I feel concerned about a thought that I had recently.

I am wondering if the current policy about duplicate questions may become problematic in the (very) long term for the health of Stack Exchange.

Let me explain: as far as I understand, questions are marked duplicate when they have already been asked on the website at some point in the past. Thus, I can't help but consider the possibility that, at some point in a far future, almost every single question would have been already answered on Stack Exchange and then marked as duplicate.

Then, what would be the difference between Wikipedia and Stack Exchange, except for completeness?

How would we be able to exchange our knowledge if 99% of questions are legitimately marked as duplicate?

Eventually, this website would just end up being a massive database of Q/A, with virtually no activity. On one hand, that would be great for readers, but on the other hand, users who love to challenge themselves, learn and share their expertise may lose interest as the activity slows down to zero.

What do we want for the future of Stack Exchange? What do you suggest? Am I taking this way too far?

EDIT: Using Chair's contribution in the comments, I produced a graph that shows the evolution over time of the percentage of questions marked as duplicates:

enter image description here

The data, obtained via this query covers the statistics of the website between august 2010 and november 2018. Between 2010 and 2015 the proportion of duplicates increases steeply, however something strange happens around 2016-2017 and then it is hard to tell if the curve is starting to increase again or stabilizing around a constant value of $7\%$.

It will be interesting to see how this curve behave in the future.

  • 23
    $\begingroup$ Essentially a duplicate of What is the ultimate purpose of physics.stackexchange?, What is the ultimate purpose of physics.stackexchange - revisited and ... $\endgroup$ – John Rennie Nov 22 '18 at 11:16
  • 7
    $\begingroup$ ... Ultimately, what will the Physics Stack Exchange Become? $\endgroup$ – John Rennie Nov 22 '18 at 11:16
  • 22
    $\begingroup$ You're assuming that every possible question will be asked on some timescale shorter than the heat death of the universe. This seems to me a questionable assumption. $\endgroup$ – John Rennie Nov 22 '18 at 11:17
  • 2
    $\begingroup$ Here's a relevant SEDE query data.stackexchange.com/physics/query/793344 $\endgroup$ – Chair Nov 22 '18 at 11:56
  • 1
    $\begingroup$ It would be good to see a graph of number of questions that have multiple duplicates. The trend I see, is that questions get asked and answered before they are marked as duplicate, especially simple questions with easy answers. That is, your worries do not seem to be valid. I see more a worry that we get a lot of different answers to same question, all slightly different and a large percentage of them plane wrong. $\endgroup$ – fishinear Nov 25 '18 at 15:35
  • 1
    $\begingroup$ If, as you say, every question will have been asked and answered, what exchange of information are you referring to? Since all the questions had been asked and answered, there can't possibly be any more information to be exchanged. $\endgroup$ – Rob Nov 27 '18 at 1:22
  • $\begingroup$ The ultimate goal of any SE is to be a complete, open, reference resource on a particular subject. Answering questions asked by other people is just a means to that end. Running out of new questions is a good thing. $\endgroup$ – OrangeDog Nov 29 '18 at 12:26
  • $\begingroup$ Keep in mind that extremely old answers are deleted from SE. Most people never notice this unless they are long time users who see their reputation strangely go down. I think that @JohnRennie has experienced this. $\endgroup$ – user400188 Nov 30 '18 at 6:13
  • 2
    $\begingroup$ The co-founder of SE has this to say - blog.codinghorror.com/… $\endgroup$ – OrangeDog Nov 30 '18 at 12:11
  • $\begingroup$ interesting ideas and nice work with data explorer. however the phrasing on this question is a bit awkward. SE network has over 100 sites now. physics is only one. there would be some expectation of similarities/ differences. newer sites probably go through different phases. sites do tend to mature and one gets more duplicates. also the maturation of wikipedia has been studied extensively and yes do believe SE has some similarities as far as "levelling off" in some key ways. how much new content is constructed? it probably declines gradually. long term sustainability is a key metric to track. $\endgroup$ – vzn Nov 30 '18 at 15:54

and then it is hard to tell if the curve is starting to increase again or stabilizing around a constant value of 7%.

I don't think this is hard. It's pretty obvious to me that the duplicate closure rate has saturated at a value of about 7%. This is vaguely high for SE standards but not something I'd be at all inclined to worry about. Here is a copy of the query Chair linked to, which can be readily switched to other sites, to look for precedent in older databases and sites with a higher throughput.

The clearest example is Mathematics Stack Exchange, whose historical duplicate closure rate looks exactly like ours but with a longer stabilized period:

enter image description here

You get a similar situation in Cross Validated:

enter image description here

There's a fair bit of variation site-to-site, but there's multiple instances of the rise-then-saturate behaviour that's evident in the Physics data,

enter image description here

so I really don't see how the current historical data is in any way a cause for concern.

To paraphrase John, when you say

I am assuming that the number [total number of non-duplicate questions] will be asked on some timescale shorter than the heat death of the universe

I find that to be an unrealistic assumption.

