96

When closing a question as a duplicate, it is becoming harder and harder to search for a good original as the pool of questions and answers grows and grows.

While the new dupe closing dialog is a huge step forward, it doesn't always work - if you've spent some time on the site, you know there are perfect answers to many new questions but you can't always find them. Or can't be bothered to, as your time is limited.

A way to further improve this could be introducing a vote for canonical answers.

It could, for example, work like this:

  • Answers get a new icon, the "canonical vote" (signified here by a cannon because I can't think of anything that makes sense right now.)

    enter image description here

    • Only users with a certain minimum rep (maybe 1k+ or 2k+) see this icon.

    • The icon would work like the favourite icon: You can toggle it on or off at any time.

    • The hover message for the icon could read e.g.

      This is a "canonical answer": It answers the question asked in the most extensive and general way, and is very likely to benefit other users with the same problem.

    • Questions that contain an answer with three or more canonical votes get special treatment:

      • They show up on the top in search results; maybe even under a special heading ("Questions with canonical answers")

      • They show up prominently in the "close as dupe" dialog (i.e. the canonical votes influence the metric that is used to determine the questions shown there).

    • There could be some rewards for users whose answers receive "canonical" votes - maybe a badge if ten of your answers get at least three canonical votes, or a counter of canonical answers in your profile. It would be really interesting as it's a badge that you can't game your way to, but this whole rewarding part is entirely optional.

    • Kobi's suggestions would be a great safeguard against over-use of this feature.

Dupe (but sans cannon):

Related:

24
  • 6
    It would be nice if there were some way of telling apart the '+1 for idea' votes from the '+1 for cannon' votes that this q will inevitably attract :)
    – AakashM
    Dec 15, 2011 at 9:31
  • 7
    Meh, doesn't work. Everybody will shoot the cannon on whatever dup they happen to run into. Now you're back to counting cannons, might as well count votes. Dec 15, 2011 at 10:21
  • 1
    @Uphill Kobi's suggestions below would help that problem.
    – Pekka
    Dec 15, 2011 at 10:22
  • I think this should be limited based on the number of upvotes a user has in the given tags, not on thier overall rep Dec 15, 2011 at 11:26
  • @Ian true, that would be an even better indicator. However, I'm not sure how often people dupe-close stuff in tags they're not active in - for me, for example, that never happens.
    – Pekka
    Dec 15, 2011 at 12:27
  • 19
    Thats a punny icon.
    – user1228
    Dec 15, 2011 at 15:05
  • Would it also be helpful to mark a question as canonical? Dec 15, 2011 at 23:20
  • @Chris not really IMO, I can't see the benefit (but I'm happy to be proven wrong). However, the question would be kinda canonical by extension if one of its answers is.
    – Pekka
    Dec 15, 2011 at 23:36
  • Aren't wiki posts canonical? Dec 16, 2011 at 3:29
  • @Ben nope, they have no characteristic that one could call canonical (like a more prominent position in search results)
    – Pekka
    Dec 16, 2011 at 9:25
  • Oh man, the line breaks in the limerick didn't work out! Rats.
    – Pekka
    Dec 29, 2011 at 20:09
  • 2
    For what it's worth, Will asked a somewhat similar question a few months ago to which Marc Gravell replied "We're playing with a few things in this area... stay tuned." Also, voting for canonical answers has been suggested before -- during the UserVoice days, no less! -- although at this point it may be better to close that one as a dupe of this one.
    – Pops
    Dec 29, 2011 at 20:31
  • 3
    -1 for offensive (and not to mention completely off-topic) bounty notice, regardless of disclaimer. Dec 30, 2011 at 11:20
  • 2
    I find a standard that thinks "bitches" is acceptable language (while "balls" is offensive) hard to take seriously.
    – Pekka
    Jan 1, 2012 at 22:47
  • 2
    Gee, it was just a joke... Jan 1, 2012 at 22:54

6 Answers 6

32

I like this idea, but I would suggest a lot more limitations on possible canonical answers. Certainly, not every answer can be considered canonical or even be allowed to voted on as such.

I would suggest only showing the cannon on answers and questions that:

  • Have many votes (for example, over 10 votes for the answer)
  • Have many views.
  • Have incoming links (via right side bar Linked), or more specifically:
  • There are closed questions which are duplicates of this question (at lease a few). IIRC, the closed as duplicate dialog already gives these question priority when suggesting duplicates.

In my opinion, only when an post satisfies these criteria (or similar), it can even be considered canonical.

