Top Banner
1 Topics in Game Theory Topics in Game Theory SS 2008 SS 2008 Avner Shaked Avner Shaked
42

1 Topics in Game Theory SS 2008 Avner Shaked. 2

Dec 14, 2015

Download

Documents

Marco Padmore
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

1

Topics in Game TheoryTopics in Game Theory

SS 2008SS 2008

Avner ShakedAvner Shaked

Page 2: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

2

http://www.wiwi.uni-bonn.de/shaked/topics/http://www.wiwi.uni-bonn.de/shaked/topics/

Page 3: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

3

K. BinmoreK. Binmore Fun & GamesFun & Games A Text on Game TheoryA Text on Game Theory D.C. Heath & Co., 1992 D.C. Heath & Co., 1992

Page 4: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

4

M. Osborne & A. RubinsteinM. Osborne & A. Rubinstein Bargaining and MarketsBargaining and Markets Academic Press, 1990Academic Press, 1990

Page 5: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

5

K. BinmoreK. Binmore Fun & GamesFun & Games A Text on Game TheoryA Text on Game Theory D.C. Heath & Co., 1992 D.C. Heath & Co., 1992

M. Osborne & A. RubinsteinM. Osborne & A. Rubinstein Bargaining and MarketsBargaining and Markets Academic Press, 1990Academic Press, 1990

Page 6: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

6

A Bargaining Problem

• S - a feasible set• d - a disagreement point

Nash Bargaining TheoryNash Bargaining TheoryNash VerhandlungstheorieNash Verhandlungstheorie

John Nash

d S s S s d , ,

2 is compact & convexS

Page 7: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

7

Nash Bargaining TheoryNash Bargaining Theory2 is compact & convexS

u2

u1

S

Page 8: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

8

Nash Bargaining TheoryNash Bargaining Theory

u2

u1

bounded

closedS

2 is compact & convexS

limn nn

x S x S

Page 9: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

9

αA+ 1 - α B

0 α 1

Nash Bargaining TheoryNash Bargaining Theory

u2

u1

A

BS

2 is compact & convexS

S

A,B S

Page 10: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

10

Nash Bargaining TheoryNash Bargaining Theory

d S s S s d , ,

u2

u1

d

S

2 is compact & convexS

Page 11: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

11

Nash Bargaining TheoryNash Bargaining Theory

d S s S s d , ,

is a bargaining problem< S,d >

is a bargaining problem{ }= < S,d > < S,d >B

u2

u1

d

S

2 is compact & convexS

Page 12: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

12

Nash Bargaining TheoryNash Bargaining Theory

d

A Nash Bargaining Solutionis a function

2:

( , )S d S

f

f

Bu2

u1

S

is a bargaining problem{ }= < S,d > < S,d >B

Page 13: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

13

Nash Bargaining TheoryNash Bargaining Theory

A Nash Bargaining Solutionis a function

2: f B

u2

u1

S

( ) ( )

( )

f S,d f S x | x d ,d

f S x | x d ,d S x | x d

( , )S d dfd

Page 14: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

14

Axioms A1-A4

A1 (Pareto)

if then x > f(S,d) x S

A2 (Symmetry)

d

S

&i i i ii x y i x y x > y

1 2 2 1( , ) ( , )x x x xα

f(S,d)

( , ) α α αf S d f S,d

S

α S

Page 15: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

15

Axioms A1-A4

A3 (Invariance to affine transformation)

A4 (Independence of Irrelevant Alternatives IIAIIA)

1 2 1 2( , ) ( , ) , 0x x x x α

( , ) α α αf S d f S,d

d S T

f T,d S f S,d f T,d

Page 16: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

16

Axioms A1-A4

A4 (Independence of Irrelevant Alternatives IIAIIA)

d S T

f T,d S f S,d f T,d

u2

u1

d

f T,dT S = f S,d

Page 17: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

17

Axioms A1-A4

A4 (Independence of Irrelevant Alternatives IIAIIA)

d S T

f T,d S f S,d f T,d

Gives f(T,d) a flavour of maximum

PastaFishMeat

IIA IIA is violated whenis violated when

Page 18: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

18

A1 ParetoA2 SymmetryA3 InvarianceA4 IIA

A1 ParetoA2 SymmetryA3 InvarianceA4 IIA

First, we show that there exists a function satisfying the axioms.

There exists a unique

satisfying A1- A4

2: f BTheorem:

Proof:

Page 19: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

19d

,S,dFor any given bargaining problem

define g S,d

S

=

= Does such a point always exist ??Is it unique ??

Yes !!!

Proof:

Page 20: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

20

?

d

g S,d

S

=

=

Proof:

does satisfy A1-A4 ?? g S,d

Pareto

Symmetry

IIA

Invariance

Page 21: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

21

(divide the $)

0

Proof:

Consider the bargaining problem ,Δ,0 f S,d g S,d

Uniqueness: If satisfies the axioms then: f S,d

(1,0)

(0,1)

1 12 2f ,0 ,By Pareto + Symmetry:

A2 (Symmetry)

1 2 2 1( , ) ( , )x x x xα

( , ) α α αf S d f S,d

( , ) α α αf 0 f ,0 f ,0

By definition: 1 12 2g ,0 ,

1 12 2f ,0 g ,0 ,

Page 22: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

22

Proof:

For a given bargaining problem ,S,d

d

S

g S,d

(a,d2)

(d1,b)

(d1, d2)

=

=

Page 23: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

23

Proof:

For a given bargaining problem ,S,d

d

S

g S,d f S,d

If is a degenerate ProblemS,d

Page 24: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

24

Proof:

For a given (nondegenerate) bargaining problem ,S,d

0 (1,0)

