Why is the close-as-duplicate policy so aggressive, and should we change it?

Question

This has been on my mind for some time, and I'd actually previously written up something to post here, but I dropped it until the discussion on the following question:

What is Tolkien trying to say in this letter?

Basically, it seems to be SF&F SE policy to close questions as a duplicate if they are anything alike, regardless of whether or not they are a duplicate, or if an answer on another question even touches on the answer to the new question, even if the actual questions are totally unrelated. @JasonBaker mentioned that this is official policy, though I was not able to find anything specific on Meta, then or now.

To be clear, I have nothing against marking questions that are actually duplicates as duplicates. But what I see over and over is marking any question, even if only tangentially related, as a duplicate of another if one answer even slightly maybe touches in passing information that might be an answer to a new question.

I don't see any benefits to this policy. And actually, to SO blog covers some of this topic pretty well.

http://blog.stackoverflow.com/2010/11/dr-strangedupe-or-how-i-learned-to-stop-worrying-and-love-duplication/

...but this issue is a little different. We're expecting users who ask a question parse out potentially unrelated questions and scroll through answers possibly consisting of mostly unrelated content that may somewhere touch on the answer to their question, rather than receiving an answer that is tailored to, and directly addresses their question. Why? The upsides seem legion: searching for answers to questions is easier, users get direct answers to their distinct questions that touch on any subtleties rightfully missing from answers to other questions, and we can still mark questions as related if they contain potentially valuable information. I've found this to be especially egregious when those voting to close aren't intimately familiar with the source material.

What are the downsides to being less aggressive about marking questions as duplicates? Why don't we move forward to at least stop marking distinct questions as duplicate? (As opposed to the current policy where distinct questions with similar answers even if those answers are not the accepted answer can be marked duplicate.)

Answer 1

Mooz wrote in a comment,

If someone asked "Who was Palpatine's last apprentice?", and someone answers "It was Anakin Skywalker" and specifically mentions that "Anakin is Luke's father, by the way", then that is what makes the question of "Who was Luke's father" a dupe of "Who was Palpatine's last apprentice?"

This is so completely wrong it hurts. If this is our policy, we might as well refer everyone to the appropriate Wiki and call it a day, because that would be just about as useful.

Duplicates are duplicates if and only if the QUESTIONS are the same. (This is how every other SE site does it, by the way.) Otherwise, we're asking users to wade through a bunch of totally irrelevant information to find the one tidbit that's useful to them. Preventing that sort of exercise in frustration was the reason StackExchange was invented, folks.

There's also the issue of reliability: how do you know that that "oh by the way" bit of information in an otherwise-unrelated answer is actually correct? The voting on the answer is likely to reflect how well it answers the question it was posted to; if it contains some not-quite-correct extraneous information, that may be worth a comment, but it may still get upvotes. If the comments then get too long and are taken to chat, the information about what is incorrect about the answer will be almost impossible to find. (In fact, for a low-rep new user who doesn't yet have chat privileges, it will be totally impossible to find.)

Answer 2

First, I want to address this issue.

We're expecting users who ask a question parse out potentially unrelated questions and scroll through answers possibly consisting of mostly unrelated content that may somewhere touch on the answer to their question, rather than receiving an answer that is tailored to, and directly addresses their question. Why? The upsides seem legion: searching for answers to questions is easier, users get direct answers to their distinct questions that touch on any subtleties rightfully missing from answers to other questions, and we can still mark questions as related if they contain potentially valuable information. I've found this to be especially egregious when those voting to close aren't intimately familiar with the source material.

I know that to some users, and especially new users, a question being closed can come across as a discipline or punishment - that we're trying to say, "Hey, pay more attention!" or "Don't do that, noob!" But, we're not (unless it becomes a repeated egregious issue). Closing a question as a duplicate isn't a personal attack, nor are duplicates inherently bad.

In fact, duplicates can be useful - even when they're closed. Two "duplicate" questions with different wordings can help other users find the information should they choose to search for it. That's a good thing.

And if the community agrees that there is enough of a distinction between the questions, or the OP can explain why the original question/answers don't cover the information they're looking for, then the question can be re-opened. Closing isn't permanent.

So why are we so "strict" about closing duplicates?

