Wednesday, February 20, 2008

Free riding in Wikipedia

Axelrod and Hamilton (1981) showed that a winning strategy in prisoner's dilemma games is tit-for-tat; that is, one should start out behaving cooperatively, and then on every subsequent move simply match what the other agent did in her previous move. This strategy, however, describes only 2-person, repeated games. In games where multiple agents interact with each other, even a small number of defectors will prompt the conditional cooperators to also defect, so that the game moves to an equilibrium of no cooperation. Thus, the model cannot explain how cooperation persists in groups with multi-agent interactions.

Experimental games have shown that cooperation can be maintained when defectors are punished. But since punishment is costly to the punisher, it would be rational for a player to let others punish the defectors--these rational players are called second-order free riders (since they are free riding by letting others punish free riders). Without rewards for the punishers, or punishment for the second-order free riders, no punishment of defectors will occur.

Panchanathan and Boyd (2003) have shown that reputation works well in models as the reward sought by punishers, who gain in reputation by helping the deserving (those with good reputations) and punishing the undeserving (those with bad reputations--i.e., persistent defectors). The ultimate reward of reputation is that others willingly cooperate with them. The reputation model requires that agents know each other's reputations, which is only realistic in small groups.

Second-order free riding is a common problem on Wikipedia. One often encounters editors who appear to be doing something wrong: inserting a strange point-of-view; moving pages without discussion; deleting good work done by others. But it is costly to battle these people. Conflict itself is unpleasant, and one is likely to break some rules when in a conflict. If an administrator observes the broken rule she is unlikely to have the time to look back at the history of the conflicting parties and figure out who is the rogue and who is the guardian, and she will simply punish whoever broke the rules. What she lacks is ready access to information on the reputations of the parties in conflict.

A mechanism to record and update reputation would encourage good editors to oppose the actions of bad editors--it would reduce the problem of second-order free riding, and make it more difficult for rogue editors to get their way.
  1. Axelrod, Robert, and William D. Hamilton. 1981. “The Evolution of Cooperation.” Science 211: 1390-1396.
  2. Panchanathan, K., and R. Boyd. 2003. “A tale of two defectors: the importance of standing for evolution of indirect reciprocity.” Journal of Theoretical Biology 224:115-126.

No comments: