The evolution of cooperation

from Wikipedia, the free encyclopedia

The Evolution of Cooperation ( english The Evolution of Cooperation ) is an 1984 in the original English and published in 1987 in the German translation book by the political scientist Robert Axelrod , based on several journal articles by the author. The work as a computer simulation models from a game theory point of view how cooperation among egoistic individuals can come about even if it is not enforced by an external authority, by morality or by laws.

content

Axelrod assumes an iterative prisoner's dilemma . In this simple two-person game, both players have the incentive to behave uncooperatively if they act rationally and selfishly - if no other aspects such as morality or mutual trust are added, and if they only engage in this once and then never again Encounter situation. The dilemma of the prisoner's dilemma is that while defection would be correct (rational) for every player from his point of view, it is a non-zero-sum game in which not only one can win, but both can maximize the overall profit through cooperation .

Axelrod investigated what happens when this situation is repeated many times; when the two players meet again and again without knowing when they will last meet (iterated prisoner's dilemma). In the first place, there is no best strategy, as the strategy always depends on the opponent's strategy. In two programming competitions , he brought together different, sometimes very sophisticated strategies. The programs are written by a large number of people from different disciplines, including mathematicians, computer scientists, social scientists, economists and psychologists. It was found that cooperative strategies prevail in an iterated prisoner's dilemma.

It turned out that a very simple strategy was most successful both times: tit for tat , won both competitions and got the most points overall, over all the games in a tournament. Tit for Tat prevailed against complex, sometimes stochastic processes, including strategies that cleverly tried to outsmart the others. Tit for Tat consists in cooperating with another player the first time they meet and then always imitating his behavior from the previous round in all subsequent rounds.

Axelrod's experiment is based on idealizing assumptions:

  • All players have the same resources. Apart from the possibility of cooperating or refusing to cooperate in the individual game rounds, they have no possibility of rewarding other players or exercising reprisals.
  • Apart from the previous history of their interaction in the previous game rounds, the players know nothing of each other.

These assumptions may seem alien. However, there are a number of real-world situations in which this multi-round prisoner's dilemma can be found. In essence, it always occurs when a group of people has the choice of exploiting a common resource for maximum self-interest - or holding back in order to maximize the common, sustainable benefit. Axelrod's work therefore became the basis for a large number of further research projects in the social, political and legal sciences.

The analysis of the two competitions revealed the following conditions for a successful strategy:

  1. The strategy has to be friendly, that means: Never be the first to fail to cooperate (to fail)! All strategies that never failed first (did not cooperate) were at the top.
  2. The strategy must be powerful and not resentful, that is, be able to repay a defection of the opponent with a defect of its own and then cooperate again. This is important in order not to be exploited. The strategy must learn from the actions of the opponent and be able to remember his moves. In order not to end up in an endless echo of mutual defection, the setback should be clear, but de-escalating.
  3. The strategy must not be jealous! If the strategy never fails first, it can only ever be as good as the opponent and never better. By answering opposing deficiencies, she always lags one step behind and has a few points less in the end. If you try to catch up, you will abandon the path of the best strategy, and overall you will do worse!
    Specifically, this means that Tit for Tat does not win the competition because it achieves a high number of individual victories, but because Tit for Tat is, on the contrary, quite easy to defeat in a direct duel and thus wins rather few duels, but these defeats are included go hand in hand with high scores.
  4. Don't be too clever, that is, the strategy has to be transparent to the opponent so that he can adapt to it and not believe it is a random strategy ( Random did not do particularly well).
  5. Any strategy only works if the probability of a future game and therefore future profit is sufficiently high. If this probability is too small, it is always better to defect and to secure the current profit. If you want to cooperate with your opponent, you have to ensure that the game returns sufficiently frequently and safely.

Of all the strategies submitted for the competition, Tit for Tat met these requirements best. One disadvantage arises from the setbacks of Tit for Tat that are not de-escalating , which means that this strategy is not error-proof. If a player incorrectly interprets the cooperative behavior of his teammate as a defection, the mutual defection continues, although both parties are basically geared towards cooperation. In politics, such misunderstandings can lead, for example, to a reluctant arms race, to reciprocal trade embargoes or to other critical developments. But not only misunderstandings, but also the exaggerated playing of tit for tat can lead to defection always being followed by defection, which would lead to a barely profitable chain.

This can be remedied by the fact that after random, i. H. For the opponent the number of repetitions cannot be estimated spontaneously, one-sided cooperation is played in order to break the cycle of confrontation, which is not provided for by Tit for Tat .

plant

Original edition:

Translation:

In a follow-up volume, Axelrod has summarized and commented on several of his specialist articles on the topic that appeared after 1984.

  • The Complexity of Cooperation: Agent-Based Models of Competition and Collaboration . Princeton University Press, Princeton, New Jersey 1997, ISBN 0-691-01568-6 .

See also

Individual evidence

  1. ↑ The book builds on the following articles:
    • Robert Axelrod: Effective Choice in the Prisoner's Dilemma . In: Journal of Conflict Resolution . tape 24 , no. 1 , 1980, p. 3-25 , JSTOR : 173932 .
    • Robert Axelrod: More Effective Choice in the Prisoner's Dilemma . In: Journal of Conflict Resolution . tape 24 , no. 3 , 1980, p. 379-403 , doi : 10.1177 / 002200278002400301 .
    • Robert Axelrod: The Emergence of Cooperation among Egoists . In: The American Political Science Review . tape 75 , no. 2 , June 1981, pp. 306-318 , doi : 10.2307 / 1961366 .
    • Robert Axelrod, William D. Hamilton : The evolution of cooperation . In: Science . tape 211 , no. 4489 , March 27, 1981, p. 1390-1396 , doi : 10.1126 / science.7466396 .