27
1 Economics & Evolution Number 2

1 Economics & Evolution Number 2. 2 Reading List

Embed Size (px)

Citation preview

Page 1: 1 Economics & Evolution Number 2. 2 Reading List

1

Economics & EvolutionNumber 2

Page 2: 1 Economics & Evolution Number 2. 2 Reading List

2

GENERAL:

1. R. DAWKINS: The Selfish Gene, OUP 1976 2. J. MAYNARD SMITH: Evolution and the Theory of Games, CUP 1982 3. T. SCHELLING: Micromotive and Macrobehavior, W.W. Norton 1978

LEARNING PROCESSES: BEST RESPONSE DYNAMICS (COURNOT), FICTITIOUS PLAY, SOCIAL LEARNING

4. K. BINMORE: Fun and Games, Heath 1992 5. D. FUDENBERG, D. K. LEVINE: The Theory of Learning in Games, MIT Press

1998 6. D. MONDERER, D. SAMET, AND A. SELA: “Belief Affirming in Learning

Processes”, JET 1997

REPLICATOR DYNAMICS, EVOLUTIONARILY STABLE STRATEGY (E.S.S.) 7. J. Weibull: Evolutionary Game Theory, MIT 1995

LOCAL INTERACTION 8. ESHEL I., L. SAMUELSON AND A. SHAKED: “ Altruists Egoists and Hooligans

in a Local Interaction Model,” American Economic review, 1998.

Reading List

Page 3: 1 Economics & Evolution Number 2. 2 Reading List

3

FURTHER READING: 1. J. HOFBAUER, K. SIGMUND: The Theory of Evolution and Dynamical Systems, CUP 1988 2. K. SIGMUND: Games of Life, Explorations in Ecology, Evolution and Behavior, Penguin 1993 3. M. Ridley: The Red Queen: Sex and the Evolution of Human Nature, Perennial, 2003. 4. R. Baker: Sperm Wars, Pan 2000 5. R. Wright: The Moral Animal, Abacus 2004 6. Kandori, Mailath, Rob: Learning, Mutation, and Long Run Equilibria in Games, Econ., 1993 7. T. BERGSTROM: “Storage for Good Times and Bad: Of Rats and Men”, Santa Barbara, mimeo,

1997 (www.econ.ucsb.edu/~tedb/Evolution/store.pdf) 8. T. BORGERS, R. SARIN: “Naive Reinforcement Learning with Endogeneous Aspirations”, JET

1998 COOPERATION: 9. R. AXELROD: The Evolution of Cooperation, Basic Books 1985 10. L. SAMUELSON, K. BINMORE: “Evolutionary Stability in Repeated Games Played by Finite

Automata”, JET 1992 11. ESHEL I., E. SANSONE AND A.SHAKED: “ The Emergence of Kinship Behavior in Structured

Populations of Unrelated Individuals,” International Journal of Game Theory, 1999. 12. ESHEL I., D. HERREINER, L. SAMUELSON, E. SANSONE AND A. SHAKED: “Cooperation, Mimesis

and Local Interaction” Sociological Methods and Research, 2000. 13. ESHEL I. AND A. SHAKED: “Partnership” Journal of Theoretical Biology, 2002.

Page 4: 1 Economics & Evolution Number 2. 2 Reading List

4

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

0

1

1

An ExampleAn Example

(B , A )(B , B )

(A , B )

(A , A )

time is spent in each quadranttime is spent in each quadrant

0 0

1t , p ,

2

1 1

1t , ,q

2,2 2

1t , p

2 =

ap t 1 -t =

bq tt

00

=ap t 1 -t

00a = t 1 - p

0

0 =bq tt

0tb =2

0 0=

t 1 - pp t 1 -

t 0=

tq t

2t

Page 5: 1 Economics & Evolution Number 2. 2 Reading List

5

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

0

1

1

An ExampleAn Example

(B , A )(B , B )

(A , B )

