# How can we discourage downvoting when inappropriate?

Note the approximately equal opinions +16/-13 on this question. Clearly, this question divides our site members into approximately equal groups of opinions for downvoting. To those who would prefer a different tenor for the text of the question, please note that democracy is supposed to bring up issues worth discussing, which is not the same as having everyone agree about everything. Surely there are users who do not feel that downvoting is broken. Those users surely would not object to tweaking downvoting with clearly better alternatives to better serve those of us who do feel it is broken.

It has been my experience on multiple answers and even on closure votes on questions that unreasonable downvoting occurs like a flock of geese landing on a pond, more of a follow-the-leader activity, as contrasted to independent, original, individual thought.

Sometimes, on our site, this crowd behavior occurs because of a word that means one thing to a statistician and something else to many other specialists. Sometimes this occurs because a minor error is committed, and even when it is corrected, the downvotes remain.

When something that is just too original a thought is being developed, and the readers are guiding the development of that thought, progress is being made. This is usually downvoted, it generates interest, draws readers to that question, and usually rewards some other answer of a more conventional, boring, and incomplete type that "fits" within the scope of the average reader's training.

This is hardly the first time this phenomenon has been noted (e.g.: see here and here).

What can we do to discourage the bandwagon effect in CV SE?

• Should we limit the number of downvotes to one per day until the user reputation is greater than 1000?
• Should we increase the +125 reputation needed for first downvoting?
• Should we add a decision tree for downvoting like we have for flagging or close voting?
• Should we modify close voting to shift some down votes to close votes?

Your thoughts, as usual are welcome. Here is another example, view it soon as these things change when attention is called to them.

Edit:

Can we do something about this problem to make downvoting more useful?

Here are some new suggestions:

• We might require commenting for downvoting but include the option for making that comment anonymous. The motive for this would be to prevent the fear of retaliation to the downvoter, while enabling the downvoter to contribute.
• We might allow the downvoter to be notified when the answer is changed, so that the downvoter has the option of regaining his(her) +2 when the perceived problem (real problems or or even a misunderstanding of the downvoter) has been addressed by changing the content of the post.

In the final analysis, downvoting fulfills a role, the question here is whether we can improve that role for everyone's sake.

Here is another example of downvoting that makes no sense, that is off topic and is reminiscent of bower-birds collecting shiny objects as part of their mating ritual.

Yet another example of nasty negative reactions. The users spent all of their time and reputation points criticizing both the question, which is now on hold, and which was naive but not ridiculous. Lots of comments on how ridiculous the question is, but no downvotes. There was even a comment about the comments that no one bothered to answer the question. Enter moi into this regurgitation of furballs, and whatever I wrote got more downvotes. Meritorious? @gung didn't think so. What is a new user to think of this nonsense? The only person who bothered to try to help rather than spew negativity got clobbered for his efforts.

EDIT on 2018-10-21 Yet another suggestion, bear with me on this. Let's make a point with some quick and dirty pseudo-science...

The graph on the left is from users with mostly very high reputations. It shows two fairly obvious clusters. The left-hand group has a wider range of percentages of question votes versus total votes, and a low percentage (<9%) of downvotes. Let's call this group 'softies.' The right hand group in this left sided figure are the 'tough-guys' they tend to downvote much more frequently (>18%). Without being totally scientific about it, the fact that there is a clear distinction between these groups (with no values between 9% and 18%) suggests to us that we users, even the highest reputation rollers we have, have inconsistent approaches to downvoting. Next, let us look search for why this is. The figure to the right shows a crude 'information content' analysis. Don't fret, I am not trying for plu-perfect stats here. It shows that there is a significant correlation between log-downvotes per reputation point and log-percentage downvotes. Sure, this correlation is not very explanatory, but note that the clustering has disappeared. I think, and do correct me if I am wrong, that this suggests that those who more readily express negative information content (opinions) do so as a relatively increased percentage of their opinions.

Now, what does this mean? For one thing, from the viewpoint of recipients of these votes, inconsistent, i.e., erratic, downvoting is problematic. That is, it more depends on who is reviewing what than what is being reviewed. If you get a softie rather than a tough-guy as a first voter, you are in luck.

Finally, the new suggestion. We could make participation badges for kind-hearted behaviour. For example, less than 10% downvotes for 400 votes (bronze), 1000 votes (silver) and 4,000 votes (gold). This might give users a motive for not being hypercritical and help balance out the perception of arbitrarily hypercritical reactions to posts. I noted unscientifically while doing this 'pseudo-analysis' that the user self-descriptions that were the most 'in your face' belonged to the most negative reviewers, and that the emotionally mature reviewers and those who said squat about themselves, tended to be kind-hearted. Some useful advice: If you have to take an oral examination, and you have a choice between having a new, unsure of themselves, assistant professor and an aging professor who has 'seen it all' examine you, take the old fart, you are more likely to be treated fairly.

Edit 2018-11-01 Just to be clear, I am trying to be helpful. There are problems with downvoting and I am trying to be helpful. Example, this post of mine was finally deleted: https://meta.stackexchange.com/questions/316262/can-we-figure-out-a-way-to-pay-moderators. That occurred following -20 votes from what appear to be accolades of the status quo. Next, is an example of unexplained downvoting for a reasonable question. The answer to that question appears to be that current notation is elliptical enough that a "+" sign between random variables does not indicate whether those are paired outcomes or single outcomes, i.e., whether in the first case, convolution is invoked or in the second case, single outcomes, arithmetic addition is invoked. Regardless, the question continues to receive downvotes (now +16/-5). Would someone care to explain what the root of the -5 votes can be attributed to? (BTW, this is a Famous Question, i.e., 10,000+ visits. But, I finally managed to figure out why someone might down vote it; the question language was informal. I then posited the question more exactly, (only) one downvote disappeared, and several up votes were added in a sort time. So it can matter how one asks a question more than what the question actually is.)

What is occurring in society at large, e.g., political correctness amounting to slander and politically motivated violence, has an effect on our site. For example, this has impacted our site with the institution of politically correct (PC) commenting lest someone be offended, which PC answers are generally twice as long as the more curt versions. Weaponized downvoting is currently poorly controlled, and may become increasingly problematic with time. In searching my own behaviour, I cannot see an altruistic motive for downvoting; something else negative always predominates whether that be strategic downvoting on competitive and incorrect answers, or whatever.

