This might sound a bit crazy, but hear me out...

There's a certain behavior among voters on SO that, while understandable, makes down-voting incorrect or unhelpful answers somewhat counter-productive: the sympathy up-vote.

Try this: watch new answers, and wait for one that's particularly useless to pop up. Make sure it isn't blatantly abusive or insulting - just completely unhelpful. Wait for a few minutes, to make sure it isn't improved by the author or a wandering editor. Note that it hasn't been up-voted.

Now, down-vote it... And wait to see how long it hangs around with a -1 score before someone up-votes it. It won't be long...

It seems there are many kind souls on SO who feel that any answer which is not blatantly offensive shouldn't be scarred by a negative score, and who will act to counter any votes that result in such a scenario. Nothing wrong with that, in and of itself - everyone is entitled to their own opinion, and so long as they don't up-vote to the point where it sorts ahead of other, correct and helpful answers, no harm done...

...except for the disparity in reputation taken from and then returned to the person who posted the answer. See, it turns out that posting benignly unhelpful answers is actually a pretty decent way to pick up a few rep points. Since it takes five down-votes to counter the rep bestowed by a single up-vote, snagging an early down-vote on your unhelpful answer will almost assuredly leave you with at least an eight-point bonus once all is said and done!

Indeed, if you were so inclined, you could garner a tidy reputation by merely being consistently fast, and only subtly wrong. It's better to be lucky than good...

That is a shame. So, I propose that, so long as the current vote score for a given post is less than 0, the rep bestowed upon the author for an up-vote should no more than match the rep taken from him for a down-vote.

One down-vote: -2, one up-vote: +2, net score for the post: 0, net score for the user: 0.

Of course, an up-vote followed by a down-vote would still result in a net +8 rep for the author. That's fine - i don't see anything wrong with the difference under normal circumstances.

(reposted from UV, now that MSO is the official feedback site...)


7 Answers 7


It would be slightly more complex to engineer, but I've often thought I'd like to see rep bonuses and penalties as a bell curve, i.e. diminishing at extremes.

My thinking here is that it's not particularly fair that an simple answer on a high visibility subject can easily gain 20+ upvotes and be no more "correct" than a detailed, insightful, genius answer on something obscure which might get just 1 or 2 upvotes.

If it was my design, I'd allow a certain potential maximum rep change (plus or minus) for any question and weight the score around zero the heaviest so +10 might represent 90% of the potential value and +20 represents 100%. Attaching rep to the score of the question rather than the individual votes would also mean the overall view was important, and up and downvotes were commutative: +5 and -5 would give you 0 rep rather than 40 as it does now.

  • 14
    Bell curve rep bonuses is an awesome idea.
    – womble
    Nov 21, 2009 at 21:29
  • 1
    Definitely agree that the rep should be a straight multiple of the net votes. Apr 28, 2011 at 7:07
  • Although, I would weight as the net votes increase, such that votes further away from 0 have more effect; it's much easier to garner the first few votes. So +5 might give 10%, +10 30%, +15 50%, and +20 100% of the potential value. Apr 28, 2011 at 7:10
  • 5
    @Software Monkey - but that would punish the low vis questions even more than currently which is the effect I would like to see mitigated.
    – bananakata
    May 10, 2011 at 20:24
  • 1
    I'm late to the party, but I love this idea! It's a little galling to find 10K+ users who have just asked/answered a few banal questions which, due to timing, resulted in thousands of rep a piece. Often they're not even constructive and should have been closed, but it happened to be posted during the era when you could get hundreds of votes for "what's your favorite programming language?".
    – Ben D
    Apr 2, 2013 at 5:56
  • I'd actually love to see a model that showed the effect on people's reps. I imagine this could be set up pretty easily and I suspect it would put people's reputation more in line with actual site participation/contribution. It might be a little hard because of the daily rep cap, but it should still be pretty do-able.
    – Ben D
    Apr 2, 2013 at 5:59

As you say, this is due to the discrepancy between the amount of rep lost for downvotes and gained for upvotes. I've never really understood this discrepancy - I would personally rather have downvotes make a difference of -10, possibly with a cost to the downvoter of -5 instead of -1. This may discourage more "frivolous" downvotes which don't achieve anything.