(A , A )

time is spent in each quadranttime is spent in each quadrant

0 0

1t , p ,

2

1 1

1t , ,q

2,2 2

1t , p

2 0 0=

t 1 - pp t 1 -

t 0=

tq t

2t

0

11

0= =

t 1 - p1p t 1 -2 t

1 0 0t = 2t 1 - p

01 0 0- 1 - 2pt t = t

Page 6: 1 Economics & Evolution Number 2. 2 Reading List

6

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

0

1

1

An ExampleAn Example

(B , A )(B , B )

(A , B )

(A , A )

0 0

1t , p ,

2

1 1

1t , ,q

2,2 2

1t , p

2

1 0 0 0-t t = 2 1 - p - 1 t

time spent in the time spent in the firstfirst quadrantquadrant

0 0=

t 1 - pp t 1 -

t 0=

tq t

2t

1 1=

t 1 - pp t 1 -

t 1 1t 1 - q

q t = 1 -t

analogously:

Page 7: 1 Economics & Evolution Number 2. 2 Reading List

7

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

0

1

1

An ExampleAn Example

(B , A )(B , B )

(A , B )

(A , A )

0 0

1t , p ,

2

1 1

1t , ,q

2,2 2

1t , p

2

time spent in the time spent in the secondsecond quadrantquadrant

1 1=

t 1 - pp t 1 -

t 1 1t 1 - q

q t = 1 -t

1

1=

1t 1 -2 t

p t 1 - = 1 -t 2t

1 1t 1 - qq t = 1 -

t

And at t2 : 2

1 1t 1 - q1 = 1-2 t

2 1 12t 2t 1 - q=

Page 8: 1 Economics & Evolution Number 2. 2 Reading List

8

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

0

1

1

An ExampleAn Example

(B , A )(B , B )

(A , B )

(A , A )

0 0

1t , p ,

2

1 1

1t , ,q

2,2 2

1t , p

2

time spent in the time spent in the secondsecond quadrantquadrant

2

1 1t 1 - q1 = 1-2 t

2 1 1t 2t 1 - q=

12 1 1- tt t 1 - 2q= 0 0 12t 1 - p 1 - 2q= (slide 2)

0

11 0

=t 1q =

2t 4 1 - p(slide 4) 2 1 0 0-t t t 1 - 2p=

2 1 1 0- -t t t t=Spending the same time in each quadrant

Page 9: 1 Economics & Evolution Number 2. 2 Reading List

9

Fictitious Play: Failure to ConvergeFictitious Play: Failure to ConvergeAn example by L. ShapleyAn example by L. Shapley

L C R

T 0 , 0 0 , 1 1 , 0

M 1 , 0 0 , 0 0 , 1

B 0 , 1 1 , 0 0 , 0

The only Nash equilibrium is (⅓, ⅓, ⅓)

Page 10: 1 Economics & Evolution Number 2. 2 Reading List

10

Fictitious Play: Failure to ConvergeFictitious Play: Failure to ConvergeAn example by L. ShapleyAn example by L. Shapley

L C R

T 0 , 0 0 , 1 1 , 0

M 1 , 0 0 , 0 0 , 1

B 0 , 1 1 , 0 0 , 0

Tp1=1

Mp2=1

Bp3=1

LC R

Lq1=1

Cq2=1

Rq3=1

TM B

History of player 1(in player’s 2 mind)

History of player 2(in player’s 1 mind)

Best Response of 2 Best Response of 1

The process does not converge, (spends longer periods in any part)

Page 11: 1 Economics & Evolution Number 2. 2 Reading List

11

Fictitious Play: Failure to ConvergeFictitious Play: Failure to ConvergeAn example by L. ShapleyAn example by L. Shapley

L C R

T 0 , 0 0 , 1 1 , 0

M 1 , 0 0 , 0 0 , 1

B 0 , 1 1 , 0 0 , 0

Tp1=1

Mp2=1

Bp3=1

LC R

Lq1=1

Cq2=1

Rq3=1

TM B

History of player 1(in player’s 2 mind)