I for one am swearing off of downvoting, take that pledge with me.

• Wow. Talk about dramatic. +1 because this was entertaining to read but calm down. The controversy badge sounds interesting, but getting a +5/-5 actually gives 40 points, so it seems like you already got rewarded for your troubles. And I don't know that slapping someone with a rioter badge because they were late to the party and weighed in on a terrible answer is such a great idea. Finally, how would limiting an individual's number of downvotes remedy this problem? You can only downvote an answer once and any individual's vote limit has nothing to do with this "mob mentality". – gammer Jan 29 '17 at 21:46
• @gammer I am not sure if I should apologize for conveying emotion, or rejoice that I did. Aristotle commented that emotional speech holds a crowd's attention. Mobs are not composed of the same individual multiple times. Usually a crowd turns to a mob on a dime--and mass behaviour; a riot, is not rational. An impassioned plea for rationality is a stark contrast, but it is what it is. – Carl Jan 29 '17 at 21:59
• OK. More related to the question, my impression is that mass downvoting of questions and answers usually happens due to extremely low quality. So, similar to what I said in the other recent meta thread from today about downvoting, I'm not sure there's a real problem here to be solved. The kind of mob behavior you wrote so eloquently about does not seem to be prevalent, as far as I can tell. In general, voting behavior on this site is idiosyncratic and, ultimately, the consequences are almost non-existent, so maybe try to brush it off. I'll be interested in what other discussants say. – gammer Jan 29 '17 at 22:10
• The more you talk the more this sounds like a rant rather than actually seeking a solution. Do you have an actual suggestion, or is your question basically "Can you all please stop voting like brainwashed idiots!?!?". Pun intended. – gammer Jan 29 '17 at 22:48
• I find 114 questions w/ a net score <=-3, out of 94,870 for 0.12%; & 160 answers w/ a net score <=-3, out of 92,626 for 0.17%. This behavior does not seem endemic to me. – gung Jan 30 '17 at 13:54
• @Carl, your answer is not deleted. It's downvoted because you made a specific proposal and people disagreed with it. That is what Meta discussions are for. – amoeba Feb 1 '17 at 11:42
• ... I can't see the evidence (consider perhaps the three answers & many comments on this thread). – Scortchi Feb 1 '17 at 13:47
• @Carl, yes, you need to provide evidence when your opinion assigns a malicious intent. Placing opinions like that on the same footing as opinions that don't assume there are mobs out to get you leads to dangerous consequences (seemingly very similar to the kind of consequences you're railing against). I hope that the political situation in the US doesn't devolve further to "provide evidence" of why what I'm saying is true :( – not_bonferroni Feb 1 '17 at 17:14
• Thanks for explaining, but I fail to grasp the argument. I don't accept that up and down should be 50:50. We don't say that 50% of people should have AIDS or red hair rather than not, given a binary classifier. I can't tie your suggestions together. You seem to be very unhappy with the way downvoting is used, but also seem to be suggesting that there should be much more downvoting. I can only imagine that CV would just fall apart with massive downvoting. I wouldn't want to be part of a forum if I got negative signals for my posts about half of the time. Am I missing your point completely? – Nick Cox Feb 3 '17 at 18:01
• (I'm very reluctant to respond further, as I suspect it will just draw out more arguments of the same character. However, I will lay out the implied context for my comments above for clarity.) From the question statement, I inferred that large numbers of downvotes is considered to be a commonly occurring problem. To see how common it is, I computed the proportion of posts (on the main site) that have a net score of -3 or lower. The proportions are very small. I take that as prima facie evidence that large numbers of downvotes do not occur very often. – gung Feb 3 '17 at 20:52
• I can honestly say that what would convince me it happened often is if the rates were high. 'High' is somewhat vague, but certainly 10% would seem high to me & <.2% seems low to me. The 1.5% rate for Qs on meta.CV is interesting, but downvoting on meta has a different meaning & we're only referring to 17 actual Qs. @NickCox makes a good point that if downvoted posts are more likely to be deleted, these would be underestimates, but I don't see good reason to believe large numbers of downvotes happen very often. I am not trying to build a classifier, so I'm not concerned that 1 might be biased. – gung Feb 3 '17 at 20:52
• @mdewey It's a puzzling thread. But I don't think it helps the assertions being made in this one. Unreasonable downvoting as a diagnosis would be the right diagnosis if there was e.g. lots of downvoting when an answer is agreed to be right (not always easy to establish); or people downvoting just because somebody else is (very difficult to establish). The weakest kind of example has the flavour of "I think I was right in that thread; but my answer was downvoted; therefore people are being unreasonable" which by itself is just assertion, not demonstration. – Nick Cox Feb 19 '17 at 20:52
• +1 I have little interest in reading the barrage of comments to this thread. Regardless, I want to add one of my own. I have to agree with Carl about the bandwagon effect. It is a widespread bias across many, many online sites and blogs. CV is neither immune nor is it an exception where "reason" rules. There is a lot of irrational behavior here, e.g., moderators who, after losing a discussion, delete the thread reflecting their mistaken line of argument, participants who revel in baiting other participants into arguments, critical comments that do nothing to suggest a solution, etc. – DJohnson Mar 20 '17 at 23:27
• @Carl It is difficult to add helpfully. But I have to suggest further that a facet here is the chicken-and-egg problem that you have done this quite often -- lash out against down-voting with colorful deprecation. These events do tend to register. Hence people may wish to keep their anonymity to avoid being identifiable targets for your scorn. That is no more than speculation, just as are the suggestions that anonymous downvoters were not thinking independently and/or not of very high reputation. There is, self-evidently, no evidence on that. – Nick Cox Apr 9 '18 at 6:13

I have only a little to add to @NickCox's answer.

Down-votes do tend to be sticky, which is a pity when the down-voted poster makes the effort to improve their post. I don't see any reason to suppose this is due to anything more than people's disinclination to keep on returning to a post they've down-voted to see whether it's been improved. There are some feature requests on Meta SE that you might be interested in:

I'd like to see something of the sort implemented. On the other hand I have the impression that for "controversial" posts with several down-votes & several up-votes it's often the case that the up-votes came after an improvement, & after the down-votes—that people are trying to rectify a now-undeservedly low score.

