Wednesday, April 1, 2009

If We Were All Better People The World Would Be A Better Place

Some of the power and meaning of game theory can be illustrated by assessing the statement "If we were all better people the world would be a better place." This may seem to you to be self-evidentally true. Or you may recognize that as a matter of logic this involves the fallacy of composition: just because a statement applies to each individual person it need not apply to the group. Game theory can give precise meaning to the statement of both what it means to be better people and what it means for the world to be a better place, and so makes it possible to prove or disprove the statement. In fact the statement is false, and this can be shown by a variation of the Prisoner's Dilemma.

Let us start with a variation on the Prisoner's Dilemma game we may call the Pride Game.

proudnot confessconfess
proud4.0, 4.05.4, 3.61.2, 0.0
not confess3.6, 5.45.0, 5.0-4.0, 10.0
confess0.0, 1.210.0, -4.01.0, 1.0

The Pride Game is like the Prisoner's Dilemma game with the addition of the new strategy of being proud. A proud individual is one who will not confess except in retaliation against a rat-like opponent who confesses. In other words, if I stand proud and you confess, I get 1.2, because we have both confessed and I can stand proud before your humiliation, but you get 0, because you stand humiliated before my pride. On the other hand, if we are both proud, then neither of us will confess, however, our pride comes at a cost, as we both try to humiliate the other, so we each get 4, rather than the higher value of 5 we would get if we simply chose not to confess. It would be worse, of course, for me to lose face before your pride by choosing not to confess. In this case, I would get 3.6 instead of 4, and you, proud in the face of my humiliation would get 5.4.

The Pride Game is very different than the Prisoner's Dilemma game. Suppose that we are both proud. In the face of your pride, if I simply chose not to confess I would lose face, and my utility would decline from 4 to 3.6. To confess would be even worse as you would retaliate by confessing, and I would be humiliated as well, winding up with 0. In other words, if we are both proud, and we each believe the other is proud, then we are each making the correct choice. Morever, as we are both correct, anything either of us learns will simply confirm our already correct beliefs. This type of situation - where players play the best they can given their beliefs, and they have learned all there is to learn about their opponents' play is called by game theorists a Nash Equilibrium.

Notice that the original equilibrium of the Prisoner's Dilemma confess-confess is not an equilibrium of the Pride game: if I think you are going to confess, I would prefer to stand proud and humiliate you rather than simply confessing myself.

Now suppose that we become "better people." To give this precise meaning take this to mean that we care more about each other, that is, we are more altruistic, more generous. Specifically, let us imagine that because I am more generous and care more about you, I place a value both on the utility I receive in the "selfish" game described above and on the utility received by you. Not being completely altruistic, I place twice as much weight on my own utility as I do on yours. So, for example, if in the original game I get 3 units of utility, and you get 6 units of utility, then in the new game in which I am an altruist, I get a weighted average of my utility and your utility. I get 2/3 of the 3 units of utility that belonged to me in the original "selfish" game, and 1/3 of the 6 units of utility that belonged to you in the "selfish" game. Overall I get 4 units of utility instead of 3. Because I have become a better more generous person, I am happy that you are getting 6 units of utility, and so this raises my own utility from the selfish level of 3 to the higher level of 4. The new game with altruistic players is described by taking a weighted average of each player's utility with that of his opponent, placing 2/3 weight on his own utility and 1/3 weight on his opponent's. This gives the payoff matrix of the Altruistic Pride Game

proudnot confessconfess
proud4.00, 4.004.8, 4.20*0.80, 0.40
not confess4.20*, 4.805.00, 5.000.67, 5.33*
confess0.40, 0.805.33*, 0.671.00*, 1.00*

What happens? If you are proud, I should choose not to confess: if I were to be proud I get a utility of 4, while if I choose not to confess I get 4.2, and of course if I do confess I get only 0.4. Looking at the original game, it would be better for society at large if when you are proud I were to choose not to confess. This avoids the confrontation of two proud people, although of course, at my expense. However, as an altruist, I recognize that the cost to me is small (I lose only 0.4 units of utility) while the benefit to you is great (you gain 1.4 units of utility), and so I prefer to "not confess." This is shown in the payoff matrix by placing an asterisk next to the payoff 4.2 in the proud column.

What should I do if you choose not to confess? If I am proud, I get 4.8, if I choose not to confess I get 5, but if I confess, I get 5.33. So I should confess. Again, this is marked with an asterisk. Finally, if you confess, then I no longer wish to stand proud, recognizing that gaining 0.2 by humiliating you comes at a cost of 1 to you. If I choose not to confess I get only 0.67. So it is best for me to confess as well.

What do we conclude? It is no longer an equilibrium for us both to be proud. Each of us in the face of the other's pride would wish to switch to not confessing. Of course it is also not an equilibrium for us both to choose not to confess: each of us would wish to switch to confessing. The only equilibrium is the box marked with two asterisks where we are both playing the best we can given the other player's play: it is where we both choose to confess. So far from making us better off, when we both become more altruist and more caring about one another, instead of both getting a relatively high utility of 4, the equilibrium is disrupted, and we wind up in a situation in which we both get a utility of only 1. Notice how we can give a precise meaning to the "world being a better place." If we both receive a utility of 1 rather than both receiving a utility of 4, the world is clearly a worse place.

The key to game theory and to understanding why better people may make the world a worse place is to understand the delicate balance of equilibrium. It is true that if we simply become more caring and nothing else happens the world will at least be no worse. However: if we become more caring we will wish to change how we behave. As this example shows, when we both try to do this at the same time, the end result may make us all worse off.

To put this in the context of day-to-day life: if we were all more altruistic we would choose to forgive and forget more criminal behavior. The behavior of criminals has a complication. More altruistic criminals would choose to commit fewer crimes. However, as crime is not punished so severely, they would be inclined to commit more crimes. If in the balance more crimes are committed, the world could certainly be a worse place. The example shows how this might work.

For those of you who are interested in or already know more advanced game theory, the Pride Game has only the one Nash equilibrium shown - it is solvable by iterated strict dominance. The Atruistic Pride Game, however, has several mixed strategy equilibria. You can compute them using the fine open source software program Gambit written by Richard McKelvey, Andrew McLennan and Theodore Turocy. One equilibrium involves randomizing between proud and confess, so is worse than the proud-proud equilibrium of the Pride game. The other is strictly mixed in that it randomizes between all three strategies. The payoffs to that equilibrium gives each player 2.31 - so while it is better than both players confessing for certain, it is still less good than the unique equilibrium of the Pride Game.

No comments:

Post a Comment