amsterdam - uni-bonn.de · ngs: set m elation ≺ on m. • all x ∈ m holds x 6≺ x. • all y...

30
Sets and Games 14. September 2004 Logic Tea, ILLC Amsterdam Stefan Bold Mathematical Logic Group Department of Mathematics University of Bonn Institute for Logic, Language and Computation University of Amsterdam This research has been supported by the DFG-NWO Bilateral Cooperation Project KO 1353/3-1/DN 61-532 Determiniertheitsaxiome, Infinit¨ are Kombinatorik und ihre Wechselwirkungen (2003-2006; Bold, Koepke, L¨owe, van Benthem)

Upload: others

Post on 21-Mar-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Sets

and

Gam

es

14.Septe

mber

2004

Logic

Tea,IL

LC

Am

sterd

am

Ste

fan

Bold

Math

em

aticalLogic

Gro

up

Depart

ment

ofM

ath

em

atics

Univ

ers

ity

ofBonn

Inst

itute

for

Logic

,Language

and

Com

puta

tion

Univ

ers

ity

ofAm

sterd

am

This

rese

arc

hhas

been

support

ed

by

the

DFG

-NW

OBilate

ralCoopera

tion

Pro

ject

KO

1353/3-1

/D

N61-5

32

Dete

rmin

iert

heitsa

xio

me,In

finitare

Kom

bin

ato

rik

und

ihre

Wechse

lwirkungen

(2003-2

006;Bold

,K

oepke,Lowe,van

Benth

em

)

Page 2: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

What

isa

Set?

Unte

rein

er

Menge

vers

tehen

wir

jede

Zusa

mm

enfa

ssung

Mvon

best

imm

ten,wohlu

nte

rschie

denen

Obje

kte

nm

unse

r

Ansc

hauung

oderunse

resD

enkens(w

elc

he

die

“Ele

mente

von

Mgenannt

werd

en)

zu

ein

em

Ganzen.

Georg

Canto

r

“A

set

Mis

unders

tood

tobe

acollection

ofcert

ain

dis-

tinguishable

obje

cts

mofourperc

eption

orourth

oughts

(whic

hare

called

“ele

ments

”ofM

).”

Page 3: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Nota

tion:

Let

Mand

Sbe

sets

.

•m∈

Mm

eans

mis

an

ele

ment

of

M.

•{a

|P}

isth

ese

tofall

asu

ch

that

pro

pert

yP

hold

sfo

ra.

•M

∪S

isth

eunio

nofth

ese

tsM

and

S.

•M

∩S

isth

ein

ters

ection

ofth

ese

tsM

and

S.

•M\S

isth

ese

twe

get

when

we

take

away

all

ele

ments

inM

that

are

also

inS.

•M

⊆S

means

that

Mis

asu

bse

tof

S.

•and

finally,

Vdenote

sth

euniv

ers

eofall

sets

.

Page 4: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Russell’s

Para

dox

:

Let

Xbe

the

set

ofall

sets

that

do

not

conta

inth

em

selv

es.

X=

{M|M

6∈M}

Then,if

Xconta

ins

itse

lf,it

does

not

conta

initse

lf.

...

On

the

oth

erhand,if

Xdoes

notconta

ins

itse

lf,it

must

conta

in

itse

lf,so

either

way

we

have

acontr

adic

tion.

Page 5: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Som

eoth

erpara

doxes

usin

gself

refe

rence:

Epid

em

es

Para

dox:

“All

cre

tians

are

alw

ays

liars

.”

Barb

er

Para

dox:

Ina

small

tow

nlives

am

ale

barb

er

who

shaves

daily

every

man

that

does

not

shave

him

self,and

no

one

else.

Acondense

dLia

rPara

dox:

“T

his

state

ment

isnot

true”

Page 6: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

The

axio

ms

ofSet

Theory

:

•Existe

nz:T

here

exists

ase

tth

at

conta

ins

no

sets

,th

eem

pty

set.

•Exte

nsionality

:T

wo

sets

are

the

sam

eif

they

conta

inexactly

the

sam

eele

ments

.

•Pairs:

Forevery

two

sets

exists

ase

tth

atconta

insexactly

those

two

sets

.