• (+1) I agree with you, and you have moved the issue forward. – Carl Feb 1 '17 at 6:23

EDIT: Here is a summary added also in reaction to various comments (some now deleted).

Downvoting is not constructive (and not intended to be). It should, mostly, express a view that a post is not useful, although whatever other reasons or motives people may hide remain indiscernible. Downvoting is not informative unless people explain why they downvote. Downvoting is part of the SE system and anyone is free to dislike it, not to use it, to use it very sparingly, or to express opinions on it, or on how it is used. How downvoting is used on CV is clearly germane for this Meta, but trying to argue against or change the system is a SE-wide matter. I don't see evidence that downvoting is out of control here on CV. I see plenty of evidence that some individuals get very upset at being downvoted, and I have some sympathy with that personally, but it is, given human nature, an inevitable side-effect of SE.

I'd be worried about "trolling", "ganging up", "mob behaviour" and "unreasonable downvoting" if ever I saw evidence of them, but I don't think we need any new mechanisms for dealing with bad practices.

I find it disconcerting that some in this thread are so ready to infer people's state of mind, personally or collectively, from the fact of downvoting. Occasionally people do leave angry or nasty comments when they downvote. Then their attitude is plain, but that is out of order and should be handled by reminding people of the need for civility and by flagging bad cases for moderator attention. (Naturally, there is scope for disagreement about precisely what is angry or nasty. I think of myself as often candid but not nasty on purpose; who does?)

But often people do not add comments. When that happens, why they're downvoting is no more than a matter for surmise.

I can tell you my typical state of mind when I downvote a question or an answer after someone else has. It's that I agree with the previous downvoting. The question or answer is just wrong, or too poor to be acceptable. And I am not surprised when other people agree in turn. The polite, parsimonious, default assumption about downvoting is that people are exercising their privilege in good faith and with good reason. Only when you see explicit evidence to the contrary is action needed.

Apart from lack of research effort as grounds for downvoting a question, the main reason for downvoting anything is that it is not useful. To me, that certainly includes being unclear; if something is unclear, it can hardly be useful. It's clearly true that downvoting by itself is not constructive, but often when a post is unclear, it will be true that a downvoter will not be in a position to suggest any specific improvements.

That's not to deny that, like almost anybody else, I do get irritated or annoyed to myself at certain posts or even at certain people.

I won't discuss @Carl's detailed proposals here. They seem to range from the serious to the facetious and it's a small imposition on Meta to be expected to work out which is which. (I don't want people to leave their sense of humour behind when they enter Meta, but proposals for change have to be serious to be taken seriously.) The more crucial point, already made by @amoeba, is that new badges or changed limits on anything are system-wide matters and need to be floated on SE meta.

I have looked at some of the threads @Carl linked to and cannot see anything untoward there. Frankly, @Carl, the pattern I see in these and some of your other posts is this:

First, several of your questions are rather long-winded and come from a very individual perspective. You're proud to be a maverick! No problem there, but even statistical experts can find it hard to see what the question is and the comments can be as or more extensive as the answers. The biggest difficulty there is likely to be your frustration at not getting the kind of discussion you want.

Second, you do write vigorously and colourfully, but seem prone to exaggerated language and occasionally snap and snarl a little. Sometimes that is very entertaining; occasionally it is not constructive. A pugnacious style is sometimes a spur: people don't always back off, but give as good as they get.

But I can't see evidence that people are downvoting you unreasonably.

A recent thread was closed and the OP protested, although not in these words, "The question is clear to me. Why is it being closed as unclear?" The point, naturally, is that a question has to be clear to others to be any use or interest.

In a forum, people will disagree and can and should express their disagreement in various ways. This happens through downvoting, which is not the tool of first preference but exists to be used when appropriate. It's designed by the system and needs to be practised to maintain a healthy archive, weeding out or limiting less fruitful questions.

(I don't like being downvoted either: who does?)

EDIT As at 10 November 2018 the question starts

Note the approximately equal opinions +13/-12 on this question. Clearly, this question divides our site members into approximately equal groups of opinions for downvoting.

This comment on the current Meta thread reflects the entire issue under discussion.

No; it doesn't show anything except that up and downvotes for the question are about equal. My answer pushes what I see as a key negative principle: Unless people explain a downvote, you can't reliably infer what lies behind it. And people aren't always candid even then. (In the case of one person no longer active and an answer of mine, they gave generally positive comments but then promptly downvoted it. I could tell this because there were no other downvotes and that person received a badge for their first downvote on the same day. This was a person who often ranted against being downvoted without explanation, but themselves gave many more downvotes than were ever explained. But that's rare.)

In this case, I can readily imagine all sorts of different reasons for up and down votes.

• Bang on!

• We should be discussing this, even if there really isn't a problem.

• I found this provocative.

• I found this entertaining.

• There isn't a problem.

• There may be a problem, but this doesn't help.

• I don't like the style of this.

• I don't agree with the specific proposals here.

And so on. All those reasons are pure speculation (but to my mind can't be ruled out either, given various answers, comments, etc. and what is evident generally from members' attitudes and behaviour).

Then again, @Carl writes

Surely there are users who do not feel that downvoting is broken. Those users surely would not object to tweaking downvoting to better serve those of us who do feel it is broken.

No "surely" about the second even granting the first, so far as I am concerned. (Surely those in favour of peace wouldn't mind a small war, given those who want one? Surely those in favour of letting people decide what they do on X wouldn't mind banning this and that aspects of X given people who disagree?)

If downvoting is not broken, there is no reason to fix it or even tweak it. In total abstraction, I can readily agree that some downvotes are not deserved, or given for reasons that aren't admirable, or given for the wrong reasons, but who is to say which they are? Almost all I know about are my own downvotes.

This thread is saved from ill-will and rancour because @Carl comes over as a well meaning and thoughtful person, perhaps despite his colourful and combative style, but we have to agree that there is evidence of inappropriate behaviour before we can decide what best to do about it.