Personally I reserve downvoting for answers which are incorrect or significantly misleading - and in those cases it seems crazy for a single upvote to counter 5 downvotes. Surely we should have a more serious form of discouragement for potentially harmful answers. (Or alternatively, you could think of it as encouragement to delete an answer which you posted in good faith but turned out to be inaccurate.)

I guess I'd like to at least hear more discussion about why downvotes are so relatively powerless.

EDIT: Thinking about this more, I suspect that one of the problems is that rep is seen in multiple ways:

  • A score in the "game"
  • An indication of how much the system trusts you (should the system trust you more for posting an answer with +1 and -4 votes than if you hadn't posted it?)
  • An indication of technical prowess / typing skills (even though it inaccurately measures these, it's still seen that way sometimes)

Having technical knowledge doesn't really mean that someone knows what's best for SO or has SO's best interests at heart - but we use it as a surrogate metric. It works pretty well most of the time, but it does make issues like this trickier to reason about.

  • 8
    I have considered increasing the downvote rep cost to -3 but I think -5 is a very large change Jul 1, 2009 at 6:52
  • Yes, that would indeed be a huge change. Possibly too big a change. It could well be that -10/-5 would have been a better pair of scores to start with, but that it's now not really a viable model to change to.
    – Jon Skeet
    Jul 1, 2009 at 6:58
  • 14
    The problem I have with that is that people would be reluctant to spend 5 rep to downvote something that's obviously wrong. Jul 14, 2009 at 20:13
  • 2
    @David: Possibly. I guess that's where it would help to have some way of having a "rebate" if you left a useful comment, for example. It's tricky.
    – Jon Skeet
    Jul 15, 2009 at 5:21

This isn't about rep, at least in my mind. It's about encouraging the effort it takes to write a question that some third person will get value out of some future day. The idea of voting for questions is that the questions that have more usefulness will be voted up, and less will be voted down. If gaming, or sympathy voting, or martians, persistently upvote awful questions, then the goal is not achieved.

  • 6
    Well, you can't really control how people vote. I don't think Sympathy Voting is a solvable problem: some folks will up-vote down-voted posts (just as others will down-vote up-voted posts), it's just the way they think. However, it should be possible to avoid the situation where some users manage to pile up rather large amounts of reputation on the site based entirely on asking scores of bad questions.
    – Shog9
    Feb 7, 2010 at 17:12

I understand your argument, Shog9. But doesn't this fine tuning make SO less understandable? We should consider giving new users not a hard time to understand the system. See also What's the single biggest barrier to entry on Stack Overflow?. Jon skeet's approach would help more, I think.

  • 3
    Down-votes don't really affect new-new-users - their rep can't drop below 1! But i would like to be able to use down-votes for their intended purpose, rather than up-voting every other answer.
    – Shog9
    Jul 1, 2009 at 8:30
  • 1
    That's not my point. I mean, if they got more rep then they will wonder why it is when reduced, increased, not increased. Depending on the order: downvote - upvote, upvote - downvote. Hard to figure out for a newbie. Jul 1, 2009 at 9:12
  • Ah, got it. Yes, good point - anything that increases the complexity of the reputation system will no doubt cause some amount of confusion, and resulting emails and questions for the SO team.
    – Shog9
    Jul 1, 2009 at 9:30
  • 3
    Complexity is precisely why rep should be based on net-votes for questions and answers. Net-votes positive == positive rep, net-votes zero == no rep, net-votes negative == negative rep. Maybe the factors for rep could be different for a negative vote score than positive, but it should be and should always have been a net-votes game. May 10, 2011 at 20:46

While the problem is well thought out, I'd be quite curious to see how true it is in practice. Honestly, getting upvotes for questions and answers (particularly questions) is not hard at all and I fail to see how someone could actually manage to get any sort of meaningful reputation out of "pity upvotes" - there are a LOT of much more powerful ways to game the system, this one just doesn't seem like a priority.

  • 4
    I think RichB and TheTXI are always on about this, basically a quantity vs. quality argument for users who ask (sometimes) HUNDREDS of bad questions that trickle in enough pity rep for them to become editors, etc. Jul 1, 2009 at 6:58

I don't think anything should be done. Simply put, rep is meaningless. It has no effect inside (except granting some extra abilities) and outside of the site. An up vote gives someone a nice feeling that what they are doing is appreciated, and a down vote with a small rep loss is an indicator to them that they may have made a mistake (and should go and fix it).

Essentially, gaining rep is what fuels users to the site. Getting more rep for an up vote provides a greater amount of positive feed back and encourages people to continue posting. The moment you increase the amount of rep lost for a down vote, the less people are going to post. You also have to remember that bad answers are punished swiftly and with great force (so a bad answer may get 6 down votes in quick succession). A 5 rep loss for active users (all those with over 10k raise your hands, I think I count at least 4 posts from people who fit that), isn't that much, but making a starting user lose 5 to 10 rep per down vote will quickly discourage that person to return. Essentially, this is all about rats pressing bars and you want them to press it as fast and as often as they can.

Even if answers are bad, you want to encourage people to post, because they will gently be moved to the bottom from negative votes. A person gaining 6 rep and losing 4 (1 up + 2 down) is little compared to that person posting a correct answer for someone else.

  • 1
    Rep fuels uses to keep posting to the site, but rep is not an indicator as to someone's abilities concerning programming. It's meaningless from the standpoint that it doesn't prove anything outside of how much someone participates on SO etc. Oct 26, 2009 at 12:49

I suspect there would be unintended consequences of this change.

Let's say we implemented your change (namely, that upvotes to negatively voted posts don't grant extra rep). Knowing that

  • a downvote is enough to deny anyone rep on a post
  • upvotes no longer give "free" rep to negatively voted posts

I have to honestly say I would be more likely to engage in "pity" upvoting than I was before.

I wasn't a fan of the practice precisely because it gave away free (and potentially unearned) extra rep and thus indirectly devalued my own. With this change implemented, you'd have single-handledly created a system where casting "pity" upvotes can be handed out without a concern.

As the OP said:

makes down-voting incorrect or unhelpful answers somewhat counter-productive: the sympathy up-vote.

This change makes sympathy up-votes far more likely to occur, as no "free" rep is given out.. so why not upvote any mean old negative votes you see? What's the harm .. other than destroying all sort orders.

  • 5
    Wait, so what's the problem then? That last sentence pretty much sums up what I was going for here...
    – Shog9
    Mar 20, 2010 at 20:27
  • 2
    So? Does it really matter if a question has a score of -25, or of -10 because of pity upvotes? I think in this case pity upvotes would only do good.. Okay, it's a bad question we get it, but there is no need to bash the user; -5 is enough to send the message. Mar 20, 2010 at 23:22
  • 2
    @shog9 so you want MORE pity upvotes to be cast? I don't... Mar 21, 2010 at 20:42
  • 4
    @Jeff: how I want other people to vote is irrelevant - that's a personal, individual decision. I'm merely suggesting that a certain voting pattern should, when it occurs, have less of an impact on reputation. You seem to think that there are people who want to vote out of sympathy but currently refrain due to the rep impact - that may be true, but there certainly seem to be enough for whom it's not a consideration...
    – Shog9
    Mar 21, 2010 at 21:24
  • @shog9 but your "solution" increases the likelihood of these votes happening. As these votes now have no effect, and it's nice to keep people in the 'let them earn rep!' category, I'd be upvoting every single negative thing I saw -- and why not? I'm not hurting anyone or devaluing my own rep any more by giving it out needlessly. That's not much of a solution in my book. Mar 28, 2010 at 5:27
  • 2
    @Jeff: again, I hold no hope that this solves the problem of sympathy votes. As much as I dislike it, I can't change the attitudes of the "plz be nice to teh newbs" crowd - they're gonna up-vote the down-voted regardless. Ultimately, this discussion gets into the realm of judging intent: are people down-voting because the post is bad, or because they don't like the author; up-voting because it's good, or because they feel bad about the negative score? Are votes being withheld now to avoid incurring sympathy? I don't know of any reliable way to answer these...
    – Shog9
    Mar 28, 2010 at 18:45
  • 7
    He's not implementing this to avoid losing a lot of rep from this answer =) Nov 26, 2010 at 23:44

You must log in to answer this question.

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