Yesterday, I caught Dan Bricklin's talk, The Cornucopia of the Commons. (He also wrote an essay, some version of which is also in O'Reilly's P2P anthology.) Dan talks about user-created databases, which is really the key value to Napster. The unique approach to the plumbing in Napster gets all the attention and was necessary for legal and cost reasons, but the functionality that added the value was the shared database of all the user's music. Other examples Dan mentions are CDDB and AmIHotOrNot. These are both great examples of user-generated, shared databases. But a distinction occurred to me, which I think is key to designing such systems: There should be a payoff to the user for entering accurate information. This exists for CDDB but not for (half) of AmIHotOrNot.
Many (most?) CD's get into CDDB because someone bought the CD and wants their computer to know what's on it, for their own use. That the information is useful to anyone else who has it is a nice side-effect, and I'm sure a major motivation to some people. But, essentially, if the user puts in good information, they are benefiting themselves (and everyone else), and if they put in bad, they are doing the opposite.
With AmIHotOrNot, there are two types of user-created data: the pictures and the votes. There is not much incentive to putting in non-real pictures. I mean, you can put in pictures that are not of you, but it doesn't really hurt the system anyway. However, there is no particular reason to put in "good" data on the other side -- i.e., vote how you really feel. Most people probably do, but there is not as strong a motivation to. In fact, something Jim Gallagher wrote, in response to Dan's talk, "He didn't even mention that it is FUN to vote high for ugly chicks," implies that some people get more entertainment (which is the main value the site provides) by putting in "bad" data than good.
One could argue this doesn't really matter for something as arbitrary and non-serious as AmIHotOrNot, but obviously if the voting results were absolutely meaningless it would be less interesting for people. It obviously works well enough, and I'm not sure how it could be better, but in general I think this idea is an important point to keep in mind. CDDB was genius in that, by benefiting yourself, you're benefiting others without any extra effort.