Showing posts with label game theory. Show all posts
Showing posts with label game theory. Show all posts

Monday, October 15, 2012

Stag hunt, deadlock, & the sickle cell anemia–malaria game

In last week's post, inspired by William Poundstone's book Prisoner's Dilemma: John von Neumann, Game Theory, and the Puzzle of the Bomb, I owned up to my years-long mistake of calling social dilemmas in general prisoner's dilemmas, and I described a distinctly different social dilemma called chicken. Today I'll describe two more dilemmas from the book.

Stag hunt

The stag hunt is like prisoner's dilemma but with mutual cooperation given the greater good. Here's the payoff table. (Again, like in last week's post, lower numbers are better.)

Cooperate Defect
Cooperate 1, 1 4, 2
Defect 2, 4 3, 3

Because in the stag hunt everyone is best off cooperating, there should in theory be no dilemma: the rational choice is to always cooperate. But that only happens on the make-believe planet inhabited in the minds of renown economists, the place where everyone is rational. In the real world the situation is more interesting because cooperating with an irrational player who defects causes you to end up with the worst possible result: a low score of 4. Thus, there's a preventative incentive to defect—just in case your opponent is thinking the same thing.

This makes stag hunt more of a tragedy than prisoner's dilemma and chicken. Whereas in those two games the players are victims of circumstance, the problems born of a stag hunt are self-made owing to a lack of trust.

A good real-world fit for a stag hunt meltdown is nearly any kind of financial bubble, whereby reason is subordinate to greed and fear. I heard more than one person in Phoenix saying after the housing bubble popped that they felt they had to buy a house during the run-up in prices because they feared otherwise becoming forever priced out of the market. This thinking follows the defect before they do destruction of a stag hunt.

Deadlock

The weakest of the four social dilemmas is deadlock.

Cooperate Defect
Cooperate 3, 3 4, 1
Defect 1, 4 2, 2

When I read Poundstone's book and first saw the payoff table for deadlock, I tried without success to imagine what this scenario describes. I should have taken a hint from the book's title: Deadlock describes nearly any attempt by two countries to agree to reduce their nuclear arsenal. Cooperation is equivalent to going along with the agreement, and defection is equivalent to breaking the agreement—presumably in secret. In such a scenario, the best outcome for any country is to secretly keep their arsenal while the other country dismantles theirs. Second best is mutual defection, in which case that country at least maintains their nuclear privilege over the have-not countries. The worst outcome is going along with the agreement when the other country defects, in which case there's still a threat of nuclear annihilation and now the country with a dismantled arsenal has no counter-threat.

As its name implies, deadlock leads rational players to always defect, just as in prisoner's dilemma.

Sickle cell anemia and resistance to malaria

So of the four games I described—prisoner's dilemma, chicken, stag hunt, and deadlock—which best describes the conflict inherent in the genetic mutation that leads to increased resistance to malaria but also to having sickle cell anemia? As you may remember, the conflict is symmetrical: any parent having (a single copy of) the mutation benefits from increased malarial resistance, but children of two parents both possessing the mutation may end up with sickle cell anemia.

Imagine the game as being played between the parents, with each parent choosing either to cooperate by not having the mutation or to defect by having the mutation. Here's the payoff table.

Cooperate Defect
Cooperate 3, 3 2, 1
Defect 1, 2 4, 4

As I've assigned the values, the best outcome for an individual is to have the mutation but one's mate not to have the mutation. But the second best outcome is switch roles so that one's children still have a chance at getting one copy of the mutation. Mutual cooperation is third best, and mutual defection, which leads to the possibility of children with sickle cell anemia, is worst.

So it turns out the sickle cell anemia–malaria game doesn't match any of the four social dilemmas I described. Indeed, I'm not sure whether it's strictly a social dilemma at all. In an iterated version, the best course would be to take turns cooperating and defecting while the other player does the opposite. In a one-shot version—which is how the game must be played in real life—the dilemma is over choosing who gets to defect, with the loser still getting the second best outcome. Because a sole defection beats mutual cooperation, this game may lack an ingredient necessary for it to be considered social.

Thursday, October 11, 2012

Prisoner's dilemma & chicken

I recently finished reading William Poundstone's book, Prisoner's Dilemma: John von Neumann, Game Theory, and the Puzzle of the Bomb, so I've got game theory on my mind, and that's what today's post is about. My apologies to reader Jill, who'll stop reading about…now.

Prisoner's dilemma

For years I've called any social dilemma where there exists a conflict between the good of the group and the good of the individual a prisoner's dilemma, but this is wrong. In the language of game theory, a prisoner's dilemma is a specific kind of social dilemma, and the other social dilemmas have their own names.

What all social dilemmas have in common is two or more people deciding, independently of others, whether to cooperate or to defect. In each dilemma, defection brings about a reward for the defector and punishment for the cooperator, though mutual defection is an undesired outcome. Prisoner's dilemma is the best known of the dilemmas because it's the most brutal: defecting is always the better choice for the individual despite mutual defection being undesirable. As a decision table, it looks like this.

