What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 134 Stack Exchange communities.

When we started Stack Overflow, we wanted to make sure that downvotes were more of a visual and psychological motivator than a punitive action. That's why

  • upvotes are +10
  • downvotes are -2, and cost the casting voter -1

However, recently, I've seen Marc Gravell and Jon Skeet both say similar things:

IMO, downvotes don't quite carry enough... -5 downvotes and +1 upvote (for sympathy) yields zero points overall; or it could even yield +10 points for a new user! I'd prefer to see, say, -4 points? -5 points?

I am starting to come around to this way of thinking. There are some users who ask a lot of questions that get downvoted heavily, but over time these users end up with a fairly sizable amount of reputation.

This doesn't feel right.

We are considering changing downvotes to -5, with the same -1 cost to the casting voter.

There really aren't that many downvotes in the system, so the net effect of this change would be to reduce rep growth for users who tend to get downvoted as much as they get upvoted. (Yes, I realize this is no panacea: over time, we'll still get rep inflation for marginal users based on sheer quantity; this is only one of several changes we are considering to help mitigate this.)

The other downside to this change, besides the obvious minor loss of rep for most users, is that it would force a global reputation recalc for every user. I think we're due for one of those anyway..

share

locked by Jeff Atwood May 4 '11 at 8:46

47  
I like this, but I would raise the weight for users with reputation > X. –  jjnguy Jul 20 '09 at 12:02
16  
I have to ask - are these 'marginal' users causing problems? How? I just want to make sure there isn't some sort of 'anti-noob' bias here. –  Michael Kohne Jul 20 '09 at 12:21
16  
Jeff - as it's a major change it would be worth making an announcement on all the sites just before it happens to explain and hopefully prevent a lot of "I just lost X rep - WTF" messages. –  ChrisF Jul 21 '09 at 10:02
9  
It would not "force" a global recalc. It would work fine without a recalc. The recalc will attract far more complaints than any extra penalty for bad questions/answers. –  Perpetual Motion Goat Jul 21 '09 at 10:49
121  
I say this only somewhat facetiously, but global recalc is unconstitutional. You can't change the law, then punish people for something they did that was legal at the time but now illegal. It's perfectly possible to write a database query that will use one calculation scheme before the cutoff date and a different one after. I support the +5/-2 proposal, but I don't support a global recalc (and my rep is only like 350, so I don't have anything to lose personally--just don't think it's right). –  devuxer Jul 22 '09 at 0:42
26  
Rather than the -2 to -5 change, I'd like to see it just not award rep points for upvotes to questions or answers with a negative score. Much the same effect, but less disruptive to users who play by the rules. –  Joel Coehoorn Jul 23 '09 at 13:25
26  
@DanThMan I'm afraid I disagree - your rep is not something you own. It only really has value relative to what everyone else's got. Everyone's rep would be recalculated, and the relative ranking afterwards would better reflect their actual kudos on the site. –  Keith Jul 23 '09 at 20:26
12  
Doesn't the EULA state that the rep points remain property of stackoverflow? –  Tom Hawtin - tackline Jul 24 '09 at 3:05
20  
It isn't a game. It is a question and answer site. –  jjnguy Aug 1 '09 at 21:10
11  
Any news on when this will be happening? –  jjnguy Aug 11 '09 at 13:29
15  
Downvoters should be forced to write a comment telling why is the downvote. This would help with this "tactic downvote" thing. –  The Disintegrator Aug 27 '09 at 3:19
10  
@Jeff: Please don't go with -5/-2. Or at least make the change to -5 first, and only increase the voter penalty to -2 if there is evidence of a problem with leaving it at -1. If you make the changes one at a time you'll be able to measure the impacts independently. –  Dan Dyer Sep 25 '09 at 23:28
6  
@Roboto: that is ridiculous. I have given 357 downvotes, and I'm sure there are users with way more than that. Forcing people to leave comments for each one would result in thousands of comments that read "asdfasdfasdf" or something equally meaningless. –  Graeme Perrow Dec 17 '09 at 16:20
11  
My forced downvote comment would default to "Take that. Hah!" for 15 character. As much as I dislike drive-by downvotes, forcing a comment is simply an invitation to uniformative or unhelpful comments –  dmckee Dec 20 '09 at 1:45
14  
Declined?! Darn, I was genuinely looking forward to this change –  Mark Henderson Mar 19 '10 at 1:54

41 Answers 41

An eye for an eye...

If a downvote inflicts -5 hit rep points, it should cost 5 mana rep points.

share
5  
Well... that would effectively kill down-voting. –  Shogging through the snow Dec 22 '09 at 2:57
1  
@Steven: Perhaps, but in that case why not just... remove it from the system. Entirely. Would simplify rep calculation, post scoring, etc. Personally, I feel that, so long as it's a part of the system, it should be useful and be used, and therefore I oppose anything that makes that less likely: increase in cost, arbitrary restrictitions on when it can be used, etc. As soon as we start treating it like some evil bastard feature that shouldn't really be touched, there's no reason for it to exist. –  Shogging through the snow Dec 22 '09 at 17:25
4  
Well, the trouble is... sometimes posts are really quite awful. Not awful in a "this doesn't belong here" sort of way, just awful in a "this is completely wrong" fashion. It's nice to down-vote them, to leave a quick and effective indicator to future readers that this isn't worth your time. Comments don't really do the job; you don't see them 'till after you've read the post they're attached to. Mass-upvoting does sorta work, but only for answers, and only when there are multiple answers, and only when at least some of those answers are at least somewhere close to being correct. –  Shogging through the snow Dec 23 '09 at 3:09
3  
@Shog, that would suck... long live the downvote! –  jmfsg Feb 4 '10 at 20:38
up vote 21 down vote accepted

Having had a few months to think about this, I'm no longer sure it's a good idea, for two reasons.

  1. Downvotes were always essentially cosmetic, with an extremely minor effect on reputation. Despite this, received downvotes are taken quite seriously by users. Almost too seriously. If we raise them to -5 they are no longer cosmetic but can be wielded as cudgels on other users. This was never the intent of a downvote, so we would be twisting it into something ugly. Users are far, far more attached to their reputation scores than I ever could have predicted, and I believe more than doubling the weight of a downvote will cause a lot of new, additional angst in the community over the occasional received downvote -- to the point that only users who want to hurt other users will cast them. Downvotes go from being cosmetic and psychological to weapons of war.

  2. If we amplify the effect of a downvote, in all fairness, we must also amplify the cost to the casting voter as well. So what was -2 to the post, with -1 to the casting voter, will become -5/-2 or even -5/-3. The only reason we had downvotes cost -1 was to make people really think about using downvotes in moderation, only when appropriate. For such an extremely minor cost, users with 20k+ rep are still hesitant to cast downvotes for fear of losing reputation! Our goal is not to prevent people from casting downvotes, and it seems that given the current observed behavior (that is, very few people cast downvotes -- though a tiny select few cast a lot) we would be causing the majority of users to cast even fewer downvotes than they do now. This is, in my mind, a HUGE downside to this proposed change. We would be effectively nerfing downvotes into oblivion because they're now twice as "expensive" to cast.

The real problem I want to solve here is "shore, there is always gold". I am no longer convinced this is the right way to do it, but I am open to hearing ideas on that thread on other ways we could do it.

share
1  
+1 I agree with this decision at this time. Each Trilogy site is as much a social experiment as it is a focused Q&A website. It's really difficult to make such a sudden fundamental change for such a big group of people. I think if we can come up with a way to test a modified reputation system with a set of voluntary users, that would allow us to make a better decision. All we have right now is forethought and hypotheticals. –  Jon Seigel Mar 19 '10 at 2:13
4  
It's sad that high rep users won't downvote becuase of the loss of reputation. I'm only on 8k on SF, but I relish the oppertunity to downvote, and the cost of -1 is negligable. –  Mark Henderson Mar 19 '10 at 2:32
7  
Downvotes go from being cosmetic and psychological to weapons of war. May be some people should stop worrying so much about reputation? –  perbert Mar 19 '10 at 3:59
2  
@John if only others were more liberal with their rep :) –  Mark Henderson Mar 19 '10 at 5:02

Rather than going through and majorly dropping all question points by half, why not just make upvotes on answers equal +15 or something? Then no one loses any abilities they've worked for.

share

So... when is this coming?

Is there some worry this will generate unwanted behavior somehow and that's why it's being postponed?

share
1  
It's coming in 6-8 weeks, haven't you heard? All joking aside, this really needs to be implemented. It would really help the community out. –  alex Feb 4 '10 at 19:59
2  
The only shameless part is, that this is still not implemented yet. –  Ladybug Killer Mar 11 '10 at 20:09
1  
I knew you'd understand me @john –  jmfsg Mar 11 '10 at 20:14

I believe that instead of changing the weights of down-votes an alternative solution could used for the reputation.

Basically instead of using every vote on the question/answer for the reputation the "vote balance" could be used.

For example if user has 4 questions:

    Vote balance    up-votes    down-votes    reputation earned
Q1:      1             1            0                +10
Q2:     -1             0            1                 -2 

Q3:      1             3            2                +24 (+30 -4)
Q4:     -1             3            4                +22 (+30 -8)

I believe that the fact that Q1 earned +10 reputation and Q2 -2 is not an issue for anyone. The problematic ones are Q3 and especially Q4.

So instead of counting each up-vote/down-vote for the reputation, we could consider the overall vote balance of a question. That would eliminate the problem the Q3 and Q4 created in the current setup.

The same situation would look as follows

    Vote balance    up-votes    down-votes    reputation earned
Q1:      1             1            0                +10
Q2:     -1             0            1                 -2 

Q3:      1             3            2                +10
Q4:     -1             3            4                 -2

It think that would be more fair than changing the weight of the down-votes.

See also Earning Reputation for Poor Questions

share
1  
@Pollyanna - I think it is more fair because you earn the reputation based on the value of the Q/A (vote balance) as opposed on number of votes on the Q/A. It believe that this number represents the value of the Q/A to the community. I think it is logical to expect when seeing a question with reputation +1 that this earned someone 10 points, and the one with -1 earned -2 points. It is quite strange that someone can gain reputation on a question that has a score of -1 which could currently be a case (see Q4 in the example above) –  kristof Feb 12 '10 at 10:26
1  
@Pollyanna for example if you look at the question that I posted meta.stackexchange.com/questions/35763 It has been closed as a duplicate and has a vote balance -1, and you now what, it actually earned me 14 points of reputation: +20 for 2 up-votes and -6 for 3 down-votes, I think that's wrong and that I should have got -2. Changing the weight on the down-vote will not eliminate the phenomena of earning reputation on answers that have score below zero. If even you apply -5 for down-votes I would have ended with +5 points here (+20 -15). –  kristof Feb 12 '10 at 10:38

As I (just now) mentioned on this thread, I think the weight of a downvote should increase non-linearly, the way that upvotes are now - only instead of starting out high and getting lower, it should start out low and get higher!

share

Why not increase the downvote value according to the user's rep?

ie, Joe N00b joins and starts downvoting everywhere - it'll be carnage. However, if his downvotes don't count much (-1 if that) then it'll be fine; If Jon Skeet decides you've gone too far (or answered incorrectly) then your rep penalty from his downvote should count more. He's an obviously more responsible user who would only downvote (hopefully) for things that really needed it.

