9
$\begingroup$

I'm not questioning about the length or quality of present answers. I want to know, if a question has one answer or more, does the number discourage someone who has never read the present answers from answering ? Someone simplement just looks at the number of answers, which is > 1?

I know if answers are bad or short, someone might invoke better answers. But not questioning about this. Do more answers discourage people?

I don't know how to research this myself. Thanks to all.

|
$\endgroup$
  • 1
    $\begingroup$ This is a curious thing to think about - it's a very easy thing to conclude as reality based on expectations of the social convention, but I'm wondering how the situations actually play out in practice - is there actually a significant contingent of material that ceases to gain new responses because of existing answers? Definitely something to poke numbers around with to see. $\endgroup$ – Grace Note Mar 6 '14 at 17:14
  • 11
    $\begingroup$ Mainly, I look at questions that I find interesting. If the question has several answers, I look through them to see if what I intended to say is already there; if not, I post a new answer. Of course, sometimes the question may not in itself be interesting, but there is a trick to address it that I like, and I see that the trick has not been made explicit yet. Sometimes, it turns out that the answers suggest a (long) comment that deserves some space, and that is what I post, explicitly not a direct answer to the question but a supplement to what is already there. $\endgroup$ – Andrés E. Caicedo Mar 6 '14 at 17:21
  • 11
    $\begingroup$ My anecdotal experience is that if there is a good deal of plausible text on the page, I feel less likely to answer. That puts the onus on me to read everything to make sure I'm not repeating stuff, and usually I'd prefer not to read a lot of stuff if there are other things left unanswered. $\endgroup$ – rschwieb Mar 6 '14 at 17:33
  • $\begingroup$ My first port of call on the site is the unanswered list and then those using my favourite tags. I feel I'm more likely to open a question if it has fewer answers, though it also depends on how difficult the problem is. If a question involves a lot of work then it's probably worth giving more than 1 view point to the OP, instead of such questions which 'follow from definitions' and which get 4 answers which are all morally identical. In those cases, unless I know of a clever trick which trumps the other approaches, I don't bother. $\endgroup$ – Dan Rust Mar 7 '14 at 11:35
  • $\begingroup$ I think that being able to truly answer your question would require an experiment: show some users the answer(s) that already exist, and hide them from others. See who's more likely to respond. $\endgroup$ – Tim S. Mar 13 '14 at 16:31
9
$\begingroup$

After the first answer, each subsequent answer is less likely than the previous one:

Answer count distribution.

The above graph shows the distribution of open questions by number of answers. Not surprisingly, people have a strong tendency to answer unanswered questions. But once the question has been answered the odds of one more answer decrease somewhat. This looks roughly like a Poisson distribution with a mean of 1.5 (the answer rate on Mathematics):

Poisson distribution mean = 1.5

Outside of the relative lack of questions with zero answers, answers are more or less distributed as we might expect from a Poisson process. If so, the odds of a new answer are independent of the number of existing answers. As a first approximation, I say that getting one answer does not artificially discourage further answers.


Another way to think of it is to consider what we'd like to avoid. It would be annoying if a person with great insight into a questions decided the existing answer was "good enough" and the site were deprived of a better solution. But remember that they have other options such as editing or leaving a comment (which sometimes prompts an edit). So even if you never get another answer posted, that doesn't mean you'll never get a better answer.

We know there is a first answer bias. On Math, it's about 0.6 votes on average. Some of that might come from the bandwagon effect. But it could also be that the initial answer legitimately tends to be better than the rest. This site boasts excellent quality, so it stands to reason that a quick answer that's upvoted by the community sets a high standard. The next answer might very well fill in some details or provide an alternate method of looking at the question. Those are useful things and we hope people offer them as answers too, but not all voters will find them as worthwhile as the initial answer.

|
$\endgroup$
  • $\begingroup$ I think that the first answer bias here is much higher than you claim. How did you calculate $\,0.6\,?\ $ What do you mean by "this site boasts excellent quality"? In the topics that I browse, frequently the first answer is not the best. $\endgroup$ – Bill Dubuque Mar 13 '14 at 18:36
  • 1
    $\begingroup$ @Bill Dubuque: Feel free to play with the query. If you change from looking at just two answers to q.AnswerCount >= 2 the bias jumps to 1.12. You probably are thinking of questions that are a) very popular and b) have a lot of answers. The bias between first and last ramps up quickly as you add more answers. $\endgroup$ – Jon Ericson Mar 13 '14 at 18:51
  • $\begingroup$ Could you please provide a link to the specific query that you used to come up with the number $\,0.6,\,$ so that I can understand how it was derived. Thanks for the effort in gathering these stats. Maybe we can refine them further to better correlate them with user experience. $\endgroup$ – Bill Dubuque Mar 13 '14 at 18:56
  • $\begingroup$ @Bill Dubuque: Here you go: data.stackexchange.com/math/query/173863/… (It says 0.59, but I rounded it off one more decimal place.) $\endgroup$ – Jon Ericson Mar 13 '14 at 19:00
  • $\begingroup$ I think that accurate measurement of first/early answer bias requires doing a much more sophisticated analysis. Now that I see the simple query I'm not at all surprised that it doesn't agree with my experience. $\endgroup$ – Bill Dubuque Mar 13 '14 at 19:11
  • $\begingroup$ @Bill: I was expecting a higher number, too. But this 0.6 is an average. My "bad" experiences with a much larger first answer bias have been cases, where a "better" answer comes much later. By a margin of hours, may be even days. It is one of my pet hard-to-stomach things. But those cases aren't typical. In many (most?) of the cases where there are many answers, the competing answers come within minutes of each other, and get nearly equal exposure. Then the bias is not unexpectedly lower. After a few hours many of the members initially interested in the problem have moved elsewhere. $\endgroup$ – Jyrki Lahtonen Mar 13 '14 at 20:00
  • $\begingroup$ @Jyrki Yes, that is but one of many factors that one would need to account for to accurately measure such bias. The linked query is far too simple to be of much use. $\endgroup$ – Bill Dubuque Mar 13 '14 at 20:11
  • $\begingroup$ I know a good number of users that search questions in their tags with answers:0 (including me), so this (unfortunately) makes sense. $\endgroup$ – user98602 Dec 14 '15 at 13:38
7
$\begingroup$

In the elder days the OP would often (be expected to) "summarize replies", which was an important and useful act of consolidating posted knowledge.

An update of that idea for multiply answered questions on SE-style forums can be to add some meta- or methodological analysis of what makes those answers work or is the "inner core" of the problem. That is an activity that could hold experts' interest even when the question is routine, since the jaded old hands would then be examining the structure of the entire solution space of the question and not the more familiar (read: boring) activity of finding one of its points.

This is, in my opinion, not done often enough on the site, and would add much value. Personally I find it more interesting to write some methodological answer than to post a first, second or third direct answer, and it is a way of creating novelty and depth around material that can feel pedestrian.

|
$\endgroup$
  • $\begingroup$ First paragraph is an awesome idea! I haven't heard of that practice before now, but I think I'll start using it :) $\endgroup$ – Eric Stucky Mar 8 '14 at 7:11

You must log in to answer this question.

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