4
  • 3
    Good suggestions, and I find the idea of automating the whole thing interesting. However, it would make it impossible to select the rare gem of an answer that hasn't achieved that status yet, and maybe never will because it is so obscure. That's a pity, because the only way to "promote" such a question would be to bookmark it, and to use it as a dupe target whenever a similar question pops up - which is a lot of work
    – Pekka
    Dec 15, 2011 at 12:23
  • 1
    Is that really an issue? Canonical answers are meant for highly FAQed subjects. Can there really be an obscure canonical answer?
    – Kobi
    Dec 15, 2011 at 13:44
  • 5
    I think so. Any new answer that outshines everything else that's on the market could become the new "best canonical" one. Also, I disagree that canonical answers are for highly FAQed subjects only - there is a lot of frequently reoccurring questions that could benefit from this.
    – Pekka
    Dec 15, 2011 at 13:48
  • Incoming duplicate links should not be necessary. In the closed, duplicate topics, there are often much better answers than in the "original" topic. That's my experience on SU at least. Jan 12, 2012 at 12:57
22

Uphill Luge noted:

Meh, doesn't work. Everybody will shoot the cannon on whatever dup they happen to run into. Now you're back to counting cannons, might as well count votes

That sums it up.

This sounds like feature creep to me.

I believe the votes are the canonical mechanism for establishing canonical answers. If the problem is with search results, lets modify the search results to show even obscure questions with good answers.

8
  • I think implementing some of Kobi's suggestions would mitigate the "everybody shooting the cannon" problem nicely. Re sorting by votes, yeah, that would be a step forward - I actually suggested that once.. Still, to me, there is something to be said for having truly canonical answers that, if all goes well, are also improved over time
    – Pekka
    Dec 15, 2011 at 22:46
  • @TheP.G.RepMiningCo.: I believe you can already sort by vote.
    – surfasb
    Dec 15, 2011 at 22:57
  • not by answer vote, no. And that's what counts for this purpose
    – Pekka
    Dec 15, 2011 at 23:03
  • @TheP.G.RepMiningCo. - I assume you mean sorting questions by answer vote. You can already sort answers within a question by votes (it's the default) except for the accepted answer which floats to the top.
    – ChrisF Mod
    Dec 15, 2011 at 23:23
  • @Scrooge yeah. I mean when searching for a dupe - the list of questions can't be sorted by answer votes
    – Pekka
    Dec 15, 2011 at 23:35
  • @TheP.G.RepMiningCo. Sorting by answer vote destroys the whole purpose of voting on answers don't you think? The votes are only relevant to the questions on hand. The quality of the answers across questions cannot be measured on a nominal scales, like number of votes. That's like saying IE9 is a better improvement over its predecessor than Firefox 8, because more people use IE9. . .
    – surfasb
    Dec 16, 2011 at 5:19
  • 2
    @surfasb I'm not sure I follow - what's wrong with additionally offering a "sort by answer vote" view when searching questions? So that when I search something, the question with the highest voted answer is on top?
    – Pekka
    Dec 16, 2011 at 8:54
  • Like I said, before, the number of votes on an answer is related to the question. An answer for a question about building a house isn't going to get the number of views as say an answer for a question about building a nuclear plant. Yet you can pretty much guarantee that people who build nuclear plants are the top experts in the field of construction.
    – surfasb
    Jan 4, 2012 at 21:21
5

Canonical answers have two measurable attributes:

  • length
  • internal links

The cannon, while super quaint, would just mirror normal votes and correlate to past traffic again. (Nevermind the ensuing gold rush.)

And I believe we rather need a blacklist for the unmindfully closevoted ones. The automated dialog suggestion algorithm is probably not easy to enhance with just another number.

4
  • The attributes are well observed, but how do internal links come about? From users searching for a duplicate. That is currently a wildly random process - about half of the dupes I flag are ones where I happen to remember a great original question. That's why I think the cannon is necessary - you can have a great, upvote-worthy answer that still isn't canonical. Upvotes are not necessarily an indicator for a good canonical answer. Still, I agree it would help to be able to search for highly voted answers which isn't possible at the moment
    – Pekka
    Dec 29, 2011 at 20:46
  • True, it's a different flag. What we actually lack is, me thinks, more featuritis in the favourites. It would be easier to uncover past reference answers if they were easier to collect. Nowadays while seeing valueable posts, I shy away from starring them, because it just keeps piling up. Would there be a favouriting tag like "canonical/reference", then it could also be displayed (gimme scrollbars!) for the closevote suggestion dialog. Making it a personal feature foremost, and allowing the suggestion algorithm to naturally improve from the actual closing, not a new count.
    – mario
    Dec 29, 2011 at 20:58
  • interesting idea. That would actually speak for implementing this feature request from an hour ago as a native SO feature
    – Pekka
    Dec 29, 2011 at 20:59
  • 1
    Sounds very much so. Either way, this is going the proof-of-concept userscript direction.
    – mario
    Dec 29, 2011 at 21:08
3

I like the idea. But I also have an other suggestion.

