First of all, some simple statistics.
26490 - questions asked 345 - rated negatively 1.3% (negative of the total)
94800 - questions asked 150 - rated negatively 0.16% (negative of the total)
I counted pages, this is why the number of questions is a multiple of 15 and 50
In absolute numbers this doesn't sound like much, but another way of seeing this is to see a tenfold increase compared to another similar site.
I participate in several other SE sites, and until today I only had this vague feeling that TeX is kinda welcoming and Codereview is not so much. I had my hunch as to what may be causing it. I don't visit CR frequently, but the general impression that I had was that, quite often the answers may be even worse than the question they were improving. I also encountered several instances of moderators making absolutely bizarre and irrelevant changes to questions, especially posted by novice users. This becomes particularly strange when someone with no expertise in a language of the subject edits the question to best match his (incorrect) intuitive understanding of what the question is asking.
All in all, my impression was that of unscrupulous and harmful attitude resulting in "Lord of the Flies" effect (aka Bystander effect). The typical scenario when the individual is harmed goes like this:
- The "victim" posts a question which to a non-expert looks as deserving bad karma.
- This opens the gate for other non-experts who hesitated to vote before.
- "Victim" with high probability never posts again.
To prove my point by ways of example, I found an interesting problem, which seems to fit my scenario very well. It sounds almost trivial, if you don't pay close attention to the problem statement. And the solution isn't really that hard to come by for an expert. And I found it reading a CS coursebook, which is almost mandatory reading for any CS graduate.
The aftermath of my test: the question is closed as "off topic" / "broken code", while it's perfectly on topic, and the code actually works! It received three good answers, one of the answers is really in-depth, and with good measure of understanding. And it was downoted at least twice. So, someone who is, apparently, an expert in the matter took his or her time to lay out an extended answer with diagrams (note that this person has very few karma points), and people who voted on closing the question are proudly displaying their achievements in SE careers profiles! :)
First of all, I believe that the reason for this kind of behavior is psychological and has very little to do with the expertise level. So, the solution needs to be psychological too. As a limited time experiment, I would suggest that CR would change the "rewards" for some activities. Most importantly, downvoting the question must deduce twice as many points to the first person downvoting the question compared to the person who asked the question. Next person downvoting the question will restore some points to the first person downvoting and will only sacrifice, say, 1.5 times as much as the original poster. The whole idea is to balance out by the time there are at least 5 downvotes. The person who "casts the first stone", however, will need to think very carefully (if they care about karma).
I didn't do the math, but if you are interested, I can do it.