Friday, January 21, 2005

MMORPGs, Jerks, and the Evolution of Cooperation

Why do some online games seem to breed so many jerks?

The players in massively multiplayer roleplaying games (MMORPGS) often seem to demonstrate one of two radically different philosophies of interactive play. On the one hand, you have players who believe that they do best for themselves by helping the entire group do well. And on the other hand, you have players who believe that the best (most fun/efficient) way to succeed is by taking advantage of other players.

Their high level of social interaction makes MMORPGs fertile ground for studying what Robert Axelrod called "The Evolution of Cooperation."


In his groundbreaking work, which has been applied in fields from evolutionary psychology to international conflict resolution, Axelrod created a computer simulation of the classic Prisoner's Dilemma game, where two players must choose whether to cooperate with each other or defect (take advantage of the other player). Axelrod's innovation was to use the computer to simulate the effects of allowing players to interact many times beyond a single interaction. This permitted the emergence of strategies for play over multiple interactions.

What Axelrod found was that the consistent winner was the simplest strategy, submitted by Anatol Rapaport, called "Tit-for-Tat." TFT's rules were simple: cooperate on the first move with another player, and subsequently do unto the other player what it does to TFT. Tit-for-Tat never did very well in individual interactions -- the strategies which defected more frequently without provocation in order to try to take advantage of "nicer" strategies tended to do best in individual interactions.

But when interactions were extended over long periods, Tit-for-Tat did best overall. By eliciting cooperative behavior from other players, TFT did just well enough for itself to win over the long term.

Axelrod then modified the simulation to create an evolutionary aspect: many interactions among many strategies (including TFT) were held simultaneously as a "generation." Based on the results of individual interactions, winning strategies were retained to play again in the next generation, while losing strategies were culled from the field of players.

This time, Tit-for-Tat not only won again as the most effective strategy over many generations, it did so even more conclusively than in the previous tournament. By being best at eliciting cooperative behavior from other players, TFT did best for itself over the long term.


In his analysis of these results, Axelrod observed that while defectors tend to do well over the short term, defining the world in which interactions occur so that several conditions are met can allow a world full of defectors to evolve over time into a world full of cooperators:

  • players must be able to recognize each other
  • players must be able to remember past interactions
  • cooperators must be able to interact frequently with each other
  • there must not be a known limit on the number of possible interactions
  • cooperation must pay more than an average defection
  • the initial population must contain a minimum percentage of cooperators

In other words, game features that support reliable recognition of, memory of, and frequent interaction with other cooperators are crucial to establishing cooperation as the primary mode of play in that gameworld, rather than defection. Knowing that other players can recognize you and remember what you did to them provides just enough of a disincentive to defection to make cooperative behavior the norm through repeated interactions of cooperators.

Conversely, anonymity and lack of grouping tools breed bad behavior. Anonymity in particular -- or, in MMORPGs, pseudonymity -- is lethal to cooperation. When the odds are high that someone with whom you interact will never know the "real" you behind your avatar, or that interactions are unlikely to be repeated, the price of defection is low enough to allow defection to survive or even thrive as a general behavior. This is why it appears that gameworlds (such as WoW or EVE) which permit large numbers of characters in one area have a much higher percentage of jerks than very small gameworlds (such as MUDs) where the players have a high frequency of recognizable interactions, and may even know each other personally in real life.

If a game is structured so that it's hard to find and group with other cooperators, or if players can easily create new characters on a server and thereby remain effectively anonymous, then it is safe to expect that game to be much more of a Hobbesian environment, full of players taking advantage of each other, safe in the knowledge that they'll probably never see that other player again. Players of games whose rules permit high levels of anonymity have less incentive to cooperate because the cost of defection is low. The typical interaction will be finding creative ways to hose your fellow players.


From a game designer point of view, Axelrod's observations offer some practical tips on how to structure an online game's rules for player interaction to either promote or discourage cooperative social behaviors. If you like the idea of a dog-eat-dog, 24x7 gankfest, then you'd encode game rules that turn the knobs way down on the bullet-pointed conditions for cooperation listed above. If instead you want to create a game that rewards trust and punishes parasites, then you'd implement game rules that crank the dial to 11 on these conditions.

Here's what I wonder: What would such games be like? Assuming other game features that make those games fun to play, would a game that deliberately violates all the conditions for cooperation be successful, either critically or commercially? Could a game that actively promotes being a bastard to every other player be enough fun for enough people to be successful?

What about the opposite: how about a game that deliberately defines the rules of the gameworld so that a very high degree of cooperation is encouraged? (Not forced, just indirectly encouraged.) Would such a game be too "angelic" to survive? Would the defectors still show up and prevent the cooperators from winning over the long term? Or would such a highly cooperative game wind up being The Game that the computer-owning but non-game-playing public are waiting for?

No comments:

Post a Comment