Kraus. Feasible formation of coalitions among autonomous
agents in nonsuperadditve environments. Computational Intelligence, 15:218
251, 1999.
[7] Richard Edwin Stearns. Convergent transfer schemes for n-person games. Trans-
actions of the American Mathematical Society, 134(3):449459, December 1968.
61
Lecture 6
The Shapley value
Thus far, we have focused on solution concepts that provide some incentives for staying
in a coalition structure. The solution concept we are about to study focuses on howwell
each agents payoff reects her contribution, i.e., a notion of fairness. This solution
concept was introduced by L.S Shapley [5] and we are going to consider two different
denitions. One is an axiomatic approach, and the other describes the idea that a player
should receive a payoff that is proportional to her contribution.
6.1 Taking the average of the contribution
A simple idea to design a payoff distribution is that an agent should receive a payoff
that is proportional to her contribution in a coalition. We recall that the marginal contri-
bution of an agent i to a coalition ( N is mc
i
(() = v((i)v((). Let us consider
that a coalition ( is built incrementally with one agent at a time entering the coalition.
Also consider that the payoff of each agent i is its marginal contribution
1
. For example,
mc
1
(), mc
2
(1), mc
3
(1, 2) is a payoff distribution for a game 1, 2, 3, v). This
payoff is efcient and it is also reasonable (from above and from below). However,
the payoff of each agent depends on the order in which the agents enter the coalition.
Let us assume that the game is convex. An agent that joins later would then have an
advantage over an agent that join the coalition earlier. As a more concrete example,
consider agents that form a coalition to take advantages of price reduction when buying
large quantities of a product. Agents that start the coalition may have to spend large
setup cost, and agents that come later benets from the already large number of agents
in the coalition. This may not be fair!
To alleviate this issue, the Shapley value averages each agents payoff over all
possible orderings: the value of agent i in coalition ( is the average marginal value
over all possible orders in which the agents may join the coalition.
1
We used this payoff distribution in the proof showing that a convex game has a non-empty core
(Theorem 2.2.3).
63
64 Lecture 6. The Shapley value
Let represent a joining order of the grand coalition N: can also be viewed as
a permutation of 1, . . . , n. We write mc() the payoff vector where agent i obtains
mc
i
((j) [ j < i). The payoff vector mc() is called the marginal vector. Let us
denote the set of all permutations of the sequence 1, . . . , n as (N). The Shapley
values can then be dened as
Sh(N, v) =
(N)
mc()
n!
.
N = 1, 2, 3
v(1) = 0 v(2) = 0 v(3) = 0
v(1, 2) = 90 v(1, 3) = 80 v(2, 3) = 70
v(1, 2, 3) = 120
1 2 3
1 2 3 0 90 30
1 3 2 0 40 80
2 1 3 90 0 30
2 3 1 50 0 70
3 1 2 80 40 0
3 2 1 50 70 0
total 270 240 210
Shapley value 45 40 35
Let y = 50, 40, 30
( e((, ) e((, y)
1 -45 0
2 -40 0
3 -35 0
1, 2 5 0
1, 3 0 0
2, 3 -5 0
1, 2, 3 0 0
This example shows that the Shapley value may not be in the core, and may not be the
nucleolus.
Table 6.1: Example of a computation of Shapley value
We provide an example in Table 6.1 in which we list all the orders in which the
agents can enter the grand coalition. The sum is over all joining orders, which may
contains a very large number of terms. However, when computing the Shapley value
for one agent, one can avoid some redundancy by summing over all coalitions and
noticing that:
There are [([! permutations in which all members of ( precede i.
There are [N((i)[! permutations in which the remaining members succede
i, i.e. (n [([ 1)!.
These observations allow us to rewrite the Shapley value as:
Sh
i
(N, v) =
CN\{i}
[([!(n [([ 1)!
n!
(v(( i) v(()) .
Note that the example from Table 6.1 also demonstrates that in general the Shapley
value is not in the core or in the nucleolus.
6.2. An Axiomatic Characterisation 65
6.2 An Axiomatic Characterisation
In the rst lecture, we provided some concepts some desirable concepts that can be
satised by a payoff distribution. The core is dened by using two of these concepts:
efciency and group rationality. In the following, we are going to use some proper-
ties, called axioms in this context, to dene a payoff distribution that guarantees some
elements of fairness.
Instead of considering properties of a payoff distribution, we will state the proper-
ties of a function that maps a game to a payoff distribution.
6.2.1. DEFINITION. [value function] Let (
N
the set of all valuation functions 2
N
R
that maps a coalition of N to a real number. A value function : N (
N
R
n
assigns an efcient allocation x to a TU game (N, v) (i.e., an allocation that satises
iN
(N, v)
i
= v(N)).
We have already seen one value function for games that have at least one imputa-
tion: the nucleolus provides a single payoff distribution for such game.
We now consider some axioms that may be desirable for such a value function. We
consider a TU game (N, v) and we simply note (N, v) as . The rst axiom uses the
denition of a dummy agent. Intuitively, there is no synergy between a dummy agent
and any other agent, in other words the marginal contribution of a dummy agent i to
any coalition is v(i).
DUM (Dummy actions) : if agent i is a dummy then
i
= v(i). In other words,
if the presence of agent i improves the value of any coalition by exactly v(i),
this agent should obtain precisely v(i).
SYM (Symmetry) : When two agents generate the same marginal contributions,
they should be rewarded equally: for i ,= j and ( N such that i / ( and
j / S, if v(( i) = v(( j), then
i
=
j
.
ADD (Additivity) : For any two TUgames (N, v) and (N, w) and their correspond-
ing value functions (N, v) and (N, w), the value function for the TU game
(N, v +w) is (N, v +w) = (N, v) +(N, w).
These three axioms appear as good properties that should be satised by a payoff
distribution. Dummy states that an agent that does not have any synergy with any
other agent should get the value of its singleton coalition. SYM provides an element
of fairness: two agents that are substitutes should get the same payoff. The last axiom
demands a value function to be additive. Shapley [5] showed that actually, there is a
unique value that satises this three axioms.
6.2.2. THEOREM. The Shapley value is the unique value that satises axioms DUM,
SYM and ADD.
These axioms are independent. One can prove that if one of the three axioms is
dropped, it is possible to nd multiple value functions satisfying the other two axioms.
To prove this results, one needs to show the existence of a value function that satises
the three axioms, and then prove the unicity of the value function. We will follow the
proof from [4]. We will need the following type of games called in the proof.
6.2.3. DEFINITION. [Unanimity Game] Let N be a set of agents and T N .
The unanimity game (N, v
T
) is the game such that ( N, v
T
(() =
1, if T (,
0 otherwise.
One can note that if i N T, i is a null player. In addition, if (i, j) T
2
, i and j
are substitutes. The following lemma will be useful for the proof.
6.2.4. LEMMA. The set v
T
(
N
[ T N is a linear basis of (
N
.
The lemma means that a TU game (N, v) can be represented by a unique set of
values (
T
)
TN\
such that ( N, v(() =
TN\
T
v
T
(().
Proof. There are 2
n
1 unanimity games and the dimension of (
N
is also 2
n
1.
We only need to prove that the unanimity games are linearly independent. Towards a
contradiction, let us assume that
TN\
T
v
T
= 0 where (
T
)
TN\
,= 0
R
2
n
1. Let
T
0
be a minimal set in T N [
T
,= 0. Then,
TN\
T
v
T
(T
0
) =
T
0
,= 0,
which is a contradiction.
We can now turn to proving that the Shapley value is the unique value that satises
SYM, DUM and ADD.
Proof. We start by proving the uniqueness. Let a feasible solution on (
N
that is
non-empty and satises the axioms SYM, DUM and ADD. Let us prove that is a
value function. Let (N, v) (
N
.
if v = 0
G
N
, all players are dummy. Since the solution is non-empty, 0
R
|N|
is the
unique member of (N, v).
otherwise, (N, v) (
N
. Let x = (N, v) and y = (N, v). By ADD,
x + y = (v v), and then since v v = 0
G
, we have that x = y is unique.
Moreover, x(N) v(N) as is a feasible solution. Also y(N) v(N). Since
x = y, we have v(N) x(N) v(N), i.e. x is efcient.
Hence, is a value function. We now show that the value function of a unanimity
game is unique. Let T N and R. Let us prove that (N, v
T
) is uniquely
dened.
Let i / T. We have trivially T ( iff T ( i. Then ( N i,
v
T
(() = v
T
(( i). Hence, all agent i / T are dummies. By DUM,
i / T,
i
(N, v
T
) = 0.
6.2. An Axiomatic Characterisation 67
Let (i, j) T
2
. Then for all ( N i, j, v(( i) = v(( j). By SYM,
i
(N, v
T
) =
j
(N, v
T
).
Since is a value function, it is efcient. Then,
iN

i
(N, v
T
) = v
T
(N) =
.
Hence, i T,
i
(N, v
T
) =

|T|
.
This proves that (N, v
T
) is uniquely dened. Since any TU game (N, v) can be
written as

TN\
T
v
T
and because of ADD, there is a unique value function that
satises the three axioms.
We now turn to show the existence of the Shapley value, i.e. we need to show
that there exist a value function that satises the three axioms. Let (N, v) a TU game
and we consider the Shapley value as dened in the previous section. We are going to
check whether this function satises the three axioms.
SYM Let i and j be substitutes, i.e., ( Ni, j, we have v((i) = v((j).
Then ( N i, j, we have
mc
i
(() = mc
j
(()
v(( i, j) v(( i) = v(( i, j) v(( j), hence, we have
mc
j
(( j) = mc
i
(( i).
Sh
i
(N, v) = Sh
j
(N, v), Sh satises SYM.
DUM Let i be a dummy agent, i.e., for all coalition ( N i, we have v(() +
v(i) = v(( i). Then, each marginal contribution of player i is v(i), and
it follows that Sh
i
(N, v) = v(i). Sh satises DUM.
ADD Sh is clearly additive.
We proved that a unique function satises all three axioms at the same time and we
found one function that satises them all, hence the Shapley value is well dened.
The axioms SYM and DUM are clearly desirable. The last axiom, ADD, is harder
to motivate in some cases. If the valuation function of a TU game is interpreted as an
expected payoff, then ADD is desirable (as you want to be able to add the value of
different states of the world). Also, if we consider cost-sharing games and that a TU
game corresponds to sharing the cost of one service, then ADD is desirable as the cost
for a joint-service should be the sum of the cost of the separate services. However, if
we do not make any assumptions about the games (N, v) and (N, w), the axiomimplies
that there is no interaction between the two games. In addition, the game (N, v + w)
may induce a behavior that may be unrelated to the behavior induced by either (N, v)
or (N, w), and in this case ADD can be questioned.
Other axiomatisations that do not use the ADD axiom have been proposed by
Young in [7] and by Myerson in [3].
6.2.5. DEFINITION. [Marginal contribution axiom] Let (N, v) and (N, v) be two TU
games. A value function satises the marginal contribution axiom iff for all i N,
if for all ( N i v(( i) v(C) = u(( i) u(C), then (u) = (v).
In other words, the value of a player depends only on its marginal contribution.
The following axiomatisation is due to Young [7].
6.2.6. THEOREM. The Shapley value is the unique value function that satises sym-
metry and marginal contribution axioms.
We refer by v i the TU game (N i, v
\i
) where v
\i
is the restriction of v to
N i.
6.2.7. DEFINITION. [Balanced contribution axiom] A value function satises the
balanced contribution axiom iff for all (i, j) N
2
i
(v)
i
(vj) =
j
(v)
j
(vi).
For any two agents, the amount that each agent would win or lose if the other
leaves the game should be the same. The following axiomatisation is due to Myer-
son [3].
6.2.8. THEOREM. The Shapley value is the unique value function that satises the
balanced contribution axiom.
6.3 Other properties
At noted before, the Shapley value always exists and is unique.
6.3.1. THEOREM. For superadditive games, the Shapley value is an imputation.
Proof. Let (N, v) be a superadditive TU game.
By superadditivity, i N, ( N i v(( i) v(() > v(i). Hence, for
each marginal vector, an agent i gets at least v(i). The same is true for the Shapley
value as it is the average over all marginal vectors.
6.3.2. THEOREM. For convex game, the Shapley value is in the core.
Proof. Let (N, v) be a convex game. We know that all marginal vectors are in the core
(to show that convex games have non-empty core, we used one marginal vector and showed
it was in the core). The core is a convex set. The average of a nite set of points in a
convex set is also in the set. Finally, the Shapley value is in the core.
When the valuation function is superadditive, the Shapley value is individually
rational, i.e., it is an imputation. When the Core is non-empty, the Shapley value may
not be in the Core. However, when the valuation function is convex, the Shapley value
is also group rational, hence, it is in the Core.
6.4. Computational Issues 69
6.4 Computational Issues
The nature of the Shapley value is combinatorial, as all possible orderings to form a
coalition needs to be considered. This computational complexity can sometimes be
an advantage as agents cannot benet from manipulation. For example, it is AT-
complete to determine whether an agent can benet from false names [6]. Neverthe-
less, some representations allow to compute the Shapley value efciently. We will
surveying few representations in a coming lecture. For now, we just concentrate on a
simple proposal.
In order to reduce the combinatorial complexity of the computation of the Shap-
ley value, Ketchpel introduces the Bilateral Shapley Value (BSV ) [1]. The idea is to
consider the formation of a coalition as a succession of merging between two coali-
tions. Two disjoint coalitions (
1
and (
2
with (
1
(
2
= , may merge when v((
1