I think answers on duplicates should be either merged or there should be a separate view in which you can see the question (probably the first) and all answers that are given to duplicates (and duplicates of those).

The view should be accessible from all questions.

3

I don't think this is a great idea for two reasons:

1. Technology changes

All technology changes; that is one fundamental quality that makes it technology. Canonical votes assert a timelessness to an answer that may only be canonical at a certain point in time.

Are people who voted for canonical really going to go back and uncheck their canonical votes as the article becomes obsolete? I kindof doubt it. Then you face perhaps better-answered competitor posts that may not be able to overcome the mountain of canonical votes given when the obsolete article was more trendy and popular.

2. Overlapping technologies

Real questions often involve overlapping technologies. How do you scope the limits of the canonical vote? Simple example, I ask about TCP performance of some library, and someone answers with an outstanding explanation of all the inner workings of TCP. Then the give an ok explanation of the performance characteristics of said library. People who like the TCP description might click canonical even though the library description was mediocre.

The canonical idea has a certain seductive appeal to it, but I don't think it's practical.

5
  • I disagree. Re 1.), the fact that things may change for some questions is not enough of an argument against having some sort of mechanism to quickly identify the best answer for a question. It doesn't need to be the solution presented here, but duplicates are a rampant problem and something needs to be done about it. Re 2.), a question with overlapping technologies most likely needs to be split up into two questions anyway.
    – Pekka
    Dec 30, 2011 at 10:57
  • @The P.G. Rep Mining Co., 1. You are knocking down a strawman. I said your idea is flawed, and I said nothing to object to the idea of duplicate detection. Other proposed duplicate detection schemes are less flawed than this one. 2. You fail to respond to the example I gave, which did not have two questions to split. Dec 30, 2011 at 11:09
  • re 1., point taken - if the canonicality of an answer fades over time, something would need to be done about it, you're right. Re 2., please give a real-world example of this. If a an answer doesn't answer the question well, it has no business becoming the canonical answer and I think the danger that people will vote it canonical is a theoretical one at this point. And even if it happens occasionally, I think it will be exceedingly rare. What you describe is not the norm.
    – Pekka
    Dec 30, 2011 at 11:25
  • @TheP.G.RepMiningCo., re 2: by definition, all discussion about canonical votes is theoretical, as there is no canonical vote to observe. However, we can observe human behavior, which votes in favor of humor, sarcasm, and off-topic discussions all the time. Dec 30, 2011 at 11:28
  • Fair enough, good counter-points. This is kind of the problem that we already have with votes. Upvoted
    – Pekka
    Dec 30, 2011 at 11:30
2

The problem with mapping duplicates is that it's a ton of work. Adding more work - more voting - isn't going to solve anything.

The best person to identify a duplicate is the person asking it. The second best is the person who answered the original. These also happen to be the two people with the most to gain from it. Giving them the tools they need to accomplish this - primarily better, faster search - has the best chance of paying off in the long run.

4
  • 3
    With a "canonical" vote, you would help power users identify outstanding dupes in a sea of mediocre dupes while searching. That would be its only use and I don't think you'll find the engaged users opposed to an additional voting metric that matters more (or at least, might matter more) than the common upvote. But if you can come up with a way to find that content without adding any new metrics, great. It's not there yet, though, not by a long shot. At the moment, you can't even search for obvious duplicates properly.
    – Pekka
    Feb 20, 2013 at 8:59
  • The C++ tag have their own take on this. That'd work as well for me. Special handling for FAQ questions
    – Pekka
    Feb 20, 2013 at 16:53
  • See my answer there, @Pekka - there's already a way to "vote" for these built-in; it's not used enough, but that doesn't give me a lot of hope for a second system.
    – Shog9
    Feb 21, 2013 at 22:59
  • 1
    A "canonical" vote would be a one-click vote that doesn't require the work of connecting two duplicates. It would be a way to say "wow, this question is well put and has excellent answers. I would like to see any mediocre duplicates point specifically to this question."
    – Pekka
    Aug 21, 2013 at 20:27

You must log in to answer this question.

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