Wednesday, March 19, 2008

The Evolution of Cooperation

The Evolution of Cooperation by Robert Axelrod

The Evolution of Cooperation was the title of a 1981 article co-authored by the political science professor Robert Axelrod and the late evolutionary biologist William D. Hamilton. Robert Axelrod expanded it into a book in 1984. It is interesting to note this book which was written more than 20 years ago is still fresh in ideas.



Some philosophers said human are born benign, but some said otherwise. The most famous answer was given over three hundred years ago by Thomas Hobbes. He was pessimistic and argued that before government existed, the state of nature was dominated by the problem of selfish individuals who competed on ruthless terms and life was solitary, poor, nasty, brutish and short. He thought cooperation could not develop without a central authority.

You may have heard, or played, the Prisoners' Dilemma. The setting was that two criminals together committed a serious crime were caught. The detective kept them in separate cells and asked each of them to confess the crime. Circumstantial evidence was weak and so if both of them cooperated and kept silence, they would only be convicted of a minor offense. The detective told each of them that if he confessed and turned into a prosecution witness against the other, he would be acquitted while the other would get the maximum sentence. If both of them confessed, they could beg leniency for a lighter sentence. The game theorists proved that the best strategy for the game was to confess and to defect against the other. This may partly prove that Thomas Hobbes was right.

The crux of the game was that the prisoners did not know the intention of the other party. They could only guess the scenarios and calculate the probability of returns. Axelrod developed a variation of the game, involving repeated game interactions between two players with the results of each interactions known to them. He gave scores to each game played: 3 points each for mutual cooperation, 1 points each for punishment of mutual defection, 5 points for defection leading to conviction of the other, and 0 point for keeping silence while being defected against. When previous performance of the other side was known, the player could take it into account and develop a strategy to take advantage of cooperation and also timely defection to gain points. There is afterall merit in cooperation sometimes.

A computer programming tournament of the iterated game was organized. Programmers engaged various strategies, ranging from algorithmic complexity, initial hostility, occasional defection, being nice or forgiving to induce cooperation, etc. The winner of the game was the program Tic-for-tac. Its strategy was to be nice and cooperative on the first move, then consistently repaying cooperation or defection according to the last move of the opponent. Analysis showed that the success of Tic-for-tac was based on four factors: being nice and never defect first, being provocable as it would retaliate by defecting whenever defected upon, being forgiving as it only retaliated once; being clear in intention as its strategy was quickly known by its opponents. Axelrod concluded that these basic characteristics were conducive to developing cooperation. Being nice and forgiving were essential attributes while making them known was also important. Being provocable was a good defense from being bullied of being nice and could raise the chance of survival.

A further conclusion by Axelrod was that Tic-for-tac was robust even in a hostile environment. In such an environment where there was no cooperation, participants would gain a few points through mutual defection and a single Tic-for-tac would die out. However, if there were a small number of Tic-for-tac in such environment, interactions of cooperation between them would bring more points. If they were a close group, then such in-group interactions, however small in number, would be better off than the always-defect majority. As a result, this Tic-for-tac group would grow in the population and cooperation would eventually evolve to be the dominating strategy. This phenomenon has been observed in the development of many civilizations.

The book also found that cooperation did not necessarily occur between friendly parties. It described the case of the trench warfare in World War I. Known as Live and Let Live, troops on both sides of the trench warfare spontaneously developed a strategy of cooperation whereby they shelled the other side according to a fixed schedule only known to the soldiers and allowing either side to minimize casualties. The generals were satisfied that the war was going on with the shelling but the combat line was neither advanced nor breached. The soldiers followed a similar strategy as Tic-for-tac that they never directly shelled bunkers nor supplies first, always retaliated with more accurate shelling when directly hit, always returned to pretentious shelling afterwards, and making sure that such behaviour was known to the enemy. This cooperative phenomenon which went on for a long time during the war saved many lives.

From the perspective of a reformer, Axelrod proposed the strategy on how to promote cooperation. 1. Enlarging the shadow of the future - Mutual cooperation can be stable if the future is sufficiently important relative to the present because the players can each use an implicit threat of retaliation against the other's defection. 2. Change the payoffs - This is usually adopted by government in raising the payoffs in terms of taxes, harsh punishment for crime and obligations to honour contracts; that is: to raise the payoffs for cooperation and vice versa for defection. 3. Teach people to care about each other - Through education, adults shape the values of children so that the preference of the new citizens will incorporate their own welfare with the welfare of others. 4. Teach reciprocity - Always reciprocating defection may not be the standard of morality for some religions, for example the teaching of always cooperative by turning the other cheek. However, reciprocity is a robust strategy in upholding cooperation and avoiding exploitation. The self-policing feature of the strategy gives an incentive to cooperation. Teaching reciprocity to those with whom one will interact will help build a mutually rewarding relationship. 5. Improve recognition abilities - The ability to recognize the other players from past interactions and to remember the relevant features of those interactions is necessary to sustain cooperation.

No comments:

Post a Comment