Natural Experiments in Synthetic Worlds

Hard sciences such as physics and chemistry tend to look down upon soft ones, such as psychology or economics, for allegedly being not scientific enough. Economics was even dubbed "the dismal science" by Thomas Carlyle. Physicists, with their frightfully expensive supercolliders smashing elementary particles to smithereens, crow that their ability to conduct and reproduce experiments is what makes their science a Science. Economists have nothing of the sort. Or do they?

Well, it appears that economists not only do have their own version of a supercollider, but that it also dwarfs to nothingness even the 27-km Large Hadron Collider physicists are building at CERN: it spans the entire globe. And unlike CERN, it cost taxpayers not a penny to set up.

This "social science supercollider" was not developed for the study of economics, however. It was "discovered" for this purpose by Edward Castronova, the foremost guru of the economics of that realm where the cyberworld meets the Earth. As he points out in his latest CESifo Working Paper, the designers of some large online computer games such as EverQuest and The Dark Age of Camelot have unwittingly provided a formidable tool for reproducing social experiments with real people under controlled conditions. To try out his insight, Mr. Castronova put a well-known social economics theory to the test in these synthetic worlds, in what may be the first time in human history that a distinct macro-social phenomenon has actually been verified experimentally.

But let's start at the beginning. Experimental verification in social sciences is enormously difficult because it is not possible to observe whole societies under controlled conditions. Worse, even if you could observe whole societies, replicating the experiment is well nigh impossible. And observing parallel societies was out of the question: they simply didn't exist.

Now, however, computer game designers have created synthetic worlds that make it possible to replicate entire societies and allow them to operate in parallel. These synthetic worlds are the stage where the online games mentioned above are played out. Hundreds of thousands of people from all over the world spend a good part of their time immersed in these cybernetic creations, interacting with other players using many of the social interaction features common to societies in the real world. What Mr. Castronova did was to exploit this heaven-sent feature to test co-ordination game theory experimentally not only under controlled conditions, but even under perfectly controlled conditions.

A bit on co-ordination
Let's see first what this co-ordination game is all about. For one thing, it is not a game: co-ordination effects are genuinely powerful in human society and are very important to a number of, well, soft sciences such as political science, economics, sociology, and anthropology. It is based on the fact that human co-ordination saves lots of time and energy.

The value of money is a case in point. A bank note actually has no value: you can't wear it, eat it or even write on it. But it has transaction value: we all expect everyone else to treat it as though it has value. What money is worth is determined by the co-ordination equilibrium: everyone agrees on what, say, one dollar is worth.

Governmental power is another example. Say, there is a revolution, and the revolutionaries broadcast through every communication channel that the New Big Man now is "the government". Thus, they instil the expectation that everyone will be treating New Big Man as the head of government (or else), and if everyone treats New Big Man as the head of government, well, then he is the head of government. By signalling to the entire populace that he is "it", he has managed a feat of co-ordination and thus, to shift the focal point from the old government to his.

But herein lies the fragility of co-ordination power. A mere shift to a new focal point condemns the old one to lose power utterly. Just think of the shift when the Berlin Wall came down.

In short, a co-ordination equilibrium, or focal point, is immensely powerful so long as everyone believes in it. Yet a slight alteration in these beliefs is all it takes to destroy the power of a focal point completely.

(Tip: If you want to learn more about co-ordination games, the best overall analysis in that field was written by Nobel laureate and CESifo Research Fellow Thomas Schelling: The Strategy of Conflict.)

So, as you can see, this powerful-fragile piece of co-ordination appears to play a significant role behind real-world social change. And yet it has never been demonstrated empirically the way it would need to be for complete conviction: under controlled conditions, at the scale of an entire society. That is, until Mr. Castronova's insight of using the "social supercollider" —i.e. large online games—to test it.

The experiment
Large online games are played by large numbers of people in synthetic worlds accessible to anyone with an internet connection. On account of their sheer size and complexity, they can be categorised as a genuine human society: some 500,000 people devote their entire attention to it for hours on end every day. While they may go about as knights, night-elves or orcs slaying monsters or each other, the players organise socially in a fashion identical to that of the real world: i.e. although synthetic, it is real societies that accumulate within these worlds. As Mr. Castronova puts it, "competitive, overlapping teamwork is just one of an uncountable number of core phenomena shared by all human societies, and, if a game gets big enough, these phenomena will appear in them with absolute fidelity."

The crucial difference of online societies is that they operate under controllable conditions. The builders of these games can vary almost everything—except human nature itself.

Given this level of control, a powerful research strategy insinuates itself: build several synthetic worlds in exactly the same way, except for some difference in a variable of interest. Attract people into the worlds, sit back, and watch what happens.

Applying this to the EverQuest world, which was replicated in 40 servers, each populated with a different, random mix of users, Mr. Castronova watched where the markets sprung up in each "world", in order to find out whether co-ordination effects really happen on a large scale in human society.

He went about it by basing partly on his observations published in an earlier CESifo paper (which by the way is the absolute number one in number of downloads from the SSRN network among economics papers). He asked where the player-to-player market was located: if I as a player want to sell certain items I have obtained during the game, where would I go to sell them? Conversely, if I was looking to buy something, where would I go to meet sellers?

The answer was surprising: more than 85% of all respondents—a statistically extremely significant finding—agreed that such-and-such location, and no other, was the trading place, even though no location was preordained by the game designers. Crucially, not the same zone was acknowledged as the trading place in every server, thus ruling out some design quirk that could push people to pick up always the same zone. Clearly, the trading zones represented a powerful shared expectation of those populations, a "focal point". Talk about co-ordination.

In terms of both the breadth and depth of the theory, the markets in this cyberworld offer rich evidence of co-ordination's existence, and its force. When the incentives are right, a whole society of real people act precisely as co-ordination theory predicts.

But that is not the really the big news here. It is the fact that large games can find these co-ordination effects.

That, it would appear, places large online games firmly in the experimental economics arena.

Edward Castronova: On the Research Value of Large Games: Natural Experiments in Norrath and Camelot, CESifo Working Paper No.1621

• Other CESifo Working Papers dealing with game theory
• Other CESifo Working Papers dealing with synthetic worlds
• Other CESifo Working Papers by Edward Castronova
• A review of his book "Synthetic Worlds: The Business and Culture of Online Games"
• Edward Castronova's website
• CESifo Working Papers published in the past three months