12
$\begingroup$

I have noticed several times that some questions are flagged by some as "duplicate" questions when the questions appear very different but have similar answers. The questions are asking about very similar problems, but in such different language or in such a different context that few (if any) would think to search for one answer in the other question.

I am not sure what to do about this, in terms of making the site as useful as possible.

$\endgroup$
  • 1
    $\begingroup$ Jeff Atwood wrote a thoughtful post on this subject a while ago, where he's in at least partial agreement with whuber. (as per my reading anyways) $\endgroup$ – AkselA May 24 '17 at 20:00
15
$\begingroup$

A posted question is a portal to information: the words and phrases it contains; the tags applied to it; the contents of any answers; and even its net votes and status (CW, etc.) are searchable. Remember, a duplicate question doesn't disappear: it remains visible and serves as a way for people to find the information they seek.

The network of pointers to duplicates enhances this "syntactic" web. By connecting apparently different questions to common answers we provide indications to the site, as well as to future users, about sometimes deep and surprising connections. This is the beginning of a semantic web.

Only people--and experienced people, working creatively towards a common goal--can create these semantics. These connections are things we should value, work to create, and curate carefully.

These considerations suggest to me that we should be spending more time identifying duplicates and building up this valuable information. Creating a single pointer to a thread with a good existing answer is far better than providing a new answer.

It is unfortunate that the SE system provides no rewards for this behavior (there are no badges or points for identifying duplicates, alas). I therefore want to thank Peter Flom for raising this issue and especially to thank all those unrecognized members of our community who have been cleaning up the tags, identifying duplicates, editing posts, and doing the hard work of building the connections that make this site ever more valuable.

$\endgroup$
  • 4
    $\begingroup$ This is the answer I would have wanted to write, if I could do so ... $\endgroup$ – kjetil b halvorsen May 17 '17 at 15:47
  • 1
    $\begingroup$ I agree that this is vital work. I'm not completely clear on how it works. Suppose that a new visitor has a question about ANOVA. He searches but finds nothing. However, there is another question, about regression, that has the same answer. The visitor, like so many, doesn't know that ANOVA and regression are related. How will the visitor find it, if the questions are merged? Should there be some pointer added to the merged question? $\endgroup$ – Peter Flom May 18 '17 at 11:13
  • 1
    $\begingroup$ I'm not completely following you, Peter, perhaps because "merging" differs from closing questions as a duplicate. Merging assembles all answers and comments into one thread, leaving only one of the questions as a stub of the thread it began. In your example, the visitor might indeed fail to find the answer; she will post her question; and a member of this community will identify the duplicate. Her question will be closed but its text will remain so that all future visitors who frame their question in a similar manner can find it and be led to the answer directly. $\endgroup$ – whuber May 18 '17 at 13:04
  • 1
    $\begingroup$ You have to handle that hypothetical duplication correctly, though. If all you do is close the ANOVA question as a duplicate to the regression one, you leave very confused (and upset) question askers, who think you're being jerks for closing an ANOVA question with a - to them - completely unrelated regression answer. -- A "true" answer to the ANOVA questions needs to explain that, in this instance, ANOVA and regression are related enough that the regression answer also applies to ANOVA. If the linked answer doesn't address the similarities, it doesn't truly answer the question. $\endgroup$ – R.M. May 29 '17 at 14:50
  • 1
    $\begingroup$ @R.M. Thank you for that cautionary advice, which is wise and helpful. On some occasions I find a duplicate very clearly answers the closed question but that it uses a different language or notation. In such cases a comment to the closed question can be helpful in pointing out the connection. If it takes more than a comment to do that, or if the comment requires expertise to understand, then probably the question should not be closed as a duplicate. $\endgroup$ – whuber May 29 '17 at 15:00
11
$\begingroup$

Where the newer post has a more canonical question (one people would search for), and the answers on the older question will work well as answers for new one, it may sometimes be better to merge into the newer post, which will move the answers across.

In some cases the best strategy may be to post a new more canonical version of the question (which has the advantage that you can tailor the question a little toward the more specific questions that already exist) and merge both existing questions into the new one.

$\endgroup$

You must log in to answer this question.

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