•Unio

n:Forevery

set

Xexists

ase

t⋃ X

thatconta

insexactly

the

ele

ments

ofth

eele

ments

of

X.

•Power

Set:

For

every

set

Xexists

the

power

setP(X

),i.e.th

ese

tth

at

conta

ins

all

subse

tsof

X.

•In

finity:T

here

exists

an

infinite

set.

•Separa

tion:Forevery

set

Xand

every

form

ula

ϕexists

ase

tth

atconta

ins

exactly

the

ele

ments

of

Xfo

rw

hic

hold

s.

•Repla

cem

ent:

Forevery

set

Xand

every

functionalfo

rmula

ϕexists

ase

tY

,s.t.

“Y

isth

eim

age

of

Xunder

ϕ.”

•Foundation:If

afo

rmula

ϕis

true

for

at

least

one

set,

then

exists

ase

tX

s.t.

ϕist

true

for

Xbut

not

for

any

ele

ment

of

X.

Page 7: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Functions

as

sets

:

Nota

tion:

(a,b

):=

{{a},{a

,b}}

,A×

B:=

{(a,b

)|a

∈A

and

b∈

B}.

Let

fbe

asu

bse

tof

B,su

ch

that

•fo

rall

a∈

Ath

ere

isa

b∈

Bs.t.

(a,b

)∈

f,

•and

if(a

,b)∈

fand

(c,b

)∈

f,th

en

a=

c

Then

we

call

fa

function

from

Ato

B.

Now

pro

pert

ies

of

functions

like

surjectivity,

dom

ain

,fu

nction

valu

e,etc

.are

just

pro

pert

ies

ofth

ecorr

esp

ondin

gse

t.

Page 8: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Num

bers

as

sets

:

Let

0:=

{}and

a+

1:=

a∪{a}.

Then

all

natu

ralnum

bers

can

be

vie

wed

as

sets

:

1=

0∪{0}=

{}∪{{}}

={{}}

,2

=1∪{1}=

{0,1},

3=

{0,1

,2},

etc

.

Now

we

can

go

on

and

form

alize

the

inte

gers

asse

ts,th

era

tional

num

bers

,th

ere

als,fu

nctions

on

reals,in

tegra

ls,derivations

...

So

all

math

em

atics

take

pla

ce

inth

euniv

ers

eof

sets

,and

set

theory

giv

es

ita

form

alfo

undation.

Page 9: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Diff

ere

nt

Infinitie

s:

There

are

counta

bly

infinitiv

em

any

natu

ralnum

bers

.T

he

size

ofth

epowerse

tP(N

)ofth

enatu

ralnum

bers

isalso

infinite,but

larg

er

than

counta

ble

infinite:

Ass

um

ewe

have

abijection

f:N

→P(N

).

Let

X:=

{z|z

∈N

and

z6∈

f(z

)}.

Then

X∈P(N

),so

there

isan

Ys.t.

X=

f(Y

).

IfY∈

f(Y

),th

en

by

definitio

nof

Xwe

have

Y6∈

X=

f(Y

),

ifY6∈

f(Y

),th

en

by

definitio

nof

Xwe

have

Y∈

X=

f(Y

),

either

way

we

have

acontr

adic

tion,so

such

abijection

can

not

exist.

Page 10: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

The

Axio

mofChoice:

Take

afinite

collection

ofnonem

pty

set.

Easily

one

can

choose

an

ele

ment

from

each

set:

Take

an

ele

ment

from

the

firs

tse

tin

the

collection,th

eone

from

the

second,and

soon.

Take

an

infinite

collection

ofnonem

pty

sets

.Now

this

pro

cedure

willnotwork

anym

ore

,atle

ast

itw

illnotcom

eto

an

end

infinite

tim

e.

The

Axio

mofChoic

est

ate

sth

atone

can

alw

ays

choose

,re

gard

-

less

ofth

esize

ofth

ecollection:

(AC)

For

every

set

C=

{Mi|i

∈I}

of

nonem

pty

sets

exists

a

function

f:C

→⋃ C

with

f(M

i)∈

Mi.

Such

afu

nction

iscalled

achoic

efu

nction.

Page 11: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Wellord

erings:

Awellord

ering

ofa

set

Mis

abin

ary

rela

tion≺

on

M,s.t.

