I'm hoping that we can bring this kind of simulator to the next level soon--a useful simulation of society as a whole. Right now we're running all sorts of different experiments on our society in education, economics, culture, etc. There's also the huge experiment being tried out in Iraq to replace a dictatorship with democracy. We have arguments over all of these and toss around theoretical analyses and historical analogies, but there's no way to test them before we try them out on real people. Even then we haven't settled anything because we don't have a control case and the arguers maintain that things are better/worse than they'd've been without the experiment, depending on where they stood before it was tried. The only solid results come from the long-term tests--democracy outcompetes feudalism, capitalism outcompetes socialism. Even then people dispute the meaning of the results.
A good society simulator would let us put different politicians' plans in and see how they'd play out over the next 15 years. I'd sure like to see that before casting my vote. Figuring out what makes a "good" simulator would be tough, but there's a good test we can give the various prototypes: put in the real world, get some predictions, and open the envelope five years later and see which one did best. A few generations of that should get us a SimSociety that reflects how real people behave. There'd be a lot of heartache over that process as various cherished theories bite the dust. If economics and political science can become testable sciences a lot of nasty arguments are going to get settled for good.
Actually trying to program something like that is a horrible problem, though. You'd wind up needing almost true artificial intelligence to simulate all the ways people could respond to changes in their environment. So there's not much chance of getting a useful SimSociety in the next few years if we want it to be pure software.
There is another possibility. Right now lots of people are paying real money to participate in simulations of complex societies. Everquest and the rest are deliberately unrealistic so they don't provide any simulation of real life, but they still have mechanisms for economics, building social institutions, dealing with predatory behavior, etc. So--if we build a SimSociety MMORPG can we find people who want to play, and if we do can we make rules that will let the game accurately reflect some portion of society?
Finding players shouldn't be a problem to judge by the Sims. If we're using this for serious research we can also pay players, either a small fee to everyone or (probably more useful) larger rewards to those who successfully push the system to its limits. What could break the simulation is getting players who want to force a particular interpretation of the world into the rules for the simulation, so they act according to their ideology rather than the current situation in the sim. If we made a SimTaxPolicy game there'd be players coming out of the woodwork to bend it to whatever agenda they were carrying. We could try jury duty type rules to keep partisans out but that also weeds out people with knowledge of the subject. A random selection of players might be best. Axe-grinders would be tolerable if they're a small fraction of the players and if they include opposed beliefs that will cancel each other out.
Once SimSociety is working scores of uses will appear. A writer recently complained that the Army's wargames of the Iraq invasion didn't look at looting, decayed infrastructure, or the mood of the people. But in the current state of the art any such simulation would just reflect the preconceptions of the organizers, which is what he wanted replaced by the sim. Until we have a good society simulation we're left with theory and history. Good for creating ideas. Lousy for testing them. Hopeless for convincing somebody his premises are wrong. We really need to do better.