For example we all pretty much now know that Bush is a pretty bad leader. But how could we tell a computer this?
I could try to model a set of objects that judge the quality of a leader but I doubt I would get far.
Or I could just use 2 rules of judging the quality of collective judgments. These rules state that people get better at making judgments as they have more experience with a subject and larger diverse populations are better at forming judgments than narrow groups.
In the case of Bush we see 2 things: the level of approval drops as people live under him longer and longer and that internationally he is unpopular. These two rules would allow a computer to be more confident of the quality judgment.
In the same way IMDB votes could be weighted by the age of the movie and the global nature of the voting. I a movie is a few years old and has a large spread between US and Non-US vote the score granted by users can be questioned, but if it is 20 yeas old and the US and no-US vote are equal a computer can see that the movie is not only popular but probably very good.