• (+1) If anything, people gang up to upvote questions and answers. I have no complaints, but it's easy to find answers that clearly deserve far more upvotes than they have received. Gosh, the world seems so unfair... . – whuber Jan 30 '17 at 15:57
• "I find it disconcerting that some in this thread are so ready to infer people's state of mind...from...downvoting." Clearly, downvote means "I do not like this." On many websites, this is "dislike." One cannot read anything else into it. In fact, that is the problem, it is ignorant. "I enjoy being a 'long-winded' maverick" Do tell. Now, who is reading what into whom, come again? – Carl Jan 31 '17 at 5:34
• Quite different points. (a) I (you, we) don't really know what lies behind a downvote that is not explained. (b) I think you express a lot about yourself in your many posts and that it is behind some of the reaction to them, including some downvotes. That's what I am trying to explain, very likely not clearly enough. You're absolutely right that "dislike" is used elsewhere but it's not part of the SE terminology (or attitude). – Nick Cox Jan 31 '17 at 8:24
• What's necessarily "ignorant" about a downvote? Someone downvoted this answer and didn't post a reason why. That's their right; but why presume or imply that they don't know why they are doing that?. I am ignorant about their reason, to be sure; that's where the ignorance is. Do I misunderstand your stance here? (FWIW, I didn't downvote either your question or your answer here; you're raising issues that concern you, which is what Meta is for; I just disagree that there is a real problem.) – Nick Cox Jan 31 '17 at 8:52
• @NickCox I was not personifying ignorance. A downvote is precisely one bit of information, that is, to my way of thinking, not enough information to justify the reputation damage because it ignores what prompted that downvote. – Carl Feb 1 '17 at 5:45
• If you can read the deleted text below which has concrete suggestions as to changes to downvoting, perhaps you will then understand why they were not given above. They were downvoted solidly, which is ironic, and an example of what I am saying about downvoting. Downvoting obviates bringing up unpopular issues, it destroys serious discussion, logic or science, it is "flat-earth". – Carl Feb 1 '17 at 5:57
• I can read your answer, which is below as I write. To me -- this may be reputation-dependent -- it is not deleted, but shown in fainter type. It was, again I have to guess, downvoted because people disagreed with it. They had to be able to read it to do that. Downvoting doesn't stop people bringing up unpopular issues: you did that and people who've voted disagree with you on balance. This isn't flat-earthism but democracy in action (and I write as someone dismayed by most recent election results). – Nick Cox Feb 1 '17 at 9:05
• I naturally do agree that it's more informative if people give their reasons for downvoting and I often wish they had, so long as they explain in a civil manner. Whether people should be obliged to do that is a repeatedly contentious issue on SE Meta, which is where this discussion really belongs. Picky point perhaps, but I think "ignorant", "ignores", etc. are poor word choices for what I take to be your view here. From the data I have seen, people reacting strongly to occasional minute reputation losses tend not to last long here. – Nick Cox Feb 1 '17 at 9:11
• Further factual detail: The answer is not currently deleted. People with high enough reputation (10000+) can vote to delete it. I won't do that; whatever the appearances, I am in no sense hostile to this issue being discussed. – Nick Cox Feb 1 '17 at 9:25
• @NickCox You said, "This isn't flat-earthism but democracy in action..." I mean no offense by characterizing that as a shallow opinion reminiscent of a slogan from some uninformed civics class teaching. Democracy is functional only when population is so sparse that a mob cannot be formed in a population. Otherwise, we form representative governments as kingdoms, republics, or dictatorships so that mob rule is filtered by, optimistically put, "sober second thought." Downvoting may cause me to rephrase a question or answer, but the downvotes do not tend to disappear, they persist. – Carl Feb 18 '17 at 18:34
• @NickCox con't Moreover, special interests are often represented directly to governments, which action can be both beneficial or damaging, but is also not something that can or should be obviated. The equivalent of special interests are the more expert opinions on this site, and, like it or not, those opinions are worth much more than a single democratic vote, and would not be properly weighted in an uninformed democracy. So, democracies evolve, first they allow issues to be presented by experts, and eventually those experts are elected to rule, and the democracy is then no longer pure. – Carl Feb 18 '17 at 18:42
• @NickCox Finally, let me offer my opinion on downvoting. It would be better if I could evolve my opinions by having downvotes removed subsequent to the evolution of those opinions. It is the permanence of those downvotes which force final opinions, often expressed in terms that embrace popular misconceptions to be presented as authoritarian "final" draft answers as contrasted to the scientific approach of floating hypotheticals for discussion and evolution. This latter is not maverick, it is experimental. The current structure supports canned wisdom without the "scaffolding"... – Carl Feb 18 '17 at 19:05
• @NickCox that Carl Gauss would erect during construction of a proof to leave only the "completed edifice." So, quotable but not process informative, authoritarian but uncreative, enviable but not emulatable, instructive but not instructing results. In the final analysis, current downvoting works against the construction of thought on this site and works to promote "current" thought. A lot goes missing in that malformed process including the joy of polishing a rough diamond, the elan of creation, the shock of discovery, and indeed, the brass ring. – Carl Feb 18 '17 at 19:19
• @Carl Describing what I said as "shallow" and "uninformed" was not a good start to encourage me to engage. You're riding your own hobby-horses in several directions: much goes far beyond CV and far, far beyond what I think typically happens on CV. It seems over the top to me to read so much deep and far-reaching significance into the fact that people here sometimes disagree with you or just can't follow your argument to find it helpful. Naturally I mean no offence by saying that. CV is just a forum bringing together people with various statistical interests, not modern democracy. – Nick Cox Feb 19 '17 at 0:20
• At the time of writing, 3 people had downvoted this answer and none left a comment saying why. Seriously, I and we don't know what their argument is or objections are because they haven't declared any. More facetiously, are these people ganging up to say that ganging up is a problem? That would be ironic, perhaps. – Nick Cox Nov 13 '18 at 8:35

I for one am swearing off of downvoting, take that pledge with me.

Just for the record: I have (long time ago) downvoted this question, two (!) of your answers, Alecos's answer, and the answer by SmallChess. I have upvoted Scortchi's and Nick's answers, and have now also upvoted your most recent answer for its nice figures.

Many other people have voted similarly and this is why we have two excellent answers appearing on the top of this page while less satisfactory answers appear below and some of them are even grayed out in the very bottom.

Thanks, StackExchange, for providing such a reasonable mechanism. Personally, I am certainly not going to stop downvoting.