So if the downvote penalty ranged between 0 and -10, a user with 2000 rep might cost you 2 points, a user with 10,000 rep might cost you -5 points.

share
1  
This has been discussed and [status-declined] before. Can't find it at the moment though... –  fretje Jan 15 '10 at 20:28
1  
But wouldn't it be cool? I think that each of my downvotes should reduce the target's rep by at least 1% of my rep (but also cost me 1% of my rep) - BAM! 280 rep gone. It would be so cool. Too bad I never downvote... –  Adam Davis Jan 15 '10 at 20:45

I always thought you chose the 5:1 ratio on purpose. Psychological studies show that we are impacted far greater by negative things than positive ones. I've heard that marriage councilors often say the magic ratio is 5 positive things for every one negative thing. I know I definitely feel down votes more than up votes and as such I thought the ratio was perfect.

share

Some effort has already been put into encouraging feedback with the down vote. Could this be an opportunity? Perhaps a system where "Anynomous" down votes count the old way, but if you want to "be on record" then your down-vote counts more.

share

This is one of the most upvoted requests, but it's still only status-planned (it's been that way for close to half a year now). Why not implement the change on Meta first? Seeing how it goes down here would give you a way to gauge the potential reaction of users on the other sites. The community is smaller, but it's the most active and vocal part of it.

After that you can then work out how the whole thing works in real-time. And there's nothing better than data you can really crunch. It's science!

Maybe with that, include an annual rep recalc, maybe twice a year even, to keep the numbers closer to being accurate. You can dress it up as an end of year present/stocktake sale.

share
8  
Bring it on! If the people hold onto their Meta rep preciously, they really need to take a step back. Bring it on Meta already. –  random Dec 17 '09 at 12:36
7  
I think recalcs should be more frequent, like maybe once a month, so they don't shock as much. –  Paul Tomblin Dec 17 '09 at 12:37
6  
The fact that we have no idea when a recalc is performed is extremely bad. Every now and then you have people popping over to meta to complain about the fact they've lost x points. If you do it at regular intervals, this thing simply goes away. –  alex Dec 17 '09 at 12:42

Allow multiple down votes by the same user at a cost to both parties of 1 point to a maximum of 10 down votes. It should always hurt to punish even when punishing the people who deserve a lot of punishment.

share

The proposed change might roughen the pleasant atmosphere I associate with Stackoverflow. So here's my 3 cents against increasing the penalty:

  1. It's good to know what's wrong is as well as what's right. An increased penalty will cause more wrong answers to be removed, along with the comments explaining why it's a wrong answer.

  2. Some answers are controversial. They are not wrong, or unhelpful, but they still get a lot of downvotes. Increasing the penalty would discourage people from expressing what they really think.

  3. Quality of information is ensured by a negative feedback loop:

    Negative feedbacks tend to dampen or buffer changes; this tends to hold a system to some equilibrium state making it more stable.

    By increasing the penalty, the constant in the feedback loop changes from 0.2 (one upvote = 5 downvotes) to 0.5 (one upvote = 2 downvotes.) This 150% increase will remove the "feedback" nature of downvotes. Downvotes will be a smack in the face rather than a gentle push in the right direction.

  4. People will be very upset if you lower their reputation retroactively. It's like saying "a penalty kick is now worth 7.5 instead of 3 points" and reordering the rankings at the end of the Rugby year.

share
1  
On (4): Should not be a problem as long as the revised formula is applied to all players equally. It's like a stock split or join. –  Lawrence Dol Nov 21 '09 at 19:41
2  
On (3): StackOverflow suffers at present from a large quantity of crap - the signal:noise ratio is too low. –  Lawrence Dol Nov 21 '09 at 19:41
1  
@Software Monkey: (4) Even if the formula is applied equally, it changes the valuation of past behavior That's almost universally acknowledged as unfair (3) Yeah, negative feedback (especially the subtle type) was reduced greatly by the "Vote to old to be changed" change. It will decrease even further when a downvote equates to -5 points. Then people will downvote only people they hate, instead of people you suspect might be wrong –  Andomar Nov 21 '09 at 20:57

So here's the problem with changing the reputation.

You want people to post on the site, even if they are not totally right. It should be ok to make mistakes and have other people correct you etc. If you get down voted a couple of points for making a mistake, that should be ok. It's a learning process. Should people be punished for making a mistake, maybe a little as a way of saying, "Hey, you made a mistake, please try and fix this." By making it higher, really what you are starting to tell the community is, "Don't be stupid, don't make mistakes about what you are posting, and don't waste our time." (wow sounds a lot like usenet). In the past, the tenet of the site was always been to be able to ask any question no matter how simple. If I remember back to the first podcasts it was always about people not having to RTFM. People need to be able to post mistakes to learn.

The point of the "game" (site) is to get people to participate and play. If people are mightily punished for trying, the immediate response is going to be to not use the site which will ultimately cause it's demise. The scoring system shouldn't be targeted towards the upper posters on the site, it should be focused on keeping the average programmer interested and coming back. The people who maybe kind of often get down votes, and try and post useful information and get some up votes too. Right now, you need only one up vote for every five people who think you made a mistake. Losing points does a lot more to someone's ego (that's what the points boost) than gaining points. Most programmers aren't experts at what they program, but they try. The reward vs. punishment factor directly affects how much they are willing to try (the greater the punishment, the less they will). There are what 9 million programmers, but there are only 400 people with a 10k + rep. Does there really need to be a greater gap between the top posters and people who ask a lot of questions to try and learn?

The amount of points lost by the recalculation isn't really that important in the big scope. It's about the person who maybe logs onto the site and asks a question or posts and answer and quickly realizes that the community suggests that "non-experts need not apply" and goes away to pay 10 dollars a month to get his questions answered elsewhere, because at least he doesn't feel dumb there. To be successful, these are the people the site needs to focus on retaining. The site is kind of like Communism, to work it has to benefit the masses.

share
3  
You don't have to be an expert to ask a decent question. You might have to put more than 15 seconds into writing it, but a fair number of people do manage this. And for the ones that don't... Surely the lame answers from people who couldn't figure out what was being asked are a bigger blow than a few lost reputation points? –  Shogging through the snow Nov 3 '09 at 22:08
3  
Some people can't write well. Some people have English as a second language. As far as I can tell Atwood's entire dev-days talk was on that, so we should punish them for having difficulty? Based on your response, you think that only experts should answer questions? People who make mistakes should be punished for trying and clearly pointing out to them they have the wrong answer, instead of nudging them in the right direction? Having a higher down vote tells new users (and existing ones) that stupidity isn't tolerated and people who try and stretch what they know should go elsewhere. –  Kevin Nov 4 '09 at 0:48
3  
You predict: "What's going to happen is that non-experts will stop posting answers" ...but then ask, "Why would people use a site that doesn't make them feel better about what they know?" ...So, which is it? Who are you afraid of scaring off: the people who don't know what they're talking about, or the people who do? Or are you just trying to beat up a strawman by putting words like "experts" in my mouth... I'm not an Expert, I'm certainly not "super smart", but i do know this: All that is required to get up-voted is the correct answer to a single question. –  Shogging through the snow Nov 4 '09 at 8:50
3  
@shog9 you have a 22 k rep, you've posted 453 answers, and you have the JavaScript specialist badge, that kinda puts you in the expert category. The site needs to be geared towards the programmers who quite frankly, don't know a lot about what they do. Occasionally they get upvoted, sometimes they get downvoted. You have to encourage them to come back and keep trying and that encouragement is affected by the reward vs punishment factor. Changing the down vote count won't encourage them to post better answers, it will encourage them to stop posting at all. –  Kevin Nov 4 '09 at 13:32
2  
The punishment button (or the downvote arrow), is a much more powerful behavioral tool than the reward button. It has to be very heavily weighted against especially considering the facts that almost everyone can use it, and it can be applied to any question or answer someone doesn't like whether it is deserved or not. –  Kevin Nov 4 '09 at 14:12
1  
@Kevin: I've answered a hundred or so JavaScript questions, and the ones that got up-voted are almost universally the ones that I put a bit of time and effort into researching. My off-the-cuff answers are usually eclipsed by someone else who knows more or put more effort into it... And that tells me the system is working - it's easy to make wild guesses, or parrot common knowledge without backing it up, but what use is that? An unhelpful answer is bad enough, but a up-voted answer can discourage the posting of additional, potentially-helpful answers - surely we don't want that! –  Shogging through the snow Nov 4 '09 at 17:47
2  
And yes, I understand that down-votes can be discouraging, but they can also be a push in the right direction, and why shouldn't we motivate users to put some effort into their answers? JavaScript answers are an example of something that is usually very easy to test (if you're using SO, you already have a browser open - just open a new page and run the script!), and yet i frequently see code in answers that won't even parse, much less work. How does encouraging these answers help anyone? We should strive for quality over quantity, rather than rewarding users for merely spamming the site. –  Shogging through the snow Nov 4 '09 at 17:52
1  
@Shog9 You are absolutely right, the site should not encourage absolutely wrong answers, but the site can't discourage people trying to post the answer they think is correct. Not all questions are black and white, answers are may have a varying degree of correctness and although it may not be correct, it's a step in the right direction. The obvious response is to maybe not down vote it, but that can't be controlled by the system, because it really depends on who is reading the answer. The only way for the system to equalize this is to minimize the severity of the down vote. –  Kevin Nov 5 '09 at 3:34
3  
The problem is that eventually if the down vote rep loss is high enough, many people will look at a question and think about posting the right answer but won't because they are "afraid" of the consequences of posting something that is incorrect. The counter to this is that people should make sure what they post is right, but many people don't have the motivation or time to do so. Some questions will get answered by someone, but others won't which detracts from using the site. The less people who try and answer questions, the more questions which will go with no real answers. –  Kevin Nov 5 '09 at 3:42

I think it shouldn't make a difference how many down votes and how many up votes there were, but that number on the side. If I get up voted 3 times, and down voted once, there would be a 2 on the left in between the arrows that let you vote up or down. It should get considered as 2 up votes, so 20 reputation, not 3 up votes (30 rep) minus 1 downvote (2 rep lost). So I think if the number is positive, it should be num × 10 reputation gained, if the number is positive, num × 4 lost (or maybe 5). The so basically a down vote in the positive area makes you lose 10 points, but you just gained those points so I think its fair.