  • $\begingroup$ "[7%]is vaguely high for SE standards" It's unsurprising that ours is about double that of mathematics: their homework policy looks like it is much more tolerant, and homework questions can't easily be exact duplicates, though there are a few which pop up frequently. $\endgroup$ – Chair Nov 24 '18 at 4:02
  • $\begingroup$ It unrealistic to suggest all newly posted questions will eventually be dupes but more likely that the pool of unanswered elementary 1st yr homework-type questions (or trivial variations thereof) is actually shrinking. Is there an easy way to separate closure for duplication of homework question from the rest? $\endgroup$ – ZeroTheHero Nov 24 '18 at 15:22
  • 1
    $\begingroup$ @Zero So long as they're tagged correctly (which isn't always the case) yes, it's an easy modification to the SQL. $\endgroup$ – Emilio Pisanty Nov 24 '18 at 15:42

As long as there are new inventions and new discoveries, there will be new questions.

For the network as a whole: find your favorite programming language that is less than ten years old, and search for questions about it on Stack Overflow. Or on Area 51.

For physics in particular: consider the timely questions about the OPERA superluminal neutrino non-result. Or look through time at the way questions about gravitational waves were changed by the LIGO detections.

If we come to a point where the scientific community has stopped coming up with new ideas to test, we will have bigger problems than the end of unique questions on Stack Exchange.

  • 2
    $\begingroup$ that sounds great/ idealistic. however for a variety of factors/ reasons physics SE is not nec highly hospitable place for content around "new discoveries". there are other examples where queries on "new physics" are quickly rejected, even in the more potentially freefrom chatroom ("theory vs practice"). this is because there is some conflict with the "no nonmainstream physics" policy. mainstream physics defined broadly is inherently conservative and skeptical of "new science". $\endgroup$ – vzn Nov 30 '18 at 15:49
  • $\begingroup$ @vzn: Yes, but surely all that means is that new science takes N years to percolate to the point it can be Physics Q&A fodder? If the site is working off "new" results that are 10 years old, that's still an ongoing font of questions. $\endgroup$ – Nathan Tuggy Dec 5 '18 at 10:30

Even though I am a new contributor to SE, I have quite a long knowledge of on-line Q&A or discussion communities on the Internet. I think that the question hits an important point which goes beyond the pure analysis of data trends.

Thus, I can't help but consider the possibility that, at some point in a far future, almost every single question would have been already answered on Stack Exchange and then marked as duplicate.

I think that this sentence is based on a dangerous implicit assumption: that for each question there is the ultimate answer and nothing more can be said once the good answer has been provided.

My experience with physics and teaching physics, but also the analysis of part of the existing SE data, tell me something different. There are answers which could be considered as final, but there are also many answers which may be improved further. My strong impression is that the possibility of improving an existing answer, is somewhat hampered by the mechanism of flagging a question as duplicate. The main reason being that new questions are much more visible than the oldest and the whole mechanism of gaining reputation from the judgement of the original poster of the question may be broken for answers to very old questions.

Then, what would be the difference between Wikipedia and Stack Exchange, except for completeness?

In my opinion, this question returns on the same point: a static view of a Q&A site makes difficult to see the difference between Wikipedia and SE. It is only the dynamic possibility of further improvement of the existing answers based on the history of the previous ones, which can be an important long term added value of SE.

How would we be able to exchange our knowledge if 99% of questions are legitimately marked as duplicate?

My 2 cents are that if people would wait, instead of hurrying up to mark as duplicate a question a few minutes after it has been asked, most of the concern about the future of SE would disappear.

  • $\begingroup$ Wikipedia also has the dynamic possiblity of further improvement. In fact, the goals of Wikipedia and SE are fundamentally the same - only the manner by which we get there is different. blog.codinghorror.com/… $\endgroup$ – OrangeDog Nov 30 '18 at 12:11
  • $\begingroup$ @OrangeDog: thanks for the link. Similarity between SE and a wiki is clear to me. However, here I am more interested in differences. I appreciate very much the self-built hierarchy involved in the reputation mechanism. This does not exclude that something could be improved further. And in my opinion, giving more time for an answer to attract better aswers than the previous ones would go in the right direction. Maybe adding some technical way to merge old and more recent Q&A. $\endgroup$ – GiorgioP Nov 30 '18 at 16:02
  • $\begingroup$ I believe moderators have the ability to move answers to a duplicate question. Likewise, absolutely everyone has the ability to manually copy them, then ask for the old to be deleted. $\endgroup$ – OrangeDog Nov 30 '18 at 16:20

The answer is, the number of duplicate questions are unlikely to reach 100%. Therefore, StackExchange is unlikely to completely halt it's Q&A activity.

Reaching 99% would cause a slowing of activity, but, as common questions are answered and marked as duplicates, non-duplicates will become increasingly advanced and complex. Also, questions which are already answered continue to be commented or alternately answered. So StackExchange will always have a purpose and have activity.

On another note, I have seen many questions marked as "duplicates" which are not truly duplicates yet are similar. A better question might be, "will the proliferation of marking questions as duplicates dissuade users from asking questions out of frustration?"


