D EFINITION I A ukasiewicz game G on c k is a tuple G = P , V , { V - - PowerPoint PPT Presentation

d efinition i
SMART_READER_LITE
LIVE PREVIEW

D EFINITION I A ukasiewicz game G on c k is a tuple G = P , V , { V - - PowerPoint PPT Presentation

ukasiewicz Games Examples Results O N UKASIEWICZ G AMES E NRICO M ARCHIONI Institut de Recherche en Informatique de Toulouse Universit e Paul Sabatier, France M ICHAEL W OOLDRIDGE Department of Computer Science University of Oxford,


slide-1
SLIDE 1

Łukasiewicz Games Examples Results

ON ŁUKASIEWICZ GAMES

ENRICO MARCHIONI

Institut de Recherche en Informatique de Toulouse Universit´ e Paul Sabatier, France

MICHAEL WOOLDRIDGE

Department of Computer Science University of Oxford, U.K.

ManyVal 2013

4-6 September 2013 Prague, Czech Republic

slide-2
SLIDE 2

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-3
SLIDE 3

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-4
SLIDE 4

Łukasiewicz Games Examples Results

OVERVIEW

◮ We introduce a compact representation of non-cooperative

games based on finite-valued Łukasiewicz logics.

slide-5
SLIDE 5

Łukasiewicz Games Examples Results

OVERVIEW

◮ We introduce a compact representation of non-cooperative

games based on finite-valued Łukasiewicz logics.

◮ Łukasiewicz Games are inspired by, and greatly extend, Boolean

games [Herrenstein et al. 2001].

slide-6
SLIDE 6

Łukasiewicz Games Examples Results

OVERVIEW

◮ We introduce a compact representation of non-cooperative

games based on finite-valued Łukasiewicz logics.

◮ Łukasiewicz Games are inspired by, and greatly extend, Boolean

games [Herrenstein et al. 2001].

◮ In Boolean games each individual player strives for the

satisfaction of a goal, represented as a classical Boolean formula that encodes her payoff;

slide-7
SLIDE 7

Łukasiewicz Games Examples Results

OVERVIEW

◮ We introduce a compact representation of non-cooperative

games based on finite-valued Łukasiewicz logics.

◮ Łukasiewicz Games are inspired by, and greatly extend, Boolean

games [Herrenstein et al. 2001].

◮ In Boolean games each individual player strives for the

satisfaction of a goal, represented as a classical Boolean formula that encodes her payoff;

◮ The actions available to players correspond to valuations that

can be made to variables under their control.

slide-8
SLIDE 8

Łukasiewicz Games Examples Results

OVERVIEW

◮ We introduce a compact representation of non-cooperative

games based on finite-valued Łukasiewicz logics.

◮ Łukasiewicz Games are inspired by, and greatly extend, Boolean

games [Herrenstein et al. 2001].

◮ In Boolean games each individual player strives for the

satisfaction of a goal, represented as a classical Boolean formula that encodes her payoff;

◮ The actions available to players correspond to valuations that

can be made to variables under their control.

◮ The use of Łukasiewicz logics makes it possible to more

naturally represent much richer payoff functions for players.

slide-9
SLIDE 9

Łukasiewicz Games Examples Results

ŁUKASIEWICZ AND GAMES

◮ Classic Game Theory:

◮ Non-cooperative games: ◮ Łukasiewicz Games on Łc

k [M. & Wooldridge]

◮ Constant Sum Łukasiewicz Games on Ł∞ [Kroupa & Majer] ◮ Cooperative games: MV-coalitions [Kroupa]

◮ Game-Theoretic Semantics:

◮ Dialogue games [Ferm¨

uller, Giles, . . . ]

◮ Evaluation games [Cintula & Majer] ◮ Ulam games [Mundici]

slide-10
SLIDE 10

Łukasiewicz Games Examples Results

DEFINITION I

A Łukasiewicz game G on Łc

k is a tuple

G = P, V, {Vi}, {Si}, {φi} where:

slide-11
SLIDE 11

Łukasiewicz Games Examples Results

DEFINITION I

A Łukasiewicz game G on Łc

k is a tuple

G = P, V, {Vi}, {Si}, {φi} where:

  • 1. P = {P1, . . . , Pn} is a set of players;
slide-12
SLIDE 12

Łukasiewicz Games Examples Results

DEFINITION I

A Łukasiewicz game G on Łc

k is a tuple

G = P, V, {Vi}, {Si}, {φi} where:

  • 1. P = {P1, . . . , Pn} is a set of players;
  • 2. V = {p1, p2, . . . } is a finite set of propositional variables;
slide-13
SLIDE 13

Łukasiewicz Games Examples Results

DEFINITION I

A Łukasiewicz game G on Łc

k is a tuple

G = P, V, {Vi}, {Si}, {φi} where:

  • 1. P = {P1, . . . , Pn} is a set of players;
  • 2. V = {p1, p2, . . . } is a finite set of propositional variables;
  • 3. Vi ⊆ V is the set of propositional variables under control of

player Pi, so that the sets Vi form a partition of V.

slide-14
SLIDE 14

Łukasiewicz Games Examples Results

DEFINITION II

  • 4. Si is the strategy set for player i that includes all valuations

si : Vi → Lk of the propositional variables in Vi, i.e. Si = {si | si : Vi → Lk}.

slide-15
SLIDE 15

Łukasiewicz Games Examples Results

DEFINITION II

  • 4. Si is the strategy set for player i that includes all valuations

si : Vi → Lk of the propositional variables in Vi, i.e. Si = {si | si : Vi → Lk}.

  • 5. φi(p1, . . . , pt) is an Łc

k-formula, built from variables in V, whose

associated function fφi : (Lk)t → Lk corresponds to the payoff function of Pi, and whose value is determined by the valuations in {S1, . . . , Sn}.

slide-16
SLIDE 16

Łukasiewicz Games Examples Results

EQUILIBRIA

◮ A tuple (s1, . . . , sn), with each si ∈ Si, is called a strategy

combination.

slide-17
SLIDE 17

Łukasiewicz Games Examples Results

EQUILIBRIA

◮ A tuple (s1, . . . , sn), with each si ∈ Si, is called a strategy

combination.

◮ s−i the set of strategies {s1, . . . , si−1, si+1, . . . , sn} not including si.

slide-18
SLIDE 18

Łukasiewicz Games Examples Results

EQUILIBRIA

◮ A tuple (s1, . . . , sn), with each si ∈ Si, is called a strategy

combination.

◮ s−i the set of strategies {s1, . . . , si−1, si+1, . . . , sn} not including si. ◮ The strategy si for Pi is called a best response whenever, fixing s−i,

there exists no strategy s′

i such that

fφi(si, s−i) ≤ fφi(s′

i, s−i).

slide-19
SLIDE 19

Łukasiewicz Games Examples Results

EQUILIBRIA

◮ A tuple (s1, . . . , sn), with each si ∈ Si, is called a strategy

combination.

◮ s−i the set of strategies {s1, . . . , si−1, si+1, . . . , sn} not including si. ◮ The strategy si for Pi is called a best response whenever, fixing s−i,

there exists no strategy s′

i such that

fφi(si, s−i) ≤ fφi(s′

i, s−i). ◮ A strategy combination (s⋆ 1, . . . , s⋆ n) is called a pure strategy Nash

Equilibrium whenever s⋆

i is a best response to s⋆ −i, for each

1 ≤ i ≤ n.

slide-20
SLIDE 20

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-21
SLIDE 21

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA I [BASU 1994]

◮ Two travelers fly back home from a trip to a remote island where

they bought exactly the same antiques.

slide-22
SLIDE 22

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA I [BASU 1994]

◮ Two travelers fly back home from a trip to a remote island where

they bought exactly the same antiques.

◮ Their luggage gets damaged and all the items acquired are

broken.

slide-23
SLIDE 23

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA I [BASU 1994]

◮ Two travelers fly back home from a trip to a remote island where

they bought exactly the same antiques.

◮ Their luggage gets damaged and all the items acquired are

broken.

◮ The airline promises a refund for the inconvenience

slide-24
SLIDE 24

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA I [BASU 1994]

◮ Two travelers fly back home from a trip to a remote island where

they bought exactly the same antiques.

◮ Their luggage gets damaged and all the items acquired are

broken.