• Привет. Useless to downvote spam compared to flagging it for deletion. My aim is to make downvoting more relevant on the correct/incorrect axis by coming up with other procedures for deleting otherwise inappropriate posts. – Carl Dec 10 '18 at 18:49
• The answer from @whuber at stats.meta.stackexchange.com/questions/5526/… specifically encourages downvoting complete garbage as well as flagging it. while emphasising that flagging is the most important thing a member not a moderator can do. – Nick Cox Dec 10 '18 at 19:58
• downvotes also works as signals to users who can VTC. I often do that after seing downvotes on the frontpage. So those downvotes are usefull – kjetil b halvorsen Dec 15 '18 at 20:13

Some research to examine downvoting was performed. The top 500 CV users with respect to reputation were examined with respect to clusters of downvotes using visual inspection then k-means for two populations performed with Mathematica v. 13.0.0.

This shows three clusters. Note that reciprocal scaling was used on the $$y$$-axis to produce a dependent variable distribution that was indistinguishable from a uniform distribution. Also, square root transformation of the $$x$$-axis data removed much of the heavy right-tailed appearance. Visual inspection was used to identify the zero percentage downvoter population as cluster analysis proved problematic for that purpose. Subsequently, k-means did a good job of isolating two populations that were relatively independent of reputation level.

Addendum Given the copious feedback, I did further processing. @NickCox suggested using folded square root of percentage downvotes, $$\sqrt{x}-\sqrt{100-x}$$, on the $$x$$-axis. This worked well in SPSS in version 13, (old but good), using two-step cluster analysis followed by k-means for a variable number of clusters with the best clustering chosen by BIC. This show the following plot. Two-step cluster analysis showed that classification by reciprocal reputation failed $$t$$-testing. The cluster data forwarded to k-means then yielded two clusters as below.

This has similar cluster results to those achieved prior. Once again the visually identified clustering at 0% downvoting is obvious, but in this case was selected post hoc. The percentage breakdown is

@NickCox suggests that downvoting is a method of removing low quality posts. Therein is a big hint as to what is going on. What's in a downvote? On a social media site based not on evidence but personal taste, voting is like/dislike. On an evidence based site, more like ours, a vote should, to my way of thinking at least, more plausibly suggest correct/incorrect material. What appears to me to be lacking, and needed on our site is some method of indicating material that is contributory/non-contributory to the site. For example, suppose a question is Why does $$2+2=4$$?. Obviously, this is not on the correct/incorrect axis but it appears to add little to the site's compendium of knowledge; it is non-contributory. Thus, currently downvoting is a mixture of two axes "like/dislike" and "correct/incorrect". That, I think, is what allowed k-means to identify the split in population at circa 12% downvoting rate. Is there any evidence to support this? The SPSS histogram resulting from the three clusters, one of them identified visually at 0% is...

Next, we ask if the non-0% clusters follow some recognizable distribution type. Mathematica's FindDistribution routine was applied to this data and it selected two Weibull (squared) distributions as the most likely fits from its extensive library of distributions. That is shown next.

Weibull distributions also best fit the original downvote data without square rooting. No matter how I tried to fit the histogram data, the results were not significantly different from Weibull distributions. In absolute terms, this cannot be a Weibull distribution as such a distribution has values beyond 100% downvoting. However, the Weibull result may be relevant as one interpretation of the Weibull distribution relates to diffusion of innovation. As per Weibull "In the context of diffusion of innovations, the Weibull distribution is a "pure" imitation/rejection model...In the context of the diffusion of innovations, this [Sic, $$\alpha>1$$ (As it is herein), where $$\alpha$$ is the Weibull shape parameter] means positive word of mouth: the hazard function is a monotonically increasing function of the proportion of adopters. The function is first convex, then concave with an inflexion point at $$1-e^{-1/\alpha}$$ [Sic, notation translated]"

Thus, there is statistical evidence for two groups of adopters beyond the 0% group, who are "non-adopters". This is evidence for an admixture of downvoting strategies that does not agree with the opinion as voiced in some comments that such a phenomenon does not exist in this data. Such an admixture of voting patterns for downvoting is opaque to the recipient of such votes. That is, although we can sort this out statistically, the individual downvote recipient has no such clues, which implies that downvoting is not efficient and not clean from the user's POV. One remedy for this could be to shunt some of this downvoting in the high downvoting rate category to a closure vote on the contributory/non-contributory opinion axis as this would provide clarity as to what a downvote means for the user while doing a more direct job of addressing material that is non-contributory for the more expert reviewer, as follows.

The selections off topic, too broad and primarily opinion based are currently being used as alternatives for not-contributory. Those alternatives, even taken together, do not span not-contributory such that current practice is not really addressing the need to prune the site for content that is just not up to par value for expert opinion; low quality, not expert question and so forth.

For answers, there is no closure voting per se. However, there is flagging which provides similar functionality, but requires review at my reputation level.

I will continue editing and collating to de-bulk and organize a bit better.

