T RUTH J USTICE A LGOS Game Theory I: Basic Concepts Teachers: Ariel Procaccia (this time) and Alex Psomas
NORMAL-FORM GAME β’ A game in normal form consists of: β¦ Set of players π = {1, β¦ , π} β¦ Strategy set π β¦ For each π β π , utility function π£ π : π π β β : if each j β π plays the strategy π‘ π β π , the utility of player π is π£ π (π‘ 1 , β¦ , π‘ π )
THE PRISONERβS DILEMMA β’ Two men are charged with a crime β’ They are told that: β¦ If one rats out and the other does not, the rat will be freed, other jailed for nine years β¦ If both rat out, both will be jailed for six years β’ They also know that if neither rats out, both will be jailed for one year
THE PRISONERβS DILEMMA Cooperate Defect -1,-1 -9,0 Cooperate Defect 0,-9 -6,-6 What would you do?
ON TV http://youtu.be/S0qjK3TWZE8
THE PROFESSORβS DILEMMA Class Listen Sleep 10 6 ,10 6 -10,0 Professor Make effort Slack off 0,-10 0,0 Dominant strategies?
NASH EQUILIBRIUM β’ In a Nash equilibrium, no player wants to unilaterally deviate β’ E ach playerβs strategy is a best response to strategies of others β’ Formally, a Nash equilibrium is a vector of strategies π = π‘ 1 β¦ , π‘ π β π π such β² β π, that for all π β π, π‘ π β² , π‘ π+1 , β¦ , π‘ π ) π£ π π β₯ π£ π (π‘ 1 , β¦ , π‘ πβ1 , π‘ π
THE PROFESSORβS DILEMMA Class Listen Sleep 10 6 ,10 6 -10,0 Professor Make effort Slack off 0,-10 0,0 Nash equilibria?
ROCK-PAPER-SCISSORS R P S R 0,0 -1,1 1,-1 P 1,-1 0,0 -1,1 S -1,1 1,-1 0,0 Nash equilibria?
MIXED STRATEGIES β’ A mixed strategy is a probability distribution over (pure) strategies β’ The mixed strategy of player π β π is π¦ π , where π¦ π (π‘ π ) = Pr[π plays π‘ π ] β’ The utility of player π β π is π π£ π π¦ 1 , β¦ , π¦ π = ΰ· π£ π π‘ 1 , β¦ , π‘ π β ΰ· π¦ π (π‘ π ) (π‘ 1 ,β¦,π‘ π )βπ π π=1
EXERCISE: MIXED NE 1 1 β’ Exercise: player 1 plays 2 , 2 , 0 , player 2 1 1 plays 0, 2 , 2 . What is π£ 1 ? 1 1 1 β’ Exercise: Both players play 3 , 3 , 3 . What is π£ 1 ? R P S R 0,0 -1,1 1,-1 P 1,-1 0,0 -1,1 S -1,1 1,-1 0,0
EXERCISE: MIXED NE Poll 1 ? Which is a NE? 1 1 1 1 1 1 1 1 1 1 1. 2 , 2 , 0 , 2 , 2 , 0 3. 3 , 3 , 3 , 3 , 3 , 3 1 1 1 1 1 2 2 1 2. 2 , 2 , 0 , 2 , 0, 4. 3 , 3 , 0 , 3 , 0, 2 3 R P S R 0,0 -1,1 1,-1 P 1,-1 0,0 -1,1 S -1,1 1,-1 0,0
NASHβS THEOREM β’ Theorem [Nash, 1950]: In any (finite) game there exists at least one (possibly mixed) Nash equilibrium β’ What about computing a Nash equilibrium? Stay tunedβ¦
DOES NE MAKE SENSE? β’ Two players, strategies are {2, β¦ , 100} β’ If both choose the same number, that is what they get β’ If one chooses π‘ , the other π’ , and π‘ < π’ , the former player gets π‘ + 2 , and the latter gets π‘ β 2 β’ Poll 2: What would you choose? 95 96 97 98 99 100
CORRELATED EQUILIBRIUM β’ Let π = {1,2} for simplicity β’ A mediator chooses a pair of strategies (π‘ 1 , π‘ 2 ) according to a distribution π over π 2 β’ Reveals π‘ 1 to player 1 and π‘ 2 to player 2 β’ When player 1 gets π‘ 1 β π , he knows the distribution over strategies of 2 is Pr π‘ 2 π‘ 1 = Pr π‘ 1 β§ π‘ 2 = π π‘ 1 , π‘ 2 Pr π‘ 1 Pr[π‘ 1 ]
CORRELATED EQUILIBRIUM β² β π β’ Player 1 is best responding if for all π‘ 1 β² , π‘ 2 ) ΰ· Pr π‘ 2 π‘ 1 π£ 1 π‘ 1 , π‘ 2 β₯ ΰ· Pr π‘ 2 π‘ 1 π£ 1 (π‘ 1 π‘ 2 βπ π‘ 2 βπ β’ Equivalently, β² , π‘ 2 ) ΰ· π π‘ 1 , π‘ 2 π£ 1 π‘ 1 , π‘ 2 β₯ ΰ· π π‘ 1 , π‘ 2 π£ 1 (π‘ 1 π‘ 2 βπ π‘ 2 βπ β’ π is a correlated equilibrium (CE) if both players are best responding β’ Every Nash equilibrium is a correlated equilibrium, but not vice versa
GAME OF CHICKEN http://youtu.be/u7hZ9jKrwvo
GAME OF CHICKEN β’ Social welfare is the sum of utilities Dare Chicken β’ Pure NE: (C,D) and 0,0 4,1 Dare (D,C), social welfare = 5 β’ Mixed NE: both 1,4 3,3 (1/2,1/2), social Chicken welfare = 4 β’ Optimal social welfare = 6
GAME OF CHICKEN β’ Correlated equilibrium: β¦ (D,D): 0 Dare Chicken 1 β¦ (D,C): 0,0 4,1 Dare 3 1 β¦ (C,D): 3 1,4 3,3 Chicken 1 β¦ (C,C): 3 16 β’ Social welfare of CE = 3
IMPLEMENTATION OF CE β’ Instead of a mediator, use a hat! β’ Balls in hat are labeled with βchickenβ or βdareβ, each blindfolded player takes a ball Poll 3 ? Which balls implement the distribution of the previous slide? 1. 1 chicken, 1 dare 3. 2 chicken, 1 dare 2. 1 chicken, 2 dare 4. 2 chicken, 2 dare
CE AS LP β’ Can compute CE via linear programming in polynomial time! find βπ‘ 1 , π‘ 2 β π, π π‘ 1 , π‘ 2 β² β π, s.t .t. βπ‘ 1 , π‘ 1 β² , π‘ 2 ) ΰ· π π‘ 1 , π‘ 2 π£ 1 π‘ 1 , π‘ 2 β₯ ΰ· π π‘ 1 , π‘ 2 π£ 1 (π‘ 1 π‘ 2 βπ π‘ 2 βπ β² β π, βπ‘ 2 , π‘ 2 β² ) ΰ· π π‘ 1 , π‘ 2 π£ 2 π‘ 1 , π‘ 2 β₯ ΰ· π π‘ 1 , π‘ 2 π£ 2 (π‘ 1 , π‘ 2 π‘ 1 βπ π‘ 1 βπ ΰ· π π‘ 1 , π‘ 2 = 1 π‘ 1 ,π‘ 2 βπ βπ‘ 1 , π‘ 2 β π, π π‘ 1 , π‘ 2 β [0,1]
Recommend
More recommend