•For

all

x∈

Mhold

sx6≺

x.

•For

all

x,y

∈M

either

x≺

yor

y≺

xor

x=

y.

•If

x≺

yand

y≺

zth

en

x≺

z.

•Every

subse

tof

Mhas

a≺-m

inim

alele

ment.

So

there

are

no

infinitiv

ely

desc

endin

gchain

sin

awellord

ering.

Page 12: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Wellord

erings

ofnum

bers

:

The

natu

ralnum

bers

are

wellord

ere

dby

<,

0// 1

// 2// 3

// 4// 5

// 6// 7

// 8// 9

// 10

// ···

where

as

the

inte

gers

are

not

wellord

ere

dby

<:

···

// −5

// −4

// −3

// −2

// −1

// 0// 1

// 2// 3

// 4// 5

// ···

But

we

can

wellord

er

the

inte

gers

,using

anoth

er

ord

er≺:

0≺

1≺−1≺

2≺−2≺

3···

···

−3

��−2

##−1

%%0

// 1cc

2aa

3\\

···

Page 13: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

AW

ellord

ering

ofth

ere

als

under

AC:

We

start

with

one

num

ber,

say

a.

' &

$ %

aX

XX

XX

XX

XX

XX

XX

XX

XX Xz

x

Page 14: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

AW

ellord

ering

ofth

ere

als

under

AC:

Then

we

choose

anum

ber

bfrom

R\{

a},

' &

$ %

a

xb

A A AA A A

A A AA A A

A A AUx

Page 15: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

AW

ellord

ering

ofth

ere

als

under

AC:

then

cfrom

R\{

a,b},

...

' &

$ %

a

xb

x

cH

HH

HH

HH

HH

HH

HH

HH

HH

HH

HH

HH

HHHj

x

≺≺

Page 16: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

AW

ellord

ering

ofth

ere

als

under

AC:

Inth

eend

we

willhave

ord

ere

dall

reals.

a≺

b≺

c≺

d≺

e≺

f≺···

Page 17: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Ord

inals:

We

desc

ribed

the

natu

ral

num

bers

as

sets

,w

here

n∈

Nwas

form

alized

as

n=

{m|m

<n}

and

m<

nm

eant

m∈

n.

Ifwe

genera

lize

this

idea

we

get

the

Ord

inalnum

bers

.Form

ally

an

ord

inalis

atr

ansitive,

wellord

ere

dse

t.T

he

collection

of

all

ord

inals

isnot

ase

t,it

is“to

big

.”

Apic

ture

ofth

euniv

ers

e:

A AA A A

A A AA A A

A A A A���

������

������

V

ON

We

write

ON

for

the

cla

ssof

ord

inals

and

for

our

purp

ose

swe

need

only

toknow

that

the

ord

inals

are

wellord

ere

dby

<.

Page 18: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

AW

ellord

ering

ofth

ere

als

under

AC:

Let

f:P(R

)→

Rbe

achoic

efu

nction

for

the

power

set

of

the

reals.

Define

the

function

F:O

N→

Rby

transfi

nite

recurs

ion

(here

α

and

βdenote

ord

inals):

F(α

):=

f(R\

⋃ {F(β

)|β

<α} )

,

undefined

if⋃ {F

(β)|β

<α}=

R.

Ifwe

now

say

a≺

biff

a=

F(α

),b=

F(β

)and

α<

β,th

en≺

is

awellord

ering

on

R.

Page 19: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Infinite

Gam

es:

Let

Xbe

acollection

ofse

quences〈x

i|i∈

N〉ofnatu

ralnum

bers

.

Call

Xth

epayo

ffse

t.

Aro

und

inth

egam

eG

Xis

ase

quence

x=

〈xi|i∈

N〉

ofnatu

ral

num

bers

,w

here

the

ele

ments

with

even

index

xI(

i):=

x2ire

pre

-

sentth

em

ovesofpla

yerIand

those

with

odd

index

xII(i

):=

x2i+

1

the

moves

of

pla

yer

II.

Ifx∈

Xth

en

pla

yer

IIw

ins,

oth

erw

ise

pla

yer

I.

Ix0

x2

x4

x6

...

(=xI)

IIx1

x3

x5

x7

...

