Since we've been having a big discussion about so-called "excessive voting" and I figured I would weigh in with a more robust idea that accomplishes the same goal, better and without the need to artificially hinder users who just happen to stumble upon a lot of bad posts: the deviant vote alarm.

Here's my idea: instead of going by absolute, arbitrary ratios, we look at votes on a weighted case-by-case manner: every vote that goes against the grain is a "deviant" vote. The more similar votes a post has, the more "deviance" the counter vote has.

In other words, if a post has ten upvotes, a downvote is deviant. Likewise, if a post has five downvotes, pity upvotes are deviant. The downvote is considered more deviant because it is more against an established norm. Votes for or against mid-ground posts cannot be considered deviant. Deviance is accumulated by users, and dissipates over time. Deviance in an of itself is not a bad thing; rather it's just a measurement of how often a user votes opposite of everyone else. There should be no automated restrictions on actions that create more deviance.

This way, both pity upvotes (against negative-voted posts) and malicious downvotes (against positive-voted posts) are captured and weighed using the same mechanism. Once a user has achieved a certain threshold of vote deviance, he is automatically flagged for a moderator to look into him (the deviant vote alarm). The moderator then uses his discretion to determine if the user is malicious or had legitimate reason for casting so many deviant votes. Singular deviant votes are not enough to trigger the alarm - thresholds need to be set so that false positives aren't commonplace.

  • There, if anyone upvotes this they are deviant! – Tyler Carter Mar 18 '10 at 17:08
  • +1 Now @Chacha is deviant! – Sampson Mar 18 '10 at 17:09
  • No, this would make the person who upvoted it after the downvotes the deviant @jon – random Mar 18 '10 at 17:09
  • @random Where was it indicated that deviousness was tied to sequence? – Sampson Mar 18 '10 at 17:11
  • First suggestion for a "Deviant" badge in 3... 2... 1... – gnostradamus Mar 18 '10 at 17:11
  • 6
    We should have a "Deviant" badge. – squillman Mar 18 '10 at 17:12
  • 1
    @squillman We could get DeviantArt.com to sponsor it! – Sampson Mar 18 '10 at 17:13
  • 1
    Strange, but weird. – mmyers Mar 18 '10 at 17:13
  • Here you go: meta.stackexchange.com/questions/42695/… – Tyler Carter Mar 18 '10 at 17:14
  • Deviance is indicated by votes coming after that go against the grain. As is spelled out in the post. @jon – random Mar 18 '10 at 17:14
  • 1
    @random So a downvote, followed by ten upvotes, isn't considered "deviant"? Seems like a design flaw. – Sampson Mar 18 '10 at 17:22
  • @Jonathan: I would say that the first downvote is not deviant, but the first upvote is. Subsequent upvotes are not deviant. Sequence is important, otherwise you could have a small vote war, with say 10 votes up and 10 votes down all around the same time (the question's vote total never goes out of the range -1 to 1), then someone makes an edit that everyone can agree on; suddenly the question has 100 upvotes. Those ten downvotes shouldn't be considered deviant, because at the time the question was still hovering around 0. – Welbog Mar 18 '10 at 17:25
  • Read the post at least once. You have it backward, it's the single vote against and following the ten in the other direction that is the deviant. @jon – random Mar 18 '10 at 17:26
  • 3
    @random: I still hold that that would be a design flaw. This assumes that malicious voting only takes place after swarms of legitimate voting. We know that to not be the case. – Sampson Mar 18 '10 at 17:29
  • 1
    @Pekka: As opposed to the current system in which Jeff just unilaterally decides that people who downvote a lot are bad users? – Welbog Mar 18 '10 at 19:09

Take this C++ question:

How can I write a macro that checks if the "this" pointer is null, that gets called in all my code?

And say 10, or 100 people respond with quite plausible solutions.

And then someone who knows what he is talking about says that if you are in a state where your implementation's this pointer is null, you are deep in undefined behaviour land, so it is pointless (meaningless in terms of the C++ Standard) to test for. And that person downvotes some of the most plausible answers (as in fact they are all wrong).

Is this person being deviant?

And this is not theoretical - this happens all the time on the C++ tags.

  • 3
    My thought too. This suggestions would make experts into deviants every time there was a truly deep questions. – dmckee --- ex-moderator kitten Mar 18 '10 at 17:34
  • That's why it's up to a moderator to figure out if the votes were legitimate or not. – Welbog Mar 18 '10 at 17:35
  • 1
    @NIfE Say what? None of the current moderators are C++ (for example) experts. How could they judge. – nb69307 Mar 18 '10 at 17:40
  • @NflE: that implies that the moderator has a deep understanding of the topic at hand. A Javascript coded isn't necessarily fit to judge a Java question, or a C++ coder a question regarding Python syntax style . – perbert Mar 18 '10 at 17:41
  • @You guys: or the moderator could e-mail the guy in question for an explanation. You don't need to be an expert to know if the guy's reply is an actual answer or some BS to support malicious voting. – Welbog Mar 18 '10 at 17:43
  • @NIfE Well "you guys" couldn't, as we don't have access to the poster's email address. And neither, I think, do the mods. And I for one do not want such access. – nb69307 Mar 18 '10 at 17:47
  • 1
    @NIfE: It seems to me that e-mail should only be used as a last resort. If I received an e-mail from a moderator asking me to explain my voting, I'd probably ignore it. – mmyers Mar 18 '10 at 17:47
  • 1
    I would have upvoted some of the above comments, except that for the first time in my SO career I have run out of comment votes! I really can't see the justification for a limit on these - do the Panjandrums (I'm reading "Anathem" at the moment) really expect us to have to switch to adding "I agree" comments instead of upvoting? – nb69307 Mar 18 '10 at 18:09
  • Guru votes should cancel deviance. e.g. 10 upvotes in a question tagged C++, I downvote -> I am deviant. Then you come and downvote too but you are a guru in the C++ tag -> You are not deviant, and I am not deviant anymore since a guru has agreed with my seemingly deviant opinion. – Daniel Daranas Mar 18 '10 at 18:30
  • 2
    @Daniel SO has no concept of a "guru" (apart from the tag with that name) and I hope it never does. – nb69307 Mar 18 '10 at 18:36
  • 3
    This scenario truly defeats my suggestion, so I am marking this as the accepted answer. I'm leaving the question up in case someone sees it and gets a better idea from it. – Welbog Mar 18 '10 at 19:20
  • Given enough answers, all questions are shallow. Or something. – Tom Hawtin - tackline Mar 18 '10 at 21:00
  • @Neil I mixed up the guru badge and the tag badges. I was talking about the tag related badge for having +400 votes. – Daniel Daranas Mar 18 '10 at 23:06
  • @Daniel So did I. I actually have two "guru" badges, but still don't think that should give me any special SO powers. – nb69307 Mar 18 '10 at 23:12
  • @Neil Butterworth: I've got both guru and C++ tag badges. Should that give me powers greater than my rep gives me? How about I retag a question as C++ myself? I think we're in complete agreement here. – David Thornley Mar 19 '10 at 13:41

The next natural step from there is a Lemming Alarm for people who always vote in lock-step with the herd.

All things in moderation, including conformity. ;)

  • 1
    I don't disagree with that, either. People who always vote with everyone else probably aren't very useful to the community, either. – Welbog Mar 18 '10 at 17:20
  • 3
    Well, SO has its fair (more than fair, some might say) share of lemmings. And I hate the horrible squeaking, suicidal, furry little creatures. – nb69307 Mar 18 '10 at 17:22
  • 1
    @Neil: the problem is that they don't commit suicide here, perpetually following the swarm of upvotes :) – perbert Mar 18 '10 at 17:39
  • 1
    You realize that whole lemming thing is a myth right? snopes.com/disney/films/lemmings.asp en.wikipedia.org/wiki/Lemming#Myths_and_misconceptions – Zoredache Mar 18 '10 at 18:29
  • 1
    I haven't finished reading your answer but it had a score of +3, so it must be good. +1. – Daniel Daranas Mar 18 '10 at 18:31
  • 1
    @Zoredache They are still horrible squeaking, furry little creatures. – nb69307 Mar 18 '10 at 18:42
  • @Zoredache: Yeah, I was aware. But they are still prone to large-scale and potentially-dangerous migrations, which is what I was referencing. – gnostradamus Mar 18 '10 at 18:48
  • @Zoredache: how do you know I wasn't talking about elizium.nu/scripts/lemmings ? ;) – perbert Mar 18 '10 at 20:06
  • I'm glad you didn't name your proposal "Lockstep Alarm". – lockstep Dec 9 '11 at 21:33

I upvoted this, because I think it takes an interesting perspective. This reminds me of controversial-patterns, but instead the controversy is a property of the vote itself. A single down vote on a post that has 3 times its weight in up votes appears to be controversial, or as you put it "deviant."

I think the idea could benefit from some discussion, but I like the direction and I think it would address much of the malicious voting patterns that take place on Stack Overflow. I often times do see posts with +5 or +10, and a single -1.

  • It does lend to some anonymity concerns, because in order to truly know if the user was being malicious or simply deviant you need to be able to see whether he voted. But the point is you cannot be malicious without first being flagged as deviant. Deviance isn't a bad thing: rather it's an indicator that something might be bad. I don't expect it to be implemented as I've written it, but I think we need to take a new perspective on this whole voting "problem", if there even is one. – Welbog Mar 18 '10 at 17:14
  • 2
    Sigh Why is a single downvote bad? I've in occasion downvoted highly regarded questions by the community, because they were either patently wrong or subtly wrong. Furthermore, that single downvote doesn't affect anyone in any meaningful way. – perbert Mar 18 '10 at 17:15
  • Perhaps it could be automated by having votes cost rep depending on their deviance: if you really must vote against the grain, you have to pay for it. Though, that might cause people to never cast deviant votes, which would be worse than malicious voting. I'm not sure. All I know is I don't like the current propositions. We need to look in a direction that isn't so Atwoody. – Welbog Mar 18 '10 at 17:17
  • @voyager: So have I. That's why one single deviant vote should never trigger the alarm. It gets triggered by systematically voting opposite what everyone else does. You'd set the deviance threshold such that only people who do it systematically will trigger it. That way the occasional against-the-grain vote won't get you flagged. – Welbog Mar 18 '10 at 17:19
  • @voyager A single downvote isn't bad. A single bad downvote is bad :) – Sampson Mar 18 '10 at 17:24
  • @Jonathan By definition anything "bad" is "bad". We are off in non sequitor land. – nb69307 Mar 18 '10 at 18:56
  • The idea of making deviant votes cost more is interesting. Courage costs, going with the flow costs less. But no serious malicious downvoter is going to be put off from downvoting by that. – Pekka Mar 18 '10 at 19:09
  • @Neil: Yes, tautologies tend to creep into conversation from time to time ;) The point of the tautology is to (again) point out that it's not "down votes" that are bad, but instead a specific kind of down vote. – Sampson Mar 18 '10 at 19:23
  • @Pekka: Okay, a moderately obscure C++ question comes up, and there's five bad answers, all of which have been voted up. I come along, and want to downvote the bad answers. How much rep is this going to cost me? – David Thornley Mar 18 '10 at 19:44
  • @David a dollar each. :) Seriously though, this has too many consequences to be workable. 1. The cost of a downvote would be subject to change if subsequent votes render it more or less deviant. 2. A checkout like "This downvote is going to cost you 4.5 reputation. Proceed?" would totally destroy the UI's flow which is one of the recipes of SO's success. – Pekka Mar 18 '10 at 20:15

You must log in to answer this question.

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