(
2
) v((
1
) + v((
2
). When they merge, the two coalitions, called founders of the
new coalition (
1
(
2
, share the marginal utility as follows: BSV ((
1
) =
1
2
v((
1
) +
1
2
(v((
1
(
2
) v((
2
)) and BSV ((
2
) =
1
2
v((
2
) +
1
2
(v((
1
(
2
) v((
1
)). This is
the expression of the Shapley value in the case of an environment with two agents.
In (
1
(
2
, each of the founders gets half of its local contribution, and half of the
marginal utility of the other founder. Given this distribution of the marginal utility, it is
rational for (
1
and (
2
to merge if i 1, 2, v((
i
) BSV (C
i
). Note that symmetric
founders get equal payoff, i.e., for (
1
, (
2
, ( such that (
1
(
2
= (
1
( = (
2
( = ,
v(( (
1
) = v(( (
2
) BSV (( (
1
) = BSV (( (
2
). Given a sequence of succes-
sive merges from the states where each agent is in a singleton coalition, we can use a
backward induction to compute a stable payoff distribution [2]. Though the computa-
tion of the Shapley value requires looking at all of the permutations, the value obtained
by using backtracking and the BSV only focuses on a particular set of permutations,
but the computation is signicantly cheaper.
Bibliography
[1] Steven P. Ketchpel. The formation of coalitions among self-interested agents. In
Proceedings of the Eleventh National Conference on Articial Intelligence, pages
414419, August 1994.
[2] Matthias Klusch and Onn Shehory. Coalition formation among rational informa-
tion agents. In Rudy van Hoe, editor, Seventh European Workshop on Modelling
Autonomous Agents in a Multi-Agent World, Eindhoven, The Netherlands, 1996.
[3] Roger B. Myerson. Graphs and cooperation in games. Mathematics of Operations
Research, 2:225229, 1977.
Press, 1994.
[5] L.S. Shapley. A value for n-person games. In H. Kuhn and A.W. Tucker, edi-
tors, Contributions to the Theory of Games, volume 2. Princeton University Press,
Princeton, NJ, 1953.
Iwasaki. Coalitional games in open anonymous environments. In Proceedings
of the Twentieth National Conference on Articial Intelligence, pages 509515.
AAAI Press AAAI Press / The MIT Press, 2005.
[7] H. P. Young. Monotonic solutions of cooperative games. International Journal of
Game Theory, 14:6572, 1985.
71
Lecture 7
A Special Class of TU games: Voting Games
The formation of coalitions is usual in parliaments or assemblies. It is therefore in-
teresting to consider a particular class of coalitional games that models voting in an
assembly. For example, we can represent an election between two candidates as a vot-
ing game where the winning coalitions are the coalitions of size at least equal to the
half the number of voters.
7.1 Denitions
We start by providing the denition of a voting game, which can be viewed as a special
class of TU games. Then, we will formalize some known concepts used in voting. We
will see how we can dene what a dictator is,
7.1.1. DEFINITION. [voting game] A game (N, v) is a voting game when
the valuation function takes only two values: 1 for the winning coalitions, 0
otherwise.
v satises unanimity: v(N) = 1
v satises monotonicity: S T N v(S) v(T).
Unanimity and monotonicity are natural assumptions in most cases. Unanimity
reects the fact that all agents agree; hence, the coalition should be winning. Mono-
tonicity tells that the addition of agents in the coalition cannot turn a winning coalition
into a losing one, which is reasonable for voting: more supporters should not harm
the coalition. A rst way to represent a voting game is by listing all winning coali-
tions. Using the monotonicity property, a more succinct representation is to list only
the minimal winning coalitions.
7.1.2. DEFINITION. [Minimal winning coalition] A coalition C N is a minimal
winning coalition iff v(C) = 1 and i C v(C \ {i}) = 0.
73
74 Lecture 7. A Special Class of TU games: Voting Games
For example, we consider the game ({1, 2, 3, 4}, v) such that v(C) = 1 when
|C| 3 or (|C| = 2 and 1 C) and v(C) = 0 otherwise. The set of winning coali-
tions is {{1, 2}, {1, 3}, {1, 4}, {1, 2, 3}, {1, 2, 4}, {1, 3, 4}, {2, 3, 4}, {1, 2, 3, 4}}. We
can represent the game more succinctly by just writing the set of minimal winning
coalitions, which is {{1, 2}, {1, 3}, {1, 4}, {2, 3, 4}}.
We can now see how we formalize some common terms in voting. We can rst
express what a dictator is.
7.1.3. DEFINITION. [Dictator] Let (N, v) be a simple game. A player i N is a
dictator iff {i} is a winning coalition.
Note that with the requirements of simple games, it is possible to have more than
one dictator! The next notion is the notion of veto player, in which a player can block
a decision on its own by opposing to it (e.g. in the United Nations Security Council,
China, France, Russia, the United Kingdom, and the United States are veto players).
7.1.4. DEFINITION. [Veto Player] Let (N, v) be a simple game. A player i N is a
veto player if N \ {i} is a losing coalition. Alternatively, i is a veto player iff for all
winning coalition C, i C.
It also follows that a veto player is member of every minimal winning coalitions.
Another concept is the concept of a blocking coalition: it is a coalition that, on its own,
cannot win, but the support of all its members is required to win. Put another way, the
members of a blocking coalition do not have the power to win, but they have the power
to lose.
7.1.5. DEFINITION. [blocking coalition] A coalition C N is a blocking coalition iff
C is a losing coalition and S N \ C, S \ C is a losing coalition.
7.2 Stability
We can start by studying what it means to have a stable payoff distribution in these
games. The following theorem characterizes the core of simple games.
7.2.1. THEOREM. Let (N, v) be a simple game. Then
Core(N, v) =
x R
n
x is an imputation x
i
= 0 for each non-veto player i
Proof.
Let x Core(N, v). By denition x(N) = 1. Let i be a non-veto player.
x(N \ {i}) v(N \ {i}) = 1. Hence x(N \ {i}) = 1 and x
i
= 0.
7.2. Stability 75
Let x be an imputation and x
i
= 0 for every non-veto player i. Since x(N) = 1,
the set V of veto players is non-empty and x(V ) = 1.
Let C N. If C is a winning coalition then V C, hence x(C) v(C).
Otherwise, v(C) is a losing coalition (which may contain veto players), and
x(C) v(C). Hence, x is group rational.
We can also study the class of simple convex games. The following theorem shows
that they are the games with a single minimal winning coalition.
7.2.2. THEOREM. A simple game (N, v) is convex iff it is a unanimity game (N, v
V
)
where V is the set of veto players.
Proof. A game is convex iff S, T N v(S) +v(T) v(S T) +v(S T).
Let us assume (N, v) is convex.
If S and T are winning coalitions, S T is a winning coalition by monotonicity.
Then, we have 2 1 + v(S T) and it follows that v(S T) = 1. The
intersection of two winning coalitions is a winning coalition. Moreover, from
the denition of veto players, the intersection of all winning coalitions is the set
V of veto players. Hence, v(V ) = 1. By monotonicity, if V C, v(C) = 1.
Otherwise, V C. Then there must be a veto player i / C, and it must be the
case that v(C) = 0. Hence, for all coalition C N, v(C) = 1 iff V C.
Let (N, v
V
) a unanimity game. Let us prove it is a convex game. Let S N and
T N, and we want to prove that v(S) +v(T) v(S T) +v(S T).
case V S T: Then V S and V T, and we have 2 2
case V S T V S T:
if V S then V T and 1 1
if V T then V S and 1 1
otherwise V S and V T, and then 0 1
case V S T: then 0 0
For all cases, v(S) + v(T) v(S T) + v(S T), hence a unanimity game is
convex. In addition, all members of V are veto players.

7.3 Weighted voting games
We now dene a class of voting games that has a more succinct representation: each
agent has a weight and a coalition needs to achieve a threshold (i.e. a quota) to be
winning. This is a much more compact representation as we only use to dene a vector
of weights and a threshold. The formal denition follows.
7.3.1. DEFINITION. [weighted voting game] A game (N, v, q, w) is a weighted voting
game when
w = (w
1
, w
2
. . . , w
n
) R
n
+
is a vector of weights, one for each voter
A coalition C is winning (i.e., (v(C) = 1) iff
iC
w
i
q, it is losing otherwise
(i.e., (v(C) = 0)
v satises monotonicity:
iN
w
i
q
The fact that each agent has a positive (or zero) weight ensures that the game is
monotone. We will note a weighted voting game (N, w
iN
, q) as [q; w
1
, . . . , w
n
]. In its
early days, the European Union was using a weighted voted games. Nowa combination
of weighted voting games are used (a decision is accepted when it is supported by 55%
of Member States, including at least fteen of them, representing at the same time at
least 65% of the Unions population).
Weighted voting games is a succinct representation of a simple game. However, not
all the simple games can be represented by a weighted voting game. We say that the
representation is not complete. For example, consider the voting game ({1, 2, 3, 4}, v)
such that the set of minimal winning coalitions is {{1, 2}, {3, 4}}. Let us assume we
can represent (N, v) with a weighted voting game [q; w
1
, w
2
, w
3
, w
4
]. We can form
the following inequalities:
v({1, 2}) = 1 then w
1
+w
2
q
v({3, 4}) = 1 then w
3
+w
4
q
v({1, 3}) = 0 then w
1
+w
3
< q
v({2, 4}) = 0 then w
2
+w
4
< q
But then, w
1
+w
2
+w
3
+w
4
< 2q and w
1
+w
2
+w
3
+w
4
2q, which is impossible.
Hence, (N, v) cannot be represented by a weighted voting game.
Not all simple games can be represented by a weighted voting game. However,
many weighted voting games represent the same simple game: two weigthed vot-
ing games may have different quotas and weights, but they may have exactly the
same winning coalitions. Two weighted voting games G = [q, w
1
, ..., w
n
] and G
=
[q
, w
1
, ..., w
n
] are said to be equivalent when C N, w(C) q iff w
(C) q
.
The denition of weighted voting games allows to choose the weights and the quota
as a real number. From a computational point of view, storing and manipulating real
number is challenging. However, one do not need to use real numbers. The following
7.4. Power Indices 77
result shows that any weighted voting game is equivalent to a weighted voting game
with small integer weights and quota.
7.3.2. THEOREM. For any weighted voting game G, there exists an equivalent weighted
voting game [q, w
1
, ..., w
n
] with q N and i N w
i
N and w
i
= O(2
nlogn
).
Without loss of generality, we can now study weighted voting games with only
integer weights and integer quota, which allows us to represent a weighted voting game
with a polynomial number of bits.
We now turn to the question about the meaning of the weight. One intuition may
be that the weight represents the importance or the strength of a player. Let us consider
some examples to check this intuition.
[10; 7, 4, 3, 3, 1]: The set of minimal winning coalitions is {{1, 2}{1, 3}{1, 4}{2, 3, 4}}.
Player 5, although it has some weight, is a dummy. Player 2 has a higher weight
than player 3 and 4, but it is clear that player 2, 3 and 4 have the same inuence.
[51; 49, 49, 2]: The set of winning coalition is {{1, 2}, {1, 3}, {2, 3}}. It seems
that the players have symmetric roles, but it is not reected in their weights.
These examples shows that the weights can be deceptive and may not represent the
voting power of a player. Hence, we need different tools to measure the voting power
of the voters, which is the goal of the following section.
7.4 Power Indices
The examples raise the subject of measuring the voting power of the agents in a vot-
ing game. Multiple indices have been proposed to answer these questions. In the
following, we introduce few of them, and we will discuss some weaknesses (some
paradoxical situations may occur). Finally, we briey describe some applications.
7.4.1 Denitions
One central notion to dene the power of a voter is the notion of being a Swing or
Pivotal Voter. Informally, when a coalition C is losing, a pivotal voter for that coalition
is a voter that makes the coalition C {i} win. The presence of the members of C is
not sufcient to win the election, but with the presence of i, C {i} wins and i can be
seen as an important voter.
7.4.1. DEFINITION. [Swing or Pivotal Voter] A voter i is pivotal or swing for a coali-
tion C when i turns the coalition from a losing to a wining one, i.e., v(C) = 0 and
v(C {i}) = 1.
In the following, w is the number of winning coalitions and for a voter i,
i
is the
number of coalitions for which i is pivotal, i.e.,
i
=
SN\{i}
v(S {i}) v(S). We
are now ready to dene some power indices.
Shapley-Shubik index: it is the Shapley value of the voting game, its interpretation
in this context is the percentage of the permutations of all players in which i is
pivotal.
I
SS
(N, v, i) =
CN\{i}
|C|!(n |C| 1)!
n!
(v(C {i}) v(C)) .
For each permutation, the pivotal player gets one more point.. One issue is
that the voters do not trade the value of the coalition, though the decision that the
voters vote about is likely to affect the entire population.
Banzhaff index: For each coalition, we determine which agent is a swing agent (more
than one agent may be pivotal). The raw Banzhaff index of a player i is
i
=
CN\{i}
v(C {i}) v(C)
2
n1
.
The interpretation is that the Banzhaff index is the percentage of coalitions for
which a player is pivotal. The raw Banzhaff index does not necessarily sum up
to one. However, for a simple game (N, v), v(N) = 1 and v() = 0, at least one
player i has a power index
i
= 0. Hence, B =
jN

j
> 0. The normalized
Banzhaff index of player i for a simple game (N, v) is dened as
I
B
(N, v, i) =

i
B
.
Coleman index: Coleman denes three indices [1]: the power of the collectivity to act
A =
w
2
n
(A is the probability of a winning vote occurring); the power to prevent
action P
i
=

i
w
(it is the ability of a voter to change the outcome from winning
to losing by changing its vote); the power to initiate action I
i
=

i
2
n
w
(it is the
ability of a voter to change the outcome from losing to winning by changing its
vote, the numerator is the same as in P, but the denominator is the number of
losing coalitions, i.e., the complement of the one of P)
We provide in Table 7.1 an example of computation of the Shapley-Schubik and
Banzhaff indices. This example shows that both indices may be different. There is
a slight difference in the probability model between the Banzhaf
i
and Colemans
index P
i
: in Banzhafs, all the voters but i vote randomly whereas in Colemans, the
assumption of random voting also applies to the voter i. Hence, the Banzhaf index can
be written as
i
= 2P
i
A = 2I
i
(1 A).
7.4. Power Indices 79
{1, 2, 3, 4} {3, 1, 2, 4}
{1, 2, 4, 3} {3, 1, 4, 2}
{1, 3, 2, 4} {3, 2, 1, 4}
{1, 3, 4, 2} {3, 2, 4, 1}
{1, 4, 2, 3} {3, 4, 1, 2}
{1, 4, 3, 2} {3, 4, 2, 1}
{2, 1, 3, 4} {4, 1, 2, 3}
{2, 1, 4, 3} {4, 1, 3, 2}
{2, 3, 1, 4} {4, 2, 1, 3}
{2, 3, 4, 1} {4, 2, 3, 1}
{2, 4, 1, 3} {4, 3, 1, 2}
{2, 4, 3, 1} {4, 3, 2, 1}
In red and underlined, the pivotal agent
1 2 3 4
Sh
7
12
1
4
1
12
1
12
winning coalitions:
{1, 2}
{1, 2, 3}
{1, 2, 4}
{1, 3, 4}
{1, 2, 3, 4}
In red and underlined, the pivotal agents
1 2 3 4
5
8
3
8
1
8
1
8
I
B
(N, v, i)
1
2
3
10
1
10
1
10
Table 7.1: Shapley-Schubik and the Banzhaff indices for the weighted voting game
[7; 4, 3, 2, 1].
7.4.2 Paradoxes
The power indices may behave in an unexpected way if we modify the game. For
example, we might expect that adding voters to a game would reduce the power of
those voters that are present in the original game, but this may not be the case.
Consider the game [4; 2, 2, 1]. Player 3 is a dummy in this game, so her Shapley
Shubik or Banzhaff indices are zero. Now assume that a voter joins the game with
a weight of 1. In the resulting game G
, player 3 becomes pivotal for a coalition

