This reminds me of some research I read a while ago, I think it was here, that said that what matters most is neither expert nor crowd, but data points. You wan

This reminds me of some research I read a while ago, I think it was here, that said that what matters most is neither expert nor crowd, but data points. You want to maximize the total data you've got to decide, but you also don't want to bias towards one data point more then another. Their conclusion was that the benefit of crowds is that it adds a lot of data points, by combining everyone's data, but it skews some way too much, because if 70% of the people are making the same decision from the same data point, it's not weighing in other data points. It was also arguing that most people would decide the same thing given the same data points.