It matters because people respond to incentives. If what you want to see encouraged is good answers to novel questions, you need a system that rewards that over copy-pasting from documentation.

What do you think happens when the two are rewarded equally, but the latter can be done much more frequently per unit of time?

I don't think what the site developers (or most users, for that matter) want to see is good answers to novel questions. I think they want to is good answers to all questions.

If I can find the answer I am after more quickly by searching and finding a hit on SO than I would by grepping documentation, how is that not providing value?

Now balance it against the more important questions that are unanswered or even unasked because SO becomes a place where there's no point to that. It's not as simple as that.

I get your argument, but it sounds like a weak version of the claim that there is a strict ordering of priorities in the world and that only the top priority should ever be worked on.

Heck, even if the site was only for "novel" questions, according to your logic, only the most important unanswered question should be visible, until it is answered. Otherwise people might be incentivized to answer the second most important question instead.

The attention of users is a limited resource. Demands on it must be balanced against one another.

Then do you agree that only the single most important unanswered question should be visible at any given time?

I'm not sure I follow you. How does erring on the side of leaving a question open leave more questions unanswered or unasked?

When the reward system is such that novel questions do not receive answers, users will eventually stop asking those questions.

I still don't see what one has to do with the other(s). Besides, go browse through SO's unanswered questions; this is the state of things today! Simple questions, the most common variety, get answers in minutes. The complex and/or specialized questions which cannot easily be answered by a large portion of the userbase tend to languish.

Indeed. The result is that SO is a lot less useful than it used to be. I have repeatedly found myself chasing an obscure error message only to dead-end on SO and be forced to backtrack.

While I've experienced something similar too, I think it may have more to do with the fact that tech has diversified in the absolute sense. An obsecure question about your processor was easily answered in the past because everyone had tinkered with it. Today a question about a bolted-on feature in <insert tool of the year> might not receive an answer because so little experts on SO use it.(?) I've got no clue if this explains it, but it makes more sense to me than the idea that SO's attention/incentive span is so short that users are only spending time on 5-minutes-typing questions.

I've had that happen repeatedly as well, both now and "back in the day". I have no idea if a higher or lower proportion of the tougher questions are being answered now than they "used to be", but my sense is that it is about the same. The tougher questions are just tougher to answer; that hasn't changed through time. What I do know for sure is that a much higher proportion of easy questions can be answered by searching google and following one of the first few SO links, which is valuable.

Some people find difficult problems more rewarding to solve. Given that internet points are meaningless, someone interested in learning a language better could spend their time regurgitating documentation (and hopefully learning along the way). Eventually, one of those regurgitated responses gets picked up by Google and becomes the top answer, and nobody even needs to go to the SE forums to ask.

I don't think they're mutually exclusive - there are multiple reward structures for multiple types of players. Both should be encouraged.

Right now, there's one reward structure. People who value and respond to that reward structure act rationally within it. This means that regurgitation gets rewarded over good answers to novel questions.

Fair point on the question of why it matters. On the question of what should be encouraged, I think "good answers to novel questions" is only one of many possibilities. In my opinion, what should be encouraged is simply "people quickly and easily finding good answers to their questions", and I think copy-pasting (or linking to) the relevant section from documentation serves that goal just fine, and that voting is a good system for indicating what answers are useful, regardless of the form they take.

My key point is that not all questions are created equal.

Further, there's the question of how much effort the questioner should be expected to put in. In a great many technical question fora, the experience is that many noobs make no attempt to search out answers.

Isn't the incentive to answer those kinds of questions mostly intrinsic? I mean, the interesting questions and answers are more likely to be given exactly because they are so. The extrinsic reward is then complementarily useful for the opposite questions: the ones that are not very interesting but important nonetheless, such as the case of a reference to a cryptic documentation.

Extrinsic motivations displace intrinsic ones. Once you have the magical internet points, the intrinsic motivations fade.

Maybe the difference of opinion largely comes down to whether you think the second sentence of your comment "sounds right". It seems false to me, but it must seem true to you, since you wrote it. I would be very interested in seeing an analysis of data relating a measure of how many "non-novel" answers are being given to how many difficult questions are being answered.

It seems like the solution is to instead of close non-novel questions, just make them not award points.

I think this would be an excellent solution. Combine it with some semantic similarity analysis, and the vast majority of newbie questions could be replaced with "Here's a link to a pre-written answer that almost certainly addresses what you want".

They already do that analysis and show similar questions while you're asking a question. You can say "well they should do it better" (and I'm sure they try), but the whole beauty of the Q&A model is combining the strengths of computers and humans. Computers are better at organizing and storing tons of information, but humans are much better at figuring out what other humans are really asking.

I could see an argument that the people more equipped to answer more technical questions may leave if they have to wade through a bunch of questions that could be answered in the docs. If those sorts of answers "take over" then you'll have to wade through them to find more novel questions to post answers to.