consisting of one of the two voters of the original game and the new player. Hence, her
index must now be positive. This situation is known as the paradox of new player.
Another unexpected behaviour may occur when a voter i splits her identity and
weight between two voters. The sum of the new identities Shapley value may be quite
different from the Shapley value of voter i. This situation is known as the paradox of
size.
increase of power by splitting identities Consider a game with |N| = n voters
[n+1; 2, 1, . . . , 1]. In this game, the only winning coalition is the grand coalition,
so I
SS
(N, v, i) =
1
n
. Now suppose that voter 1 splits into two voters of weight
one. We have a new game game with n+1 voters [n + 1; 1, . . . , 1]. Using a
similar argument, the Shapley Shubik index for each voter is
1
n+1
. Hence, the
joint power of the new identities is
2
n+1
, almost twice the power of agent 1 by
herself!
decrease of power by splitting identities Consider an n-voter voting game in
which all voters have a weight of 2 and the quota is 2n 1, i.e., we have the
game [2n1; 2, . . . , 2]. All the players being symmetric, the Shapley value is
1
n
.
If player 1 splits into two voters of weight 1, each of her identities has a Shapley
value of
1
n(n+1)
in the new game. Hence, the sum of the Shapley values of the
two identities is smaller than the value in the original game, by a factor of
n+1
2
.
7.4.3 Applications
When designing a weighted voting game, for example to decide on the weights for
a vote for the European Union or at the United Nations, one needs to choose which
weights are to be attributed to each nation. The problem of choosing the weights so
that they corresponds to a given power index has been tackled in [2]. If the number
of country changes, you do not want to re-design and negotiate over a new game each
time. Each citizen vote for a representative and the representatives for each country
vote. It may be desirable that each citizen, irrespective of her/his nationality, has the
same voting power. If
x
is the normalized Banzhaf index for a person in a country i
in EU with population n
i
, and
i
is the normalized Banzhaf index of a representative
for country i, then Felsenthal and Machover have shown that
x

i
2
n
i
. Thus the
Banzhaf index of each representative
i
should be proportional to n
i
for each person
in the EU to have equal power.
7.4.4 Complexity
The computational complexity of voting and weighted voting games have been studied
in [3, 4]. For example, the problem of determining whether the core is empty is poly-
nomial. The argument for this result is the following theorem: the core of a weighted
voting game is non-empty iff there exists a veto player. When the core is non-empty,
the problem of computing the nucleolus is also polynomial, otherwise, it is an NP-
hard problem.
Bibliography
[1] James S. Coleman. The benets of coalition. Public Choice, 8:4561, 1970.
[2] Bart de Keijzer, Tomas Klos, and Yingqian Zhang. Enumeration and exact design
of weighted voting games. In Proc. of the 9th Int. Conf. on Autonomous Agents
and Multiagent Systems (AAMAS-2010), pages 391398, 2010.
[3] Xiaotie Deng and C H Papadimitriou. On the complexity of cooperative solution
concetps. Mathematical Operation Research, 19(2):257266, 1994.
[4] Edith Elkind, Leslie Ann Goldberg, Paul Goldberg, and Michael Wooldridge.
Computational complexity of weighted threshold games. In Proceedings of the
Twenty-Second AAAI Conference on Articial Intelligence (AAAI-07), pages 718
723, 2007.
81
Lecture 8
Representation and complexity
We have studied different solution concepts, mainly looking at their different proper-
ties. One natural question from the point of view of a (theoretical) computer scientist is
how hard it is to compute a solution, or to check whether the set of solutions is empty
or not. This is the question we are going to consider in this lecture
8.1 Naive representation
Let us assume we want to write a computer program for computing a solution concept.
The rst question that comes to mind is how to represent the input of a TU game.
A straighforward representation is by enumeration: we can use a array, each entry
represents a coalition and contains the value of that coalition (and one can use the
binary representation of a number to encode which agents are members of a coalition,
e.g. 21 = 10101 corresponds to coalition {1, 3, 5}). This requires storing 2
n
numbers,
which may be problematic for large values of n. Typically, computer scientists are
made happier with an input of polynomial length.
The complexity of an algorithm is measured in terms of the input size. If we use
enumeration, many algorithms may appear good as they manipulate an exponential
input. To properly speak about complexity issues, one need to nd a polynomial rep-
resentation of the game, which we will call a compact or succinct representation.
In general, there is a tradeoff between how succinct the representation is and how
easy or hard the computation is. The idea is that to represent the game compactly, one
needs to encode a lot of information in a smart way, which may make it difcult to
manipulate the representation to compute something interesting. Hence we look for a
balance between succinctness and tractability.
83
84 Lecture 8. Representation and complexity
8.2 Representations that are good for computing the
Shapley Value
The nature of the Shapley value is combinatorial, as all possible orderings to form
a coalition need to be considered. This computational complexity can sometimes be
an advantage as agents cannot benet from manipulation. For example, it is NP-
complete to determine whether an agent can benet from false names [15]. Neverthe-
less, some representations allow to compute the Shapley value efciently, and we are
surveying few representations.
8.2.1 Bilateral Shapley Value
In order to reduce the combinatorial complexity of the computation of the Shapley
value, Ketchpel introduces the Bilateral Shapley Value (BSV ) [13]. The idea is to
consider the formation of a coalition as a succession of merging between two coali-
tions. Two disjoint coalitions C
1
and C
2
with C
1
C
2
= , may merge when v(C
1

C
2
) v(C
1
) + v(C
2
). When they merge, the two coalitions, called founders of the
new coalition C
1
C
2
, share the marginal utility as follows: BSV (C
1
) =
1
2
v(C
1
) +
1
2
(v(C
1
C
2
) v(C
2
)) and BSV (C
2
) =
1
2
v(C
2
) +
1
2
(v(C
1
C
2
) v(C
1
)). This is
the expression of the Shapley value in the case of an environment with two agents.
In C
1
C
2
, each of the founders gets half of its local contribution, and half of the
marginal utility of the other founder. Given this distribution of the marginal utility, it is
rational for C
1
and C
2
to merge if i {1, 2}, v(C
i
) BSV (C
i
). Note that symmetric
founders get equal payoff, i.e., for C
1
, C
2
, C such that C
1
C
2
= C
1
C = C
2
C = ,
v(C C
1
) = v(C C
2
) BSV (C C
1
) = BSV (C C
2
). Given a sequence of succes-
sive merges from the states where each agent is in a singleton coalition, we can use a
backward induction to compute a stable payoff distribution [14]. Though the computa-
tion of the Shapley value requires looking at all of the permutations, the value obtained
by using backtracking and the BSV only focuses on a particular set of permutations,
but the computation is signicantly cheaper.
8.2.2 Weigthed graph games
[8] introduce a class of games called weighted graph games: they dene a TU game
using an undirected weighted graph G = (V, W) where V is the set of vertices and
W : V V is the set of edges weights. For (i, j) V
2
, w
ij
is the weight of the edge
between the vertices i and j. The coalitional game (N, v) is dened as follows:
N = V , i.e., each agent corresponds to one vertex of the graph.
the value of a coalition C N is the sum of the weights between any pairs of
members of C, i.e. v(C) =
(i,j)C
2
w
ij
.
8.2. Representations that are good for computing the Shapley Value 85
1
2
3
4 5
w
12
w
23
w
34
w
45
w
15
w
13
w
14
w
24
v({1, 2, 4}) = w
12
+ w
34
+ w
14
Figure 8.1: Example of a graph with 5 agents
This representation is succinct as we only need to provide n
2
values to represent the
entire game. However, it is not a complete representation as some TU games cannot
be represented this way (e.g., it is not possible to represent a majority voting game).
If we add some restrictions on the weights, we can further guarantee some properties.
For example, when all the weights are nonnegative, then the game is convex, and then
the game is guaranteed to have a non-empty core.
8.2.1. PROPOSITION. Let (V, W) be a weighted graph game. If all the weights are
nonnegative then the game is convex.
Proof. Let S, T N, we want to prove that v(S) + v(T) v(S T) + v(S T).
v(S) + v(T) =
(i,j)S
2
w
ij
+
(i,j)T
2
w
ij
=
(i,j)S
2
(i,j)T
2
w
ij
+
(i,j)(ST)
2
w
ij
(i,j)(ST)
2
w
ij
+
(i,j)(ST)
2
w
ij
= v(S T) + v(S T)
One other nice property of this representation is that the Shapley value can be
computed in quadratic time, as shown in the following theorem.
8.2.2. THEOREM. Let (V, W) be a weighted graph game. The Shapley value of an
agent i is given by
i
(N, v) =
(i,j)N
2
,i=j
w
ij
.
One simple proof of this theorem uses the axioms that dene the Shapley value.
Proof. Let (V, W) a weighted graph game. We can view this game as the sum of
the |W| games: each game G
ij
is associated to an edge (i, j) in the graph as follows
G
ij
= (V, {w
ij
}) such that w
ij
kl
= w
ij
if i = k and j = l and 0 otherwise.
For a game G
ij
corresponding to edge (i, j):
the agents i and j are substitutes (this is clear).
all other agents k = i, j are dummy agents (this is also clear).
Using the symmetry axiom, we know that Sh
i
(G
ij
) = Sh
j
(G
ij
). Then, using the
dummy axiom, we also know that Sh
k
(G
ij
) = 0. This tells us that Sh
i
(G
ij
) =
1
2
w
ij
.
Since (V, W) is the sum of all games, by the additivity axiom, we obtain Sh
k
=
i,j
Sh
k
(G
ij
) =
k,i
w
ij

8.2.3 Multi-issue representation
Conitzer and Sandholm [7] analyse the case where the agents are concerned with mul-
tiple independent issues that a coalition can address. For example, performing a task
may require multiple abilities, and a coalition may gather agents that work on the same
task but with limited or no interactions between them. A characteristic function v can
be decomposed over T issues when it is of the form v(C) =
T
t=1
v
t
(C), in which, for
each t, (N, v
t
) is a TU game.
8.2.3. DEFINITION. [Decomposition]The vector of characteristic functions v
1
, v
2
, . . . , v
T
,
with each v
i
: 2
N
R, is a decomposition over T issues of characteristic function
v : 2
N
R if for any S N, v(S) =
T
i=1
v
i
(S).
Using this idea, we can represent any TU game (we can express a TU game using
a single issue).
The Shapley value for agent i for the characteristic function v is the sum of the
Shapley values over the t different issues:
i
(N, v) =
T
t=1