(=xII)

Page 20: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

The

Axio

mofD

ete

rmin

acy:

An

infinite

gam

eG

Xis

dete

rmin

ed,

ifone

of

the

pla

yers

has

a

win

nin

gst

rate

gy,

i.e.

ifth

ere

isa

function

that

tells

one

pla

yer

what

todo

next,

dependin

gon

the

pre

vio

us

moves,

s.t.

inth

e

end

that

pla

yer

win

sth

egam

ere

gard

less

of

the

moves

of

his

opponent.

So

ast

rate

gy

for

pla

yer

Iwould

be

afu

nction

ofth

efo

rm

f:

⋃ {N2n|n

∈N}→

N.

The

Axio

mofD

ete

rmin

acy

now

issim

ply

the

state

ment

(AD)

All

infinite

gam

es

are

dete

rmin

ed.

Page 21: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Counta

ble

Choice

under

AD

:

AD

isnot

com

patible

with

full

choic

e,

but

itim

plies

counta

ble

choic

efo

rcollections

ofse

quences

from

N.

Let

M=

{Sj|j

∈N}

be

acounta

ble

collection

ofnonem

pty

sets

ofsu

ch

sequences:

Counta

ble

many

Sj

︷︸︸

︷S1

S2

S3

······

x1

=〈x

1(i

)|i∈

N〉

y1

=〈y

1(i

)|i∈

N〉

z 1=

〈z1(i

)|i∈

N〉

······

x2

=〈x

2(i

)|i∈

N〉

y2

=〈y

2(i

)|i∈

N〉

z 2=

〈z2(i

)|i∈

N〉

······

. . .. . .

. . .······

=〈x

α(i

)|i∈

N〉

=〈y

α(i

)|i∈

N〉

z α=

〈zα(i

)|i∈

N〉···

. . .. . .

. . .···

Page 22: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Counta

ble

Choice

under

AD

:

We

now

const

ruct

the

payo

ffse

tX

from

M:

Foreach

j∈

Nwe

define

Xjto

be

the

collection

ofall

sequences

xth

atst

art

with

jand

where

xII∈

Sj.

Let

Xbe

the

unio

nofth

e

Xj.

X=

⋃ {Xj|j∈

N}

︷︸︸

︷X

1X

2X

3······

〈1,x

1(1

),�

,x1(2

),�

,···〉

〈2,y

1(1

),�

,y1(2

),�

,···〉

·········

. . .. . .

·········

〈1,x

2(1

),�

,x2(2

),�

,···〉

〈2,y

2(1

),�

,y2(2

),�

,···〉

······

. . .. . .

. . .······

〈1,x

3(1

),�

,x3(2

),�

,···〉〈2

,y3(1

),�

,y3(2

),�

,···〉

···

. . .. . .

. . .. . .

Page 23: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Counta

ble

Choice

under

AD

:

Then

pla

yer

Icannot

have

aw

innin

gst

rate

gy

inth

egam

eG

X:

Anyth

ing

he

does

aft

er

his

firs

tm

ove

jhas

no

conse

quences

on

the

outc

om

e,and

pla

yerII

only

needsto

pla

yone

ofth

ese

quences

from

the

set

Sj

tore

fute

him

.

So

by

AD

pla

yer

IIm

ust

have

aw

innin

gst

rate

gy.

Let

τbe

II’s

win

nin

gst

rate

gy

and

let

(x∗

τ) I

Ibe

the

sequence

of

pla

yerII’s

moves

inth

egam

ew

here

pla

yerIpla

ys

xand

pla

yerII

pla

ys

accord

ing

toτ.

Then

F(S

j):=

(〈j|i∈

N〉∗

τ) I

Idefines

achoic

efu

nction

for

M.

Page 24: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Fin

ite

gam

es:

Every

finite

exte

nsive

gam

eofperfectin

form

ation

thatis

aW

in/Lose

-gam

eis

dete

rmin

ed.