• Nice job getting the data, but I don't see any evidence for two distinct clusters. Apart from a spike at 0%, the rest of the distribution looks unimodal to me. – amoeba Dec 5 '18 at 11:34
• Strong and sincere upvote (!) because you are throwing some data at the question. But your interpretation is, as you will recognise, one of many possible. Lots of people never downvote: I guess it's their personal rule not to be negative; I admire that in individuals, but they are doing nothing thereby to clear CV of poor questions and answers. Like @Amoeba I can't see strong clustering here, just a continuum, no surprise given impressions of people generally. Technical note: I am a great fan of transformations: here folded root sqrt(x) - sqrt(100 - x) is an alternative for your x axis. – Nick Cox Dec 5 '18 at 13:22
• @amoeba Unfortunately, I couldn't get my SPSS to work which would allow choice of number of k-means clusters using least BIC. I did however try clustering in a dozen ways including methods for which the number of clusters is a variable. A single cluster was achieved only once, and only when the population at 0% downvotes was included. Obviously, cluster identification is a bit artsy, for example, with less data and higher reputation, look at my question EDIT on 2018-10-21, clustering is more obvious. A single cluster histogram has a dip in it at ~10% downvotes, and sometimes software... – Carl Dec 5 '18 at 20:22
• ...sees something the eye does not. That is why I left the choice up to k-means rather than my intuition. Even if one does not "see" three clusters, I think it would be less explanatory to mix the moderate and high downvote rates together, the medians are different by a factor of 6.8 times. – Carl Dec 5 '18 at 20:29
• @Carl "a downvote is rather easier to attribute to a random event than a rational one" it is a bit absurd to argue that downvotes have nothing to do with the post and everything with the random selection of people that happens to review a post. This would mean that votes received follow a Poisson distribution and do not cluster at people and/or posts. You should check data.stackexchange.com/stats/query/941119/most-downvoted-users – Martijn Weterings Dec 6 '18 at 0:44
• also, if I look at down-voted posts than most often they are bad quality. This is why most down-voted posts are eventually deleted (I will check this statement one day by adding up all downvotes on posts that actually exist) – Martijn Weterings Dec 6 '18 at 0:45
• how-much-downvotes-get-deleted Total downvotes that have been given (sum downvotes from users). Plus total downvotes that are not deleted (by summing downvotes over posts that actually still exist). More than half of the downvotes are still remaining. Of the 48k downvotes that have been given only 17k are actually still on existing posts. This means 31k downvotes have been on posts that eventually got deleted (thus deemed unimportant by the people on this website). – Martijn Weterings Dec 6 '18 at 1:15
• also note that some of those user's in your graph, with high percentage of downvotes, actually vote very little in absolute numbers. There are a few bulk voters (half the downvotes come from ten users), among which community somehow does 6k votes or almost 1/6 of the total. Among these 10 users are some high rep users but they would fall, as you say, in the below 10% group (also in terms of votes per day they do not score high, they are just the dinosaurs of this website). The second highest downvoter with 4k downvotes is a suspended account. – Martijn Weterings Dec 6 '18 at 1:30
• @Carl, sure one may experience the downvotes that one receives as a random events. But that does not make the downvotes less rational, as if people are randomly making downvotes without reason. It is also very unclear how you see this as a consequence of the large variance/spread among the percentage of down-votes among the user's votes. Without this spread it is already obvious that voting has a strong random component. Posts do not get as many votes as views, and they have some probability to be voted up or down depending on a complex combination of factors, it's still rational however. – Martijn Weterings Dec 6 '18 at 8:39
• @Carl, regarding the downvotes that you personally get. Posts like this one is how many of your post go. You put a lot of information out there, but the conclusion is small and sometimes weak and without a good connection with the stuff that you write down. While the complex thoughts behind the conclusion may be amusing and interesting it is also very verbose, confusing and all over the place. In this post, the first half is some cluster analysis and second half is a comments to Nick Cox, the conclusion does not connect well to the rest of the post and is difficult to distill out of your text. – Martijn Weterings Dec 6 '18 at 8:57
• I did not get that from your answer at all. So your conclusion is that we should have some close voting on answers in a similar way as we have on questions. You should make that much more salient in your answer. The cluster analysis (which is dubious) takes this out of sight. You are right that this 'prevents' down-votes. Or at least for me if I feel that if a post/question needs correction, then I vote to close instead of downvote, partly as well because of the stickiness of downvotes. – Martijn Weterings Dec 6 '18 at 12:10
• We could turn all this round and point to variations in fractions of upvotes and opine that some voters are being too generous and should rein in their positive inclinations. It's interesting that @MartijnWeterings raised the analogy of grading because he captured my feelings about grading. A failed submission is a personal tragedy but being a nice person, not wanting to disappoint or upset, etc., is neither here nor there; consistency requires that low marks be given to work that deserves them. – Nick Cox Dec 6 '18 at 15:24
• Folded square root isn't log(x) - log(100-x) (which is just logit on percents). As your results start at -10, corresponding to 0, it's clear that you did use folded square roots and the references to log are typos. – Nick Cox Dec 7 '18 at 9:51
• @amoeba I now present stronger evidence for multimodality. Beyond the 0% peak the data is not unimodal, but approximately bimodal Weibull distributed. – Carl Dec 8 '18 at 10:25
• See for instance this image obtained from data with this query. The users that make relatively more up-votes are also users that make relatively more comments. Possibly you can add more variables and time (to track a change in tendencies and general behavior) only then can you really make stronger conclusions with this data. But, at least this image shows some alternative (sidenote: the fraction downvotes looks like a logit normal). – Martijn Weterings Dec 11 '18 at 17:15

I was one of those @Carl mentions. When I see a downvoted answer, I assume it is a stupid answer. I downvoted @Carl's answer on simple linear regression in machine learning before I even read it. I did that simply because I saw -2 to his answer.

• I applaud your honesty in saying that's what happened, but do please try to avoid this behavior - downvote because you've been able to conclude that the answer has a problem (i.e. you should be able to explain why you think the answer is wrong or otherwise problematic - and ideally, actually do so when you downvote) – Glen_b Mar 21 '17 at 8:28
• I think telling other people how to vote on questions is a total waste of energy. That applies to the entire premise of this question, and to the comment from @Glen_b. You can't control how people vote (both in a general sense, and in a practical sense). Let's just accept the vote counts on this site for what they are - a difficult-to-interpret point system based on a mixture between rating the actual quality of the content and more base motives - and move on. – GoF_Logistic Mar 21 '17 at 15:42
• @GoF_Logistic My comment relates to the top answer to the FAQ question "When should I vote?". It says "Try not to let other factors like the vote score and other answers influence your vote" and "...try to ignore the existing vote score when voting." and goes on to list reasons to upvote and downvote. Explaining what the policies are is part of the moderator's tasks, whether you regard it as futile or not. – Glen_b Mar 21 '17 at 15:58
• Sounds like more of a suggestion than a policy. I'm sure everyone will try to keep it in mind. – GoF_Logistic Mar 21 '17 at 16:20
• What this site could be is a place to develop new ideas, and it is missing that opportunity by supporting conformity over disturbing novelty. In my case, this site tries my patience at times because when I identify certain points of dogma as flaky, and they often are, I wind up dealing with irrelevant asides. e.g., outrage, and other ill-considered rants. Missing in that is positing dogma hypothetically, i.e., the scientific method. – Carl Mar 24 '17 at 20:19
• (-1). I am surprised this answer got four upvotes, probably because the OP was sincere and the answer is indeed relevant to what was asked here. I downvoted it based on the assessment that its content shouldn't be followed/rewarded. This is because it reflects an automatic voting pattern, like the one of a robot. Obviously, voting like this can lead to good content being downvoted and also the opposite (as suggested in this question). I agree (from other meta posts) that how people vote is personal, but I support at least reading a post before voting for it. – Andre Silva Dec 28 '17 at 14:46
• @AndreSilva Don't be that serious with voting. They're fake points... – SmallChess Dec 28 '17 at 14:47
• @AndreSilva: I up-voted this answer based on the fact that it shows candour in admitting having engaged in that action, and thereby provides evidence that people sometimes do this. It is useful information in relation to the topic. – Ben Apr 27 '18 at 0:30