i
(N, v
t
). When a small
number of agents is concerned about an issue, computing the Shapley value for the
particular issue can be cheap. For an issue t, the characteristic function v
t
concerns
only the agents in I
t
when C
1
C, C
2
C such that I
t
C
1
= I
t
C
2
v
t
(C
1
) =
v
t
(C
2
). When the characteristic function v is decomposed over T issues and when |I
t
|
agents are concerned about each issue t [1...T], computing the Shapley value takes
O(
T
t=1
2
|I
t
|
).
8.2.4 Marginal Contribution Networks (MC-nets)
Ieong and Shoham propose a representation in which the characteristic function is
represented by a set of rules [11]. A rule is composed by a pattern and a value: the
pattern tells which agent must be present or absent from a coalition so that the value
of the coalition is increased by the value of the rule. This representation allows to
represent any TU game.
More formally, each player is represented by a boolean variable and the charac-
teristic vector of a coalition is treated as a truth assignment. Each rule associates
a pattern and a weight w R. The pattern is a formula of propositional logic
containing variables in N. A positive literal represents the presence of an agent in a
8.2. Representations that are good for computing the Shapley Value 87
coalition, whereas a negative literal represents the absence of an agent in the coalition.
The value of a coalition is the sum over the values of all the rules that apply to the
coalition.
8.2.4. DEFINITION. [Rule] Let N be a collection of atomic variables. A rule has a
syntactic form (, w) where is called the pattern and is a boolean formula containing
variables in N and w is called the weight, and is a real number.
Let us consider that there are two variables a and b and here are two rules:
(a b, 5): each coalition containing both agents a and b increase its value by 5
units.
(b, 2): each coalition containing b increase its value by 2.
8.2.5. DEFINITION. [Marginal contribution nets (MC-net)] An MC-net consists of a
set of rules {(p
1
, w
1
), . . . (pk, w
k
)} where the valuation function is given by
v(C) =
k
i=1
p
i
(e
C
)w
i
,
where p
i
(e
C
) evaluates to 1 if the boolean formula p
i
evaluates to true for the truth
assignment e
C
and 0 otherwise.
The valuation function of the MC net {(a b, 5), (b, 2)} is the following one:
v() = 0 v({b}) = 2
v({a}) = 0 v({a, b}) = 5 + 2 = 7
We can use negations in rules, and negative weights. Let consider the following
MC-net: {(a b, 5), (b, 2), (c, 4), (b c, 2)} In this case, the valuation function is
v() = 0 v({b}) = 2 2 = 0 v({a, c}) = 4
v({a}) = 0 v({a, b}) = 5 + 2 2 = 5 v({b, c}) = 4 + 2 = 6
When negative literals are allowed or when the weights can be negative, MC-nets
can represent any TU-game, hence this representation is complete. When the patterns
are limited to conjunctive formula over positive literals and weights are nonnegative,
MC-nets can represent all and only convex games (in which case, they are guaranteed
to have a non-empty core).
Using this representation and assuming that the patterns are limited to a conjunction
of variables, the Shapley value can be computed in time linear to the size of the input
(i.e. the number of rules of the MC-net).
8.2.6. THEOREM. Given a TU game represented by an MC-net limited to conjunctive
patterns, the Shapley value can be computed in time linear in the size of the input.
Proof. (sketch) we can treat each rule as a game, compute the Shapley value for the
rule, and use ADD to sum all the values for the overall game. For a rule, we cannot
distinguish the contribution of each agent, by SYM, they must have the same value. It
is a bit more complicated when negation occurs (see Ieong and Shoham, 2005).
8.3 Some references for simple games
The computational complexity of voting and weighted voting games have been studied
in [8, 9]. For example, the problem of determining whether the core is empty is poly-
nomial. The argument for this result is the following theorem: the core of a weighted
voting game is non-empty iff there exists a veto player. When the core is non-empty,
the problem of computing the nucleolus is also polynomial, otherwise, it is an NP-
hard problem.
8.4 Some interesting classes of games from the compu-
tational point of view
We want to briey introduce some classes of games that have been studied in the AI
literature. Some of these classes of games can be represented more compactly than by
using 2
N
values, one for each coalition, using an underlying graph structure. In some
restricted cases, some solution concepts can be computed efciently.
minimum cost spanning tree games. A game is V, s, w where V, w is as in a
graph game and s V is the source node. For a coalition C, we denote by (C) the
minimum cost spanning tree spanning over the set of edges C {s}. The value of a
coalition V \ {s} is given by
(i,j)(C)
w
i,j
.
This class of game can model the problem of connecting some agents to a central
node played by the source node s. Computing the nucleolus or checking whether the
core is non-empty can be done in polynomial time.
Network ow games. A ow network V, E, c, s, t is composed of a directed
graph (V, E) with a capacity on the edge c : V
2
R
+
, a source vertex s and a sink
vertex t. A network ow is a function f : E R
+
that satises the capacity of an
edge ((i, j) E, f(i, j) c(i, j)) and that is conserved (except for the source and
sink), i.e., the total ow arriving at an edge is equal to the total ow leaving that edge
(j V ,
(i,j)E
f(i, j) =
(j,k)E
f(j, k)). The value of the ow is the amount
owing out of the sink node.
In network ow game [12], V, E, c, s, t, the value of a coalition C N is the
maximum value of the ow going through the ow network C, E, c, s, t.
This class of games can model a situation where some cities share a supply of water
or some electricity network. [12] proved that a network ow game is balanced, hence
it has a non-empty core. [3] study a threshold version of the game and the complexity
of computing power indices.
Afnity games. The class of afnity games is a class of hedonic games introduced
in [5, 6]. An afnity game is dened using a directed weigthed graph V, E, w where
V is the set of agents, E is the set of directed edges and w : E R is the weight of
the edges. w(i, j) is the value of agent i when it is associated with agent j. The value
of agent i for coalition C is v
i
(C) =
jC
w(i, j).
8.4. Some interesting classes of games from the computational point of view 89
Some special classes of afnity games have a non-empty core (e.g. when the
weights are all positive or all negative). In this games, there may be a trade-off be-
tween stability and efciency (in the sense of maximizing social welfare) as the ratio
between an optimal CS and a stable CS may be innite.
Skill games. This class of games, introduced by [4] is represented by a triplet
N, S, T, u where N is the set of agents, S is the set of skills, T is the set of tasks, and
u : 2
T
R provides a value to each set of tasks that is completed. Each agent i has a
set of skills S(i) S, each task t
i
requires a set of skills S(t
i
) S. A coalition C can
perform a task t when each skill needed for the task is the skill of at least a member
of C (i.e. s S(t), i C such that S(i) = s). The value of a coalition C is u(T
C
)
where T
C
is the set of tasks that can be performed by C.
This representation is exponential in the number of agents, but variants of the repre-
sentation lead to polynomial representation. For example when the value of a coalition
is the number of tasks it can accomplish, or when each task has a weight and the value
of a coalition is the sum of the weights of the accomplished tasks. In general, com-
puting the solution concepts with these polynomial representation is hard. However,
in some special cases, checking whether the core is empty or computing an element of
the core can be performed in polynomial time. The problem of nding an optimal CS
is studied in [2].
Some more papers are studying the computational complexity of some subclasses
of games, e.g. in [1, 10] to name a few. We do not want to provide a full account of
complexity problem in cooperative games as it could be the topic of half a course.
Bibliography
[1] Haris Aziz, Felix Brandt, and Paul Harrenstein. Monotone cooperative games and
their threshold versions. In Proceedings of the 9th International Conference on
Autonomous Agents and Multiagent Systems: volume 1 - Volume 1, AAMAS 10,
pages 11071114, Richland, SC, 2010. International Foundation for Autonomous
Agents and Multiagent Systems.
[2] Yoram Bachrach, Reshef Meir, Kyomin Jung, and Pushmeet Kohli. Coalitional
structure generation in skill games. In Proceedings of the Twenty-Fourth AAAI
Conference on Articial Intelligence (AAAI-10), pages 703708, July 2010.
[3] Yoram Bachrach and Jeffrey Rosenschein. Power in threshold network ow
games. Autonomous Agents and Multi-Agent Systems, 18:106132, 2009.
10.1007/s10458-008-9057-6.
[4] Yoram Bachrach and Jeffrey S. Rosenschein. Coalitional skill games. In Proc. of
the 7th Int. Conf. on Autonomous Agents and Multiagent Systems (AAMAS-08),
pages 10231030, May 2008.
[5] Simina Brnzei and Kate Larson. Coalitional afnity games. In Proceedings
of the International Joint Conference on Articial Intelligence (IJCAI-09), pages
13191320, May 2009.
[6] Simina Brnzei and Kate Larson. Coalitional afnity games. In Proc of the 8th
Int. Conf. on Autonomous Agents and Multiagent Systems (AAMAS 2009), pages
7984, July 2009.
[7] Vincent Conitzer and Tuomas Sandholm. Computing shapley values, manipu-
lating value division schemes, and checking core membership in multi-issue do-
mains. In Proceedings of the 19th National Conference on Articial Intelligence
(AAAI-04), pages 219225, 2004.
[8] Xiaotie Deng and C H Papadimitriou. On the complexity of cooperative solution
concetps. Mathematical Operation Research, 19(2):257266, 1994.
91
92 Bibliography
[9] Edith Elkind, Leslie Ann Goldberg, Paul Goldberg, and Michael Wooldridge.
Computational complexity of weighted threshold games. In Proceedings of
the Twenty-Second AAAI Conference on Articial Intelligence (AAAI-07), pages
718723, 2007.
[10] Gianluigi Greco, Enrico Malizia, Luigi Palopoli, and Francesco Scarcello. On
the complexity of compact coalitional games. In Proceedings of the 21st inter-
national jont conference on Artical intelligence, pages 147152, San Francisco,
CA, USA, 2009. Morgan Kaufmann Publishers Inc.
[11] Samuel Ieong and Yoav Shoham. Marginal contribution nets: a compact repre-
sentation scheme for coalitional games. In EC 05: Proceedings of the 6th ACM
conference on Electronic commerce, pages 193202, New York, NY, USA, 2005.
ACM Press.
[12] E. Kalai and E. Xemel. Totally balanced games and games of ow. Mathematics
of Operations Research, 7:476478, 1982.
[13] Steven P. Ketchpel. The formation of coalitions among self-interested agents. In
Proceedings of the Eleventh National Conference on Articial Intelligence, pages
414419, August 1994.
Lecture 9
Non-Transferable Utility Games (NTU games)
An underlying assumption behind a TU game is that agents have a common scale to
measure the worth of a coalition. Such a scale may not exist in every situations, which
leads to the study of games where the utility is non-transferable (NTU games). We
start by introducing a particular type of NTU games called Hedonic games. We chose
this type of games for the simplicity of the formalism. Next, we provide the classical
denition of a NTU game, which can represent a lot more situations.
9.1 Hedonic Games
In an hedonic game, agents have preferences over coalitions: each agent knows whether
it prefers to be in company of some agents rather than others. An agent may enjoy more
the company of members of (
1
over members of (
2
, but it cannot tell by how much
it prefers (
1
over C
2
. Consequently, it does not make sense to talk about any kind of
compensation when an agent is not part of its favorite coalition. The question that each
agent must answer is which coalition to form?.
More formally, let N be a set of agents and A
i
be the set of coalitions that contain
agent i, i.e., A
i
= ( i [ ( N i. For a CS o, we will note o(i) the coalition
in o containing agent i.
9.1.1. DEFINITION. [Hedonic games] An Hedonic game is a tuple (N, (_
i
)
iN
) where
N is the set of agents
_
i
2
N
i
2
N
i
is a complete, reexive and transitive preference relation for agent
i, with the interpretation that if S _
i
T, agent i prefers coalition T at most as
much as coalition S.
In a hedonic game, each agent has a preference over each coalition it can join. The
solution of a hedonic game is a coalition structure (CS), i.e., a partition of the set of
93
94 Lecture 9. Non-Transferable Utility Games (NTU games)
agents into coalition. A rst desirable property of a solution is to be Pareto optimal: it
would not be possible to nd a different solution that is weakly preferred by all agents.
The notion of core can be easily extended for this type of games. Given a current
CS, no group of agents should have an incentive to leave the current CS. As it is the
case for TU games, the core of an NTU game may be empty, and it is possible to
dene weaker versions of stability. We now give the denition of stability concepts
adapted from [2]. For the core, it is a group of agents that leave their corresponding
coalitions to form a new one they all prefer. In the weaker stability solution concepts,
the possible deviations feature a single agent i that leaves its current coalition (
1
i
to join a different existing coalition (
2
N i or to form a singleton coalition i.
There are few scenarios one can consider, depending on the behavior of the members
of (
1
and (
2
. For a Nash stable o, the behavior of (
1
and (
2
are not considered at
all: if i prefers to join (
2
, it is a valid deviation. This assumes that the agents in (
2
will accept agent i, which is quite optimistic (it may very well be the case that some
or all agents in (
2
do not like agent i). For individual stability, the deviation is valid if
no agent in (2 is against accepting agent i, in other words the agents in (2 are happy
or indifferent about i joining them. Finally, for contractual individual stability, the
preference of the members of (1 the coalition that i leaves are taken into account.
Agents in (1 should prefer to be without i than with i. The three stability concepts
have the following inclusion: Nash stability is included in Individual stability, which is
included in contractual individual stability
1
. We now provide the corresponding formal
denitions.
Core stability: A CS o is core-stable iff ( N [ i (, ( ~
i
o(i).
Nash stability: A CS o is Nash-stable iff (i N) (( o ) o(i)
i
( i. No player would like to join any other coalition in o assuming the other
coalitions did not change.
Individual stability A CS o is individually stable iff (i N) (( o
) [ (( i ~
i
o(i)) and (j (, ( i
j
(). No player can move to
another coalition that it prefers without making some members of that coalition
unhappy.
Contractually individual stability: ACS o is contractually individually stable
iff (i N) (( o ) [ (( i ~
i
o(i)) and (j (, ( i
j
() and (j o(i) i, o(i) i
j
o(i)). No player can move to a coalition
it prefers so that the members of the coalition it leaves and it joins are better off.
Let us see some examples.
1
This ordering may appear counter-intuitive at rst, but note that the conditions for being a valid de-
viation are more difcult to meet from Nash stability to Contractual individual stability, and the stability
concepts are dened as CS such that no deviation exists
9.1. Hedonic Games 95
Example 1
1, 2 ~
1
1 ~
1
1, 2, 3 ~
1
1, 3
1, 2 ~
2
2 ~
2
1, 2, 3 ~
2
2, 3
1, 2, 3 ~
3
2, 3 ~
3
1, 3 ~
3
3
Let us consider each CS one by one.
1, 2, 3 1, 2 is preferred by both agent 1 and 2, hence it is not Nash stable.
1, 2, 3
1, 2, 3 is preferred by agent 3, so it is not Nash stable,
as agents 1 and 3 are worse off, it is not a possible move for individual stability.
no other move is possible for individual stability.
1, 3, 2
agent 1 prefers to be on its own (hence the CS is not Nash stable,
and then, neither individually stable).
2, 3, 1
agent 2 prefers to join agent 1,
and agent 1 is better off, hence the CS is neither Nash stable nor
Individually stable.
1, 2, 3
agents 1 and 2 have an incentive to form a singleton,
hence the CS is neither Nash stable, nor Individually stable.
As a conclusion:
1, 2, 3 is in the core and is individually stable.
There is no Nash stable partitions.
Example 2
1, 2 ~
1
1, 3 ~
1
1, 2, 3 ~
1
1
2, 3 ~
2
1, 2 ~
2
1, 2, 3 ~
2
2
1, 3 ~
3
2, 3 ~
3
1, 2, 3 ~
3
3
Again, let us consider all the CSs.
1, 2, 3 1, 2, 1, 3, 2, 3 and 1, 2, 3 are blocking. As a result, the CS is
not Nash stable
1, 2, 3 2, 3 is blocking, and since 2 can leave 1 to form1, 2, it follows that
the CS is not Nash stable
1, 3, 2 1, 2 is blocking, and since 1 can leave 3 to form1, 2, it follows that
2, 3, 1 1, 3 is blocking, and since 3 can leave 2 to form 1, 3,it follows that
1, 2, 3 1, 2, 1, 3, 2, 3 are blocking, but it is contractually individually
stable
As a conclusion, we obtain that
The core is empty.
1, 2, 3 is the unique Nash stable partition, unique individually stable parti-
tion (no agent has any incentive to leave the grand coalition).
Example 3
1, 2 ~
1
1, 3 ~
1
1 ~
1
1, 2, 3
2, 3 ~
2
1, 2 ~
2
2 ~
2
1, 2, 3
1, 3 ~
3
2, 3 ~
3
3 ~
3
1, 2, 3
For this game we can show that
The core is empty (similar argument as for example 2).
There is no Nash stable partition or individually stable partition. But there are
three contractually individually stable CSs: 1, 2, 3, 1, 3, 2,2, 3, 1.
For example, we can consider all the possible changes for the CS 1, 2, 3:
agent 2 joins agent 3:
In this case, both agents 2 and 3 benet from forming 1, 2, 3, hence
1, 2, 3 is not Nash or individually stable. As agent 1 is worse off it is
not a legal deviation for contractual individual stability.
agent 1 joins agent 3, but agent 1 has no incentive to join agent 3, hence this is
not a deviation.
Agent 1 or 2 forms a singleton coalition, but neither agent has any incentive to
form a singleton coalition. Hence, there is no deviation.
In Examples 2 and 3, we see that the core may be empty. The literature in game
theory focuses on nding conditions for the existence of the core. In the AI literature,
Elkind and Wooldridge have proposed a succinct representation of Hedonic games [4]
and Brnzei and Larson considered a subclass of hedonic games called the afnity
games [3].
9.2 NTU games
We now turn to the most general denition of an NTU game. This denition is more
general than the denition of hedonic gamess. The idea is that each coalition has
the ability to achieve a set of outcomes. The preference of the agents are about the
outcomes that can be brought about by the different coalitions. The formal denition
is the following:
9.2.1. DEFINITION. [NTU Game] A non-transferable utility game (NTU Game) is de-
ned by a tuple (N, X, V, (~
i
)
iN
) such that
9.2. NTU games 97
N is a set of agents;
X is a set of outcomes;
V : 2
N
2
X
is a function that describes the outcomes V (() X that can be
brought about by coalition (;
~
i
is the preference relation of agent i over the set of outcomes. The relation is
assumed to be transitive and complete.
Intuitively, V (() is the set of outcomes that the members of ( can bring about by
means of their joint-actions. The agents have a preference relation over the outcomes,
which makes a lot of sense. This type of games is more general that the class of hedonic
games or even TU games, as we can represent these games using a NTU game.
9.2.2. PROPOSITION. A hedonic game can be represented by an NTU games.
Proof. Let (N, (_
H
i
)
iN
) be a hedonic game.
For each coalition ( N, create a unique outcome x
C
.
For any two outcomes x
S
and x
T
corresponding to coalitions S and T that con-
tains agent i, We dene _
i
as follows: x
S
_
i
x
T
iff S _
H
i
T.
For each coalition ( N, we dene V (() as V (() = x
C
.
9.2.3. PROPOSITION. A TU game can be represented as an NTU game.