share
1  
How would few points be lost as a consequence of a spam post? If there are only down votes, it would be no different then the current system. –  Mk12 Sep 1 '09 at 20:56
2  
And if you got 1 downvote after multiple upvotes, or just one, you would still never lose any reputation points from what you originally had... I don't think you understood my suggestion. –  Mk12 Sep 5 '09 at 16:52

I think that the penalty for downvotes should be less when the downvoter leaves a comment. Maybe -2 without comment and -1 with a comment.

share

Yes, definitely -5 for a down-vote for the down-votee, but then I would also "up" the cost to -2 for the down-voter.
The higher cost will reduce the "just feels like it" down-vote behaviour some people are displaying.

share
5  
I agree with this. –  michelemarcon Jul 20 '09 at 10:19
76  
And this might also reduce the number of people who down-vote all answers that aren't their own –  a_m0d Jul 20 '09 at 10:58
10  
Yes, -5/-2 sounds like a reasonable trade-off. –  Matt Hamilton Jul 20 '09 at 11:10
9  
I reckon a downvote should take half the value of an upvote, and cost half that again. Given you need a global recalc to do this effectively, I like the idea of it being powers of two, so +8 for an upvote, and -4 (at a -2 cost) for a downvote. You could even throw in a +1 for every ten times a comment gets up voted :) –  Rowland Shaw Jul 20 '09 at 12:15
62  
The only reason to up the penalty to -2 is, if there is a problem that too many people are downvoting stuff, the stats do not seem to back that up. –  waffles Jul 20 '09 at 21:43
8  
There are some people ( you know who ), who have more down-votes than up-votes, who would lose a lot of reputation if the cost of casting down-votes increased. –  Brad Gilbert Jul 21 '09 at 15:48
12  
@Brad: Don't thinks that's so bad. As I see it, that behaviour (having more down-votes than upvotes) is generally 'frowned upon' by the community. –  fretje Jul 21 '09 at 16:28
4  
Important: have a look at the impact of this change in my post below. -2 will heavily punish serial downvoters (RichB will lose almost half his rep). Problem users do not seem to downvote much. Further more, problem users get a lot less sympathy votes than people think. –  waffles Aug 14 '09 at 2:26
2  
There are a lot of knee-jerk downvotes. I've gotten a few for perfectly valid responses, where people were agreeing with me in the comments, but others were downvoting without giving any explanations. –  alex Sep 14 '09 at 14:41
5  
I disagree. Changing to -5 is a step in the right direction (I favour -10 to make up votes and down votes equivalent). However, I am strongly against changing the penalty for the voter to -2. In fact I'm strongly against any penalty for down-voting. People should not be penalised for casting a deserved down-vote. I understand why it was introduced, but I think there are better solutions to that problem. –  Dan Dyer Sep 25 '09 at 18:02
13  
maybe the -2-1 cost for downvoting is void if you supply a comment? –  Kevin Peno Nov 5 '09 at 8:06
4  
@Kevin I actually like that idea considering how frustrating it can be to get downvoted with no reason why. "Explain yourself or pay up" is a good motto for downvoting. –  statenjason Nov 7 '09 at 18:13
2  
@Kevin & statenjason that will lead to the comment: I don't want to lose any rep. –  johnc Dec 22 '09 at 4:51
1  
@lagerdalek. True, but you'd see who was doing the downvotes. –  Gary Dec 22 '09 at 6:05
3  
Oh wow, 10 people down voted this ... I should down vote it too! Sorry, but mob mentality comes into play. Just forcing a comment for a down vote would suffice. If you can't articulate a reason for a down vote, well, why are you down voting? SO is very good at catching retribution in votes, what is the issue in enforcing mandatory comments? –  Tim Post Mar 16 '10 at 20:08

I just had a look at my stats and the effect of this change would be really small for me. I stand to lose maybe 100 rep.

I think its interesting to look at the list of people who lose most out of this (restricting to people with less than 10k rep) people with more really should not care:

select top 100 UserId, UserName, (t.DownVotes * -4) + (u.DownVotes * -1)  as Loss1, t.DownVotes * -4 as Loss2, u.Reputation  from UserTotals t
join Users u on t.UserId = u.Id
where u.Reputation < 10000
order by Loss2 asc

Results (people most impacted from the change - -5 instead of -1 for a downvote):

User            Loss  Reputation 

Rich B          -1156 7773       
Mastermind      -944  6542       
theman_on_vista -568  670        
Thomas Hansen   -552  2430       
Joan Venge      -540  4844       
mgb             -468  9772       
unknown         -452  918        
Masi            -448  4193       
Paul Nathan     -444  5653       
mP              -440  1606       
Unkwntech       -440  8838       
Malfist         -420  3332       
Bombe           -412  8322       
Click Upvote    -412  5683       
mson            -396  2773       
StingyJack      -392  4907       
GateKiller      -388  7190       
warren          -388  6602       
Kirsh           -340  3235       
Bill K          -336  9121       
Kevin           -332  2795       
David Arno      -324  5905       
ck              -320  6869       
Grauenwolf      -320  4016       
Filip Ekberg    -316  3935       
Justice         -312  7653       
Shore           -312  1500       
BCS             -308  6844       
Jon Harrop      -304  1342       
Teifion         -304  7439       
unknown         -304  4590       
samoz           -304  5816       
Joshua          -300  2509       
Gold            -300  1056       
Claudiu         -300  8830       
Johanna         -300  638        
DannySmurf      -296  6650       
TraumaPony      -296  4192       
Cody Brocious   -292  8965       
dacracot        -288  3991       
Alex Fort       -288  5676       
tim             -288  8798       
Rob Cooper      -284  9021       
Thanks          -284  4350       
David Dorward   -280  8342       
Javier          -280  9524       
gs              -280  7802       
CodeToGlory     -276  2885       
gbjbaanb        -276  8978       
Ben Hoffstein   -276  7138       
Esteban Araya   -272  4793       
DrPizza         -272  4342       
acidzombie24    -272  4359       
Jason Punyon    -272  6670       
Ali A           -268  7412       
Luca Matteis    -268  4468       
workmad3        -264  6243       
Geo             -260  5079       
SpliFF          -260  4337       
SilentGhost     -252  7726       
jrockway        -248  6781       
Blankman        -248  4546       
Joshxtothe4     -248  911        
mattlant        -248  4414       
BobbyShaftoe    -248  9724       
David W. Fenton -244  2468       
hasen j         -244  6373       
Dev er dev      -244  4831       
Rob Wells       -240  9495       
Daniel A. White -240  6145       
shahkalpesh     -240  5147       
Chris Ballance  -236  7509       
Unknown         -236  8945       
MarkusQ         -236  8882       
Brad Wilson     -232  8417       
Rich Bradshaw   -232  5318       
Robert S.       -232  9875       
Diodeus         -228  6360       
KM              -228  4843       
lomaxx          -228  9666       
EBGreen         -228  6827       
strager         -228  9242       
troelskn        -228  7736       
Sir Psycho      -224  1193       
17 of 26        -224  7972       
Skizz           -224  6272       
unknown         -224  695        
Will Dean       -220  7464       
Nathan Campos   -220  746        
LFSR Consulting -220  7074       
Jian Lin        -220  4016       
R. Bemrose      -216  8240       
Mike Stone      -216  9750       
bpapa           -216  2974       
Spencer Ruport  -216  8374       
Dillie-O        -216  8139       
raj             -216  5790       
Omar Abid       -212  842        
Charles Bretana -212  7565       
James Anderson  -212  2868                 

People most affected if the change is an extra -1 for a downvote and -5 for being downvoted.

User            Loss  Reputation 

Rich B             -3398 7773       
David Dorward      -996  8342       
Mastermind         -988  6542       
Unkwntech          -859  8838       
Click Upvote       -850  5683       
DJ                 -822  4910       
hop                -809  3473       
Joan Venge         -756  4844       
Rob                -703  8979       
Brian Knoblauch    -667  4004       
Bombe              -660  8322       
Rob Cooper         -652  9021       
Juan Manuel        -624  5192       
jrockway           -624  6781       
TheSoftwareJedi    -617  6218       
LFSR Consulting    -609  7074       
Robert S.          -600  9875       
Paul Nathan        -600  5653       
GateKiller         -589  7190       
hasen j            -582  6373       
gbjbaanb           -580  8978       
Cody Brocious      -577  8965       
Malfist            -575  3332       
SilentGhost        -570  7726       
theman_on_vista    -568  670        
Thomas Hansen      -561  2430       
rp                 -549  4776       
StingyJack         -537  4907       
warren             -513  6602       
Grauenwolf         -512  4016       
Suma               -508  3590       
mP                 -506  1606       
DannySmurf         -499  6650       
Outlaw Programmer  -499  3820       
starblue           -495  7913       
Masi               -490  4193       
David Arno         -483  5905       
mgb                -482  9772       
tim                -479  8798       
Will Dean          -470  7464       
Jason Punyon       -469  6670       
mson               -462  2773       
unknown            -457  918        
David W. Fenton    -456  2468       
Alex Fort          -450  5676       
bdukes             -445  8246       
Ed Swangren        -443  5628       
bzlm               -442  1515       
Samuel             -438  6273       
Stu Thompson       -436  4212       
gs                 -436  7802       
Josh Stodola       -427  3955       
Jon Harrop         -422  1342       
Kevin              -420  2795       
TraumaPony         -418  4192       
Bill K             -413  9121       
Apocalisp          -410  4848       
Jason Bunting      -405  7121       
EBGreen            -404  6827       
Filip Ekberg       -403  3935       
Kirsh              -394  3235       
Graeme Perrow      -392  6515       
Sklivvz            -392  6722       
Jimmy              -390  8532       
Kristopher Johnson -389  7058       
stepancheg         -389  550        
J-P                -389  4867       
Geo                -386  5079       
Software Monkey    -383  7697       
mausch             -381  6271       
Justice            -375  7653       
01                 -374  2571       
samoz              -371  5816       
Chris Ballance     -370  7509       
blowdart           -369  6120       
porneL             -364  5767       
superjoe30         -364  2701       
DrPizza            -363  4342       
BCS                -361  6844       
Rob Wells          -358  9495       
Greg Dean          -353  4573       
Esteban Araya      -350  4793       
lomaxx             -350  9666       
Ngu Soon Hui       -348  4187       
dacracot           -345  3991       
lothar             -345  5577       
Pesto              -344  6185       
ck                 -341  6869       
Patrick            -339  4852       
unknown            -339  4590       
JohnFx             -338  4428       
x-x                -337  3626       
recursive          -337  5151       
Dan                -335  698        
Javier             -334  9524       
strager            -334  9242       
Ben Hoffstein      -332  7138       
mattlant           -332  4414       
tloach             -328  3981       
Svante             -328  6437            

Observations:

  • "Problem users" do not seem to downvote that much, so it pointless making it more expensive to downvote. (however it seems rich b gets a lot of retribution votes)
  • Many of the "bad users" will keep their editing rights
  • It is incredibly rare for anyone to lose more than 400 rep.
share
5  
yes, this is pretty much the same query I looked at when trying to tell what kind of impact this would have, and on which types of users. One caveat -- you should disallow community wiki posts from counting (although the relative percentages stay pretty much the same for most users, even with CW downvotes incorrectly included) –  Jeff Atwood Jul 20 '09 at 9:49
2  
I don't think that there is any flag for CW in the data dumps yet, but there have been requests for it (see the question re: what should be included in the data dump) –  a_m0d Jul 20 '09 at 11:02
1  
Great info! –  jjnguy Aug 14 '09 at 18:36
3  
-436? I'll get over it. –  Stu Thompson Nov 3 '09 at 21:27

TBH, I'd rather prefer to see the -1 on downvotes go, maybe also only for people with more than X reputation. That could encourage a lot more people (incl. me) to actually use the feature, because at the moment I feel like I'm getting punished for trying to keep low quality answers/questions down, which is why I rarely ever downvote something. (and yes, it's just one small tiny friggin point and I have plenty, but still, it feels like a punishment against me or asking me to "pay" for something that has no value to me)

share

I don't know - this really feels wrong some how. Even at -2 we already have a lot of people complaining about getting downvoted. -5 would make that much worse.

A subtler change that would have almost the same effect is just don't award rep points for upvotes to questions or answers with a negative score at the time of the vote. If the complaint is that some users are slowly accumulating a lot of rep as a result of "pity voting", than just target the pity votes rather than all votes.

Unfortunately, it seems like the new -5/-2 change is already decided :(


As an aside, I was opposed to changing this at all in the past because it looked like a solution in search of a problem. So a few users gain a little extra rep score. So what?

If they're using SO that much they've still invested something of themselves in the site and therefore still earned a certain amount of trust. Beyond that point is doesn't matter; rep is just a toy. Trying to make rep meaningful beyond that is ultimately futile. Who cares if they're gaming the system? Providing good answers is still a much faster way to earn rep, and so they're never going to pass any of the big contributors.

There might be some concern these users don't really understand the system, and so will abuse the abilities they gain from higher rep. But if they don't even care enough to post thoughtful content to their own questions, they're unlikely to use these powers much if at all.

I've recently realized this view was wrong. On one hand, I'm saying these users have "invested something of themselves in the site." On the other, I acknowledge they "don't even care enough to post thoughtful content to their own questions." So which is it? Well, both. They do care. But their actions demonstrate that they don't care as much as I thought, or care about things that don't align with the site's values. Therefore they have not really demonstrated that they've earned any trust, and something should be done to correct this. Whether or not they would actually abuse the trust or use it all, even correctly, isn't really relevant until they actually earn it.

share
12  
why can't we do both of these things? –  Jeff Atwood Jul 23 '09 at 14:12
1  
I suppose you could, but is one really necessary if you also do the other? And of the two, I think this is really the better fix. –  Joel Coehoorn Jul 23 '09 at 14:43
4  
@fretje: everything starts out at 0 - if it's bad, then a down-vote will put it below. The problem then is the knee-jerk reaction to <0 scores often triggers up-votes, regardless of the post's content - in some instances this makes bad posts generate positive reputation for their authors. It's not the end of the world... it's just weird! –  Shogging through the snow Jul 23 '09 at 15:40
1  
@shog9: Yes indeed. I'm very aware of bad posts potentially generating positive reputation for their authors... that's why I advocate to NOT down-vote below 0, then they won't get any pity votes as well. –  fretje Jul 23 '09 at 16:10
4  
Well, then you've just hamstrung the voting system - you're faced with having to up-vote every other answer (for instance) to direct users away from a potentially-harmful solution. –  Shogging through the snow Jul 23 '09 at 16:29
4  
@fretje: but that action carries with it implicit approval of every other answer... Which isn't necessarily what i'm going for. Picture one flat-out wrong answer, two accurate but unhelpfully brief and incomplete answers, and one reasonably good answer: i want to up-vote the good one, down-vote the bad one, and leave the two lack-luster answers alone. –  Shogging through the snow Jul 24 '09 at 3:33
3  
Or potentally- make the pity upvotes RETURN points from downvotes, so a +1 on a negative post = +2 rep (or 5 after the change) –  gnarf Aug 12 '09 at 8:29

I think a downvote on a question should be -5 and downvote on an answer should stay -2.

If the goal is to prevent people gaming the downvote system by asking lots of silly questions with the hope of sympathy votes, you can just adjust the points just for question asking.

I'd also like to see downvotes on answers scale like this

1 -2
2 -2
3 -3
4 -3
5 -4
6 -5

So starting with the 6th downvote, each downvote is -5

share

I would only change the value of the UpVote to only +2 when the Question/Answer is below zero.

Examples:

-------------------------------------------------------------------
| Downvotes  |  Upvotes  | Rep (New system)  |  Rep (Today system) |
-------------------------------------------------------------------
     3             2            -2                 14
     1             3            20                 28
     4             4             0                 32

This way the poster do not gain any rep unless there are more people thinking that the post is right than the people who think is wrong.

share
1  
Yes, this is one of the big problems on SO -- upvotes are not weighed the same as downvotes, so controversial posts that have a net < 0 score can still amass lots of reputation for the poster. –  Ether Mar 18 '10 at 23:30

Don't increase the penalty-points per downvote. Instead consider using one of the following:

  • increase incrementally the penalty-points per consecutive downvotes

2 downvotes received one after the other causes the weight of the penalty-points carried by the 2nd downvote to increment by say +1 (i.e. 3) and so forth. This increment gets reset to 0 whenever an upvote is received.