◮ The airline promises a refund for the inconvenience ◮ Both travelers must write on a piece of paper a number between

0 and 100 corresponding to the cost of the antiques.

slide-25
SLIDE 25

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA II [BASU 1994]

◮ If they both write the same number x, they both receive x − 1.

slide-26
SLIDE 26

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA II [BASU 1994]

◮ If they both write the same number x, they both receive x − 1. ◮ If they write different numbers, say x < y, the one playing x will

receive x + 2.

slide-27
SLIDE 27

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA II [BASU 1994]

◮ If they both write the same number x, they both receive x − 1. ◮ If they write different numbers, say x < y, the one playing x will

receive x + 2.

◮ The other player will receive x − 2.

slide-28
SLIDE 28

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA II [BASU 1994]

◮ If they both write the same number x, they both receive x − 1. ◮ If they write different numbers, say x < y, the one playing x will

receive x + 2.

◮ The other player will receive x − 2. ◮ Travelers’ payoff is given by the functions:

f1(x, y) =    max (x − 1, 0) x = y min (min(x, y) + 2, 100) x < y max (min(x, y) − 2, 0) y < x ; f2(x, y) =    max (x − 1, 0) x = y min (min(x, y) + 2, 100) y < x max (min(x, y) − 2, 0) x < y .

slide-29
SLIDE 29

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA: PAYOFF MATRIX

T1 T2 1 2 3 · · · 97 98 99 100 0, 0 2, 0 2, 0 2, 0 · · · 2, 0 2, 0 2, 0 2, 0 1 0, 2 0, 0 3, 0 3, 0 · · · 3, 0 3, 0 3, 0 3, 0 2 0, 2 0, 3 1, 1 4, 0 · · · 4, 0 4, 0 4, 0 4, 0 3 0, 2 0, 3 0, 4 2, 2 · · · 5, 0 4, 0 4, 0 4, 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 0, 2 0, 3 0, 4 0, 5 · · · 96, 96 99, 95 99, 95 99, 95 98 0, 2 0, 3 0, 4 0, 5 · · · 95, 99 97, 97 100, 96 100, 96 99 0, 2 0, 3 0, 4 0, 5 · · · 95, 99 96, 100 98, 98 100, 97 100 0, 2 0, 3 0, 4 0, 5 · · · 95, 99 96, 100 97, 100 99, 99

slide-30
SLIDE 30

Łukasiewicz Games Examples Results

TRAVELER’S DILEMMA AS A ŁUKASIEWICZ GAME

OVER Łc

100 Let G = {T1, T2}, {p, q}, {p}1, {q}2, {φ1(p, q), φ2(p, q)}, where the payoff formulas are:

φ1(p, q) :=

  • ∆ (p ↔ q) ∧
  • p ⊖

1 100

  • ¬∆ (q → p) ∧
  • p ⊖

2 100

  • ¬∆ (p → q) ∧
  • q ⊕

2 100

  • ,

φ2(p, q) :=

  • ∆ (p ↔ q) ∧
  • p ⊖

1 100

  • ¬∆ (p → q) ∧
  • q ⊖

2 100

  • ¬∆ (q → p) ∧
  • p ⊕

2 100

  • ,
slide-31
SLIDE 31

Łukasiewicz Games Examples Results

OTHER EXAMPLES

◮ Auctions. ◮ Coordination Games. ◮ Matching Pennies. ◮ Weak-Link Games.

slide-32
SLIDE 32

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-33
SLIDE 33

Łukasiewicz Games Examples Results

MAIN THEOREM

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

slide-34
SLIDE 34

Łukasiewicz Games Examples Results

MAIN THEOREM

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium
slide-35
SLIDE 35

Łukasiewicz Games Examples Results

MAIN THEOREM

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium
  • 2. n

i=1 Bi = ∅.

slide-36
SLIDE 36

Łukasiewicz Games Examples Results

MAIN THEOREM

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium
  • 2. n

i=1 Bi = ∅.

  • 3. EG is satisfiable.
slide-37
SLIDE 37

Łukasiewicz Games Examples Results

MAIN THEOREM

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium
  • 2. n

i=1 Bi = ∅.