Proof. Let (N, v) be a TU game.
We dene X to be the set of all allocations, i.e., X = R
n
.
For any two allocations (x, y) X
2
, we dene _
i
as follows: x _
i
y iff x
i
y
i
.
For each coalition ( N, we dene V (() as V (() = x R
n
[

iN
x
i

v((). V (() lists all the feasible allocation for the coalition (.
First, we can note that the denition of the core can easily be modied in the case
of NTU games.
9.2.4. DEFINITION. core(V ) = x V (N) [ ( N, y V ((), i ( y ~
i
x
An outcome x X is blocked by a coalition ( when there is another outcome y X
that is preferred by all the members of (. An outcome is then in the core when it can
be achieved by the grand coalition and it is not blocked by any coalition. As is the case
for TU game, it is possible that the core of an NTU game is empty.
As for TU games, we can dene a balanced game and show that the core of a
balanced game is non-empty.
9.2.5. DEFINITION. [Balanced game]A game is balanced iff for every balanced col-
lection B, we have
CB
V (() V (N)
9.2.6. THEOREM (THE SCARF THEOREM). The core of a balanced game is non-empty.
9.2.1 An application: Exchange Economy
For TU games, we studied market games and proved such games have a non-empty
core. We now consider a similar game without transfer of utility. There is a set of con-
tinuous goods that can be exchanged between the agents. Each agent has a preference
relation over the bundle of goods and tries to obtain the best bundle possible.
Denition of the game
The main difference between the exchange economy and a market game is that the
preference is ordinal in the exchange econmomy whereas it is cardinal in a market
game.
9.2.7. DEFINITION. An exchange economy is a tuple (N, M, A, (_
i
)
iN
) where
N is the set of n agents
M is the set of k continuous goods
A = (a
i
)
iN
is the initial endowment vector
(_
i
)
iN
is the preference prole, in which _
i
is a preference relation over bun-
dles of goods.
Given an exchange economy (N, M, A, (_
i
)
iN
), we dene the associated ex-
change economy game as the following NTU game (N, X, V, (_
i
)
iN
) where:
The set of outcomes X is dened as X =
(x
1
, . . . , x
n
) [ x
i
R
k
+
for i N
.
Note that x
i
= x
i1
, . . . , x
ik
represents the quantity of each good that agent i
possesses in an outcome x.
The preference relation for an agent i is dened as: for (x, y) X
2
x _
i
y
x
i
_
i
y
i
. Each player is concerned by its own bundle only.
9.2. NTU games 99
The value sets are dened as ( N,
V (() =
x X
iC
x
i
=
iC
a
i
x
j
= a
j
for j N (
.
The players outside ( do not participate in any trading and hold on their ini-
tial endowments. When all agents participate in the trading, we have V (N) =
x X [

iN
x
i
=
iN
a
i
.
Solving an exchange economy
Let us assume that we can dene a price p
r
for a unit of good r. The idea would be to
exchange the goods at a constant price during the negotiation.
Let us dene a price vector p R
k
+
. The amount of each good that agent i possesses
is x
i
R
k
+
. The total cost of agent is bundle is p x
i
=
k
r=1
p
r
x
i,r
. Since the initial
endowment of agent i is a
i
, the agent has at his disposal an amount p a
i
, and i can
afford to obtain a bundle y
i
such that p y
i
p a
i
.
We can wonder about what an ideal situation would be. Given the existence of the
price vector, we can dene a competitive equilibrium, the idea is to make believe to
each player that it possesses the best outcome.
9.2.8. DEFINITION. [Competitive equilibrium] The competitive equilibrium of an ex-
change economy (N, X, V, (_
i
)
iN
) is a pair (p, x) where p R
k
+
is a price vector
and x
(x
1
, . . . , x
n
) [ x
i
R
k
+
for i N
such that

iN
x
i
=
iN
a
i
(the allocation results from trading)
i N, p x
i
p a
i
(each agent can afford its allocation)
i N y
i
R
k
+
(p y
i
p a
i
) x
i
_
i
y
i
Among all the allocations that an agent can afford, it obtains one of its most favorites
outcomes.
This competitive equilibrium seems like an ideal situation, and surprisingly, Arrow
and Debreu [1] proved such equilibrium is guaranteed to exist. This is a deep theorem,
and we will not study the proof here.
9.2.9. THEOREM (ARROW & DEBREU, 1954). Let (N, M, A, (_
i
)
iN
) be an exchange
economy. If each preference relation _
i
is continuous and strictly convex, then a com-
petitive equilibrium exists.
The proof of the theorem is not constructive, i.e., it guarantees the existence of the
equilibrium, but not how to obtain the price vector or the allocation. The following
theorem links the allocation with the core:
9.2.10. THEOREM. If (p, x) is a competitive equilibriumof an exchange economy, then
x belongs to the core of the corresponding exchange economy game.
Proof. Let us assume x is not in the core of the associated exchange economy game.
Then, there is at least one coalition ( and an allocation y such that i ( y ~
i
x. By
denition of the competitive equilibrium, we must have p y
i
> p a
i
. Summing over
all the agents in (, we have p
iC
y
i
> p
iC
a
i
. Since the prices are positive, we
deduce that
iC
y
i
>
iC
a
i
, which is a contradiction.
It then follows that if each preference relation is continuous and strictly convex,
then the core of an exchange economy game is non-empty. In an economy, the out-
comes that are immune to manipulations by groups of agent are competitive equilib-
rium allocation.
Bibliography
[1] Kenneth J. Arrow and Gerard Debreu. Existence of an equilibrium for a competi-
tive economy. Econometrica, 22:265290, July 1954.
[2] Anna Bogomolnaia and Matthew O. Jackson. The stability of hedonic coalition
structures. Games and Economic Behavior, 38(2):201230, 2002.
[3] Simina Brnzei and Kate Larson. Coalitional afnity games. In Proc of the 8th
Int. Conf. on Autonomous Agents and Multiagent Systems (AAMAS 2009), pages
7984, July 2009.
[4] Edith Elkind and Michael Wooldridge. Hedonic coalition nets. In Proc. of the
8th Int. Conf. on Autonomous Agents and Multiagent Systems (AAMAS-09), pages
417424, May 2009.
101
Lecture 10
Outside of the traditional games
10.1 Games with a priori unions
a different interpretation of a coalition structure
So far, a coalition has represented a set of agents that worked on its own. In a CS,
the different coalitions are intended to work independently of each other. We can
also interpret a coalition to represent a group of agent that is more likely to work
together within a larger group of agents (because of personal or political afnities).
The members of a coalition do not mind working with other agents, but they want to
be together and negotiate their payoff together, which may improve their bargaining
power. This is the idea used in games with a priori unions. Formally, a game with a
priori unions is similar to a game with CS: it consists of a triplet (N, v, S) when (N, v)
is a TU game and S is a CS. However, we assume that the grand coalition forms. The
problem is again to dene a payoff distribution.
10.1.1. DEFINITION. [Game with a priori unions] A game with a priori unions is a
triplet (N, v, S), where (N, v) is a TU game, and S is a particular CS. It is assumed
that the grand coalition forms.
Owen [8] proposes a value that is based on the idea of the Shapley value. The
agents forms the grand coalition by joining one by one. In the Shapley value, all
possible joining orders are allowed. In the Owen value, an agent i may join only when
the last agent that joined is a member of is coalition or when the last agents (j
1
, . . . , j
k
)
that joined before formed a coalition in S. This is formally captured using the notion
of a consistency with a CS:
10.1.2. DEFINITION. [Consistency with a coalition structure] A permutation is con-
sistent with a CS S when, for all (i, j) C
2
, C S and l N, (i) < (l) < (j)
implies that l C.
103
104 Lecture 10. Outside of the traditional games
We denote by
S
(N) the set of permutations of N that are consistent with the
CS S. The number of such permutations is m
CS
|C|! where m is the number of
coalitions in S. The Owen value is then dened as follows:
10.1.3. DEFINITION. Owen value Given a game with a priori union (N, v, S), the
Owen value O
i
(N, v, S) of agent i is given by
O
i
(N, v, S) =
S
(N)
mc()
|
S
(N)|
In Table 10.1, we present the example used for the Shapley value and compute
the Owen value. The members of the coalition of two agents improve their payoff by
forming an union.
N = {1, 2, 3}
v({1}) = 0 v({2}) = 0 v({3}) = 0
v({1, 2}) = 90 v({1, 3}) = 80 v({2, 3}) = 70
v({1, 2, 3}) = 120
S
2
= {{1, 2}, {3}} S
2
= {{1, 3}, {2}}
1 2 3
1 2 3 0 90 30
1 3 2
2 1 3 90 0 30
2 3 1
3 1 2 80 40 0
3 2 1 50 70 0
total 220 200 60
Owen value O
i
(N, v, S
1
) 55 50 15
1 2 3
1 2 3
1 3 2 0 40 80
2 1 3 90 0 30
2 3 1 50 0 70
3 1 2 80 40 0
3 2 1
total 220 80 180
Owen value O
i
(N, v, S
2
) 55 20 45
Table 10.1: Example of the computation of an Owen value
10.2 Games with externalities
A traditional assumption in the literature of coalition formation is that the value of a
coalition depends solely on the members of that coalition. In particular, it is indepen-
dent of on non-members actions. In general, this may not be true: some externalities
(positive or negative) can create a dependency between the value of a coalition and
the actions of non-members. [10] attribute these externalities to the presence of shared
resources (if a coalition uses some resource, they will not be available to other coali-
tions), or when there are conicting goals: non-members can move the world farther
from a coalitions goal state. [9] state that a recipe for generating characteristic func-
tions is a minimax argument: the value of a coalition C is the value C gets when the
10.2. Games with externalities 105
non-members respond optimally so as to minimise the payoff of C. This formulation
acknowledges that the presence of other coalitions in the population may affect the
payoff of the coalition C. As in [4, 9], we can study the interactions between different
coalitions in the population: decisions about joining forces or splitting a coalition can
depend on the way the competitors are organised. For example, when different compa-
nies are competing for the same market niche, a small company might survive against
a competition of multiple similar individual small companies. However, if some of
these small companies form a viable coalition, the competition signicantly changes:
the other small companies may now decide to form another coalition to be able to suc-
cessfully compete against the existing coalition. Another such example is a bargaining
situation where agents need to negotiate over the same issues: when agents form a
coalition, they can have a better bargaining position, as they have more leverage, and
because the other party needs to convince all the members of the coalition. If the other
parties also form coalition, the bargaining power of the rst coalition may decrease.
Two main types of games with externalities are described in the literature, both are
represented by a pair (N, v), but the valuation function has a different signature.
Games in partition function form [11]: v : 2
N
S
n
R. This is an extension of
the valuation function of a TU game by providing the value of a coalition given
the current coalition structure (note that v(C, S) is meaningful when C S).
Games with valuations : v : N S
n
R. In this type of games, the valuation
function directly assigns a value to an agent given a coalition structure. One
possible interpretation is that the problem of sharing the value of a coalition to
the members has already been solved.
The denitions of superadditivity, subadditivity and monotonicity can be adapted
to games in partition functions [3]. As an example, we provide the denition for su-
peradditivity.
10.2.1. DEFINITION. [superadditive games in partition function] A partition function
v is superadditive when, for any CS S and any coalitions C
1
and C
2
in S, we have
v(C
1
C
2
, S \ {C
1
, C
2
} {C
1
C
2
}) v(C1, S) + v(B, S).
The partition function may also have some regularities when two coalition merge:
either they always have a positive effect on the other coalition, or they always have a
negative one. More precisely, a partition function exhibits positive spillovers when for
any CS S and any coalitions C
1
and C
2
in S, we have v(C, S \ {C
1
, C
2
} {C
1
C
2
})
v(C, S) for all coalitions C = C
1
, C
2
in S.
We now turn to considering solution concepts for such games. The issue of extend-
ing the Shapley value has a rich literature in game theory. We want the Shapley value
to represent an average marginal contribution, but there is a debate over which set of
coalition structures. Michalak et al. [5] provide references on different solutions and
present three solutions in more details.
106 Lecture 10. Outside of the traditional games
Airiau and Sen [2] considers the issue of the stability of the optimal CS and dis-
cusses a possible way to extend the kernel for partition function games. In [1], they
consider coalition formation in the context of games with valuations and propose a
solution for myopic agents (an agent will join a coalition only when it is benecial,
without considering long-terms effect).
Michalak et al. [7] tackle the problem of representing such games and propose three
different representations that depends on the interpretation of the externalities. The rst
representation considers the value of a coalition in a CS: the value of a coalition can
be decomposed into on term that is free of externality and another term that models
the sum of the uncertainty due to the formation of the other coalitions. The two other
representations consider that the contribution of a coalition in a CS: either by providing
the mutual inuence of any two coalitions in a CS (outward operational externalities)
or by providing the inuence of all the other coalitions on a given coalition (inward
operational externalities). Michalak et al. (in [5] and [6]) extend the concept of MC-
nets to games with partition function.
Bibliography
[1] Stphane Airiau and Sandip Sen. A fair payoff distribution for myopic rational
agents. In Proceedings of the Eighth International Conference on Autonomous
Agents and Multiagent Systems (AAMAS-09), May 2009.
[2] Stphane Airiau and Sandip Sen. On the stability of an optimal coalition struc-
ture. In Proceedings of the 19th European Conference on Articial Intelligence
(ECAI-2010), pages 203208, August 2010.
[3] Francis Bloch. Non-cooperative models of coalition formation in games with
spillover. In Carlo Carraro, editor, The endogenous formation of economic coali-
tions, chapter 2, pages 3579. Edward Elgar, 2003.
[4] Sergiu Hart and Mordecai Kurz. Endogenous formation of coalitions. Economet-
rica, 51(4), July 1983.
[5] Tomasz Michalak, Talal Rahwan, Dorota Marciniak, Marcin Szamotulski, and
Nicholas R. Jennings. Computational aspects of extending the shapley value to
coalitional games with externalities. In Proceeding of the 2010 conference on
ECAI 2010: 19th European Conference on Articial Intelligence, pages 197
202, Amsterdam, The Netherlands, The Netherlands, 2010. IOS Press.
[6] Tomasz Michalak, Talal Rahwan, Dorota Marciniak, Marcin Szamotulski, Peter
McBurney, and Nicholas R. Jennings. Alogic-based representation for coalitional
games with externalities. In Proceedings of the 9th International Conference on
Autonomous Agents and Multiagent Systems: volume 1 - Volume 1, AAMAS 10,
pages 125132, Richland, SC, 2010. International Foundation for Autonomous
Agents and Multiagent Systems.
[7] Tomasz Michalak, Talal Rahwan, Jacek Sroka, Andrew Dowell, Michael
Wooldridge, Peter McBurney, and Nicholas R. Jennings. On representing coali-
tional games with externalities. In Proceedings of the 10th ACM conference on
Electronic Commerce 09 (EC09), 2009.
107
108 Bibliography
[8] Guilliermo Owen. Values of games with a priori unions. In O. Moeschlin R. Hein,
editor, Mathematical Economics and Game Theory: Essays in Honor of Oskar
Morgenstern. Springer, New York, 1977.
[9] Debraj Ray and Rajiv Vohra. A theory of endogenous coalition structures. Games
and Economic Behavior, 26:286336, 1999.
[10] Tuomas Sandholm and Victor R. Lesser. Coalitions among computationally
bounded agents. AI Journal, 94(12):99137, 1997.
[11] R. M. Thrall and W. F. Lucas. N-person games in partition function form. Naval
Research Logistics Quarterly, 10(1):281298, 1963.
Lecture 11
Coalition Structure Generation problem and
related issues
In the previous sections, the focus was on individual agents that are concerned with
their individual payoff. In this section, we consider TU games (N, v) in which agents
are concerned only about the societys payoff: the agents goal is to maximise utilitar-
ian social welfare. The actual payoff of the agent or the value of her coalition is not
of importance in this setting, only the total value generated by the population matters.
This is particularly interesting for multiagent systems designed to maximise some ob-
jective functions. In the following, an optimal CS denotes a CS with maximum social
welfare. This may model multiagent systems that are designed to optimise an objective
function.
More formally, we consider a TU game (N, v), and we recall that a coalition struc-
ture (CS) s = {S
1
, , S
m
} is a partition of N, where S
i
is the i
th
coalition of agents,
and i = j S
i
S
j
= and
i[1..m]
S
i
= N. S denotes the set of all CSs. The goal
of the multiagent system is to locate a CS that maximises utilitarian social welfare, in
other words the problem is to nd an element of argmax
sS
Ss
v(S).
The space S of all CSs can be represented by a lattice, and an example for a
population of four agents is provided in Figure 11.1. The rst level of the lattice
consists only of the CS corresponding to the grand coalition N = {1, 2, 3, 4}, the
last level of the lattice contains CS containing singletons only, i.e., coalitions con-
taining a single member. Level i contains all the CSs with exactly i coalitions. The
number of CSs at level i is S(|N|, i), where S is the Stirling Number of the Second
Kind
1
. The Bell number, B(n), represents the total number of CSs with n agents,
B(n) =
n
i=0
S(n, k). This number grows exponentially, as shown in Figure 11.2,
and is O(n
n
) and (n
n
2
) [15]. When the number of agents is relatively large, e.g.,
n 20, exhaustive enumeration may not be feasible.
The actual issue is the search of the optimal CS. Sandholm et al. [15] show that
given a TU game (N, v), the nding the optimal CS is an NP-complete problem. In
the following, we will consider centralised search where a single agent is performing
1
S(n, m) is the number of ways of partitioning a set of n elements into m non-empty sets.
109
110 Lecture 11. Coalition Structure Generation problem and related issues
Level 1
Level 2
Level 3
Level 4 {1}{2}{3}{4}
{1, 2}{3}{4} {1, 3}{2}{4} {1, 4}{2}{3} {1}{2, 3}{4} {1}{2, 4}{3} {1}{2}{3, 4}
{1, 2, 3}{4} {1, 2}{3, 4} {3}{1, 2, 4} {1, 4}{2, 3} {2}{1, 3, 4} {1, 3}{2, 4} {1}{2, 3, 4}
{1, 2, 3, 4}
Figure 11.1: Set of CSs for 4 agents.
the search as well as the more interesting case of decentralised search where all agents
make the search at the same time on different parts of the search space. Before doing
so, we review some work where the valuation function v is not known in advance.
In a real application, these values need to be computed; and this may be an issue on
its own if the computations are hard, as illustrated by an example in [14] where the
computation of a value requires to solve a traveling salesman problem.
11.1 Sharing the computation of the coalition values
Thus far, when we used a TU game, the valuation function was common knowledge.
For a practical problem though, one needs to compute these values. We said that the
value of a coalition was the worth that could be achieved through cooperation of the
coalitions members. In many cases, computing the value of a coalition will be an
optimization problem: nd the optimal way to cooperate to produce the best possible
worth. In some cases, such a problem may be computationaly hard. The following
example is given by Sandholm and Lesser [14]: we are in a logistics application and
the computing the value of a coalition requires to solve a travelling salesman problem,
a problem known to be NP-complete. Before being able to compute an optimal CS,
one needs to compute the value of all coalitions. Since agents are cooperative (i.e. they
want to work together to ensure the best outcome for the society), we are interested in
a decentralised algorithm that computes all the coalition values in a minimal amount
of time, and that requires minimum communication between the agents.
Shehory and Kraus were the rst to propose an algorithm to share the computation
11.1. Sharing the computation of the coalition values 111
100000
1e+10
1e+20
1e+30
1e+40
0 5 10 15 20 25 30 35 40 45 50
l
o
g
(
n
u
m
e
r
)
Number of agents
Number of coalitions and coalition structures
number of coalition structures
number of coalitions
Figure 11.2: Number of CSs in a population of n agents.
of the coalition values [19]. In their algorithm, the agents negotiate which computa-
tion is performed by which agent, which is quite demanding. Rahwan and Jennings
proposed an algorithm where agents rst agree on an identication for each agent par-
ticipating in the computation (an index between 1 and n the number of agents). Then,
each agent use the same algorithm that determines which coalition values they need
to compute, removing the need of any further communication, except announcing the
result of the computation. The index is used to compute a set of coalitions and ensures
that the values of all the coalitions are computed exactly once. This algorithm, called
DCVC [7] outperforms the one by Shehory and Kraus. To minimize the overall time of
computation, it is best to balance the work of all the agents. The key observation is that
in general, it should take longer to compute the value of a large coalition compared to a
small coalition (i.e., the computational complexity is likely to increase with the size of
the coalition since more agents have to coordinate their activities). Their method im-
proves the balance of the loads by distributing coalitions of the same size to all agents.
By knowing the number of agents n participating in the computation an index number
(i.e., an integer in the range {0..n}), the agents determine for each coalition size which
coalition values to compute. The algorithm can also be adapted when the agents have
different known computational speed so as to complete the computation in a minimum
amount of time.
11.2 Searching for the optimal coalition structure
Once the value of each coalition is known, the agents needs to search for an optimal
CS. The difculty of this search lies in the large search space, as recognised by existing
algorithms, and this is even more true in the case where there exists externalities (i.e.,
when the valuation of a coalition depends on the CS). For TU games with no external-
ities, some algorithms guarantee nding CSs within a bound from the optimum when
an incomplete search is performed. Unfortunately, such guarantees are not possible for
games with externalities. We shortly discuss these two cases in the following.
11.2.1 Games with no externalities
Anytime algorithms
Sandholm et al. [15] proposed a rst algorithm that searches through a lattice as
presented in Figure 11.1. Their algorithm guarantees that the CS found, s, is within
a bound from the optimal s
when a sufcient portion of the lattice has been visited.

To ensure any bound, it is necessary to visit a least 2
n1
CSs (Theorems 1 and 3 in
[15]) which corresponds to the rst two levels of the lattice, i.e., the algorithm needs
to visit the grand coalition and all the CSs composed of 2 coalitions. Let S
be the
best CS found in the rst two levels, then we have v(s
) n v(S
). To see this,
let C
max
a coalition with the highest value (i.e. C
max
argmax
{CN}
v(C). It is clear
that v(s
) n v(C
max
) as each coalition forming the CS s
has a most the value of

v(C
max
) and there are at most n coalitions in s
. Since all coalitions are part of these

levels, it is clear that we have v(C
max
) v(S
). Finally, we have v(s
) n v(S
),
which was what we wanted.
The bound improves each time a new level is visited. An empirical study of differ-
ent strategies for visiting the other levels is presented in [4]. Three different algorithms
are empirically tested over characteristic functions with different properties: 1) sub-
additive, 2) superadditive, 3) picked from a uniform distribution in [0, 1] or in [0, |S|]
(where |S| is the size of the coalition). The performance of the heuristics differs over
the different type of valuation functions, demonstrating the importance of the proper-
ties of the characteristic function in the performance of the search algorithm.
The algorithm by Dang and Jennings [3] improves the one of [15] for low bounds
from the optimal. For large bounds, both algorithms visit the rst two levels of the
lattice. Then, when the algorithm by Sandholm et al. continues by searching each
level of the lattice, the algorithm of Dang and Jennings only searches specic subset of
each level to decrease the bound faster. This algorithm is anytime, but its complexity
is not polynomial.
These algorithms were based on a lattice as the one presented in Figure 11.1 where
a CS in level i contains exactly i coalitions. The best algorithm to date has been
developed by Rahwan et al. and uses a different representation called integer-partition
(IP) of the search space. It is an anytime algorithm that has been improved over a series
11.2. Searching for the optimal coalition structure 113
of paper: [11, 12, 8, 9, 13]. In this representation the CSs are grouped according to
the sizes of the coalitions they contain, which is called a conguration. For example,
for a population of four agents, the conguration {1, 3} represents CSs that contain a
coalition with a singleton and a coalition with three agents. A smart scan of the input
allows to search the CSs with two coalitions the grand coalition and the CS containing
singletons only. In addition, during the scan, the algorithm computes the average and
maximum value for each coalition size. The maximum values can be used to prune the
search space. When constructing a conguration, the use of the maximum values of a
coalition for each size permits the computation of an upper bound of the value of a CS
that follows that conguration, and if the value is not greater than the current best CS,
it is not necessary to search through the CSs with that conguration, which prunes the
search tree. Then, the algorithm searches the remaining congurations, starting with
the most promising ones. During the search of a conguration, a branch and bound
technique is used. In addition, during the search, the algorithm is designed so that no
CS is evaluated twice. Empirical evaluation shows that the algorithm outperforms any
other current approach over different distributions used to generate the values of the
coalitions.
dynamic programming
Another approach is to use dynamic programming technique. The key idea is pro-
vided in the following lemma: in order to compute the optimal value of a CS, it sufces
to consider partitions of N into two disjoints coalitions and apply the argument recur-
sively. To help us, let us recall the denition of the supeadditive cover (N, v) of a
TU game (N, v). The valuation function v is v(C) = max
PS
C
TP
v(T)
for all
C N \ and v() = 0. The set of optimal CSs can now be noted argmax v(N). Let
us now state the key lemma:
11.2.1. LEMMA. For any C N, we have
v(C) = max {max { v(C
) + v(C
) | C
= C C
= C
, C
= } , v(C)} .
Proof. Clearly, v(C) v(C). Take two disjoint non-empty coalitions C
and C
such
that C
= C. Let S
and S
be two partitions of C
and C
such that v(C
) = v(S
)
and v(C
) = v(S
). Then S
is a CS over C with v(S
) = v(S
) +v(S
), so
we must have v(C) v(C
) + v(C
).
Now, let S be a partition of C such that v(C) = v(S). If S = {C}, then we are
done. Otherwise, let C
be a coalition in S, C
= C \ C
and S
be S \ {C
}. Since S
is a CS over C
, we have v(C
) v(S
) = v(S) v(C
). On the other hand, we have