An interesting idea would be to publish or have a periodic magazine containing the most popular questions and most popular answers.


I think there ought to be a reasonable time limit (3 month?, as a first bid) after which a question may be putten again without incurring a "duplicate".

  • 5
    $\begingroup$ A duplicate is a duplicate whether it's been a day or a decade. $\endgroup$ – Kyle Kanos Nov 25 '18 at 18:25
  • 1
    $\begingroup$ I find this happens anyway in practice because no one can remember or find the duplicate. This explains why (by my reckoning) about 30% of questions currently asked are duplicates but only 7% get marked as duplicates. $\endgroup$ – Bruce Greetham Nov 25 '18 at 19:00
  • $\begingroup$ @Kyle Anos ... And a tautology is tautology. The older a question gets, the less likely it is that it will be seen. Asking it afresh opens an opportunity for the answering of it by someone who probably would never see it were it not asked afresh. $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 19:24
  • $\begingroup$ @Bruce Greetham ... that's good as far as it goes - that there is a bit of slack in the system as it stands - and it's also evidence tending to support my assertion, as duplication policing is shown to be a flaky & precarious process ... yet, by reason of the same flakiness & precariosity, the said slack in it is not something that ought to be relied on ... wherefore I think there ought to be a cutoff in the timescope of policing. $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 19:30
  • 1
    $\begingroup$ @AmbretteOrrisey that's one theory. Another one is that linking to the dupe will allow other answerers to post their answer to the original, rather than the copy. $\endgroup$ – Kyle Kanos Nov 25 '18 at 19:37
  • $\begingroup$ @Kyle Anos ... I'm sure I'm not the only one who browses the questions. I certainly do not on all occasions of my using this here stackexchange operate it strictly according to a flowchart of think of question - check whether it's been asked before - if not then ask it. And even if we require that people use it strictly only in this way, which would be a crazy extreme of stiffneneckedness, do we have a search engine that could comprise all possible wordings of a question in its search logic? $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 19:53
  • $\begingroup$ @Noöne in particular. I would have not only duplicates freely permitted, without any restriction, but all the effort that currently goes into duplicate policing diverted into attaching links to all duplicates deemed such within reasonable measure of tolerance - which proviso is necessary, as questions can be very different in the nuances of their purport. And then I should not have the frustration of devising an answer to a question only to find that I cannot in fact dispense it, by reason of the question having been blocked, without trawling through the archive to find the original. $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 20:06
  • 1
    $\begingroup$ @AmbretteOrrisey yes, we actually do expect people to follow that flow chart; it's part of the "show some work" aspects to posting. And Google is actually pretty good at finding the right posts given the phrasing, though it's certainly not perfect. $\endgroup$ – Kyle Kanos Nov 25 '18 at 20:10
  • $\begingroup$ OK ... maybe not without restriction absolutely: obviously there would have to be some way of putting a curb on outright trolling by the gratuitous posing of duplicate questions. $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 20:11
  • $\begingroup$ Really ... there is an expectation that that flowchart be followed!!? Well I am certainly going to continue my browsing of the questions from which practice I have gotten colossal resource, and a map of the concerns of people at large ... obviously one filtered in that it is of the concerns of those whom the spirit moves to use this facility, rather than a strictly unbiased sample of the populace. I don't know who actually does the duplicate policing, or whose is the final decision to interdict a question, but I can only hope that they will not utterly be ruled by the imperative $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 20:23
  • $\begingroup$ to impose so very extreme a constraint of 'lawful purport' of those who use this facility as you propose. $\endgroup$ – AmbretteOrrisey Nov 25 '18 at 20:26
  • $\begingroup$ Anyone can proffer a dupe closure, but either a single mod or 5 users with rep $\geq3k$ (or any combination of the two) is needed to fully close as a dupe. How active this requirement is is probably a bit subjective, but it is generally an expectation that you do some work to resolve your own issues, rather than just dumping it off to us without thinking about it. $\endgroup$ – Kyle Kanos Nov 26 '18 at 3:18
  • $\begingroup$ @AmbretteOrrisey unfortunately there are already so many duplicates that search engines can return too many versions of the same core question. What is needed are users who actually search before posting. $\endgroup$ – ZeroTheHero Nov 26 '18 at 5:37
  • $\begingroup$ @ZeroTheHero Users don't usually change their behavior to match the wishes of the software, rather the other way around, if we're being realistic. $\endgroup$ – nunya Nov 27 '18 at 0:44
  • 1
    $\begingroup$ I think soon we may have a real answer to this ages-old problem: AI. We can't control human nature, and our software responds to humans imperfectly. But AI moderation at scale? Imagine having a team of 1,000,000 dedicated moderators, organizing, tagging, and identifying dupes. This is probably closer than we think, given that I'm now getting AI-oriented developer marketing around reusable third-party frameworks and APIs, which was not the case 10-15 years ago, when AI was also being promised. $\endgroup$ – nunya Nov 27 '18 at 1:13

You must log in to answer this question.

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