There are plenty of resources that "define" good questions, e.g. every site's FAQ, here and here. The heuristic seems to be that good questions correlate with good answers.

That does not seem to be true in all cases. For instance, a (scientifically) interesting question -- that is one that has a good answer which might be unknown -- can be "hijacked" just because (for instance) it is near topics that invite for battles of opinion, causing lots of "bad" answers to be given. This has arguably happened to varying degrees in several questions:

Two questions naturally arise:

  • What constitutes a good answer?
  • What to do with bad (=not good) answers?

For reference, check out these guidelines on Skeptics Stack Exchange.

  • $\begingroup$ I think you have to define good for this question. You define good for the questions at SE, but you do not define good for the answers at CS:SE. This question is about good answers at CS:SE. As I currently understand CS:SE this is to help others learn about CS. Before an answer can be good it has to be correct. The problem is that at SE in general it is upvotes that decide, even when the upvote is for an incorrect answer. For me the real question here is how this site deals with wrong answers regardless of votes. Who is qualified to decide? Who is given power to decide? $\endgroup$
    – Guy Coder
    Apr 1 '12 at 12:44
  • $\begingroup$ If this site continues to allow incorrect answers, the quality of the site will decline, and then the usefulness of the site, and eventually the site will become useless. A programmer can tell if a code answer is incorrect because the code will fail if properly tested. How is a student able to tell if an answer here is incorrect? If they convert it to code and then try it they may find something. But if they are teaching themselves and take the answer as a truth and build upon that they have built a house of cards. $\endgroup$
    – Guy Coder
    Apr 1 '12 at 12:49
  • $\begingroup$ @GuyCoder: Downvoting (or at least few upvotes) should serve in this capacity; Quality control is a community effort on SE. There is no single authority here! If mods were to delete all answers they thought were wrong, they would make mistakes and the site would fail. It is true that checking a proof is harder than checking code (in the sense that fewer can do it), but we will have to live with that and do our best: experts have to check what others write, and non-experts have to keep asking until they understand, thusly exposing unclear and wrong posts. $\endgroup$
    – Raphael Mod
    Apr 1 '12 at 12:56

My first thought on this matter was: this is a computer science site, so questions and answers should to some extent adhere to the scientific method. For answers, that means they should make only falsifiable claims and back them with solid evidence, which can be a (good) reference, (significant) empirical data or a (clear) formal proof (idea).

Any answer that does not provide any of the three is likely to be not useful.

Notable exceptions are answers to questions that explicitly ask for recommendations or experience.

To answer the second question, the software allows not many shades of grey. If salvagable, the answer should be commented with inquiries for evidence. Otherwise and if evidence is not provided in a timely manner, the answer can be downvoted and/or flagged. Ultimately, if enough people agree that the answer is no good, all that can be done is to delete it.
And, of course, we should try to post better answers.


The question Can a dynamic language like Ruby/Python reach C/C++ like performance? makes me uneasy.

On the one hand, it is a reasonable question. It might need a bit of editing, but I don't see a reason to close it.

On the other hand, it's already attracted several poor, but highly upvoted answers. These answers aren't necessarily wrong or uninteresting, but they are opinion pieces or anecdotes. They are not what I expect to see on a site about science: statements backed by mathematical proofs or experimental results (and remember, “data” is not any old plural of “anecdote”).

This is clearly a “bikeshed question”: everybody has an opinion on the color of the bikeshed. Yet there is a science question in there. I've gone ahead and edited the question look less of a call for opinions and more of a science. I don't know what to make of the answers.

The other two questions you cite (Why has research on genetic algorithms slowed?; Why are there so many programming languages?) feel a bit different: they are not really questions about computer science, but questions about the (recent) history of computer science. I find the answers less inappropriate.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .