58

The official reasoning behind doubling the reward for question upvotes from +5 to +10 was - when I understood it correctly - SE (inc.) wants to encourage the community to generate more "good" questions. But assumed the increased upvote reward really generates more questions in total, following the same logic, shouldn't we expect also to get more bad questions?

Which leads straightforward to the question, why SE did not double the reputation loss as well from -2 to -4 for downvotes on questions? Or at least to -3?

  • 4
    I completely agree: see this question of mine for a verification: meta.stackexchange.com/questions/338293/…. It now takes 5 downvotes to "undo" an upvote. Personally I think downvotes should be symmetric: that would encourage authors to spruce up their questions. By all means raise the reputation threshold for downvoting. – Bathsheba Nov 15 at 17:11
  • 26
    There's no evidence that people generated questions that were any better before we reduced rep for questions from +10 to +5. The original rep reduction was, in fact, motivated by the desire to improve question quality. The real reason for the current change is to encourage more new users to engage. – user102937 Nov 15 at 17:13
  • 4
    @Rob Mo' usas, mo' $$ :-P ... – πάντα ῥεῖ Nov 15 at 17:20
  • 4
    @RobertHarvey: so you are saying the SE blog post is trying to mask a hidden agenda? Well, I have to admit, I was thinking quite in the same line of direction. Hope SE won't block us both here for a week for writing such "snarky comments". – Doc Brown Nov 15 at 17:21
  • 1
    @Doc Not hidden, follow what Mr. Chandrasekar has to say. – πάντα ῥεῖ Nov 15 at 17:22
  • 23
    The next step is obvious: cancel downvotes altogether. Downvotes are evil and unwelcoming to the people who Stack Exchange cares about. (Hint: not us) – Shadow The Princess Wizard Nov 15 at 17:28
  • 1
    Because the change was about making reputation for questions equal to that for answers, which is now is for positive as well as negative votes. – Chris says Reinstate Monica Nov 15 at 17:29
  • 3
    @ShadowThePrincessWizard That would be to obvious. I would be "easier" to first re-add the "don't show downvotes" feature that prevents questions from becoming negative and then cancel downvotes when the question is at 0. And after the backlash has settled, then cancel downvotes altogether, because "good questions gain upvotes and bad question can stay at 0" (we had that argument on several feature-requests already). – Tom Nov 15 at 17:31
  • 1
    @Tom couple of years ago this might be true. But now they land the punches all at once, not one by one, and without letting the community to recover in between. – Shadow The Princess Wizard Nov 15 at 17:34
  • 3
    @Shadow And probably change the naming of votes to kudos or karma :-P .. Oh tempora, o mores. – πάντα ῥεῖ Nov 15 at 17:36
  • 3
    @πάνταῥεῖ when there won't be downvotes anymore? It's more than obvious: one will be able only to Like question or answer. And no more silly arrow too, but rather either a heart or thumbs up icon. – Shadow The Princess Wizard Nov 15 at 17:47
  • While I don't like the new question ratio, I would note it is the same as the answer ratio has been this whole time. Myself, I would prefer a 10:4 upvote:downvote ratio for both Q & A, but I doubt that will ever happen :) – De Novo supports GoFundMonica Nov 15 at 17:51
  • 1
    @Shadow Yes, like on all those "social media" sites. Correctness and non subjectivity (technical) would be thrown overboard in a hurry. – πάντα ῥεῖ Nov 15 at 17:52
  • @πάνταῥεῖ those were thrown overboard already, long ago. – Shadow The Princess Wizard Nov 15 at 17:57
  • 4
    @KevinB: That's not entirely untrue, but we believe improving the ask page will do far more to increase participation long-term than this one-time change. Making it easier to ask good questions also helps increase answer participation which is a concern we have within the company. – Jon Ericson Nov 15 at 19:30
31

In our way of thinking, this change was a rollback of a decision made in 2010. That change didn't adjust the weight of downvotes. We think it did not change people's behavior and the problem it attempted to solve (people spamming the site with meh questions in hopes of getting enough upvotes to make up for downvotes) was better solved with rolling rate limits. In any case, the payout change didn't seem to do much to discourage bad questions.

The basic principle here is we don't intend to remove reputation from anyone because of loss aversion. So downvoting a question will continue to cost nothing, downvoting an answer will cost one reputation point and receiving a downvote on either type of post will continue to cost two reputation.