(0,1)

d

S

g S,d

(a,d2)

(d1,b)

(d1, d2)

Consider the bargaining problem ,Δ,0Find an affine transformation α

α

α

α

Page 25: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

25

Proof:

0 (1,0)

(0,1)

d

S

g S,d

(a,d2)

(d1,b)

(d1, d2)

Find an affine transformation α

α

α

α

α x, y = μx + ν, μy + ν

1 2α 0,0 = ν,ν = d ,d 2α 1,0 = μ + ν ν = a,d, 1α 0,1 = ν, μ ν = d+ ,b

,1 2

1 2

ν,ν

μ, μ

= d ,d

a - d b - d

,1 2a - d b - d > 0,0

Page 26: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

26

Proof:

0 (1,0)

(0,1)

d

S

g S,d

(a,d2)(d1, d2)

α

α

α Δ = S

S

αf Δ,0 = f α ,α 0

f S,d 1 12 2α ,

1 12 2,

==

(d1,b)α

g S,d??

Page 27: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

27

Proof:

0 (1,0)

(0,1)

d

S

g S,d

(a,d2)(d1, d2)

α

α

S

αf Δ,0 = f α ,α 0

f S,d 1 12 2α ,

1 12 2,

==

(d1,b)α

g S,d??

1 12 2α , 1 1

2 2α 1,0 + 0,1

1 1 1 12 12 2 2 2α 1,0 + α 0,1 a,d d ,b

= g S,d

g S,d f S,d S

Page 28: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

28

Proof:

d

S

g S,d f S,dS

g S,d f S,d S

By IIA f S,d f S,d

g S,d f S,d

f S,d

Page 29: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

29

Proof:

d

S

g S,d f S,d

g S,d f S,d

f S,d

satisfying

g f,

! f A1 - A4

end of proof

Page 30: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

300

A Generalization

Changing A2 (Symmetry)

A2 (nonsymmetric)

for some f(Δ,0) = α,1 - α 0 < α < 1.

Δ

(1,0)

(0,1)

A

B

α measures the strength of Player 1

α,1 - α

A

B

1 - α=

α

Page 31: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

31d

,S,dFor any given bargaining problem

define g S,d

SA

B

With the new A2, define a different g S,d

A

B

1 - α=

α

Page 32: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

32d

g S,dS

Does such a point always exist ??Is it unique ??

Yes !!!

A

B

Following the steps of the previous theorem,

g S,d is the unique function satisfying the 4 axioms.

Yes !!!Yes !!!

A

B

1 - α=

α

Page 33: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

33

A brief mathematical InterludeA brief mathematical Interlude

Consider the (implicit) functionα 1-αx y = K

Find a tangent at a point (x0,y0) on the curve

α-1 1-α α -ααx y + 1 - α x y y = 0

0 0αy + 1 - α x y = 0

0

0

αyy = -

1 - α x

differentiating

y

x

Page 34: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

34

A brief mathematical InterludeA brief mathematical Interlude

Find a tangent at a point (x0,y0) on the curve

0

0

αy-

1 - α x0

0

y - y=

x - x

x

y

The tangent’s equation:

The intersections with the axis (x=0, y=0)

0 0x y,0 , 0,

α 1 - α

Page 35: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

35

A

B

A brief mathematical InterludeA brief mathematical Interlude

x

y

0 00 0

x yx , y = α ,0 + β 0,

α 1 - α

(x0/α,0)

(0, y0/(1-α))

(x0 , y0)

A

B

1 - α=

α

Page 36: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

36

A

B

A brief mathematical InterludeA brief mathematical Interlude

x

y

α 1-αx y = KAny tangent of the function

is split by the tangency point in the ratio 1 - α

α

Page 37: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

37

A brief mathematical InterludeA brief mathematical Interlude

x

y

α 1-α

(x,y) Smax x y

For any convex set S, by maximizing

S

end of mathematical Interludeend of mathematical Interlude

We find the unique point in S in which the tangent is split

in the ratio 1 - α

α

Page 38: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

38

α 1-α

1 2(x,y) Smax x - d y - d

To find the Nash Bargaining Solution of a bargaining problem ,S,d

S

d

Nash Bargaining Solution

Page 39: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

39

All axioms were used in the proof

But are they necessary?

All axioms were used in the proof

But are they necessary?

A1. Without Pareto, h(S,d) = dsatisfies the other axioms.

A2. Without Symmetry,

satisfies the other axioms.

α 1-α

1 2(x,y) Smax x - d y - d

Page 40: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

40

All axioms were used in the proof

But are they necessary?

All axioms were used in the proof

But are they necessary?

A3. Without Invariance,

satisfies the other axioms.

(x, y) S

h(S,d) = max x + y

S

(0,1)

(0,0) (2,0)d

(x, y) Smax xy

h(S,d)

(1, 0.5)

Nash Bargaining Solution

Page 41: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

41

All axioms were used in the proof

But are they necessary?

All axioms were used in the proof

But are they necessary?

S

d

A4. Without IIA, the following function satisfies the other axioms.

The Kalai Smorodinsky solution

(S,d)K

1 2max max, (S,d)K

Page 42: 1 Topics in Game Theory SS 2008 Avner Shaked. 2

42

All axioms were used in the proof

But are they necessary?

All axioms were used in the proof

But are they necessary?

S

d

A4. Without IIA, the following function satisfies the other axioms.

The Kalai Smorodinsky solution

(S,d)K

1 2max max,(S,d)K

is on the Pareto front of S

1

2

max

max

(S,d) = (x, y)

x=

y

K is on the Pareto front of S

1

2

max

max

(S,d) = (x, y)

x=

y

K