  • increase incrementally the penalty-points per downvote based on total number of downvotes

After 9 (total, not consecutive) downvotes are received then the following downvotes would each carry an incremented weight of say +10 (or whatever) penalty-points, which will again be incremented by +10 on reaching 100 downvotes and so forth

This could also be applied to serial downvoters. The more they downvote (either consecutively or in total or a combination thereof), the heavier the penalty-points.

You may also have to keep in mind the total points each user already possesses, as it would really hurt newcomers to be hit by a tsunami of downvoters. Maybe a little red pop-up box could remind them that when they withdrew their answer/response, the weight of the negative points would be withdrawn as well. This way they can keep the answer up until they get sufficient explanation via comments as to why their answer was considered incorrect or inappropriate. Then they can defend their answer, or edit their answer to clear up any misunderstandings, or just delete it (and put in a new answer).

share

Just let users with large amounts of reputation down-vote for free. Also you could limit the amount of free down votes for these users (maybe 10-20 per day). This way users who have proven themselves over time will get a tiny bit closer to being moderators.

share

Disagree. I've always voted with the understanding that it hurts only a little. I've never felt bad about down-voting something I just didn't feel I liked. Increasing the penalty would discourage me from down-voting at all.

I have wanted to down-vote something twice in a row though, and found I can't. Considering I pay for each down-vote, I felt that I should be able to say: this is really bad. So, there'd be 1 down-vote = -2:-1 and 2 down-votes = -4:-2. You could limit it there, or let people keep going.

Now I tend to vote without regard to the cost to myself really, it's something I forget about until after the "this is not worth reading" feeling has passed and my vote is cast. Then I notice, oh yeah, I lose a little for down-votes. That might have to be made more obvious if you're increasing the cost, or changing the limit.

Definitely the rep-recalc is something of a "yikes" that could really leave people feeling stung. So that's one big reason not to do it.

share

I think the downvote mechanism should encourage useful feedback.

So how about the voter gets -3 when they downvote unless they either:

  1. Comment on the answer/question that they're voting down.
  2. Vote on someone else's comment.

If they did either of these things the voter would only get -1 rep.

This would encourage useful feedback on downvotes. It might also discourage downvoting where the reason for the downvote is trivial.

I know that the one thing that I always want to know when I get voted down is why?

share
2  
This would be very helpful. There are also cases where one person's answer is simply outdated, and they get downvoted for being wrong. This happened to me yesterday. I received a downvote, and even though the guy left a pretty generic comment, I was able to go back and forth with him and fix my answer. It would also be nice if more people would go back and revise their votes once an answer is corrected. –  rob Aug 11 '09 at 18:46
2  
For point 1, I could just add any junk comment to avoid the penalty. There's no way of enforcing that the voter leaves useful feedback. For point 2, there may not be another answer yet that deserves an up vote. –  Dan Dyer Sep 25 '09 at 18:12

I think if there are a lot of downvotes AND upvotes then that tends to indicate that it is a religious or other contentious issue. Why penalize them overall for particular questions?

I think it is a serious mistake to keep the -1 for the down voter.

Leaving it with -1 only makes it MORE enticing to do "strategic" downvoting or retribution voting.

If you're going to ding someone for 5 hitpoints then it should probably cost you a little too.

One of the reasons why there may be so many downvotes (there are pile-ons - it would be interesting to see the statistics on downvotes) is that they are essentially "free". Maybe the problem really is that it is too easy to cast downvotes.

share
4  
I'm pretty sure we'll go with -5/-2 –  Jeff Atwood Jul 23 '09 at 0:22
1  
that works better than 5/-1 I think. –  tim Jul 23 '09 at 2:21

Some users end up down voting an answer simply because they don't like the advice given by the answerer. Adding more weight would probably make these people think twice before doing so.

share

Sliding Scale Based on Voting Record

I'd like to see some kind of sliding scale that would reduce the weight of people who are super downvoters (Just IMHO, I give less weight to the people that seem to downvote a lot - YMMV of course!).

Some scale like

2-ln(1.3*dn/(up+dn))

gives a range with a nice long tail, but the exact parameters can be tuned as desired.

95% 5
75% 3
50% 2
 5% 2
share

I think part of the problem is that the effect of voting is +10/-2, but the number next to the question/answer is +1/-1. Say an answer gets 2 upvotes and 5 downvotes, they've gained 10 rep, but it their question/answer is at -3, so it's unlikely anyone else will vote it down unless it's trolling. Don't really know how you'd tackle that short of subtracting .2 for each downvote and then rounding.

share

I agree with the idea of increasing the weight of downvotes. If that goes in, I would also like to see a way to let the downvoters be made aware that the post has been updated to reflect their input. If on my 'Recent Activity' page I was alerted to modifications to answers that I downvoted, I would have the opportunity to go back, review and maybe even un-downvote if I thought the answerer redeemed him/herself.

share

You must log in to answer this question.

protected by Jon Seigel Feb 10 '11 at 18:07

Thank you for your interest in this question. Because it has attracted low-quality answers, posting an answer now requires 10 reputation on this site.

Would you like to answer one of these unanswered questions instead?

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