How can we change the motivation system to encourage people to spend more time curating and looking for pearls, rather the wallowing in the sand?

We have heard that "Questions are the sand, and great answers are the pearls." (See https://blog.stackoverflow.com/2011/06/optimizing-for-pearls-not-sand/)

  • We want to optimize for great answers because they end up helping many people, not just one.
  • We want people to have an easy time finding answers to their questions.
  • Sometimes, duplicate questions will be asked, since the duplicate is not immediately obvious.
  • Pearls exist already, but they are not obvious.
  • Our reward system is supposed to optimize for pearls, but it doesn't always do so.

Some similar critiques/issues: Why do we reward fastest answers?, Improve tools for closing as duplicate, Some people just do not know how to search, Would it be useful to be able to vote for Canonical answers?

There was a comment here - https://meta.stackexchange.com/a/108584/187896, saying that we should self-curate. Would this work? It seems like the reputation system is optimized for the opposite! I get many more points for answering quickly and badly than for re-writing, fixing, clarifying, etc.

We are drowning in new, previously asked questions, and we're giving out points for failing to help!


4 Answers 4


It strikes me that those with the most to gain by closing duplicates are those writing the canonical answers. Perhaps nudging this with some badges would help...

  • Acknowledged - top-scoring answer on a question referenced by 5 duplicates.
  • Cited - top-scoring answer on a question referenced by 10 duplicates.
  • Authoritative - top-scoring answer on a question referenced by 20 duplicates.


  • This might be somewhat tricky to track.
  • Could become a perverse incentive to close things that are not actually duplicates.

(note that this differs from the suggestions on Give an incentive for finding duplicate questions, as I'm not recommending a badge for closing)

  • Would you want some kind of minimal score on the answer as well? (At least for the upper tiers) Sure it's an answer that might answer all of those questions, but that doesn't necessarily make it a good canonical answer.
    – Troyen
    Jun 14, 2012 at 23:12
  • 6
    Hey, this is a cool idea. I don't see the "perverse incentive" problem as it will take 5 to close, and only 1 of them will benefit. However, it will make you select your authoritative answer over one that might be better... I'd say your own closings shouldn't count toward your badge. But overall this reeks of a great idea
    – Pekka
    Jun 14, 2012 at 23:23
  • 2
    @Pekka: I've seen a few vote-to-close-as-dupes where the first voter wrote their own question and answered it later than the question they were closing in an effort to build the "canonical question and answer".
    – user7116
    Jun 15, 2012 at 0:09
  • @Pekka: probably simpler to just discount questions where the close votes were split among more than one dup-destination.
    – Shog9
    Jun 15, 2012 at 0:19
  • 1
    @Pekka: But all my CSS answers are the authoritative answers!!! (Disclaimer: not really, I'm not actually a CSSWG member.) Jun 15, 2012 at 4:31
  • Hulk is about to smash your abuse of <kbd> but the badges seem like a good idea Jun 15, 2012 at 4:47
  • @BoltClock hmm, right - seeing as you're probably one of the most prolific closevoters in the tag, that would be a real problem as it would discourage you from closing. I'm not sure about Shog's discounting idea though, users pointing out two different (legitimate) canonical answers seems pretty normal and having your answer discounted because of that would feel pretty unfair.
    – Pekka
    Jun 15, 2012 at 7:12

We are drowning in new, previously asked questions, and we're giving out points for failing to help!

For a site as aggressively programmed as SO (in terms of sidebar widgets, rollover popups, etc) the search capabilities are shockingly thin. I would suggest that to reduce duplicate questions and to serve up those "pearls", some serious work on the search UI is needed.

Perhaps the minimal search capabilities/minimal search UI is a design/site philosophy - I recall seeing some meta posts which implied such: prioritizing breadth over depth. This is also present in the design decision to not allow a user to limit the display of questions to the 'favorite' tags they have chosen.

At any rate, improving search (e.g. enabling tag search through a UI rather than entering [tag], since many users may not even realize that is an option) might help reduce the clutter and noise.