Because a question should have a single repository of answers, and because often times, the relevant necessary discussions have already been had. How many questions do we really need asking how the infection started in The Walking Dead, or how many do we need about why time turners were never used again after Prisoner of Azkaban?

But those are the issues of exact duplicate questions. The lines get fuzzier when we start talking about defining duplicates based on answers. And this is something that has been generally controversial over the years.

I believe the most recent discussion on this was Why are questions considered duplicates although they ask different things?

The same reasoning really applies here. We've already posted the information being searched for. Answering it again and again in multiple places makes the whole thing confusing as we start finding contradicting answers and information across the site (look at questions about "The Rule of Two" for an example of this). Users looking for the information get confused when there are more than one "right" answers or sources of information.

Admittedly, this is less of an issue when it comes to answer-dupes, but this is how the community has decided so far. Personally, I see no particular reason to change the policy as of now, but am not vehemently opposed to doing so. The community has changed its mind about things in the past, and will likely do so again.

Some other pertinent reading:

• Should questions tagged as duplicates because of an answer be treated different?
• Is clarification needed on what makes a duplicate question?
• Are multiple variations of a question with subtle distinctions duplicate?
• When voting to close, MUST the “already answered” information come from the accepted answer?

Answer 3

What do I do?

Since I'm the one who more-or-less kicked off this discussion, I thought I'd weigh in on how I decide whether two questions are dupes.

Mandatory disclaimer: I vote according to my own conscience, based partly on how I've seen other high-rep users on this site vote, partly from my own understanding of this site's policies and practices, and partly on my own (evolving) understanding of what it means for a question to be a dupe. Despite being a pretty high-rep user, I'm still fairly new to the community (less than a year), so I might be wrong - I'm often wrong, in fact - but this is the measuring stick I use.

So let's say I have questions A and B. In my mind, B is a dupe of A if:

1. B is asking the same question as A. This is pretty uncontroversial; When does the Doctor actually tell River his name? and How does River Song know The Doctor's name? are pretty transparently asking the same question (there may be better examples of this, but this was a recent one that was easy to find).

2. B is asking a different question to A, but the answer to A is the answer to B. This is a little controversial, but I don't think it's terribly so. Let's look at some examples:

   • Did Bellatrix really love her husband? and Did Bellatrix Lestrange have any affection towards Voldemort? are slightly different questions. The first is about Bellatrix's relationship to Rodolphus, the second is about her relationship to Voldemort. However, these two questions have basically the same answer; alexwlchan and I use the same quote, but frame it in slightly different context.

     However, the framing context has no bearing on whether or not the answers are the same. The existence of my answer does not benefit the site. It also doesn't actively harm it, which is part of the reason I'm not freaking out trying to get it deleted¹

3. Any answer to B would also be an answer to A. This is probably the most controversial, since it doesn't rely on existing answers, and I suspect I'm in a minority for feeling this way. I know that I've done this before, but I can't think of any other examples off the top of my head.

   However, the reason I vote this way is because (and I know there's a meta discussion on this, but can I find it? Nooo.) the fact that A doesn't have useful answers doesn't affect whether or not B is a dupe. That's actually part of the reason I VTC'd What is Tolkien trying to say in this letter?. Wad, the OP on that question, pointed out that the answers to the dupe (Was Elrond, in Tolkien's opinion, more inherently powerful than Galadriel?) weren't satisfying because they were based on opinion rather than from authorial intent. This is completely true.

   However, the top two upvoted answers (and arguably the third as well) would both be perfectly valid answers to Wad's question. Would they be satisfying? No. But they'd be upvote-worthy, because they would be an earnest attempt to answer the question.

   Equally true, suppose we found a way to get an answer to Wad's question based on authorial intent. That answer would also be a correct answer to the dupe question. It would in fact be more correct than the other answers to that question, for the same reason it would be an answer to Wad's question: we actually find out what the author thought.

Why do I do it?

What I've found to be true about these answer-dupe questions is that, when you essentialize them, they are the same question; Izkata in comments relates this to the notion of a "parent" question, and you can apply this rule to every example I've given:

• The Doctor Who examples in my first point are both different ways of asking "How does River know the Doctor's name?"²

• The Bellatrix questions in my second point are just different ways of asking "Who does Bellatrix have feelings for?" (essentialization courtesy of Izkata, who got be thinking about this)

• The Tolkien questions in my third point, the questions that sparked this discussion, are both different ways of asking "Did Tolkien think that Elrond was more powerful than Galadriel?"

