How do we determine reliability?

I have found the discussion surrounding Wikipedia to be immensely interesting. As I have stated before, I like Wikipedia for school use, as well as acknowledge that it has limitations and potential problems. It is, however, quite difficult to make (and defend) the mental leap required to accept that Wikipedia can be reliable despite its disregard for conventional ideas of authority. I feel the need to couch my statements in this regard (as in my reply to Jody's posting on July 5), and the statements of others (such as the representatives from Britannica in the Bernstein article on the extreme end) seem to support this difficulty.

Personally, I think it's incredibly compelling to think that the way we conventionally define authority is not the only way attribute reliability to information. In this age of information (I cringe to use a buzz word, but it's a very telling description of our time) where data points and users are so numerous, we can count more heavily on probability. The iterations are so many that we have a statistically significant sample. As such, if we say that most people will do conscientious work on Wikipedia, then we can rely on most articles being conscientiously written. If many people are assessing, correcting, and evaluation articles, then we can rely on most errors being caught. With so many articles, maybe we can count on these probabilities enough to define a new kind of reliability. Maybe, we can depend on "surely someone will notice". I'm not saying that this is necessarily true, but we are redefining so many ways of thinking, and wouldn't it be fascinating? Wouldn't it be fascinating if this system that depends on millions iterations could be as reliable as "authority" as defined by Britannica?