  I just finished James Surowiecki's terrific The Wisdom of Crowds . The thesis is that large groups, in many different situations, can provide better judgments than the best individuals. The archetypal example is the Jelly Bean count: have a group of people guess the number of jelly beans in a jar.
On average, the average guess of the group will be better than the "best" member's guess, and will be fairly close to the correct number. The "best" member will not be consistent. From this he deduces that there are no real experts, and that the crowd is better for getting the decision right. I've been exploring an alternative explanation. Assume the following: there's no way of knowing how many beans are in the jar; however, there are maximum, minimum, and "somewhere around here" that normal people can figure out based on their experience; individual guesses within #2 are random with a bell-like curve distribution based on #2.
If this is true, the fact that the group gets close to the right number isn't "magic", it's just that if you get lots of people guessing, based on some boundaries, you'll get a nice smooth distribution with an average near the right number. In theory, if you could get people to make their implicit knowledge of #2 explicit, you could just have a computer Monte Carlo simulation guess just as well 