What we're doing by closing these questions as answer-dupes is not fundamentally different from what other SE sites suggest doing: canonical questions (or answers). The difference in our approach is twofold:

1. We let every question potentially be a canonical question
2. We don't wait for a question to become a "problem" question

What are the benefits?

phantom42 already laid them out. Go read his answer.

What are the downsides?

This list is largely going to be based on comments I've seen other users making in these discussions, and my responses to them. But I'll try to throw in some originality too

• If question A and B are different, the part of A's answer (unrelated to question A) that addresses B may be incorrect. This is definitely a potential problem, but it's a problem related to all answers. No information in any answer, no matter how tangential or unrelated to the question its been posted under, should be immune from the community vetting process; incorrect information should be corrected.

  For example, consider my answer to Is MODOK planned to appear in the MCU? (I've linked to a specific revision to make my point better). I initially wrote this answer as:

  In the game, [MODOK is] an enhanced version of Aldrich Killian, the main antagonist from Iron Man 2 and founder and CEO of Advanced Idea Mechanics (A.I.M.).

  I wrote that line from memory, which was incorrect. If somebody else had asked "Who was the main antagonist of Iron Man 2?", and if that question was closed as a dupe of this MODOK question, and my answer was left as it was, that would be a problem. Without that hypothetical question, my answer being left as it was would still be a problem.

  Of course what happened was that I had to leave for a family engagement, and I was unable to correct the information. Someone commented (I don't know who it was; I only know the comment was made because I received a notification on the mobile app, which I part of, but the comment was deleted before I could address it) that this was incorrect, and I was unable to redress it. So Keen, a subject-matter expert on the Marvel Cinematic Universe, corrected it for me.

  That's an example of the system working properly. Another example of the system working properly is on https://scifi.stackexchange.com/a/92282/31051, and an answer which has since been deleted. The answer read:

  I would just add that while the movie makes Fili and Kili die at the Battle of the Five Armies, they in fact die in the Moria (as per JRR Tolkien), during the failed attempt from Balin to reclaim it from the Orc.

  The last pages of the diary of the colony of dwarves of Moria have been written by Kili, which has a writing elfish in nature (perhaps why he has been chosen by Peter Jackson to be the subject of the invented Love Story with Tauriel...).

  This information is obviously incorrect, and it so incorrect that the answer is beyond salvaging; it could not be edited without substantially deviating from the answer's intentions. The correct course of action is to downvote the answer, and eventually delete it from the site altogether, which is exactly what happened; the answer attracted 13 downvotes and no upvotes, and was deleted by its owner 14 hours after it was posted.

  Other users have commented in this discussion that sometimes incorrect information gets called out in comments, and then never corrected. This is absolutely a problem, but it's a problem with our policy on editing posts, not on our dupe policy. If this is a problem that we're currently grappling with (and it might be; I've not seen it myself but what does that matter?), then we should be dealing with it on all answers.

• Questions might be closed as dupes based on a throwaway one-liner in an answer to an unrelated question. My other comments in this discussion notwithstanding³, I actually agree with this one. Questions should not be closed-as-dupe based on insignificant, tangential parts of answers to other questions; and you'll notice that none of my points above involve this.

  Picking on myself again, consider my answer to How could Treebeard call for war without an Entmoot?. In that answer, I have a footnote which reads:

  [Quickbeam, an Ent is] so named because he once replied "Yes" before another Ent had finished asking the question.

  Imagine if someone asked "Why is Bregalad named Quickbeam?" That would be a terrible question because it's answered immediately following the character's introduction, but imagine. If that question were then closed as a dupe of How could Treebeard call for war without an Entmoot? based on my answer, I would have a problem with that; if I took that single line, in isolation, and made it an answer, I would be disappointed in myself. That line is not an answer.

  However strongly I may agree with our current policy of closing questions based on answers, I do agree with this point; we should apply the same scrutiny to those answers as we would to answers on the question we're closing.

