Should duplicate questions be deleted?

Question

Lately, I came across several suggestions for deletion of questions simply because they are duplicates. In my opinion:

• If the question is somehow a bad question, then it ought to be deleted, and so should the original one.
• Otherwise, there is no reason for deleting it. Being marked as a duplicate is enough.

I would like to know whether someone thinks otherwise and, if so, why.

Answer 1

If a duplicate will improve search then it should be left. There are questions on the site whose title bears little relation to the general form of the question it answers, or to a number of different ways of asking, but nevertheless constitute a canonical answer.

As a result, it is next to impossible for users to find that answer and users ask it again, often wording their title differently.

The duplicate question functionality exists for precisely the reason of signposting users who may have searched anticipating a title in the form of the duplicate, to the canonical answer.

Answer 2

No. If the original question is a bad question on its own merits, then of course it should be deleted, whether that's recognized today or long ago. Then technically, the duplicate no longer has an original to be a duplicate of.

But if the original is a good question, let's give the duplicate asker the benefit of the doubt and assume that they tried searching but for whatever reason could not find the original.

In my opinion, only a word-for-word duplicate should be deleted (e.g., if both the original asker and the duplicate asker are quoting the same textbook exercise).

The textual differences between the original and the duplicate might be enough to encourage a previously unthought of approach to the problem. At the very least, the duplicate brings attention to an old question that might languish forgotten.

Answer 3

As a rule I am in favor of deleting duplicates. Admittedly this is a relatively recent change of heart. The reasons for my change of opinion are the following.

• Nowadays we have several human answering machines who simply refuse to search before posting their answers - often their answers are also essential duplicates of the old posts.
• This practice makes the site more difficult to navigate and therefore less useful for future visitors.
• This is also related to general quality control problems plaguing the site. There is a significant overlap between the set of users who answer low quality questions and the set of users who don't check for dupes.
• Therefore I want to disincentivize the practice of dupe-answering. Merely closing the question as a dupe won't erase the wrongfully gained rep (=the incentive to increase site entropy), but deletion does work. In other words, only deletion is a sufficient deterrent encouraging A) the askers, and B) the answerers to search first and only then post.
• Observe that deleting the new version of the question does not stop anyone so inclined from adding a new answer to the older version. Whenever an answerer can add a new angle they can do so easily - deletion does not stop that! They can post their answer to the older version of the question! That action will even bump the old thread allowing voters to see the new addition and evaluate it! I don't see any reason not to post the answer to the older thread - unless the answerer is specifically after that green checkmark from the asker of the new version.
• As an alternative to deletion we have the mechanism of merging two variants of the same question. Unfortunately this is not always optimal. After all, two variants of the same question may use slightly different notation, and naturally the answerers often copy the askers notation, raising post-merger compatibility issues.
• There are also those cases when the new version of the question is decidedly better than the older version. May be the newer version is slightly more general, or the fresh asker added more context? In those cases we sometimes close the older version as a dupe of the new one.

---

Judging from the other answer there may be a misunderstanding. Also, my thinking is colored by those questions that are "obvious" dupes. I put the obvious in scare quotes, because its meaning varies from person to person. Largely addressing users who have spent enough time on the site to earn, say 20k of rep, or who have held an academic position (implying that they have taught a section or three of calculus and/or intro to abstract algebra). Such people cannot possibly think that certain kinds of questions would NOT have been handled on our site ages ago. And I'm not talking about questions themed like "irrationality of $\sqrt2$", "$0.999\ldots$". At the other end, there are truly accidental dupes, I'm not talking about those. What I have in mind are questions like:

• Find the indefinite integral $\int x\sin x\,dx$.
• Show that $2^n>n^3$ for $n\ge10$.
• Stars and bars, edition #117.
• $x^p-x-a$ is irreducible over $\Bbb{F}_p$.
• $\cdots$ (you can find such examples from all the undergrad courses and core 1st year grad courses)

That is, questions that appear so frequently in relevant courses. Yes, they are fine exercises to a noob. They demonstrate certain techniques to a student just fine. But, anyone who's seen them KNOWS that they are soooo standard that there's snowball's chance in hell that we haven't already covered them on this site. I'm talking about questions that smell like dupes, talk like dupes, and walk like dupes. MAY BE THEY ARE DUPES! LET's CHECK BEFORE ANSWERING!

I don't have any qualms about somebody who wants to add a new proof to an old problem using a different idea, like here. What I'm maintaing is that people who

• should (from their experience either in real life or on this site) suspect that this question has been handled here earlier, and
• yet post a copy/paste textbook answer without searching

are guilty of either sloth or greed (for rep points).

I am not saying that all the dupes should be deleted, but I decidedly oppose the logic that because a dupe may be ok in such and such a situation, we should never delete dupes.

I guess we are back to the old rule of

Judge these matters case-by-case.