  • 3. EG is satisfiable.
  • 4. There exists a satisfiable normalized game G′ equivalent to G.
slide-38
SLIDE 38

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-39
SLIDE 39

Łukasiewicz Games Examples Results

NORMALIZED GAMES I

◮ Two games

G = P, V, {Vi}, {Si}, {φi} and G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

are equivalent whenever:

slide-40
SLIDE 40

Łukasiewicz Games Examples Results

NORMALIZED GAMES I

◮ Two games

G = P, V, {Vi}, {Si}, {φi} and G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

are equivalent whenever:

  • 1. P = P′,
  • 2. V = V′,
  • 3. For each i, Vi = V′

i and Si = S′ i,

  • 4. (s⋆

1, . . . , s⋆ n) is a NE for G if and only if (s⋆ 1, . . . , s⋆ n) is a NE for G′.

slide-41
SLIDE 41

Łukasiewicz Games Examples Results

NORMALIZED GAMES I

◮ Two games

G = P, V, {Vi}, {Si}, {φi} and G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

are equivalent whenever:

  • 1. P = P′,
  • 2. V = V′,
  • 3. For each i, Vi = V′

i and Si = S′ i,

  • 4. (s⋆

1, . . . , s⋆ n) is a NE for G if and only if (s⋆ 1, . . . , s⋆ n) is a NE for G′.

◮ A game G is normalized whenever each payoff formula

φi(p1, . . . , pm) contains all the variables from V.

slide-42
SLIDE 42

Łukasiewicz Games Examples Results

NORMALIZED GAMES II

◮ An Łc k-formula φ(p1, . . . , pw) has an equivalent extension in

{q1, . . . , qv} if there exists a formula φ♯(p1, . . . , pw, q1, . . . , qv) such that, for every {a1, . . . , aw} ∈ Lk fφ(a1, . . . , aw) = f ♯

φ(a1, . . . , aw, b1, . . . , bv)

for all {b1, . . . , bv} ∈ Lk.

slide-43
SLIDE 43

Łukasiewicz Games Examples Results

NORMALIZED GAMES II

◮ An Łc k-formula φ(p1, . . . , pw) has an equivalent extension in

{q1, . . . , qv} if there exists a formula φ♯(p1, . . . , pw, q1, . . . , qv) such that, for every {a1, . . . , aw} ∈ Lk fφ(a1, . . . , aw) = f ♯

φ(a1, . . . , aw, b1, . . . , bv)

for all {b1, . . . , bv} ∈ Lk.

◮ Every Łc k-formula φ(p1, . . . , pw) has an equivalent extension in

{q1, . . . , qv} by taking φ(p1, . . . , pw) ⊕

v

  • j=1
  • qj ⊙ ¬qj
  • .
slide-44
SLIDE 44

Łukasiewicz Games Examples Results

NORMALIZED GAMES II

◮ An Łc k-formula φ(p1, . . . , pw) has an equivalent extension in

{q1, . . . , qv} if there exists a formula φ♯(p1, . . . , pw, q1, . . . , qv) such that, for every {a1, . . . , aw} ∈ Lk fφ(a1, . . . , aw) = f ♯

φ(a1, . . . , aw, b1, . . . , bv)

for all {b1, . . . , bv} ∈ Lk.

◮ Every Łc k-formula φ(p1, . . . , pw) has an equivalent extension in

{q1, . . . , qv} by taking φ(p1, . . . , pw) ⊕

v

  • j=1
  • qj ⊙ ¬qj
  • .

◮ Every game is equivalent to a normalized game.

slide-45
SLIDE 45

Łukasiewicz Games Examples Results

BEST RESPONSE SETS

◮ We assume that every game is normalized.

slide-46
SLIDE 46

Łukasiewicz Games Examples Results

BEST RESPONSE SETS

◮ We assume that every game is normalized. ◮ For each i, let

xi be tuple of variables controlled by i.

slide-47
SLIDE 47

Łukasiewicz Games Examples Results

BEST RESPONSE SETS

◮ We assume that every game is normalized. ◮ For each i, let

xi be tuple of variables controlled by i.

◮ The slice of fφi at s−i, denoted as

σs−1(fφi), is the function fφi( xi, s−1).

slide-48
SLIDE 48

Łukasiewicz Games Examples Results

BEST RESPONSE SETS

◮ We assume that every game is normalized. ◮ For each i, let

xi be tuple of variables controlled by i.