Several points made here. Currently, there is little oversight for downvoting, perhaps this is not as constructive as it could be. For example,

(1) We routinely cast closure votes based upon criteria, and closures can be reversed. If we were to institute something along those lines for downvoting, the quality of downvoting might approach that of question closure, but might have to be individualized for each downvoter. Therein, the down-voter would choose an approved reason for down-voting. Then, we could thus have a review cue for edited Q/A for down-vote reversal by a third party or parties. For example, this could be required for users with less that 500 reputation, so as to train users to down vote appropriately. With or without a reputation threshold for review, this may offer some remedy for 'sticky' down votes, and oversight for 'strategic' down voting, e.g., down voting someone else's answer for spite of the 'How dare you compete with my answer?' type, or down voting to vandalize posts as a form of entertainment.

(2) Expert opinion is more needed for down voting for cause than recognition of cause for up voting, for example, currently we can up vote with 15+ reputation and down vote with 125+ reputation. The thinking on this is that it may not be 'enough' of a difference, but at least it is 'something'. (This is human nature, for example, people recognize more concepts (or words) than they are capable of using properly.) I would like to see this 125+ increased if we do not review downvotes. As 200+ is the threshold for being in the leagues, I do not see why someone with so low a reputation as to not be in the leagues should be trusted to issue a down vote that is never reviewed.

(3) There is currently no limit to the percentage of downvotes one can issue. How would the reader feel about a user who never upvotes? Would that be someone who the reader would be comfortable with to review the reader's posts? Perhaps we should consider behaviour limits of some type, for example, not allowing more than 20% down votes. Down vote quota exceeded. You must up vote 4 more Q/A to down vote again.

(4) Downvote avoidance strategies. (a) I note that the use of flippant language, personification, mixed metaphors, and other forms of informal language, even when it is intended to be merely entertaining, or less sterile than formal expression, that it can attract down-votes. (b) I sometimes lose longer posts that are in the process of being edited. Use of the back arrow (above) may be the remedy for this rather than posting incomplete text to obviate incomplete text, non-meritorious, down-voting.

• It is pointless to suggest here any changes to Stack Exchange core functionality. Such suggestions should be raised on the main Meta (and even then, have very low probability of being implemented). There is no sense in discussing your suggestions (1), (2), and (3) here on our Meta. They will never ever be implemented for Cross Validated alone. As such, I see this answer as "not useful" in this thread and am tempted to downvote 8-[] – amoeba Nov 14 '18 at 8:10
• @amoeba Understood. However, I care about this site more, and, pitching to main meta without understanding how things work here would not be taking our needs under advisement, would it? Moreover, I have discovered by looking at other sites, for example, that the reputation needed for down voting is variable from site to site, so our opinion is not as useless as one might think. – Carl Nov 14 '18 at 9:35
• The premiss is that downvoting is negative and to be avoided at almost any cost. When I downvote I do that because (a) an answer is wrong, so confused as to be useless, or quite irrelevant or (b, more commonly) a question is some mix of very poor, offensive or spam. I don't feel happy that anyone may be disappointed, dismayed, even maddened by receiving a downvote; almost always I have no sense of who they are and where they are on a spectrum from timid to obnoxious. I am downvoting what they wrote and trying to improve the site thereby. That's the spirit, I think, in which it is to be used. – Nick Cox Nov 14 '18 at 10:09
• @NickCox Down voting is negative. Negative reinforcement without explanation given is ineffectual as behaviour modification for the recipient. So the effect is against the post, and for spam, a flag deletion request works better. It may be that you feel that you down vote meritoriously, not every one is your equal. – Carl Nov 14 '18 at 13:46
• Naturally nothing I say rules out downvoting being accompanied by specific explanation. I still remember a mark of 1/15 for a question circa 1968 together with the reason why. SE isn't a help line; questions or answers that aren't good aren't welcome either. – Nick Cox Nov 14 '18 at 13:57
• @NickCox Just got two more down votes. So, you are suggesting that stats.stackexchange.com/a/320218/99274 and stats.stackexchange.com/a/376264/99274 were questions or answers that are not good and are not welcome just because someone decided to down vote? I see no rationality for either down vote, and if you do, would you mind explaining it? – Carl Nov 14 '18 at 19:01
• No; as already said more than once, I can't infer anyone else's reasons for their downvotes unless they explain them. Moreover, my support of (down)voting as a mechanism doesn't imply that I think all votes are right any more than my preference for democracy over dictatorship means that I think all elections produce the right answer. I don't recollect studying either of those posts. That questions or answers that aren't good aren't welcome is abstract principle; it doesn't imply that it's always easy to tell which they are. – Nick Cox Nov 14 '18 at 19:19
• @NickCox Would it not be better then to give a concrete multiple choice selection of reasons for down voting, and to train users to down vote for cause as contrasted to expecting down voting to be performed according to a hypothetical abstract principle of untested general acceptance? I think here that you may be projecting your own principled behaviour as a common pathway for users. Historically, I have down voted significantly less frequently than you, I generally avoid it; e.g., I can edit things that are sloppy, and vote to close questions that are senseless. – Carl Nov 14 '18 at 20:06
• Not a contest, but I’ve also edited a great deal and often contributed by voting to close. As many have said, if you want a different set-up, you need to air proposals on SE Meta. – Nick Cox Nov 14 '18 at 21:19
• @NickCox I know, but this has to be done in very simple sentences and in baby steps because the capacity for misinterpretation is apparently rather large see link. – Carl Nov 14 '18 at 23:12
• @amoeba The point of posting here is that my colleagues on this site are most interested in sharing thoughts on downvoting, and on SE meta, they are not, as I have discovered through bitter experience, willing to discuss the matter. Our site meta is much better structured than SE meta, and our users a lot more knowledgeable for statistical analysis of whatever data is presented. If we can come to some agreement, it may be possible for a request to be presented on SE meta sensibly. The SE meta problem is that downvoting is frequently complained about, about 16,000 times... – Carl Dec 6 '18 at 17:28
• @amoeba ...such that anything other than a well documented and organized presentation on that site is likely to be dismissed out of hand, despite the obvious that downvoting really is problematic, or there would not be 16 k Q/A in which it is mentioned. This inexplicable reticence to evolve a better approach to downvoting I can only attribute to an irritation of the developers rather than an intellectual response. Sort of a 'don't bother us with that old saw' which is actually a 'flat earth' type of response. Confirmation bias at work. – Carl Dec 6 '18 at 17:34