The way to think of these penalties is that they are tokens rather than true costs. Getting -1 reputation makes you think twice about, say, downvoting other answers to make your answer look better. Getting -2 makes you think twice about posts that tend to garner downvotes. Most people don't do the mental calculation to determine if the risk of downvotes is outweighed by the benefit of upvotes.

  • "the payout change didn't seem to do much to discourage bad questions." <- Do you mean back then, or now? – einpoklum - reinstate Monica Nov 15 at 18:10
  • @reinstateMonica: Back then. – user102937 Nov 15 at 18:14
  • 5
    @einpoklum-reinstateMonica: Back then. (But probably this time too. ;-) – Jon Ericson Nov 15 at 18:17
  • 3
    The "payout" for asking a question should be getting a useful answer, not rep farming. They should reduce question asking to +0 points because asking questions is easy, giving knowledgeable answers takes expertise and work. – SurpriseDog Nov 15 at 19:07
  • 5
    @SurpriseDog: This is exactly what a plurality of questions have earned before and after the change. But don't you think that some questions are better than others and that the extra effort put into those questions might be worth rewarding? Also the value of questions is that they prompt answers that would not otherwise exist. If self-answering were common, I think I could imagine a world in which upvoted questions have no value. But until then it's short-sighted to say they don't. – Jon Ericson Nov 15 at 19:21
  • This is a tangent, but do you have any data on voting patterns for self-answered questions? My anecdotal experience is that people are happy voting on those questions but often reluctant to vote on those answers, even when the answer arrives days later after a bunch of research. I think seeing someone own both the question and answer can be interpreted as attempting to rep-farm, even in cases where it's not. – Troyen Nov 15 at 22:12
  • @Troyen: It certainly seems your intuition is correct. There's some odd things about that query, including the higher than expected average question scores. I think it's the result of highly voted questions being answered more often and questions be counted once per answer. Average answer scores seem high too, however. – Jon Ericson Nov 15 at 22:38
  • @JonEricson Thanks for the query, the results are interesting. My anecdote came from Arqade where those ratios are a little closer together. There's a lot about each site's voting culture that I'd have to factor in for a more comprehensive study. I might see later if I can build off that query to look at e.g. highest-voted answers only (to reduce the multiple-answer muddling of the answer score, or maybe break out by number of answers for more direct comparison) and differences between tags (minecraft and non-minecraft can seem like very different voting cultures). – Troyen Nov 15 at 23:21
  • @JonEricson Questions don't get upvoted based on effort or quality. They get upvoted by how many people click them and how many people like the Title. The body of the question has nothing to do with it. That's why HNQ do so well. – SurpriseDog Nov 16 at 16:57
  • @SurpriseDog: I'd love to see the research that leads you to that conclusion. – Jon Ericson Nov 16 at 17:23
24

This was about returning parity to the value of questions-vs-answers (like it was years ago before the "question value" was reduced to +5). It's a statement that well-asked questions are just as valuable as well-received answers.

So it wasn't about "doubling the reward" (1) per se; that's just how the math worked out.

(1) Incidentally, we didn't "cut the reputation loss in half" back then, either.

4

SE wanted to encourage SE to generate more good questions, which is different from discouraging bad questions.

On a separate note, from the same article:

Our goal is to have the question asking process be painless and beneficial for new people

New users often feel unwelcome when they don't read any of the FAQs, ask a duplicate or unfit question & get downvoted to heck. Increasing the value of a downvote could make this worse.

  • 3
    Why should we care if lazy people feel unwelcome? A meritocracy is perfectly welcoming: everybody is welcome, as long as they do their homework first. – Eric Duminil Nov 22 at 13:42
2

The official answer to this is: Because of loss aversion SE does not implement a change that can actually take away some reputation from the users. So even when you have had a negative net reputation from questions (which is pretty hard, but perfectly possible), they do not want to make the situation worse.

  • 2
    "SE does not implement a change that can actually take away some reputation from the users" ... they already did that when decreasing the amount of reputation gained after editing a documentation page. So, wouldn't be the first time. – Tom Nov 15 at 17:24
  • @Tom It is their stance for this time. – jknappen - Reinstate Monica Nov 15 at 17:25
  • 5
    @Tom: We considered Documentation a beta feature, so it seemed reasonable to adjust the reputation system as we discovered problems. Even so, we allowed people to keep their Documentation reputation when we ended the beta because it's not sound to remove something a person earned for reasons completely outside of their control. I'm not saying these are easy decisions or even obvious. But there are behavior economics reasons for making them. – Jon Ericson Nov 15 at 18:28

You must log in to answer this question.

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