  • An idea that I think would help improve searching for Good Answers is Pekka's Cannons
    – sarnold
    Jun 14, 2012 at 23:19
  • 5
    If people won't do a web search what makes you think they'll do a search on SO Jun 14, 2012 at 23:50
  • 3
    Try asking a question and see how many "hey, how about this!" prompts you get as you compose the question. Really go through the motions, write an actual question. Note that new users (by rep) also get an unskippable EULA page with a giant search box directly in the middle of it. See for yourself: stackoverflow.com/questions/how-to-ask Jun 15, 2012 at 4:50
  • 3
    @jeff atwood: No, I am not high. Are you jaded? Smug? Too used to SO to even see it? Seems so. The prompts that are presented can be mere word hits if one doesn't include a tag as a filter. They are not ranked by number of votes, by number of answers or by date (that's a separate tab). There is no UI for advanced search - just a separate page with suggestions. If you want users to 'do the right thing' you make it easy. But whatever.
    – spring
    Jun 15, 2012 at 5:05
  • @skinny I don't think you actually know how the site works. 1) Try entering "java generics" in the search and see what happens. 2) Try asking a question as a new user (open an anonymous browser page) and see what happens. Jun 15, 2012 at 5:09
  • 2
    @jeff atwood: So I don't know hw the site works? Sorry but that is blaming the user, too often the cause of poor UI design. fwiw, I have been using the site for over a year.
    – spring
    Jun 15, 2012 at 5:23
  • 1
    @skinny can you answer the two questions I asked in the above comment, please? I numbered them so it's clear which is which. Thanks. Jun 15, 2012 at 5:27
  • 3
    and just so that we're clear, searching for "java generics" (without the quotes) is in fact a tag search equivalent to entering "[java] generics". This is completely automatic and true for the top 60 tags, as documented on the stackoverflow.com/search page. That's why your statement that "users need to enter [tag]" is totally incorrect, at least for the 60 most common tags. Jun 15, 2012 at 5:31
  • 4
    I did do that prior to my last comment. The results were exactly the same and no, I did not see your point and it was not obvious that a tag search was automatic. I see that now. My point however remains: search here is weak in terms of UI/usability (as mentioned, no sort, no filter). I don't get why that should be. Example: do a search for "UIImage crop" (no quotes). Loads of hits in no apparent order. Click on the 'votes' tab - is that votes of my search result? Appears not. Why not? Click on 'newest' or 'active' and it appears to be hits only for 'UIImage'. Confusing/ not helpful.
    – spring
    Jun 15, 2012 at 6:13
  • 1
    @Jeff I find SO's search largely helpful. SkinnyTOD has a point in one aspect though - the advanced search is incredibly hard and counter-intuitive to discover. Have you considered adding a simple link next to the box? (I know you have, but I can't find the discussion in question)
    – Pekka
    Jun 15, 2012 at 14:43
  • @Pekka Whoa, didn't know about that. Perhaps "Advanced Search" as a search suggestion (pops up when you select the search box) would be more aesthetic? Jul 23, 2012 at 6:31

Maybe linking to duplicate or related posts should be votable for rep (perhaps +2 for each upvote, -1 for each downvote; or maybe just +/-1 to keep it simple) based on how relevant it is to the current post. This would actively encourage people to sift the sand for pearls instead of just piling on more sand.

Let's assume a priori that many questions are at least vaguely related. We already have the automated system providing Related links, but there's no real substitute for human curators. Let us also assume (because it's pretty self-evident) that many users "just ask" rather than search for similar posts first; that being the case, there needs to be some incentive for people to go searching for them, other than the joy of voting to close.

This system would also have a salutory effect on "fastest gun" syndrome. It would reward people who took the time to find pearls with some easy rep. While that won't stop people from posting duplicate or related answers for quick rep, it will at least bias the system more towards post-reuse than it currently is now, while simultaneously improving cross-linking and the closing of valid duplicates.

  • I think that we need reputation rewards for linking to a similar question to be at least as much as for answering. Upvotes on a badly written, hasty, but basically correct answer get you at least +10. Jun 14, 2012 at 23:39
  • 1
    Something like this - which effectively changes "finding duplicates" into "answering with links to duplicates" - could accomplish this, provided you made the "duplicate" answers generate reputation in response to votes. It does have the advantage of capturing an existing [anti-]pattern and turning it into something useful, but may also encourage dup-spamming if not handled carefully.
    – Shog9
    Jun 15, 2012 at 0:43

Some Ideas

  • Answering a duplicate question (typically less well than the original answers) should be less rewarded than finding the duplicate. We want to encourage re-use of pearls. Maybe we should take away half of the reputation earned on a question that is closed?
  • Editing should be more encouraged, to improve the answers that other people made that are good, but could be great. Maybe edits can be voted on, and if sufficiently good, could get reputation from future upvotes, or a proportion of the upvote reputation. (How do we measure sufficiently good?)
  • Possibly create some method for nominating the best re-writes of the day/week? This might be a better gamelike mechanism.)
  • Finding the best answers, and then clarifying the questions by generalizing them, should be rewarded with reputation. This might be done by giving a proportion of the sum of the reputation of the question and the current answers to people who improve the question significantly. (How do we measure significant contributions?)
  • Some type of point bonus for finding repeated questions, and closing them as duplicates, may be important. This will help by eliminating sand that, while not useless, does not increase the pearl population.
  • 3
    I don't agree with Finding the best answers, and then clarifying the questions by generalizing them, should be rewarded with reputation.. You don't want easy rep hunters ruining nice answers. Jun 14, 2012 at 22:02
  • I'm not clear on how rep hunters could use this as a peverse incentive to ruin nice answers... Jun 15, 2012 at 11:33

You must log in to answer this question.

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