I was talking to some friends who are also MO patrons, and we noticed that shared the following phenomenon. Our decision whether to upvote a question or answer seemed to be based on different criterion, based on how many upvotes the question has received. There seemed to be three basic domains:

A) If the question or answer had received no upvotes, then we didn't want to upvote it unless we felt comfortable `vouching' for it; that is, either stating that it was a reasonable question or that the answer was correct. This was particularly strong when it was an answer either had details that were too involved to check, or referenced external articles we lacked the time to check.

B) If the question or answer had a few upvotes, we voted for it if it seemed like it could be correct and was well-written. This kind of upvote seems motivated by logic of the form "well, if thats correct, it should be worth 5 or so votes", drawing upon some sort of internal metric of how many votes a given question or comment SHOULD have.

C) If the question or answer already had many upvotes (say, 10 or more), we based our decision entirely on how much we liked it: whether it was cool, clever, exciting, had a pretty picture, made us laugh, etc. The idea here is presumably that the suitability of the question/answer had already been established, and that any other upvotes were a form of kudos or applause.

So, I guess my question is two-fold. First, whether this kind of phenomenon is as generic as it appeared to us; and second, whether this kind of thing should be encouraged or discouraged. Often I'll see questions that seem to be good questions in fields I am not an expert in, but I am discouraged from upvoting them because they have a low number of votes and I don't want to vouch for their suitability. Is this `correct'? If so, is it worth mentioning in the FAQ?

I think this is sort of inevitable. It is difficult to check a long and involved argument so few people will do it, even if those are some of the best answers on MO and the ones we should be working hardest to encourage. So instead we rely on various heuristics.

The best way to combat this, in my opinion, is for people to endorse answers they think are really good in the comments.

I agree with Qiaochu that to an extent this phenomenon is inevitable (and I also think that it's not especially pernicious.)

But there are some things one can do to combat it.

If I see a very specific, technical question, then I tend to vote it up if it is coherently worded and nontrivial, especially if it has few upvotes. An upvote here doesn't necessarily mean that I myself am interested in the question, but rather that it has the form of questions that I would like to see more of on MO. We've gotten pretty good at discouraging questions like "What kind of mouthwash do mathematicians use?" We could be a little better at encouraging questions like "Under what conditions does this spectral sequence degenerate?"

If I have taken the time to read through a somewhat complicated (correct) answer, then I may leave a comment indicating that I like it (and, implicitly, that I am vouching for it). If someone else that I respect does the same thing, then I may sometimes upvote an answer that sounds good even if I haven't checked it. This goes somewhat against my ingrained instincts, but an anonymous upvote is certainly not a certificate of correctness, and this site gets enough attention so that if there is an error in an answer it will probably come out sooner rather than later.

For what it's worth, I think the problem is much more severe on math.SE. In general that site seems seriously undervoted. My guess is that many of the "answerers" who are active on that site are not in the habit of voting up a question that are well-formed and demonstrate thought and effort on the part of the questioner, whereas most of the "askers" on the site vote up questions that they can understand. As a result, if you ask a question about how to evaluate a particular indeterminate form, you are likely to receive more upvotes than if you ask a question about class field theory which would give most professional number theorists some pause.

Here's a math.SE example for you: my (as of yet, not up-voted) answer to: http://math.stackexchange.com/questions/11066

It took me about half a day to work out the mathematics details, write and debug the code and double-check the answer wherever possible. Although, I'm not too upset about this, it was quite an interesting exercise.

Hey, I didn't realise! Thanks for pointing that out. (:

Mr Hardy, please do not make unwarranted personal
attacks here.

Alex, I would distinguish between hot heads and thin skins. There are some of each on MO, and it can be unpleasant when they meet.

@Pete, I agree. Indeed, I see rather less of it here than in some other corners of the internet, and am pleased that for the most part we keep a civilized tone here and do not descend into personal attacks.