◮ The slice of fφi at s−i, denoted as

σs−1(fφi), is the function fφi( xi, s−1).

◮ The set

Bi =

  • (si, s−i) | argmax

s′

i ∈Si

(σs−i(fφi)) = si

  • ,

is called the best response set for i.

slide-49
SLIDE 49

Łukasiewicz Games Examples Results

EXAMPLE

Take the game G = {A1, A2}, {p, q}, {p}1, {q}2, {φ1(p, q), φ2(p, q)}, where φ1(p, q) := (p → q), φ2(p, q) := (q → p), and their associated functions are fφ1(x, y) = min(1 − x + y, 1) fφ2(x, y) = min(1 − y + x, 1).

slide-50
SLIDE 50

Łukasiewicz Games Examples Results

EXAMPLE: PAYOFF MATRIX

T1 T2 1 2 3 · · · 8 9 10 10, 10 10, 9 10, 8 10, 7 · · · 10, 2 10, 1 10, 0 1 9, 10 10, 10 10, 9 10, 8 · · · 10, 3 10, 2 10, 1 2 8, 10 9, 10 10, 10 10, 9 · · · 10, 4 10, 3 10, 2 3 7, 10 8, 10 9, 10 10, 10 · · · 10, 5 10, 4 10, 3 . . . . . . . . . . . . . . . ... . . . . . . . . . 8 2, 10 3, 10 4, 10 5, 10 · · · 10, 10 10, 9 10, 8 9 1, 10 2, 10 3, 10 4, 10 · · · 9, 10 10, 10 10, 9 10 0, 10 1, 10 2, 10 3, 10 · · · 8, 10 9, 10 10, 10

slide-51
SLIDE 51

Łukasiewicz Games Examples Results

EXAMPLE: fφ1

0.5 1 0.5 1 0.5 1 x y fφ2(x, y)

slide-52
SLIDE 52

Łukasiewicz Games Examples Results

EXAMPLE: THE SLICE OF fφ1 AT 0

0.5 1 0.5 1 0.5 1 x y fφ1(x, y)

slide-53
SLIDE 53

Łukasiewicz Games Examples Results

EXAMPLE: THE SLICE OF fφ1 AT 0.1

0.5 1 0.5 1 0.5 1 x y fφ1(x, y)

slide-54
SLIDE 54

Łukasiewicz Games Examples Results

EXAMPLE: THE SLICE OF fφ1 AT 0.2

0.5 1 0.5 1 0.5 1 x y fφ1(x, y)

slide-55
SLIDE 55

Łukasiewicz Games Examples Results

EXAMPLE: THE SLICE OF fφ1 AT 0.3

0.5 1 0.5 1 0.5 1 x y fφ1(x, y)

slide-56
SLIDE 56

Łukasiewicz Games Examples Results

EXAMPLE: DEFINING Bi

0.5 1 0.5 1 0.5 1 x y fφ1(x, y) {(0, 0)}

slide-57
SLIDE 57

Łukasiewicz Games Examples Results

EXAMPLE: DEFINING Bi

0.5 1 0.5 1 0.5 1 x y fφ1(x, y) {(0, 0), (0, 0.1), (0.1, 0.1)}

slide-58
SLIDE 58

Łukasiewicz Games Examples Results

EXAMPLE: DEFINING Bi

0.5 1 0.5 1 0.5 1 x y fφ1(x, y) {(0, 0), (0, 0.1), (0.1, 0.1), (0, 0.2), (0.1, 0.2), (0.2, 0.2)}

slide-59
SLIDE 59

Łukasiewicz Games Examples Results

EXAMPLE: DEFINING Bi

0.5 1 0.5 1 0.5 1 x y fφ1(x, y)

{(0, 0), (0, 0.1), (0.1, 0.1), (0, 0.2), (0.1, 0.2), (0.2, 0.2), (0, 0.3), (0.1, 0.3), (0.2, 0.3), (0.3, 0.3)}

slide-60
SLIDE 60

Łukasiewicz Games Examples Results

EXAMPLE: INTERSECTION OF BEST RESPONSE SETS

0.2 0.4 0.6 0.8 1 0.2 0.4 0.6 0.8 1 x y

slide-61
SLIDE 61

Łukasiewicz Games Examples Results

BEST RESPONSE SETS AND EQUILIBRIA

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium.
  • 2. n

i=1 Bi = ∅.