To

see

whic

hof

the

two

pla

yers

Aand

Bhas

the

win

nin

gst

rate

gy,

we

use

asim

ple

backtr

ackin

gm

eth

od:

Let

this

tree

repre

sent

such

agam

e,th

epla

yers

take

turn

sand

pla

yer

Ast

art

s:

������

����

����

����

���

&&NNNNNNNNNNNNNNNNNNNNNNNNNNNNNNN

A

������

����

����

����

���

��////////////////•

������������������

##GGGGGGGGGGGGGGGGGGGGGGGG

B

������������������

��////////////////• ��

��������������������

����>>>>>>>>>>>>>>>>>>>

• ��

A

••

••

••

Page 25: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Win

nin

gstr

ate

gie

sfo

rfinite

gam

es:

First

we

labelth

eend-n

odes: •

~~}}}}

}}}}

}}}}

}}}}

}}}}

''OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

A

~~ ~~~~

~~~~

~~~~

~~~~

~~~~

��0000000000000000•

������������������

$$IIIIIIIIIIIIIIIIIIIIIIIIII

B

������������������

��////////////////• ��

������

����

����

����

���

����@@@@@@@@@@@@@@@@@@@@

• ��

A

AA

BB

AB

A

Page 26: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Win

nin

gstr

ate

gie

sfo

rfinite

gam

es:

The

last

move

isA’s,we

labelth

epre

vio

us

nodes

accord

ingly

:

~~}}}}

}}}}

}}}}

}}}}

}}}}

''OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

A

~~ ~~~~

~~~~

~~~~

~~~~

~~~~

��////////////////•

������������������

$$HHHHHHHHHHHHHHHHHHHHHHHHH

B

A

���������������

��3333333333333B ��

A

~~ ~~~~

~~~~

~~~~

~~~~

��!!DDDDDDDDDDDDDDDDD

A ��

AA

BB

AB

A

Page 27: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Win

nin

gstr

ate

gie

sfo

rfinite

gam

es:

Now

itis

B’s

turn

:

~~ }}}}

}}}}

}}}}

}}}}

}}}}

''OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO

A

B

|| yyyy

yyyy

yyyy

yyyy

yy

��3333333333333A

���������������

%%LLLLLLLLLLLLLLLLLLLLLLLL

A

���������������

��3333333333333B ��

A

~~ ~~~~

~~~~

~~~~

~~~~

��""EEEEEEEEEEEEEEEEEE

A ��

AA

BB

AB

A

Page 28: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Win

nin

gstr

ate

gie

sfo

rfinite

gam

es:

And

inth

eend

we

see

that

pla

yer

Ahas

aw

innin

gst

rate

gie

:

A

~~ }}}}

}}}}

}}}}

}}}}

}}}}

'''g

'g'g

'g'g

'g'g

'g'g

'g'g

'g'g

'g'g

'g'g

'g'g

'g'g

'g

A

B

~~}}}}

}}}}

}}}}

}}}}

}}}}

��0000000000000000A

���G�G�G�G�G�G�G�G�G�G

$$$d

$d$d

$d$d

$d$d

$d$d

$d$d

$d$d

$d$d

$d

B

A

������������������

��0000000000000000B ��

A

��������������������

���O�O�O�O�O�O�O�O�O

AAAAAAAAAAAAAAAAAAAAA ���O�O�O�O�O�O�O�O�O

A

AA

BB

AB

A

Page 29: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

Sim

ilarities

betw

een

AC

and

AD

:

•Both

axio

ms

gra

nt

the

existe

nce

offu

nctions.

•Both

imply

rest

ricte

dvers

ions

ofth

eoth

er.

•Both

are

exte

nsions

offinite

princip

les.

•Both

are

consist

ent

with

ZF

(under

som

eass

um

ptions)

.

•T

hey

contr

adic

teach

oth

er.

Page 30: Amsterdam - uni-bonn.de · ngs: set M elation ≺ on M. • all x ∈ M holds x 6≺ x. • all y ∈ M r x ≺ y r y ≺ x r x = y. • If x ≺ y and y ≺ z then x ≺ z. • of

and

Diff

ere

nces:

•Under

AC

exists

awellord

ering

ofth

ere

als,butnotunder

AD.

•Under

AD

every

subse

tofth

ere

als

isLebesg

ue

measu

rable

,

not

sounder

AC.

•Under

AC

every

set

can

be

wellord

ere

d,but

not

under

AD.

•Under

AD

infinite

part

itio

npro

pert

ies

hold

,butnotunder

AC.

•···