2
$\begingroup$

Inspired by this post, I had the idea that browsing questions would be more fun if there was a feature that allowed sorting/filtering/searching based on the number of up-votes a question has received from high karma/reputation users exclusively.

The issue is that a non-trivial number of deep questions are only accessible to an advanced audience, and therefore they don't receive the up-votes and "favorites" that experts might feel they deserve. Don't get me wrong, I think we want to leave question-voting/scoring unchanged. I'm glad newbies can vote. But I think this is an important second-order problem.

We could add this feature without making it super-prominent if we want to keep the UI clean.

This feature would make CV and other SE sites more interesting to experts, and attracting experts has benefits that trickle down to all users.

Thanks to Nick Stauner for suggesting I make this an official feature-request. I encourage all of you to offer more specific suggestions for how it might be implemented.

$\endgroup$
9
  • 6
    $\begingroup$ I think this is not possible on SE, because it violates the assumption of voting anonymity. $\endgroup$
    – Andre Silva
    Feb 23 '14 at 0:18
  • 1
    $\begingroup$ I think voting anonymity could be maintained quite easily. Say the threshold for whether a user's votes counted was whether they were in the top 25% of the karma distribution. I don't see that someone could deduce much about who voted for what by being able to sort by that. $\endgroup$
    – Michael Bishop
    Feb 23 '14 at 19:40
  • 4
    $\begingroup$ If by "karma", you mean reputation, the user at the 75th percentile is presumably on this page, w/ rep = 101. Realistically, very close to 100% of all votes likely come from the top 25% of CV users by reputation. Moreover, only 275 users have cast >= 100 votes. To narrow your set of users down (ie, below 25%) to a point where this would be useable would diminish the anonymity beyond what I believe the devs had in mind. $\endgroup$
    – gung - Reinstate Monica
    Feb 23 '14 at 21:28
  • $\begingroup$ @gung 'karma' is what reddit calls its version of 'fake internet points' (and perhaps, some other sites). $\endgroup$
    – Glen_b
    Feb 23 '14 at 22:43
  • 2
    $\begingroup$ @MichaelBishop: 25% is WAY too much. I think a reasonable cutoff for what you are suggesting would be around 1000 rep, and there are only ~250 users passing this threshold, i.e. only ~1%. $\endgroup$
    – amoeba
    Feb 23 '14 at 23:37
  • 1
    $\begingroup$ Based on the helpful stats provided by @gung I guess we would need a higher cutoff than the 75th percentile in reputation points. I don't think it would need to be as high as 1000 rep to be meaningful. If the top 250-500 users qualify, that seems to me to be sufficient anonymity. Additional measures to preserve anonymity could be taken, like adding a variable time lag to the vote count and hide the number of votes and only use rank to sort. $\endgroup$
    – Michael Bishop
    Feb 24 '14 at 1:19
  • 1
    $\begingroup$ I guess that ranking answers by the top 250-500 users' votes would yield something very close to this. $\endgroup$
    – Andre Silva
    Feb 24 '14 at 4:13
  • 2
    $\begingroup$ Anonymity is far from the only difficulty with this proposal. The principle on this site is that all votes are of equal standing; I doubt the powers that be will do anything to undermine that, even indirectly. I don't think it's even necessarily true that high-reputation users are more likely to vote for questions more appealing to experts, but my guess is based largely on unsystematically watching what the highest-reputation users comment on and answer; as the question implies, we don't have the data. $\endgroup$
    – Nick Cox
    Feb 24 '14 at 9:15
  • $\begingroup$ @AndreSilva, I agree that the top of the ranked list would probably be close to identical with what we've already got. I predict the difference would be the type of questions that one would find in the middle of the pack. Admittedly, I don't think this feature would get constant use, even by power-users. And to Nick Cox, I think you're right that I'm fighting an uphill battle... but I would argue that be a serious violation of the principle that all votes are of equal standing. Vote counts would remain unchanged, it's just a proposal to aid question search/filtering. $\endgroup$
    – Michael Bishop
    Feb 24 '14 at 13:01
9
$\begingroup$

As @AndreSilva notes, this cannot be done within the SE ecosystem. It is important that voting be anonymous. As a result, who has voted on what cannot be accessed by anyone except the developers. You cannot get this information either directly or indirectly from the SE data dump, for instance. (For example, it is not possible to query and find out your progress towards the sportsmanship badge, because this would make it possible for someone else to figure out how you voted.) If the system you suggest were to exist, it would be at least possible to narrow down who is voting on what.

I'm not sure what you mean by a "high-karma user" exactly, but anyone whose reputation is >75 can promote a particular thread by setting a bounty. There are a variety of reasons related to your idea that can be cited, including to draw more attention to the question, to get a canonical answer, or to reward an outstanding answer. Questions that have active bounties are listed on the featured tab of the main page, and do get more attention. You can also navigate to the user page of whomever's opinion you particularly respect and see what bounties they have offered in the past, by clicking on bounties -> offered. We probably should encourage people to offer more bounties for questions and answers that they feel are especially noteworthy. I'll admit I feel remiss for not doing this more often. At present, the reigning champions seem to be @Glen_b and @chl.