• Closing based on answers means that the user who asked the closed question has to go hunting for the answer to their question. This is true, but I think it's more a limitation of how we're applying this policy, not of the policy itself. Obviously this situation would be improved if the "this question is a dupe" dialogue pointed to the particular answer, but that's not something that's going to happen. One thing I often try to do is leave a comment saying "This is a dupe of question X, and in particular look at user Y's answer", which may help the questioner but may not help future travellers (since comments are ephemeral and subject to deletion at any time).

  Ultimately this isn't a bad point, but I feel like it's directed at the wrong policy. We already have systems in place that encourage questioners to seek out and read related questions to find the answer, rather than asking a dupe question: the existence of the duplicate close function tends to act like this, and can be perceived by new users (and old ones) as a slap on the wrist; we downvote posts that show a lack of research effort, although that's thankfully limited to questions that are answered by a simple Google search; we have "similar questions" boxes that pop up when you're creating a new question, showing you possible examples of similar questions that have already been asked (these boxes are generally pretty awful, but that's a different issue).

  If the objection to the current policy is that it makes users look harder for the answer to their question, then I think we should be having a broader conversation about whether we should ever be closing questions as duplicates, because that seems like an unavoidable result of the system in general.

• The answer to the old question hasn't been properly vetted as an answer to the new question. This objection hinges primarily on the fact that closing a question only requires five votes from 3k+ users, but up- or downvoting an answer is open to the entire community (who have at least 15 rep for upvoting or 125 rep for downvoting).

  I'll leave aside the fact that many correct answers get fewer than 5 upvotes (there are currently 4100 accepted answers with scores of 5 or lower, accounting for 30% of all accepted answers, although that excludes correct answers that aren't accepted).

  As with the last objection I discussed, this is something that is true in a broader context. It only takes five votes (or one modhammer, or one gold tag badge for dupe votes) to close a question for any reason, and to argue that those five people may not be correct is an argument better levelled at the entire closing policy.

  It is absolutely possible for close voters to get it wrong, on any kind of close vote. That happens, because we're human and make mistakes, and because we have different ideas about what each close reason means. But there is recourse for users who disagree with this, of all rep levels:

  • Any user can comment on their own posts, and users with 50 or more rep can comment on any. It's somewhat regrettable that comments don't bump the question (although on balance it's probably better that they don't), because that does limit the exposure of these questions, but recent questions that are still near the top of the list are going to get viewed by plenty of users who can cast votes anyway

  • You only need 15 rep to flag a post for moderator intervention. The modhammer can open a closed question with only a single vote.

  • 250+ users (of which there are about 2000) nominate their own question for reopening, and 3k+ users (of which there are currently over 200) can cast reopen votes on any question. It may take five votes to close a question, but it also only takes five votes to reopen it

  • You only need 5 rep to participate in Meta at which point you can bring your case to that subset of the community (mostly long-time, experienced users) who participate here.

  And bear in mind as well that a new question is going to get seen by plenty of users who are able to flag, comment, or vote on it to get it reopened, and casting reopen votes sends it into the queue where it can potentially be seen by any 3k+ user (with a time delay so you don't get the same users who VTC'd it immediately voting to leave it closed).

  If after all of this the question remains closed, unless most 3k+ users just don't care about voting (which is a much larger problem), I submit to you that the expert community has vetted the duplicate answer.

• If the dupe question doesn't have satisfying answers, the asker of the closed question doesn't have many options for getting them. This is absolutely true, and is a side-effect of duplicate closing in general. On more than one occasion I've flagged questions as dupes of other questions that have no upvoted answers, because the questions are dupes.

  I agree that this is a problem, which I don't have a good solution to (the typical answer is "post a bounty", but you need 75 rep to do that, at which point the minimum bounty would basically wipe out your site privileges; it's not something that works for all users). I'd be open to discussing ways to solve this problem, but as it stands I feel (and the community both on SFF ans the broader SE network seem to agree) that the benefits of a duplication policy outweigh this problem

---

¹ Sidenote: Yes, my answer probably should be deleted. I jumped the gun, answering before I'd checked to see if it was a dupe, and only found the dupe after my answer had been accepted and could no longer be deleted. C'est la vie.

² Of course that's the actual title of one of the questions, which is convenient for us; but I maintain that it's not strictly necessary

³ I got a little upset, and said some things I didn't mean. I'm better now