  • 3. EG is satisfiable.
  • 4. There exists a satisfiable normalized game G′ equivalent to G.
slide-62
SLIDE 62

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-63
SLIDE 63

Łukasiewicz Games Examples Results

THE FORMULA EG (I)

◮ We want to define an Łc k-formula EG whose satisfiability encodes

the existence of equilibria.

slide-64
SLIDE 64

Łukasiewicz Games Examples Results

THE FORMULA EG (I)

◮ We want to define an Łc k-formula EG whose satisfiability encodes

the existence of equilibria.

◮ EG should not require additional constants (apart from the payoff

formulas).

slide-65
SLIDE 65

Łukasiewicz Games Examples Results

THE FORMULA EG (I)

◮ We want to define an Łc k-formula EG whose satisfiability encodes

the existence of equilibria.

◮ EG should not require additional constants (apart from the payoff

formulas).

◮ For every variable p and every valuation v : {p} → Lk there exists

a formula ψ(p) such that v(p) = i k IFF v(ψ(p)) = 1.

slide-66
SLIDE 66

Łukasiewicz Games Examples Results

THE FORMULA EG (I)

◮ We want to define an Łc k-formula EG whose satisfiability encodes

the existence of equilibria.

◮ EG should not require additional constants (apart from the payoff

formulas).

◮ For every variable p and every valuation v : {p} → Lk there exists

a formula ψ(p) such that v(p) = i k IFF v(ψ(p)) = 1.

◮ This means that every strategy combination (s1, . . . , sn) can be

encoded by a formula ψ( p1, . . . , pn) so that fψ(s′

1, . . . , s′ n) = 1

IFF si = s′

i

for all i.

slide-67
SLIDE 67

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-68
SLIDE 68

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-69
SLIDE 69

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-70
SLIDE 70

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-71
SLIDE 71

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-72
SLIDE 72

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-73
SLIDE 73

Łukasiewicz Games Examples Results

THE FORMULA EG (II)

EG :=

  • s∈(Lk)

n i=1 mi

  • n
  • i=1
  • ψα1i (x1i) ∧ · · · ∧ ψαmi (xmi)

n

  • i=1
  • si∈(Lk)mi
  • ψβ1i
  • y

β1i 1i

  • ∧ · · · ∧ ψβmi
  • y

βmi mi

(φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , y

β1i 1i , . . . , y βmi mi , . . .

x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn) → φi(x11, . . . , xm1, . . . , x1i−1, . . . , xmi−1, . . . , x1i, . . . , xmi, . . . x1i+1, . . . , xmi+1, . . . x1n, . . . , xmn))

slide-74
SLIDE 74

Łukasiewicz Games Examples Results

SATISFIABILITY AND EQUILIBRIA

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium.
  • 2. n

i=1 Bi = ∅.

  • 3. EG is satisfiable.
  • 4. There exists a satisfiable normalized game G′ equivalent to G.
slide-75
SLIDE 75

Łukasiewicz Games Examples Results

OUTLINE

Łukasiewicz Games Basic Definitions Examples Traveler’s Dilemma Results Theorem Best Response Sets Equilibrium Formula Satisfiable Games

slide-76
SLIDE 76

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (I)

◮ A game G is called satisfiable if there exists a strategy combination

(s1, . . . .sn) such that for every i, φi is satisfied under (s1, . . . .sn).

slide-77
SLIDE 77

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (I)

◮ A game G is called satisfiable if there exists a strategy combination

(s1, . . . .sn) such that for every i, φi is satisfied under (s1, . . . .sn).

◮ Every satisfiable game admits a NE.

slide-78
SLIDE 78

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (I)

◮ A game G is called satisfiable if there exists a strategy combination

(s1, . . . .sn) such that for every i, φi is satisfied under (s1, . . . .sn).

◮ Every satisfiable game admits a NE. ◮ Every φi is satisfiable under (s1, . . . .sn), so no player can

unilaterally improve her payoff.

slide-79
SLIDE 79

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (II)

◮ Take the first-order theory Th(Lk) of the finite MV-chain Lk in the

language of MV-algebras.

slide-80
SLIDE 80

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (II)

◮ Take the first-order theory Th(Lk) of the finite MV-chain Lk in the

language of MV-algebras.

◮ We want to show that there exists a sentence EG that encodes the

existence of equilibria.

slide-81
SLIDE 81

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (II)

◮ Take the first-order theory Th(Lk) of the finite MV-chain Lk in the

language of MV-algebras.

◮ We want to show that there exists a sentence EG that encodes the

existence of equilibria.

◮ Define the formula EG:

∃ x1, . . . , xn∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • where each

xi, yi refers to the tuple of variables assigned to player i.

slide-82
SLIDE 82

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (II)

◮ Take the first-order theory Th(Lk) of the finite MV-chain Lk in the

language of MV-algebras.

◮ We want to show that there exists a sentence EG that encodes the

existence of equilibria.

◮ Define the formula EG:

∃ x1, . . . , xn∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • where each

xi, yi refers to the tuple of variables assigned to player i.

◮ A game G admits a NE iff EG holds in Th(Lk).

slide-83
SLIDE 83

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (III)

◮ EG holds in Th(Lk) iff the set defined by E′ G

∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • is not empty.
slide-84
SLIDE 84

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (III)

◮ EG holds in Th(Lk) iff the set defined by E′ G

∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • is not empty.

◮ Th(Lk) has quantifier elimination in the language of MV algebras

slide-85
SLIDE 85

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (III)

◮ EG holds in Th(Lk) iff the set defined by E′ G

∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • is not empty.

◮ Th(Lk) has quantifier elimination in the language of MV algebras ◮ There exists a quantifier-free Efree G

logically equivalent to E′

G that

defines the same set as E′

G.

slide-86
SLIDE 86

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (III)

◮ EG holds in Th(Lk) iff the set defined by E′ G

∀ y1, . . . , yn n

i=1

  • φi(

x1, . . . , xi−1, yi, xi+1, . . . , xn) ≤ φi( x1, . . . , xi−1, xi, xi+1, . . . , xn)

  • is not empty.

◮ Th(Lk) has quantifier elimination in the language of MV algebras ◮ There exists a quantifier-free Efree G

logically equivalent to E′

G that

defines the same set as E′

G. ◮ There exists an Łc k-formula ǫG that is satisfiable off so is Efree G .

slide-87
SLIDE 87

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (IV)

◮ Given a game

G = P, V, {Vi}, {Si}, {φi}

slide-88
SLIDE 88

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (IV)

◮ Given a game

G = P, V, {Vi}, {Si}, {φi}

◮ Define a new game

G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

where:

slide-89
SLIDE 89

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (IV)

◮ Given a game

G = P, V, {Vi}, {Si}, {φi}

◮ Define a new game

G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

where:

  • 1. P = P′,
  • 2. V = V′,
  • 3. For each i, Vi = V′

i and Si = S′ i,

  • 4. φ′

i := ǫG ∨ φi.

slide-90
SLIDE 90

Łukasiewicz Games Examples Results

SATISFIABLE GAMES (IV)

◮ Given a game

G = P, V, {Vi}, {Si}, {φi}

◮ Define a new game

G′ = P′, V′, {V′

i}, {S′ i}, {φ′ i}

where:

  • 1. P = P′,
  • 2. V = V′,
  • 3. For each i, Vi = V′

i and Si = S′ i,

  • 4. φ′

i := ǫG ∨ φi.

◮ G′ is a normalized satisfiable game equivalent to G.

slide-91
SLIDE 91

Łukasiewicz Games Examples Results

SATISFIABLE GAMES AND EQUILIBRIA

Let G be any Łukasiewicz game on Łc

  • k. Then there exists a formula EG
  • f Łc

k so that the following statements are equivalent:

  • 1. G admits a pure strategy Nash Equilibrium.
  • 2. n

i=1 Bi = ∅.

  • 3. EG is satisfiable.
  • 4. There exists a satisfiable normalized game G′ equivalent to G.
slide-92
SLIDE 92

Łukasiewicz Games Examples Results

WORK IN PROGRESS

◮ Games with costs and efficiency. ◮ Classes of games. ◮ Complexity and tractable games. ◮ Games with external influence. ◮ Games with mixed strategies. ◮ And more...

slide-93
SLIDE 93

Łukasiewicz Games Examples Results

THANKS!