v(C
) v(C
). Hence v(C
) + v(C
) v(S) = v(C).
More recently, [17, 18] designed an algorithm that uses dynamic programming and
that guarantees a constant factor approximation ratio r in a given time. In particular,
the latest algorithm [17] guarantees a factor of
1
8
in O(2
n
).
Other approaches
Some algorithms are now trying to combine an anytime approach and an dynamics
programming. Other researchers try to use different techniques. For example, Silaghi
et al [20] propose to use a different representation, assuming that the value of a coali-
tion is the optimal solution of a distributed constraint optimization problem (DCOP).
The algorithm uses a DCOP solver and guarantees a bound from the optimum.
The algorithms above assume that the TUgame is represented in a naive way. There
exists some algorithms that take advantage of compact representation. For example,
[6] proposes algorithms in the case where the game is represented using an MC-nets
and in the case where the synergy coalition group is used. Another example is [1] for
skill games.
11.2.2 Games with externalities
The previous algorithm explicitly uses the fact that the valuation function only depends
on the members of the coalition, i.e., has no externalities. When this is not the case,
i.e., when the valuation function depends on the CS, it is still possible to use some al-
gorithms, e.g., the one proposed in [4], but the guarantee of being within a bound from
the optimal is no longer valid. Sen and Dutta use genetic algorithms techniques [16] to
perform the search. The use of such technique only assumes that there exists some un-
derlying patterns in the characteristic function. When such patterns exist, the genetic
search makes a much faster improvement in locating higher valued CS compared to
the level-by-level search approach. One downside of the genetic algorithm approach
is that there is no optimality guarantee. Empirical evaluation, however, shows that the
genetic algorithm does not take much longer to nd a solution when the value of a
coalition does depend on other coalitions.
More recently, Rahwan et al. and Michalak et al. consider the problem for some
class of externalities and modify the IP algorithm for the games with externalities [5,
10], however, they assume games with negative or positive spillovers. [2] introduce a
representation to represent games in partition function games using types: each agent
has a single type. They make two assumptions on the nature of the externalities (based
on the notions of competition and complementation) and they show that games with
negative or positive spillovers are special cases. They provide a branch and bound
algorithm for the general setting. They also provide a worst-case initial bound.
Bibliography
[1] Yoram Bachrach, Reshef Meir, Kyomin Jung, and Pushmeet Kohli. Coalitional
structure generation in skill games. In Proceedings of the Twenty-Fourth AAAI
Conference on Articial Intelligence (AAAI-10), pages 703708, July 2010.
[2] Bikramjit Banerjee and Landon Kraemer. Coalition structure generation in multi-
agent systems with mixed externalities. In Proceedings of the 9th International
Conference on Autonomous Agents and Multiagent Systems: volume 1 - Volume
1, AAMAS 10, pages 175182, Richland, SC, 2010. International Foundation
for Autonomous Agents and Multiagent Systems.
[3] Viet Dung Dang and Nicholas R. Jennings. Generating coalition structures
with nite bound from the optimal guarantees. In Proceedings of the third
International Joint Conference on Autonomous Agents and Multiagent Sys-
tems(AAMAS04), 2004.
[4] Kate S. Larson and Tuomas W. Sandholm. Anytime coalition structure gener-
ation: an average case study. Journal of Experimental & Theoretical Articial
Intelligence, 12(1):2342, 2000.
[5] Tomasz Michalak, Andrew Dowell, Peter McBurney, and Michael Wooldridge.
Optimal coalition structure generation in partition function games. In Proceeding
of the 2008 conference on ECAI 2008, pages 388392, Amsterdam, The Nether-
lands, The Netherlands, 2008. IOS Press.
[6] Naoki Ohta, Vincent Conitzer, R. Ichimura, Y. Sakurai, and Makoto Yokoo.
Coalition structure generation utilizing compact characteristic function represen-
tations. In Proocedings of the 15th International Conference on Principles and-
Practice of Constraint Programming (CP09), pages 623638, 2009.
[7] Talal Rahwan and Nicholas R. Jennings. An algorithm for distributing coali-
tional value calculations among cooperating agents. Articial Intelligence, 171(8-
9):535567, 2007.
115
116 Bibliography
[8] Talal Rahwan and Nicholas R. Jennings. Coalition structure generation: Dynamic
programming meets anytime optimization. In Proceedings of the 23rd conference
on articial intelligence (AAAI-08), pages 156161, 2008.
[9] Talal Rahwan and Nicholas R. Jennings. An improved dynamic programming al-
gorithm for coalition structure generation. In Proceedings of the 7th international
conference on Autonomous Agents and Multi-Agent Systems (AAMAS-08), 2008.
[10] Talal Rahwan, Tomasz Michalak, Nicholas R. Jennings, Michael Wooldridge,
and Peter McBurney. Coalition structure generation in multi-agent systems with
positive and negative externalities. In Proceedings of the 21st International Joint
Conference on Articial Intelligence (IJCAI-09), 2009.
[11] Talal Rahwan, Sarvapali D. Ramchurn, Viet Dung Dang, Andrea Giovannucci,
and Nicholas R. Jennings. Anytime optimal coalition structure generation. In
Proceedings of the Twenty-Second Conference on Articial Intelligence (AAAI-
07), pages 11841190, 2007.
[12] Talal Rahwan, Sarvapali D. Ramchurn, Viet Dung Dang, and Nicholas R. Jen-
nings. Near-optimal anytime coalition structure generation. In Proceedings of the
Twentieth International Joint Conference on Articial Intelligence (IJCAI07),
pages 23652371, January 2007.
[13] Talal Rahwan, Sarvapali D. Ramchurn, Nicholas R. Jennings, and Andrea Gio-
vannucci. An anytime algorithm for optimal coalition structure generation. Jour-
nal of Articial Intelligence Research, 34:521567, 2009.
[15] Tuomas W. Sandholm, Kate S. Larson, Martin Andersson, Onn Shehory, and
Fernando Tohm. Coalition structure generation with worst case guarantees. Ar-
ticial Intelligence, 111(12):209238, 1999.
[16] Sandip Sen and Partha Sarathi Dutta. Searching for optimal coalition structures.
In ICMAS 00: Proceedings of the Fourth International Conference on Multi-
Agent Systems (ICMAS-2000), page 287, Washington, DC, USA, 2000. IEEE
Computer Society.
[17] Travis Service and Julie Adams. Approximate coalition structure generation.
In Proceedings of the Twenty-Fourth AAAI Conference on Articial Intelligence
(AAAI-10), pages 854859, July 2010.
[18] Travis Service and Julie Adams. Constant factor approximation algorithms for
coalition structure generation. Autonomous Agents and Multi-Agent Systems,
pages 117, 2010. Published online February 2010.
Bibliography 117
[19] Onn Shehory and Sarit Kraus. Methods for task allocation via agent coalition
formation. Articial Intelligence, 101(1-2):165200, May 1998.
[20] Suguru Ueda, Atsushi Iwasaki, Makoto Yokoo, Marius Calin Silaghi, Katsutoshi
Hirayama, and Toshihiro Matsui. Coalition structure generation based on dis-
tributed constraint optimization. In Proceedings of the Twenty-Fourth AAAI Con-
ference on Articial Intelligence (AAAI-10), pages 197203, July 2010.
Lecture 12
Issues for applying cooperative games
We now highlight issues that have emerged from the different types of applications
(e.g. resource or task allocation problem or forming a buying group). Some of the
issues have solutions while others remain unsolved, for example, dealing with agents
that can enter and leave the environment at any time in an open, dynamic environment.
None of the current protocols can handle these issues without re-starting computation,
and only few approaches consider how to re-use the already computed solution [6, 13].
12.1 Stability and Dynamic Environments
Real-world scenarios often present dynamic environments. Agents can enter and leave
the environment at any time, the characteristics of the agents may change with time,
the knowledge of the agents about the other agents may change, etc.
The game-theoretic stability criteria are dened for a xed population of agents and
the introduction of a new agent in the environment requires signicant computation to
update a stable payoff distribution. For example, for the kernel, all the agents need to
check whether any coalition that includes the new agent changes the value of the max-
imum surplus, which requires re-evaluating O(2
n
) coalitions. Given the complexity
of the stability concept, one challenge that is faced by the multiagent community is to
develop stability concepts that can be easily updated when an agent enters or leaves
the environment.
In addition, if an agent drops during the negotiation, this may cause problems for
the remaining agents. For example, a protocol that guarantees a kernel stable payoff
distribution is shown not to be safe when the population of agents is changing: if an
agent i leaves the formation process without notifying other agents, the other agents
may complete the protocol and nd a solution to a situation that does not match the
reality. Each time a new agent enters or leaves the population, a new process needs to
be restarted [9].
In an open environments, manipulations will be impossible to detect: agents may
use multiple identiers (or false names) to pretend to be multiple agents, or the other
119
120 Lecture 12. Issues for applying cooperative games
way around, multiple agents may collude and pretend to be a single agents, or agents
can hide some of their skills. Hence, it is important to propose solution concepts that
are robust against such manipulations. We will come back later to some of the solution
that have been proposed: the anonymity-proof core [44] and anonymity-proof Shapley
value [35].
12.2 Uncertainty about Knowledge and Task
In real-world scenario, agents will be required to handle some uncertainty. Different
sources of uncertainty have been considered in the literature:
the valuation function is an approximation [38] and agents may not use the same
algorithm. Hence, the agents may not know what is the true value.
agents may not know some tasks [9] or the value of some coalitions. In such
cases, the agents play a different coalitional game that may reduce the payoff of
some agents compared to the solution of the true game.
some information is private, i.e., an agent knows some property about itself, but
does not know it for other agents. In [28], it is the cost incurred by other agents
to perform a task that is private. In [16, 17], agents have a private type, and the
valuation function depends on the types of the coalitions members.
uncertainty about the outcome of an action [16]: when a coalition makes an
action, some external factors may inuence the outcome of the actions. This can
be captured by a probability of an outcome given the action taken and the type
of the members of the coalition.
there are multiple possible worlds [24], which models the different possible out-
comes of the formation of a coalition. Agents know a probability distribution
over the different worlds. In addition, an agent may not able to distinguish some
worlds as it lacks information and they know a partition of the worlds (called
information sets), each set of the partition represent worlds that appears as indis-
tinguishable.
Some authors also consider that there is uncertainty in the valuation function with-
out modeling a particular source, for example in [25], each agent has an expectation
of the valuation function. In [10, 11] fuzzy sets are used to represent the valuation
function. In the rst paper, the agents enters bilateral negotiations to negotiate Shapley
value, in the second paper, they dene a fuzzy version of the kernel.
In the uncertainty model of [24], the denition of the core depends on the time
one reasons about it. They proposed three different denitions of the core that de-
pend on the timing of the evaluation: before the world is drawn or ex-ante, not much
information can be used; after the world is drawn but before it is known, also called
12.3. Safety and Robustness 121
ex-interim, an agent knows to which set of its information set the real world belongs,
but does not know which one; nally when the world is announced to the agent or
ex-post, everything is known.
The model of [16] combines uncertainty about the agent types and uncertainty
about the outcome of the action taken by the coalition. Each agent has a probabilistic
belief about the types of the other agents in the population. Chalkiadakis and Boutilier
propose a denition of the core, the Bayesian core (introduced in [14]) in which no
agent has the belief that there exists a better coalition to form. As it may be difcult to
obtain all the probabilities and reason about them, [17] propose to use a point belief:
an agent guesses the type of the other agents and reason with these guesses. The paper
analyses the core, simple games (proving that the core of a simple game is non-empty
iff the game has a veto player) and some complexity result in this games with belief.
12.3 Safety and Robustness
It is also important that the coalition formation process is robust. For instance, com-
munication links may fail during the negotiation phase. Hence, some agents may miss
some components of the negotiation stages. This possibility is studied in [9] for the
KCA protocol [27]: coalition negotiations are not safe when some agents become
unavailable (intentionally or otherwise). In particular, the payoff distribution is not
guaranteed to be kernel-stable. [6] empirically studies the robustness of the use of a
central algorithm introduced in [5]: the cost to compute a task allocation and payoff
distribution in the core is polynomial, but it can still be expensive. In the case of agent
failure, the computation needs to be repeated. Belmonte et al. propose an alterna-
tive payoff division model that avoids such a re-computation, but the solution is no
longer guaranteed to be in the core, it is only close to the core. There is a trade-off
between computational efciency and the utility obtained by the agent. They conclude
that when the number of agents is small, the loss of utility compared to the optimal is
small; hence, the improvement of the computational efciency can be justied. For a
larger number of agents, however, the loss of utility cannot not justify the improvement
in computational cost.
12.3.1 Protocol Manipulation
When agents send requests to search for members of a coalition or when they accept to
form a coalition, the protocol may require disclosure of some private information [36].
When the agents reveal some of their information, the mechanism must ensure that
there is no information asymmetry that can be exploited by some agents [7]. To protect
a private value, some protocol [9] may allow the addition of a constant offset to the
private value, as long as this addition does not impact the outcome of the negotiation.
Belmonte et al. study the effect of deception and manipulation of their model
in [6]. They show that some agents can benet from falsely reporting their cost. In
some other approaches [9, 20], even if it is theoretically possible to manipulate the
protocol, it is not possible in practice as the computational complexity required to
ensure higher outcome to the malevolent agent is too high. For example, [20] show that
manipulating marginal-contribution based value division scheme is NP-hard (except
when the valuation function has other properties, such as being convex).
Other possible protocol manipulations include hiding skills, using false names, col-
luding, etc. The traditional solution concepts can be vulnerable to false names and to
collusion [44]. To address these problems, it is benecial to dene the valuation func-
tion in terms of the required skills instead of dening it over the agents: only skills,
not agents, should be rewarded by the characteristic function. In that case, the solution
concept is robust to false names, collusion, and their combination. But the agents can
have incentive to hide skills. A straight, naive decomposition of the skills will increase
the size of the characteristic function, and [45] propose a compact representation in
this case.
12.4 Communication
While one purpose of better negotiation techniques may be to improve the quality of
the outcome for the agents, other goals may include decreasing the time and the number
of messages required to reach an agreement. For example, learning is used to decrease
negotiation time in [41]. The motivation Lermans work in [30] is to develop a coalition
formation mechanism that has low communication and computation cost. In another
work, the communication costs are included in the characteristic function [42].
The communication complexity of some protocols has been derived. For instance,
the exponential protocol in [40] and the coalition algorithm for forming Bilateral Shap-
ley Value Stable coalition in [26] have communication complexity of O(n
2
), the nego-
tiation based protocol in [40] is O(n
2
2
n
), and it is O(n
k
) for the protocol in [39] (where
k is the maximum size of a coalition). The goal of [37] is to analyse the communica-
tion complexity of computing the payoff of a player with different stability concepts:
they nd that it is (n) when either the Shapley value, the nucleolus, or the core are
used.
12.5 Scalability
When the population of heterogeneous agents is large, discovering the relevant agents
to perform a task may be difcult. In addition, if all agents are involved in the coalition
formation process, the cost in time and computation will be large. To alleviate this
scalability issue, a hierarchy of agents can be used [1]. When an agent discovers a
task that can be addressed by agents below this agent in the hierarchy, the agent picks
the best of them to perform the task. If the agents below cannot perform the task, the
agent passes the task to the agent above it in the hierarchy and the process repeats. The
12.6. Long Term vs. Short Term 123
notion of clans [22] and congregations [12], where agents gather together for a long
period have been proposed to restrict the search space by considering only a subset of
the agents (see Section 12.6).
12.6 Long Term vs. Short Term
In general, a coalition is a short-lived entity that is formed with a purpose in mind
and dissolve when that need no longer exists, the coalition ceases to suit its designed
purpose, or critical mass is lost as agents depart [23]. It can be benecial to consider
the formation of long term coalitions, or the process of repeated coalition formation
involving the same agents. [43] explicitly study long term coalitions, and in particular
the importance of trust in this content. [12] refer to a long term coalition as a con-
gregation. The purpose of a congregation is to reduce the number of candidates for
a successful interaction: instead of searching the entire population, agents will only
search in the congregation. The goal of a congregation is to gather agents, with similar
or complementary expertise to perform well in an environment in the long run, which
is not very different from a coalition. The only difference is that group rationality is
not expected in a congregation. The notion of congregation is similar to the notion of
clans [22]: agents gather not for a specic purpose, but for a long-term commitment.
The notion of trust is paramount in the clans, and sharing information is seen as another
way to improve performance.
12.7 Fairness
Stability does not necessarily imply fairness. For example, let us consider two CSs S
and T with associated kernel-stable payoff distribution x
S
and x
T
. Agents may have
different preferences between the CSs. It may even be the case that there is no CS
that is preferred by all agents. If the optimal CS is formed, some agents, especially if
they are in a singleton coalition, may suffer from the choice of this CS. [3] propose
a modication of the kernel to allow side-payment between coalitions to compensate
such agents.
[2] consider partition function games with externalities. They consider a process
where, in turns, agents change coalition to improve their immediate payoff. They
propose that the agents share the maximal social welfare, and the size of the share is
proportional to the expected utility of the process. The payoff obtained is guaranteed
to be at least as high as the expected utility. They claim that using the expected utility
as a base of the payoff distribution provides some fairness as the expected utility can
be seen as a global metric of an agent performance over the entire set of possible CSs.
12.8 Overlapping Coalitions
It is typically assumed that an agent belongs to a single coalition; however, there are
some applications where agents can be members of multiple coalitions. For instance,
the expertise of an agent may be required by different coalitions at the same time,
and the agent can have enough resources to be part of two or more coalitions. In
a traditional setting, the use of the same agent i by two coalitions C
1
and C
2
would
require a merge of the two coalitions. This larger coalition U is potentially harder to
manage, and a priori, there would not be much interaction between the agents in C
1
and C
2
, except for agent i. Another application that requires the use of overlapping
coalition is tracking targets using a sensor networks [21]. In this work, a coalition is
dened for a target, and as agents can track multiple targets at the same time, they can
be members of different coalitions.
The traditional stability concepts do not consider this issue. One possibility is for
the agent to be considered as two different agents, but this representation is not satis-
factory as it does not capture the real power of this agent. Shehory and Kraus propose
a setting with overlapping coalition [39]: Each agent has a capacity, and performing a
task may use only a fraction of the agents capacity. Each time an agent commits to a
task, the possible coalitions that can perform a given task can change. A mapping to
a set covering problem allows to nd the coalition. However, the study of the stability
is not considered. Another approach is the use of fuzzy coalition [8]: agents can be
members of a coalition with a certain degree that represents the risk associated with
being in that coalition. Other work considers that the agents have different degree of
membership, and their payoff depends on this degree [4, 31, 34]. The protocols in [29]
also allow overlapping coalitions.
More recently, [19]
1
have studied the notion of the core in overlapping coalition
formation. In their model, each agent has one resource and the agent contributes a
fraction of that resource to each coalition it participates in. The valuation function v
is then [0, 1]
n
R. A CS is no longer a partition of the agents: a CS S is a nite
list of vectors, one for each partial coalition, i.e., S = (r
1
, . . . , r
k
). The size of S
is the number of coalitions, i.e., k. The support of r
C
S (i.e., the set of indices
i N such that r
C
i
= 0) is the set of agents forming coalition C. For all i N
and all coalition C S, r
C
i
[0, 1]
n
represents the fraction of resource that agent i
contributes to coalition C; hence,
CS
r
C
i
1 (i.e., agent i cannot contributes more
than 100% of its resource). A payoff distribution for a CS S of size k is dened by
a k-tuple x = (x
1
, . . . , x
k
) where x
C
is the payoff distribution that the agents obtain
for coalition C. If an agent is not in the coalition, it must not receive any payoff for
this coalition, hence (r
C
i
= 0) (x
C
i
= 0). The total payoff of agent i is the sum of
its payoffs over all coalitions p
i
(CS, x) =
k
C=1
x
C
i
. The efciency criterion becomes
r
C
S,
iN
x
C
i
= v(r
C
). An imputation is an efcient payoff distribution that is
also individually rational. We denote by I(S) the set of all imputations for the CS S.
1
An earlier version is [18]
12.9. Trust 125
We are now ready to dene the overlapping core. One issue is the kind of permis-
sible deviations: when an agent deviates, she can completely leave some coalitions,
reduce her contribution in other coalitions, or contributes to new coalitions. If she
stills contribute to a coalition containing non-deviating agents, how should they be-
have? They rst may refuse to give any payoff to the deviating agent, as she is seen as
not trustworthy. Agents that are not affected by the deviation may, however, consider
that the deviators agents did not fail them, and consequently, they may continue to
share payoffs with the deviators. A last case occurs when the deviators are decreasing
their implication in a coalition. This coalition may no longer perform the same tasks,
but it can still perform some. If there is enough value to maintain the payoff of the
non-deviators, the deviators may be allowed to share the surplus generated. Each of
these behaviors give raise to different types of deviations, and consequently, different
denition of a core: the conservative core, the rened core and the optimistic core.
The paper also provides a characterization of conservative core, properties of the dif-
ferent core, including a result showing that convex overlapping coalitional games have
a non-empty core.
12.9 Trust
The notion of trust can be an important metric to determine whom to interact with. This
is particularly important when the coalition is expected to live for a long term. In [7],
an agent computes a probability of success of a coalition, based on a notion of trust
which can be used to eliminate some agents from future consideration. This probability
is used to estimate the value of different coalitions and help the agent in deciding which
coalition to join or form. In [43], the decision to leave or join a coalition is function of
the trust put in other agents. In this paper, the concept of trust is dened as a belief that
agents will have successful interaction in the future; hence, trust is used to consider a
subset of the entire population of agents for the formation of future coalitions. Trust is
used to compute coalitions, but agents do not compute a payoff distribution. Another
work that emphasises trust is [22] which introduces the concept of clans. A clan is
formed by agents that trust each other with long-term commitments. Given the trust
and an estimate of local gain, agents can accept to join a clan. The idea behind this
work is that agents that trust each other will be collaborative. Moreover, when an
agent needs to form a coalition of agents, it will only search partners in the clan, which
reduces the search space. Trust can therefore be very effective for scaling up in large
society of agents.
12.10 Learning
When agents have to repeatedly form coalitions in the presence of the same set of
agents, learning can be used to improve performance of the coalition formation process
both in terms of speed of the process and in terms of better valuation.
A basic model of iteratively playing many coalitional games is presented in [32]:
at each time step, a task is offered to agents that are already organised into coalitions.
The task is awarded to the best coalition. The model is made richer in [33] where the
agents can estimate the value of a coalition and have a richer set of actions: as the
agents can re members from a coalition, join a different coalition, or leave a coalition
to replace some agents in a different coalition. However, in both works, the agents are
not learning, they have a set of static strategies. Empirical experiments compare the
results over populations using either the same strategy or a mix of strategies.
Chalkiadakis and Boutilier also consider a repeated coalition formation problem[14,
15, 16]. The setting is a task allocation problem where agents know their own types
(i.e., skill to perform some type of tasks), but do not know the ones of other agents in
the population. Each time a coalition is formed, the agents will receive a value for that
coalition. From the observation of this value, the agents can update a belief about the
types of other agents. When an agent is reasoning about which coalition to form, it
uses its beliefs to estimate the value of the coalition. This problem can be formulated
using a POMPD (Partially observable Markov Decision Process) where the agents are
maximising the long-termvalue of their decision over the repetition of the coalition for-
mation process. Solving a POMPD is a difcult task, and the POMPD for the coalition
formation problem grows exponentially with the number of agents. In [14], a myopic
approach is proposed. More recently, Chalkiadakis and Boutilier propose additional
algorithms to solve that POMPD, and empirically compare the solutions [15].
Bibliography
[1] Sherief Abdallah and Victor Lesser. Organization-based cooperative coalition
formation. In IAT 04, 2004.
[2] Stphane Airiau and Sandip Sen. A fair payoff distribution for myopic rational
agents. In Proceedings of the Eighth International Conference on Autonomous
Agents and Multiagent Systems (AAMAS-09), May 2009.
[3] Stphane Airiau and Sandip Sen. On the stability of an optimal coalition struc-
ture. In Proceedings of the 19th European Conference on Articial Intelligence
(ECAI-2010), pages 203208, August 2010.
[4] J.-P. Aubin. Mathematical Methods of Game and Economic Theory. North-
Holland, 1979.
[5] Mara-Victoria Belmonte, Ricardo Conejo, Jos-Luis Prez de-la Cruz, and Fran-
cisco Triguero Ruiz. A stable and feasible payoff division for coalition formation
in a class of task oriented domains. In ATAL 01: Revised Papers from the 8th
International Workshop on Intelligent Agents VIII, pages 324334, London, UK,
2002. Springer-Verlag.
[6] Mara-Victoria Belmonte, Ricardo Conejo, Jos-Luis Prez de-la Cruz, and Fran-
cisco Triguero Ruiz. A robust deception-free coalition formation model. In Pro-
ceedings of the 2004 ACM symposium on Applied computing (SAC 04), pages
469473, New York, NY, USA, 2004. ACM Press.
[7] Bastian Blankenburg, Rajdeep K. Dash, Sarvapali D. Ramchurn, Matthias
Klusch, and Nicholas R. Jennings. Trusted kernel-based coalition formation. In
Proceedings of the fourth international joint conference on Autonomous agents
and multiagent systems, pages 989996, New York, NY, USA, 2005. ACM Press.
[8] Bastian Blankenburg, Minghua He, Matthias Klusch, and Nicholas R. Jennings.
Risk-bounded formation of fuzzy coalitions among service agents. In Proceed-
ings of 10th International Workshop on Cooperative Information Agents, 2006.
127
128 Bibliography
[9] Bastian Blankenburg and Matthias Klusch. On safe kernel stable coalition form-
ing among agents. In Proceedings of the third International Joint Conference on
Autonomous Agents and Multiagent Systems, 2004.
[10] Bastian Blankenburg and Matthias Klusch. Bsca-f: Efcient fuzzy valued stable
coalition forming among agents. In Proceedings of the IEEE International Con-
ference on Intelligent Agent Technology (IAT). IEEE Computer Society Press,
2005.
[11] Bastian Blankenburg, Matthias Klusch, and Onn Shehory. Fuzzy kernel-stable
coalitions between rational agents. In Proceedings of the second international
joint conference on Autonomous agents and multiagent systems (AAMAS-03).
ACM Press, 2003.
[12] Christopher H. Brooks and Edmund H. Durfee. Congregation formation in mul-
tiagent systems. Autonomous Agents and Multi-Agent Systems, 7(1-2):145170,
2003.
[13] Philippe Caillou, Samir Aknine, and Suzanne Pinson. A multi-agent method for
forming and dynamic restructuring of pareto optimal coalitions. In AAMAS 02:
Proceedings of the rst international joint conference on Autonomous agents and
multiagent systems, pages 10741081, New York, NY, USA, 2002. ACM Press.
[14] Georgios Chalkiadakis and Craig Boutilier. Bayesian reinforcement learning for
coalition formation under uncertainty. In Proceedings of the third International
Joint Conference on Autonomous Agents and Multiagent Systems(AAMAS04),
2004.
[15] Georgios Chalkiadakis and Craig Boutilier. Sequential decision making in re-
peated coalition formation under uncertainty. In Proc. of the 7th Int. Conf. on
Autonomous Agents and Multiagent Systems (AAMAS-08), May 2008.
[16] Georgios Chalkiadakis and Craig Boutilier. Sequentially optimal repeated coali-
tion formation. Autonomous Agents and Multi-Agent Systems, to appear, 2010.
Published online November 2010.
[17] Georgios Chalkiadakis, Edith Elkind, and Nicholas R. Jennings. Simple coali-
tional games with beliefs. In Proceedings of the 21st international jont conference
on Artical intelligence, pages 8590, San Francisco, CA, USA, 2009. Morgan
Kaufmann Publishers Inc.
[18] Georgios Chalkiadakis, Edith Elkind, Evangelos Markakis, and Nicholas R. Jen-
nings. Overlapping coalition formation. In Proceeedings of the 4th Interna-
tional Workshop on Internet and Network Economics (WINE2008), pages 307
321, 2008.
Bibliography 129
[19] Georgios Chalkiadakis, Edith Elkind, Evangelos Markakis, and Nicholas R. Jen-
nings. Cooperative games with overlapping coalitions. Journal of Articial In-
telligence Research, 39:179216, 2010.
[20] Vincent Conitzer and Tuomas Sandholm. Computing shapley values, manipu-
lating value division schemes, and checking core membership in multi-issue do-
mains. In Proceedings of the 19th National Conference on Articial Intelligence
(AAAI-04), pages 219225, 2004.
[21] Viet Dung Dang, Rajdeep K. Dash, Alex Rogers, and Nicholas R. Jennings. Over-
lapping coalition formation for efcient data fusion in multi-sensor networks. In
Proceedings of the Twenty-First Conference on Articial Intelligence (AAAI-06),
pages 635640, 2007.
[22] Nathan Grifths and Michael Luck. Coalition formation through motivation and
trust. In Proceedings of the second international joint conference on Autonomous
agents and multiagent systems (AAMAS03), New York, NY, USA, 2003. ACM
Press.
[23] Bryan Horling and Victor Lesser. A survey of multi-agent organizational
paradigms. The Knowledge Engineering Review, 19:281316, 2004.
[24] Samuel Ieong and Yoav Shoham. Bayesian coalitional games. In Proceedings of
the Twenty-Second AAAI Conference on Articial Intelligence (AAAI-08), pages
95100, 2008.
[25] Steven P. Ketchpel. Forming coalitions in the face of uncertain rewards. In Pro-
ceedings of the Eleventh National Conference on Articial Intelligence, pages
414419, August 1994.
[27] Matthias Klusch and Onn Shehory. A polynomial kernel-oriented coalition algo-
rithm for rational information agents. In Proceedings of the Second International
Conference on Multi-Agent Systems, pages 157 164. AAAI Press, December
1996.
[28] Sarit Kraus, Onn Shehory, and Gilad Taase. Coalition formation with uncertain
heterogeneous information. In Proceedings of the second international joint con-
ference on Autonomous agents and multiagent systems, pages 18. ACM Press,
2003.
[29] Hoong Chuin Lau and Lei Zhang. Task allocation via multi-agent coalition
formation: Taxonomy, algorithms and complexity. In 15th IEEE International
130 Bibliography
Conference on Tools with Articial Intelligence (ICTAI 2003), pages 346350,
November 2003.
[30] Katia Lerman and Onn Shehory. Coalition formation for large-scale electronic
markets. In Proceedings of the Fourth International Conference on MultiAgent
Systems (ICMAS-2000). IEEE Computer Society, July 2000.
[31] M. Mares. Fuzzy cooperative games: cooperation with vague expectations. Stud-
ies in fuzziness and soft computing, 72, 2001.
[32] Carlos Mrida-Campos and Steven Willmott. Modelling coalition formation over
time for iterative coalition games. In Proceedings of the third International Joint
Conference on Autonomous Agents and Multiagent Systems(AAMAS04), 2004.
[33] Carlos Mrida-Campos and Steven Willmott. The effect of heterogeneity on
coalition formation in iterated request for proposal scenarios. In Proceedings
of the Forth European Workshop on Multi-Agent Systems (EUMAS 06), 2006.
[34] I. Nishizaki and M. Sakawa. Masatoshi. Fuzzy and multiobjective games for
conict resolution. Studies in Fuzziness and Soft Computing, 64, 2001.
[35] Naoki Ohta, Vincent Conitzer, Yasufumi Satoh, Atsushi Iwasaki, and Makoto
Yokoo. Anonimity-proof shapley value: Extending shapley value for coalitional
games in open environments. In Proc. of the 8th Int. Conf. on Autonomous Agents
and Multiagent Systems (AAMAS-09), May 2009.
[36] Michal P echou cek, Vladimr Mark, and Jaroslav Brta. A knowledge-based ap-
proach to coalition formation. IEEE Intelligent Systems, 17(3):1725, 2002.
[37] Ariel D. Procaccia and Jeffrey S. Rosenschein. The communication complexity
of coalition formation among autonomous agents. In Proceedings of the Fifth
International Joint Conference on Autonomous Agents and Multiagent Systems
AAMAS-06, May 2006.
[39] Onn Shehory and Sarit Kraus. Methods for task allocation via agent coalition
formation. Articial Intelligence, 101(1-2):165200, May 1998.
[40] Onn Shehory and Sarit Kraus. Feasible formation of coalitions among au-
tonomous agents in nonsuperadditve environments. Computational Intelligence,
15:218251, 1999.
[41] Leen-Kiat Soh and Costas Tsatsoulis. Satiscing coalition formation among
agents. In AAMAS 02: Proceedings of the rst international joint conference
on Autonomous agents and multiagent systems, pages 10621063, New York,
NY, USA, 2002. ACM Press.
Bibliography 131
[42] Fernando Tohm and Tuomas Sandholm. Coalition formation processes with
belief revision among bounded-rational self-interested agents. Journal of Logic
and Computation, 9(6):793815, 1999.
[43] Julita Vassileva, Silvia Breban, and Michael Horsch. Agent reasoning mechanism
for making long-term coalitions based on decision making and trust. Computa-
tional Intelligence, 18(4):583595, 2002.
Iwasaki. A compact representation scheme for coalitional games in open anony-
mous environments. In Proceedings of the Twenty First National Conference on
Articial Intelligence, pages . AAAI Press AAAI Press / The MIT Press, 2006.