$\endgroup$
5
  • 3
    $\begingroup$ I'd probably "reward an existing answer" substantially more frequently if there was the possibility of awarding a 20 to 30 point bonus for that particular purpose; I do worry about skewing things too much. I'd probably also do it more if I didn't have to wait until the question had been there two days; I often forget a post I meant to do that on. $\endgroup$
    – Glen_b
    Feb 23 '14 at 7:03
  • $\begingroup$ @Glen_b, I could have sworn it was possible to award lower amounts, eg 25, but evidently that is the amount automatically awarded by the system when someone offered 50 & didn't award it manually. Asking the devs for more options is a potential meta.SO question, but I doubt it would go anywhere. $\endgroup$
    – gung - Reinstate Monica
    Feb 23 '14 at 14:22
  • $\begingroup$ "a potential meta.SO question, but I doubt it would go anywhere" - yep; I debated trying some time ago, but I concluded the same thing. It would require an odd exception to the usual operation of bounties. $\endgroup$
    – Glen_b
    Feb 23 '14 at 14:58
  • 3
    $\begingroup$ In light of this thread, an interesting idea would be to be able to award bounties to questions as well as answers, however. That might be worth asking, but it also probably wouldn't go anywhere. Any time you want to ask the devs to change the SE system, that is a high threshold to cross. $\endgroup$
    – gung - Reinstate Monica
    Feb 23 '14 at 15:09
  • 1
    $\begingroup$ Bounties are relevant, so I'm glad you brought them up. But based on current usage, I don't think they can help sufficiently for the particular use-case I'm considering - when an expert wants to find questions of interest only to other experts. As I mentioned in reply to Andre's comment, I don't see a meaningfully degree of anonymity is lost with this idea. $\endgroup$
    – Michael Bishop
    Feb 23 '14 at 20:54
-2
$\begingroup$

I was thinking some such sorting method could even be invisible or implemented via user-localized, potentially customizable scripts. I'm out of my league on the programming side here, so I don't know how complex or dangerous (e.g., exploitable) an implementation of these ideas would be, but FWIW:

  1. Links to data feeds could be provided on meta-site posts.
    • This would preclude necessitating a main site UI overhaul.
    • Would facilitate limited testing – could even proceed by email invite in early stages.
  2. Downloadable computer scripts could generate organized hyperlinks in a separate program dedicated to displaying the proposed list sorting structure.
    • This could connect to the aforementioned data feed, which could be raw to cut down on UI development work.
    • I'm not sure how SE feels about web crawler bots, but this could produce some objections.
  3. SE as a whole could set up thresholds or more continuous weighting systems for votes by rep.
    • E.g., sort by votes received from users with above-average rep score, or weight by score.
    • Users could customize their own thresholds / weighting schemes, ideally.
      • To preserve anonymity, SE could restrict modifications to rough, e.g., decile increments
  4. Alternatives to weighting by rep could be considered, such as possession of badges like:
$\endgroup$
7
  • $\begingroup$ Any comments from the downvoter, or anyone else for that matter? I don't mind if you don't like these ideas, but I would like to address any specific critiques, and I don't see any completely intractable problems implied by the other answer(s for posterity?) or comments here. $\endgroup$
    – Nick Stauner
    Feb 26 '14 at 22:17
  • 2
    $\begingroup$ I didn't downvote this. But I can't support it. As a matter of principle, even implying in a limited way that some votes are more cogent than others seems contrary to the ethos here. (I know we have reputation, clearly.) As a matter of practice, your proposal hinges on what seem fairly arbitrary criteria. Again, we have arbitrary criteria all over the site but that doesn't make me fond of them. Finally, trying to quantify "interesting" in any sense seems doomed to failure. People should just decide what they personally think interesting and search accordingly. $\endgroup$
    – Nick Cox
    Feb 27 '14 at 9:36
  • 3
    $\begingroup$ Bear in mind, if you haven't already, that "On posts tagged feature-request, voting indicates agreement or disagreement with the proposed change rather than just the quality or usefulness of the post itself". To me it just doesn't seem worth the candle: like @Andre & Nick, I'd guess that high-rep users vote on much the same kinds of post as low-rep users (they only get one vote per post, & tend to vote on plenty), while I can already search for posts by user, tag, or key-words; there are other things I'd rather the SE developers spent their time on. $\endgroup$
    – Scortchi - Reinstate Monica Mod
    Feb 27 '14 at 10:10
  • $\begingroup$ Of course, and no objections to disagreement; just interested in the basis for it. Now that I've got some, I'd welcome more downvotes from anyone who agrees with your points more than mine. It's a good point about the limits on SE developers' time; easy to forget how small SE still is behind the scenes. I suppose this factors into my interest in helping work out the kinks in the idea for them ahead of time. $\endgroup$
    – Nick Stauner
    Feb 27 '14 at 17:43
  • $\begingroup$ Other @NickCox: I agree, and would not want to change the site-wide core ethos. I would only propose to facilitate the preferences of those whose ethea(?) differ with it, and preferably in a very limited, experimental way. Simply having access to summary stats of some kind (let alone the sorting ability in question) might provide interesting insights about whether differences in vote cogency or "interestingness" are quantifiable and useful. I respect your pessimism, and don't mean to cry "perfect solution fallacy" yet, but some see potential here. $\endgroup$
    – Nick Stauner
    Feb 27 '14 at 17:57
  • $\begingroup$ Without the data you want we are all speculating. My impression, and it is no more than that, is that SE developers won't buy tinkering of this kind on all sorts of grounds, and being too busy is near the top of the list. (Ethos is an abstract singular noun, but it can clearly differ from group to group.) $\endgroup$
    – Nick Cox
    Feb 27 '14 at 18:32
  • $\begingroup$ Agreed. Still, it's a shame to deny statisticians data to tinker with! I guess this is one of those ideas that's bound to languish in limbo unless it picks up popular interest someday (or gets a status-declined stamp in the meantime). $\endgroup$
    – Nick Stauner
    Feb 27 '14 at 18:47

You must log in to answer this question.

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