Prisoner's Dilemma

From P2P Foundation
Jump to navigation Jump to search

Prisoner's Dilemma = famous thought experiment in Game Theory, about selfishness vs. altruism in the context of human cooperation.


Discussion

Axelrod

From http://bostonreview.net/BR23.6/bowles.html:

"A convenient starting point in tracing the birth of Homo reciprocans is a tournament involving differing strategies of the play in the "prisoner's dilemma" game undertaken two decades ago by Robert Axelrod at the University of Michigan. The prisoner's dilemma requires each of two players to choose simultaneously one of two actions, "cooperate" or "defect." The way the payoffs work is that both players do better by cooperating than defecting, but whatever one player does, the other player does better by defecting. For example, the payoff to mutual cooperation is 10 for each, the payoff to mutual defecting is 5 for each, but the payoff to defecting when the other player cooperates is 15 for the defector and 0 for the cooperator. The iterated prisoner's dilemma is simply repeated play of the game with "winners" being those with high cumulative scores over however many rounds are played.

Axelrod asked a number of game theorists, economists, political scientists, sociologists, and psychologists to submit computer programs giving complete strategies for playing the game successive rounds of which were repeated with the same partner. Each program was pitted against every other program, as well as itself and a program that randomly chose to cooperate and defect. Surprisingly, the winner among the fourteen strategies submitted was the simplest, called "tit-for-tat" (submitted by game theorist Anatol Rappoport). Tit-for-tat cooperates on the first round, and then does whatever its partner did on the previous round.

Following up on this result, Axelrod held a second tournament in which a larger number of participants, including the original contributors, were told of the success of tit-for-tat and asked to submit another program for playing the iterated prisoner's dilemma. Knowing that tit-for-tat was the strategy to beat did not help the players: once again Rappoport submitted tit-for-tat, and once again, it won.

Speculating on the strong showing of tit-for-tat, Axelrod noted that this strategy for cooperation has three attributes that are essential for successful cooperation. The first is that tit-for-tat is nice: it begins by cooperating, and it is never the first to defect. Second, tit-for-tat is punishing: it retaliates relentlessly once the other party defects. Finally, tit-for-tat is forgiving: as soon as a defecting partner returns to cooperating, tit-for-tat does the same." (http://bostonreview.net/BR23.6/bowles.html)


C. Ford Runge

"The theoretical justification of Hardin's "Tragedy of the Commons" reasoning was also challenged in this period. That justification modeled the tragedy as a Prisoner's Dilemma game, where the rational strategy is to be "greedy" even though the long-term benefits of being "fair," though "irrational," are much greater. This model was challenged because in a prisoners' dilemma game, the players are limited to a one-shot trial and are not allowed to communicate with each other. But if the players of the commons game can communicate and can have many trials it is easily shown that Hardin's conclusions do not hold. Indeed, the comparison between the prisoners' dilemma game and the typical common situation is far-fetched.


C. Ford Runge pointed out this absurdity in a series of papers in the 1980s according to this account:

…most users of a common-pool resource-at least in developing countries-live in the same village where their families had lived for generations and intend to live in the same villages for generations to come. Given the level of poverty facing many villagers, their dependence on natural resources, and the randomness they all face in the availability of natural resources, Runge argued that it is implausible to assume that individuals have a dominant strategy of free riding. He suggested that users of common-pool resources in developing countries faced a repeated coordination game rather than a one-shot prisoners' dilemma game. In such situations, all users would prefer to find ways of limiting their own use so long as others also committed themselves to stinting (Dietz et al. 2002: 12)." (http://www.globaljusticecenter.org/papers/caffentzis.htm)


More Information

  1. See for a good review of the Prisoner’s Dilemma, this entry of the Wikipedia, at http://en.wikipedia.org/wiki/Prisoner's_dilemma
  2. In this essay, defending the possibility of ‘anarchist cooperation’ as a model of society, the author reviews the objections of the classic Prisoner’s Dilemma, arguing that they are superseded in case of repeated re-iteration, see http://www.spunk.org/texts/misc/sp000161.txt
  3. The following text is a good defense of the use of game theory in scientific research, at http://www.geocities.com/hmelberg/elster/AR82MFGT.HTM
  4. The Prisoner's Dilemma is often contrasted to the Assurance Game and the Chicken Game.
  5. Video treatment in The Trap, 3-part BBC documentaries by Adam Curtis, http://video.google.com/videoplay?docid=404227395387111085&q=the+trap++adam+curtis&ei=eryGSKTTMJKuwgPO8LiJCA
  6. The Prisoner's dilemma in The Dark Knight movie, http://thepiratesdilemma.com/changing-the-game-theory/the-jokers-trap