In the answer of user @scortchi we read

Down-votes do tend to be sticky, which is a pity when the down-voted poster makes the effort to improve their post. I don't see any reason to suppose this is due to anything more than people's disinclination to keep on returning to a post they've down-voted to see whether it's been improved.

This is exactly how we treat a crime: we punish the wrong-doer for the crime (bad post) and we also hope that punishment will contribute to better future behavior (future posts). In the factual world this may be realistic since in most cases an action that has exceeded the threshold for a crime cannot be altered/undone. But SE questions can - and by not following up on our downvotes, rewarding by removal of the downvote those posts that were improved, we forego the opportunity to beneficially exploit this additional flexibility.

When individuals (users) use rights & granted powers (downoting) in such an incomplete way, and there is no prospect for a change-of-mentality on the horizon, usually the only solution in sight is to have community representatives take up the duty to clean up (this is a large part of what public sectors are tasked with doing across the globe). In our case, our public sector is the moderators, and it would mean moderators being able and tasked with monitoring downvoted posts and removing downvotes from those that were improved.

• This is not exactly how we treat a crime. Crimes are adjudicated by an expert judge or an instructed jury of enough people to decrease the likelihood of false conviction. If two people agree, there is no conflict. Two people disagreeing is a conflicting opinion. Up votes require a rep of 15+ and down votes from approximately 100+ to 150+ on StackExchange sites, which reflects the need for more expert opinion to determine wrongness. – Carl Nov 8 '18 at 4:01
• The spirit of this is that downvotes that are no longer deserved should be reversed. So far, so good, but it's not reasonable to expect moderators to do this for us. It's the downvoter's responsibility. I still remember a courteous reaction from a moderator very early in my CV career when I urged moderator action because an answer was very wrong: The reply: it's not a moderator's job to decide what's right or wrong. Nor is it their job to decide what's "improved". So, downvote, ideally explain why if no one has already, and post a better answer or comment if you can. – Nick Cox Nov 8 '18 at 15:16
• @NickCox I totally agree with you, but note the "When individuals (users) use rights & granted powers (downoting) in such an incomplete way, and there is no prospect for a change-of-mentality on the horizon" point in my answer. – Alecos Papadopoulos Nov 8 '18 at 21:12
• When I cast a down-vote I usually check the post for subsequent edits about a couple of times over the next week or so, if I remember. I think it's cheering for all concerned when a downvote becomes an upvote. But I don't have a bone to pick with anyone who simply down-votes what they don't find clear or useful at the time they happen to read it, & leaves it to future readers to judge the clarity & utility of any later versions that may appear. – Scortchi Nov 13 '18 at 11:58
• As @NickCox points out, your proposal would extend the purview of a moderator's role into new areas: I don't think there's a serious enough problem to justify that; I don't see unequivocally good posts languishing with negative scores owing to down-votes clearly accrued because of now-amended faults. – Scortchi Nov 13 '18 at 11:58
• I up voted your answer because I agree with the problem as you posited it. However, the solution you proposed is only one possible of many a few others of which I put in as a new answer. – Carl Nov 13 '18 at 22:06
• @Scortchi Certainly, as always with most problems, it matters how severe they are, meaning in our case how many/often are "...unequivocally good posts languishing with negative scores owing to down-votes clearly accrued because of now-amended faults", to repeat this marvelous sentence of yours. And I do not argue that the problem is severe. – Alecos Papadopoulos Nov 14 '18 at 23:33

Currently, and do correct me if I am wrong, it costs (-1) to downvote, and the downvote creates a (-2) reputation for the recipient. One alternative is increasing the penalty to the downvoter to at least (-2) to make it more altruistic. Right now, the small penalty encourages trolling (It hurts my target more than me; HaHaHa!), ganging up in a bloodfest as in monkey see monkey do, grudge holding, and other very discouraging behaviors that negatively impact people who are are honestly attemptng to help others. Another alternative for better altruism would be to make it (-1) for the recipient and (-2) for the downvoter.

• @gammer I will take a break from this, and see how people feel. At least on meta, downvoting does not close a good question or answer, which you have to admit, can destroy a lot of work and sincere effort. – Carl Jan 30 '17 at 2:45
• Downvoting does not close an answer anywhere. It can get deleted if enough high rep people cast delete votes, but that's a whole other thing than the type of voting you're talking about. Regarding changing the cost of casting a downvote to -2, sure, why not? But I have a hard time imagining that would actually change the intent of someone dead set on trolling (which you seem to think most/all downvotes are). – gammer Jan 30 '17 at 2:49
• I will just remark that we can discuss this as long as we want but any such change would require a change in the core StackExchange-wide policies and this will not happen, like it or not. This is not for us (as a CrossValidated community) to change. – amoeba Jan 30 '17 at 7:55
• Down-voting questions is free: see stats.stackexchange.com/help/whats-reputation. -1 to this suggestion because I don't see any positive evidence of a problem that needs fixing. – Scortchi Jan 30 '17 at 14:12
• I don't think that this reputation stuff will change the down-voting. A down-vote has little effect on reputation whether it is -1 or -2, or even if you would set it to -10. The much bigger impact is that a down vote equals an up vote in the 'ranking' of answers and questions. To me this is what matters much more than reputation, the gratification for having given an answer. Scortchi's answer mentioning ideas about notifications for edit's on down-voted answers would be much better in this direction. In fact, it would allow me to down-vote more since this is one reason I would not down-vote. – Martijn Weterings Apr 9 '18 at 2:40
• @MartijnWeterings I like your thinking on this because you have proposed a solution. – Carl Apr 9 '18 at 2:55