History of player 2(in player’s 1 mind)

Best Response of 2 Best Response of 1

They (almost) always play either (0,1) or (1,0)

Page 12: 1 Economics & Evolution Number 2. 2 Reading List

12

Fictitious Play: Failure to ConvergeFictitious Play: Failure to ConvergeAn example by L. ShapleyAn example by L. Shapley

L C R

T 0 , 0 0 , 1 1 , 0

M 1 , 0 0 , 0 0 , 1

B 0 , 1 1 , 0 0 , 0

Tp1=1

Mp2=1

Bp3=1

LC R

Lq1=1

Cq2=1

Rq3=1

TM B

History of player 1(in player’s 2 mind)

History of player 2(in player’s 1 mind)

Best Response of 2 Best Response of 1

Exercise: What happens when they start at the following points ???

Page 13: 1 Economics & Evolution Number 2. 2 Reading List

13

A proof of the non convergence of Fictitious Play in the Shapley game.

Monderer, Samet, Sela: ‘Belief Affirming in Learning Processes’JET, vol 73, April 1997

player belief at time about player

the other player's action at time

it

i it ti

-it

-i it -1 t -1i

t

i's t j

t

γ

σ = BR γ

s

s + t - 1γ =

t

γ

ˆ = i i i it t t tiU γ π σ ,γ

i -i i ii t -1 t -1 i t -1 t -1i i i i

t t ti i t -1

π σ ,s + t - 1 π σ ,γπ σ ,γ π σ ,γ

t

Page 14: 1 Economics & Evolution Number 2. 2 Reading List

14

i -i i ii t -1 t -1 i t -1 t -1i i i i

t t ti i t -1

π σ ,s + t - 1 π σ ,γπ σ ,γ π σ ,γ

t

i -i i ii t -2 t -2 i t -2 t -2i i i i

i t -1 t -1 i t -t - 12

π σ ,s + t - 2 π σ ,γπ σ ,γ π σ ,γ

t - 1

i -i i -i i ii t -1 t -1 i t -2 t -2 i t -2 t -2i i

t ti

π σ ,s + π σ ,s + t - 2 π σ ,γπ σ ,γ

t

etc. etc.

i -i i ii -i i -ii t -1 t -1 i t -2 t -2 i 1 1i i

t tii 1 1+π σ ,s + π σ , πs + ...+ π σ ,s

π σ ,γ,γ

t

σ

ex ante ex post

ˆ i i it t tU γ U ˆ -i -i -i

t t tU γ U ˆ i i

1 1U γ

Page 15: 1 Economics & Evolution Number 2. 2 Reading List

15

ˆ ˆi i -i -i i -it t t t t tU γ +U γ U +U

If the process converges it must be that:

ˆ ˆ i -i i i -i -it t t t t t

1= = =

3γ γ U γ U γ

= 2

3

Impossible !!!!

So the process cannot converge.

Page 16: 1 Economics & Evolution Number 2. 2 Reading List

16

A Social Interpretation to Fictitious play dynamics

• Two populations of size N, meet at random and play a

2X2 game.

• At time t, p(t) of the the first population play the second

strategy [q(t) for the second population]

• Players die and are replaced by new ones.• The newly born learn to play the best response against

the other population at the time of their birth.

Page 17: 1 Economics & Evolution Number 2. 2 Reading List

17

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3

The game: (same as in the previous fictitious game example)

Page 18: 1 Economics & Evolution Number 2. 2 Reading List

18

A B

A 0 , 2 3 , 0B 2 , 1 1 , 3 q

0

1

(A)

(B)

(A) p 1(B)

As long as ( p(t) , q(t) ) is in the first quadrant,

the best responses are: ( B , A ).

