Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Strategic Behavior – Spring 2012 Homework Assignment 4 - Answers Due: Wednesday, April 4, 2012 (Please write legibly; if in doubt, type your assignment.) This homework assignment is based on the material in chapters 7 and 8 of the course text. 1. (Question 8, chapter 7) It is the closing seconds of a football game, and the losing team has just scored a touchdown. Now down by only one point, the team decides to go for a twopoint conversion that, if successful, will win the game. The offense chooses between three possible running plays: run wide left, run wide right, and run up the middle. The defense decides between defending against a wide run and a run up the middle. The payoff to the defense is the probability that the offense does not score, and the payoff to the offense is the probability that it does score. Find all pure and mixed-strategy Nash equilibria. This question is not easy; think carefully. Offense Run wide left Run up middle Run wide right Defense Defend against wide run .6,.4 .4,.6 .6,.4 Defend against run up the middle .3,.7 .5,.5 .3,.7 Let l be the probability that Offense chooses run wide left, r be the probability that it chooses run wide right, and 1 – l – r be the probability of middle. The defense is indifferent between its two strategies when: l x .6 + (1 – l – r) x .4 + r x .6 = l x .3 + (1 – l - r) x .5 + r x .3 l + r = .25 Let d be the probability that Defense picks defend against wide run. Offense’s expected payoff from its three pure strategies is: Wide left: d x .4 + (1 – d) x .7 = .7 - .3d Middle: d x .6 + (1 – d) x .5 = .5 - .1d Wide right: d x .4 + (1 – d) x .7 = .7 - .3d For the offense to be indifferent among them, it must be true that: .7 - .3d = .5 + .1d d = .5 This means a Nash equilibrium is any strategy pair in which the defense defends against the outside run with probability .5 and the offense runs up the middle with probability .75. The offense can divide up .25 between left and right outside in any way; it is not correct to force the offense to divide equally between left and right. This percentage must be specified to allow flexibility. [While (.5; .125) is a NE, it is not the ONLY: (.5; .1, .75) is also.] Answer: There is one Nash equilibrium; it is in mixed strategies: (1/2, q2 = 3/4), or (.5, q2 = .75), or (1/2, 1- q2 = .25). 2. (Question 4, chapter 8) Benjamin Franklin once said, “Laws too gentle are seldom obeyed; too severe, seldom executed.” To flush out what he had in mind, the game [below] has three players: a lawmaker, a (typical) citizen, and a judge. The lawmaker chooses among a law with a gentle penalty, one with a moderate penalty, and a law with a severe penalty. In response to the law, the citizen decides whether or not to obey it. If she does not obey it, then the judge decides whether to convict and punish the citizen. Using subgame perfect Nash equilibrium, find values for the unspecified payoffs (those with letters, not numbers) that substantiate Franklin’s claim by resulting in lawmaker’s choosing a law with a moderate penalty. Lawmaker Gentle Moderate Severe Citizen Obey Judge Do not obey Obey Do not obey Convict 8 a c Do not convict 6 b d Do not obey 10 4 8 10 4 8 10 4 8 Obey Convict 8 e g Do not convict 6 f h Convict Don’t 8 i k The objective is to find values so that a subgame perfect Nash equilibrium has the following properties: 1) When the law is gentle, the citizen disobeys the law even though he will be convicted 6 j l 2) When the law is moderate, the citizen obeys the law because he will be convicted 3) When the law is severe, the citizen disobeys the law because he will not be convicted. Answer: For 1) to occur, we need c > d (so the judge will convict the citizen) and a > 4 (so the citizen prefers to disobey the law and be convicted than to obey the law). For 2) to occur, we need g > h (so the judge will convict the citizen) and 4 > e (so the citizen prefers to obey the law rather than disobey and be convicted). For 3) to occur, we need l > k (so the judge will not convict the citizen and j > 4 (so the citizen prefers to disobey the law and not be convicted than obey it). By backward induction, the lawmaker then faces these three alternatives: Choose a gentle law and receive a payoff of 8 because the citizen disobeys the law and is convicted Choose a moderate law and receive a payoff of 10 because the citizen obeys the law Choose a severe law and receive a payoff of 6 because the citizen disobeys the law and is not convicted. The lawmaker optimally chooses a moderate law. 3. (Question 6, chapter 8) In 1842, the Sangamo Journal of Springfield, Illinois, published letters that criticized James Shields, the auditor of the State of Illinois. Although the letters were signed “Rebecca,” Shields suspected that it was state legislator Abraham Lincoln who penned the letters. As shown in [the figure below], Shields considered challenging Lincoln to a dual, and, as history records, Shields did challenge Lincoln. In response to a challenge, Lincoln could avoid the duel, or; if he chose to meet Shield’s challenge, he had the right to choose the weapons. We will also allow Lincoln to decide whether to offer an apology of sorts. If he decides to go forward with a duel, then Lincoln has four choices: propose guns, propose guns and offer an apology, propose swords, or propose swords and offer an apology. In response to any of the four choices, Shields must decide to either go forward with the duel or stop the duel (and accept the apology if offered). Find all subgame perfect Nash equilibria. Shields 5 6 Do not challenge Challenge Lincoln Avoid Guns Guns & apology Swords Swords & apology Shields 10 1 Stop 0 5 Go 8 2 Stop 6 4 Go Stop 8 2 0 5 Go 2 3 Stop 6 4 Go 2 3 Begin with Shield’s final four decision nodes. If Lincoln proposes guns, then Shields will go forward with the dual, as his payoff is 8, while it is 0 if he were to back down. If Lincoln proposes guns with an apology, then Shields still goes forward with the duel, as his payoff is 8 again, while it is 6 if he were to back down. If Lincoln proposes swords, then Shields will go forward with the duel as his payoff is 2, while it is 0 if he were to back down. Finally, if Lincoln proposes swords but with an apology, then Shields will stop the duel and accept the apology, as the payoff is 6, while it is 2 if he were to actually battle Lincoln using swords. Substituting that part of the game with the associated payoffs derived using backward induction gives the game shown below. Shields Do not challenge 5 6 Challenge Lincoln Avoid Shields 10 1 Guns 8 2 Guns & apology Swords 8 2 2 3 Swords & apology 6 4 Lincoln chooses between his five choices in the event that Shields has challenged him to a duel. Lincoln’s optimal choice is to accept the challenge, propose that they sue swords, and offer an apology. In response to that choice, Shields will accept the apology and the duel will be avoided, which means a payoff of 4 for Lincoln. If Lincoln makes any of the other three choices that involve going forward with the duel, his payoff is no higher than 3, as the duel will commence. Avoiding the duel is even worse, with a payoff of 1. Moving up to Shields’s initial decision node, his payoff is 5 from not challenging Lincoln and is 6 from doing so. Thus, Shields challenges Lincoln. Answer: The subgame perfect Nash equilibrium strategy profile is [challenge/go/go/go/stop; swords and apology]. 4. Suppose you are analyzing a simple two-stage bargaining game, in which the first move is a proposal of how to divide $4, which must be either $3 for the proposer and $1 for the responder or $2 for each. The responder sees the proposal and either accepts (which implements the split) or rejects the offer, which results in $0 for each. The responder’s strategy must specify a reaction to each possible proposal. First, indicate the pure strategy Nash equilibrium(ia). Then indicate how the concept of subgame perfect Nash equilibrium can eliminate seeming irrationality within a Nash equilibrium. Your answer should explain the value of the concept with respect to the usefulness of game theory and Nash equilibrium in general. Proposer P gets $3; R gets $1 Each gets $2 Responder Accept Reject 3 1 Accept 2 2 0 0 Reject 0 0 Responder Proposer A/A A/R R/A R/R P = $3; R = $1 3, 1 3,1 0,0 0,0 P = $2; R = $2 2,2 0,0 2,2 0,0 Answer: There are three Nash Equilibria: (P = $3; R = $1, A/A), which is SPNE (P = $3; R = $1, A/R) (P = $2; R = $2, R/A) It would not be optimal for the Proposer to offer $2, $2, knowing he could instead offer $3, $1 and get a higher payoff of $3. While in strategic form offering $2, $2 appears as a NE, the extensive form clearly indicates that $3, $1 is optimal since the Proposer gets $3 rather than $2. SPNE adds predictive power by ruling out equilibria that are sustained by a player making an irrational choice. 5. Carefully write the first section of your paper. This should include the introduction to your paper generally, the scenario you are modeling, the players, and their objectives. It also should include your extensive form diagram and your strategic form diagram. This part of your homework assignment should be typed (your extensive and strategic form diagrams may be handwritten and submitted with the problems) and sent to me via e-mail as a word document.