Lectures in Cooperative Game Theory

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lectures in Cooperative Game Theory

Uploaded by

Copyright:

Available Formats

Lecture 1

Introduction and Denition of TU games

1 when |C| >

, i ,= j, the cost of connecting i and j by the edge e

such that for <

Hence, (Q, z) is a counter-objection.

Using similar arguments, we nd a counter-objection for each other objections (you

the reordering of l in decreasing order. In the ex-

is the maximum excess: the agents involved in the corre-

| x Imp} is a non-empty compact subset of R

| x Imp} is a non-empty compact subset of R

| (x Imp) (y Imp e(x)

)} is non-empty. From this, it

| x Imp} is a non-empty compact subset of

is a continuous function and Imp is a non-empty and compact subset of R

} has at most one element.

}. We now prove that |M

, we have (x) = (y), hence x = y. This contradicts the fact that

} has at most one element.

. Note that for any coalition S N such

be the minimal value of i 1, . . . , M such that there is a coalition (

then // payment should be individually

76 Lecture 7. A Special Class of TU games: Voting Games

, player 3 becomes pivotal for a coalition

9.2.3. PROPOSITION. A TU game can be represented as an NTU game.

when a sufcient portion of the lattice has been visited.

has a most the value of

. Since all coalitions are part of these

). Finally, we have v(s

such that v(C

is a CS over C with v(S

). On the other hand, we have

You might also like