What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 134 Stack Exchange communities.

It's hardly news that the share of duplicates is going up. As the sites accumulate more information and more participants, the effort required to confirm that something hasn't already been answered increases at the same pace.

Currently they are simply closed if discovered by the small fraction of users (2%) having a "Vote to Close" privilege but it doesn't happen very often since finding of duplicates is not rewarded.

Although some users discourage referral answers, I don't mind upvoting those which point to a question with better insight - it gives the most useful duplicate a chance to bubble up. To me it seems like a good idea but what does the community think? Are there any drawbacks to it?

share|improve this question
    
@Gnome: See the last comment to dmckee's answer. –  user151803 Nov 12 '10 at 18:27
    
@Gnome: The current consensus is that dupes should be pointed out by a comment. I am telling you that having dupes pointed out by an answer motivates people to find the best match, not just any. –  user151803 Nov 12 '10 at 18:44
1  
@Saul: Ah, you want to reward them with rep for pointing out duplicates that way? I strongly disagree, it will lead to too much noise in blatant non-answers. And too much worrying about rep by those users, in the first place. –  Gnome Nov 12 '10 at 18:47
    
(Cleaned up the comments that were about my confusion.) –  Gnome Nov 12 '10 at 19:07
    
@Gnome: Answers on duplicates are noise anyway. The proposal is to convert that noise into a signal by having low-value duplicates refer to the most useful version. –  user151803 Nov 12 '10 at 19:25
    
I disagree. Suggested duplicates are not always duplicates — even not counting questions that fail to reach 5 close votes but "really" are duplicates. Plus, an especially confused or new user may benefit from an answer that is just a tad more specific to them — just including their variable names could be enough. If you see users answering when you think they should answer on another question, comment that "this would be a great answer on [link], no one else has yet provided this info there". –  Gnome Nov 12 '10 at 19:38
    
@Gnome: I was more referring to answers that link to another question. The issue of duplicates resolving the ambiguity of natural language is already covered in dmckee's post. –  user151803 Nov 13 '10 at 10:45

3 Answers 3

up vote 4 down vote accepted

As the sites accumulate more information and more participants, the effort required to confirm that something hasn't already been answered increases

Gotta say I disagree with this on two levels:

  1. As a practical matter, with respect to oft repeated questions the more content the site accumulates the more likely it is that any user who searches will find their question asked with their choice of words.

    If gets easier to find existing instances over time which is the whole point of not deleting the duplicates: they serve to resolve the ambiguity of natural language.

  2. From a philosophical POV, there is a subtext here that the goal is to ask questions.

    The real goal is to get an answer to a question that is causing you trouble or delay. I repeat: to get an answer and to get it quickly. When the answer is already on the site is should generally be faster to find it than to compose a well asked question.

share|improve this answer
    
Okay. But what about the pros and cons of referring to a duplicate question using an answer? –  user151803 Oct 29 '10 at 18:27
2  
@Saul why is that necessary? In my eyes, an answer pointing to a dupe is nothing more than noise. What should be done is, the reference to the "original" question should be made more prominent in the site's layout, finding a dupe should yield reputation, and people should be able to point out duplicates from early on. –  Pëkka Oct 29 '10 at 19:22
    
@Pekka: "By rewarding good cross-links one gives an incentive to find the best (duplicate) match". Why is that bad? –  user151803 Oct 30 '10 at 10:06
    
@Saul again: Why is it necessary? It just creates noise. Why not simply vote to close as a duplicate of that question? As I said, I'm totally on board with awarding that with reputation, but all suggestions to that effect have been declined in the past.... –  Pëkka Oct 30 '10 at 10:07
    
@Pekka: All answers are noise before getting any upvotes and closing a question removes the mechanism that surfaces the "signal". Not to mention that currently only 2% of users have the privilege to vote for a close. –  user151803 Oct 31 '10 at 12:34
1  
@Saul I do see your point about that this could be a nice way of rewarding the finding of duplicates. I still think the answer format is not good for this, especially because such answers pointing to duplicates will not lead to the question getting closed. The "close as dupe" facility needs to be improved instead. If there are not enough users able to vote to close, then the threshold needs to be lowered. –  Pëkka Oct 31 '10 at 16:37
    
@Pekka: Not just finding any duplicate. It's about finding the most useful duplicate. I think having a high threshold for a closing vote is good since a trigger-happy closing team could go a long way in taking the fun out of StackOverflow. –  user151803 Oct 31 '10 at 16:50
    
@dmckee: I am not advocating for deletion of duplicates. The goal is to introduce a mechanism that results in a link to the highest-quality duplicate in every less-prominent duplicate. –  user151803 Nov 1 '10 at 14:25
2  
@Saul: I didn't think that you were. As you noted this post doesn't really "answer" your question at all, but it was too long and substantive for a comment. You bring a subtext to this discussion that is somewhat at odds with the existing philosophy. I'd note that we had duplicate identification in the answers with votes (by default, not by design) before comments were implemented (and I most have collected a score or so votes for that kind of thing) and it was ugly and awkward. I thought that using comments for the purpose was an improvement. –  dmckee Nov 1 '10 at 14:49

As said in the comments, I like the spirit of the idea, but I'm against the practical implementation.

Pointing out potential duplicates in answers creates too much noise, and doesn't count towards actually closing the question as such a duplicate.

Improve the existing duplicate closing process instead. Make finding dupes yield reputation; lower the thresholds for voting to close if necessary; and make answers to duplicates CW (Anybody who has something new to contribute, can do so in the original question).

share|improve this answer
    
The goal is not to inspire closing but to motivate the connecting of duplicates with the most insightful one amongst them. –  user151803 Oct 31 '10 at 17:17
    
@Saul but that would result in dozens of identical questions having similar, but partly different answers. If you need an answer to the question, you need to browse every single copy to find the best answer - his is a problem already. I think there is no way around identifying one canonical "original" - no matter by what metric - and concentrating all efforts to focus on that. Everything else, if you ask me, results in chaos –  Pëkka Oct 31 '10 at 17:19
    
@Pekka: It's not so much about having one canonical "original" as it is about usefulness - which can be identified by the amount of upvotes on the referral. Having different answers is okay in my opinion, as long as the community can indicate which ones are beneficial. –  user151803 Oct 31 '10 at 18:12
1  
@Saul I believe that only canonical questions are useful in the long term. Look at this question from today. It is pure noise and adds zero value to the question base. To me, it needs to get closed ASAP. I have no problem with duplicates as such, and I also don't mind people gaining reputation from them - I have answered a lot of dupes in my time without realizing it. But the goal must be having one question where one can find all the relevant information on a topic. The "add answers to point out duplicates" approach won't work for this. –  Pëkka Nov 1 '10 at 12:35
2  
@Pekka: Your long term goal is similar to what I had in mind. Apparently we disagree on premises and on the method. Right now there's only stick. There should also be a carrot. My proposal is to "outsource" the hard work of consolidation to everday users instead of the elite. For example the question you linked still lacks a reference to the most useful duplicate. Why? Because there is no incentive to provide it. –  user151803 Nov 1 '10 at 13:53
    
@Saul the latter is the central point, I totally agree. There should be easy to find, high-quality, high-voted references to point to. I also agree that finding dupes should be incentivized. We just disagree on the execution (although I admit that your idea is the only way to achieve rep gain for pointing out dupes without introducing completely new functionality to the system) –  Pëkka Nov 1 '10 at 14:00
    
I need to blog about this, (working title: "Dr. Strangedupe: Or, How I Learned to Stop Worrying and Love Duplication") but there is a continuum here. 10+ duplicates is obviously too many. But having one "perfect" form of a question that contains every possible answer to every slight variation of that question is a myth at best and actively harmful at worst. No, what we want is on the order of 4-5 similar-but-not-quite-the-same duplicates to cover all possible search terms and permutations of the question. More than that, is bad, but so is less than that. –  Jeff Atwood Nov 14 '10 at 4:38
    
@Jeff that's very interesting (seriously, not snarkily) but that doesn't invalidate the point of creating easier ways to identify dupes, does it? –  Pëkka Nov 15 '10 at 10:15
    
@pekka some users think that any atom of duplication in the universe is too much... –  Jeff Atwood Nov 15 '10 at 12:41
    
@Jeff yeah, and the point you raise is a good one, no doubt. But finding better ways to deal with the questions of which new duplicates come in on a daily basis is still very much an issue IMO. –  Pëkka Nov 15 '10 at 12:45

Pros

  • This incentivizes searching for duplicates for new questions, which is a desired behavior
  • This may reduce scattering of information, depending on whether the people who are searching for dupes are the same people who would have previously been composing duplicate answers

Cons

  • This disincentivizes searching for duplicates before asking a new question, which is also a desired behavior
  • Reputation was not intended to represent a user's ability to use search tools
share|improve this answer
    
Reputation was intended to represent a totally arbitrary number in the first place. So your last point does not count. –  Ladybug Killer Oct 29 '10 at 19:20
    
@Ladybug, I can't tell if you're joking, but reputation isn't supposed to be completely arbitrary; if it were, everyone would just get assigned a random number at registration time. It's good as a rough measurement of two things: knowledge of the topic and knowledge of how the site mechanics work. –  Pops Oct 29 '10 at 23:30
    
@Popular Demand: I agree with the first con but I think reputation is rather intended to represent the amount of useful information the user has given. At least that's what the SO team has stressed all along. Why should useful information obtained by search be less qualified? –  user151803 Oct 30 '10 at 10:15
    
@Saul, there's a subtle but significant difference between generating new content and helping people find good existing content. –  Pops Oct 30 '10 at 19:07
1  
@Popular Demand: I agree. What I am struggling to understand is how does that difference change the relevance of information. After all, the alt text for the upvoting button says "This answer is useful". It's simply beyond me why should undiscovered content be less useful than new content. –  user151803 Oct 31 '10 at 16:03
1  
@Pop: Without joking I can tell you, that reputation has nothing to do with knowledge. One of our most famous trolls calls it EXPerience and he is much closer to the truth than everyone else. SO is a game. If you have ever played a role-playing game, you should know the fascination of arbitrary numbers. –  Ladybug Killer Oct 31 '10 at 20:35

You must log in to answer this question.