Page 19: 1 Economics & Evolution Number 2. 2 Reading List

19

In a short interval

boys die and the newly born learn to play strategy

girls die and the newly born learn to play strategy .

2

τ, λτNλτN 1

Boys playing strategy

=2 :

Np t + τ N - λτN p t + λτN

=p t + τ - p t λτ 1 - p t

=•p λ 1 - p t

•p+ λp t = λ λteX -λtp t = 1 -ae

For girls:

=Nq t + τ N - λτN q t

=-•q λq t -λtq t = be

Page 20: 1 Economics & Evolution Number 2. 2 Reading List

20

-λtp t = 1 -ae -λtq t = be

1- p t abq t

=

,Compared with the fictitious game:

the social process is faster.

a bp t = 1 - q t =t t

Page 21: 1 Economics & Evolution Number 2. 2 Reading List

21

The Replicator Dynamics and

Evolutionarily Stable StrategiesE.S.S.

• Replicator: gene, phenotype• It replicates, according to how well it did.• It determines the behaviour, the strategy.• The replicators play, replicate and then die so that the

population remains of a fixed size.

Page 22: 1 Economics & Evolution Number 2. 2 Reading List

22

D H

D 1 ,1 0 , 2

H 2 , 0 -1 , -1

Chicken game, or Dove & Hawk

The population plays: 1-p strategy D, and p strategy H.

The fitness of a player (the no. of his offsprings):

D

H

f p = U + 1- p

f p = U + 2 1- p - p

In a short time interval τ: the total number of D,H players after replication:

D DN 1- p + N 1- p τf p N 1- p 1+ τf= p

HNp 1+ τf pH players:

Page 23: 1 Economics & Evolution Number 2. 2 Reading List

23

The new proportion of H players:

H

H D

Np 1+ τf p

Np 1+ τf p N 1- p 1+p + τ =

τft

p

H

H D

Np 1+ τf p

N + Nτ pf p 1 - p fp

pt + τ =

HNp 1+ τf p

Np t + τ =

+ Nτf

H D

HNp 1+ τf p

N + Nτ pf p 1 - p fp

pt + τ =

H=

p 1+ τf p

1+ τf

H=p p f p - f

H D= p 1- p f p - f p

Page 24: 1 Economics & Evolution Number 2. 2 Reading List

24

H=p p f p - f

H D= p 1- p f p - f p

We have not used the particular numbers of the Chicken game.The above equation holds for all 2x2 games,

And it is independent of U.

The Replicator Dynamics

Now apply it to Chicken: D H

D 1 ,1 0 , 2

H 2 , 0 -1 , -1

D

H

f = 1 1 - p + 0p

f = 2 1 - p + -1 p

= 1 - p

= 2 - 3pH Df - f = 2 - 3p

Page 25: 1 Economics & Evolution Number 2. 2 Reading List

25

H=p p f p - f

H D= p 1- p f p - f p

D H

D 1 ,1 0 , 2

H 2 , 0 -1 , -1

H Df - f = 2 - 3p

=p p 1- p 1 - 2p

t

1/2

D

H

p < 1/2

p

Page 26: 1 Economics & Evolution Number 2. 2 Reading List

26

C D

C 2 ,2 0 , 3

D 3 , 0 1 , 1

D Cf - f = 1

tC

D

The Prisoners’ Dilemma

C

D

f = 2 1 - p + 0p = 2 - 2p

f = 3 1 - p + 1p = 3 - 2p

D C=p p 1- p f p - f p

p

=p p 1- p

Page 27: 1 Economics & Evolution Number 2. 2 Reading List

27

L R

L 1 ,1 0 , 0

R 0 , 0 2 , 2

R Lf - f = 3p - 1

tL

R

p < 1/3

Coordination Game

L

R

f = 1 - p

f = 2p

R L=p p 1- p f p - f p

p

=p p 1- p 3p - 1

1/3

p > 1/3

Spontaneous Order, (no one maximizes)Spontaneous Order, (no one maximizes)