clipped from www.wired.com

Constance Steinkuehler -- a game academic at the University of Wisconsin -- was spending 12 hours a day playing Lineage, the online world game. She was, as she puts it, a "siege princess," running 150-person raids on hellishly difficult bosses. Most of her guild members were teenage boys.

But they were pretty good at figuring out how to defeat the bosses. One day she found out why. A group of them were building Excel spreadsheets into which they'd dump all the information they'd gathered about how each boss behaved

Then they'd develop a mathematical model to explain how the boss worked -- and to predict how to beat it.

Often, the first model wouldn't work very well, so the group would argue about how to strengthen it.

That's when it hit her: The kids were practicing science.

They'd think of a hypothesis -- This boss is really susceptible to fire spells -- and then collect evidence to see if the hypothesis was correct. If it wasn't, they'd improve it until it accounted for the observed data