Cooperate Defect
Cooperate 3, 3 1, 4
Defect 4, 1 2, 2

The way to read the table goes as follows: in each cell, the first number represents the payoff for the player whose choice determines which row to use, and the second number represents the payoff for the player whose choice determines the column. Each payoff is ranked from 1 to 4, with higher numbers being better for that person. For example, if the row player (Player 1) defects, and the column player (Player 2) cooperates, then we follow the defect row and the cooperate column and see that Player 1 is rewarded with a 4 (top score) and Player 2 is punished with a 1 (lowest score).

As you can see from the table, a prisoner's dilemma player is always better off defecting than cooperating even though mutual cooperation beats mutual defection. To see this, imagine you're Player 1, so you must select a row. Player 2 has already made their choice, though you don't know what that choice is. Should Player 2 have decided to defect, then you ought to select the best cell for you in the second column. Your choices are to cooperate and score a 1 or to defect and score a 2, so you're better off defecting. Whereas, should Player 2 have decided to cooperate then you should select the best cell in the first column. Your choices then would be to cooperate and score a 3 or to defect and score a 4, so again you would be better off defecting. The dilemma here is that Player 2 will likely arrive at the same conclusion and thus defect, too. Therefore, in a prisoner's dilemma two rational players will both defect, bringing about a score of 2 for both players. Whereas, two irrational players might have both cooperated and each scored a 3.

That's prisoner's dilemma. What are the other dilemmas?

Chicken

Another dilemma is called chicken. Here's its decision table.

Cooperate Defect
Cooperate 3, 3 2, 4
Defect 4, 2 1, 1

Chicken gets its name from any number of popular uses, one being a game where two people each drive a car straight at the other to see who swerves out of the way first. The winner is the macho player who doesn't swerve (score 4) and the loser is the coward who does swerve (score 2), though the catastrophic event here is if both players win, in which case both macho drivers are dead on impact (score 1).

Chicken is a useful scenario for describing a situation whereby everyone involved needs someone to commit to a sacrificial action, but everyone has incentive not to be that person. One example is stopping a well armed and suicidal hijacker; someone needs to stop the hijacker for the good of the group, but individually the best course of action is for someone else to be the hero. In some ways this is worse than the prisoner's dilemma because with prisoner's dilemma there's a fixed rational choice—always defect—but in chicken there's no fixed rational choice: sometimes a player is better off defecting and sometimes they're better off cooperating. Indeed, in a version of chicken where your behavior can affect the other player's decision (though this is not allowed in the strict game-theory version of the game, where both players must make their decision independently of the other), one rational course of action is to convince the other player of your own irrationality, thereby making the other player more fearful of the looming catastrophe of mutual defection and thus more likely to cooperate. But of course the other player can try the same trick on you! Just as with prisoner's dilemma, chicken is a dilemma with no clear, ideal solution.

Next post, I'll describe two more social dilemmas from the book.

Thursday, September 20, 2012

Underwhelmed by conscious choice

The genetic mutation that causes sickle-cell anemia also increases one's resistance to malaria. But whereas sickle-cell anemia is a recessive trait—meaning that to get the disease you must have two copies of the mutation—increased resistance to malaria is a dominant trait that comes from having only one copy of the mutation. Therefore, the optimal strategy for individuals living in an area with a high incidence of malaria is to have a single copy of the sickle-cell mutation, thus gaining resistance to malaria without the early death brought on by sickle-cell anemia.

The problem with this strategy is that if too many individuals pursue it then individuals suffer as a group. People who have a single copy of the sickle-cell mutation are carriers of the disease, and an offspring of two mated carriers has a one-in-four chance of having the disease and a one-in-two chance of being a carrier. An offspring of a carrier mated with a non-carrier has no chance of having the disease. So even though it's better for an individual to be a carrier and benefit from increased resistance to malaria, it's better for the group to have a mix of carriers and non-carriers, thus reducing the incidence of sickle-cell anemia. As with prisoner's dilemma, there's a best solution for the group that's in conflict with the best solution for the individual. How many other prisoner's dilemmas are lurking in our DNA?

For most of our species' history, we have benefited from mindless Nature solving our genetic prisoner's dilemmas for us. When the optimal solution is a mixed strategy, such as in tropical zones with regards to the sickle-cell mutation, Nature does an OK job of selecting for the mixed strategy. Without a mind to overthink the problem, or to pursue fashions or Faustian gains, Nature finds a balance between two diseases.

So when I hear that designer babies are on the way, Gattaca-style, and that some couples are already choosing the sex of their babies, I'm not struck with overwhelming confidence that this is a smart move over the long run—not until we're consciously able to deal with prisoner's dilemmas as well as mindless entities do.