Should questions have votes?

We’ve talked a lot about votes and scoring on answers, with an apparent consensus. We’ve brushed past the question of voting on questions along the way.

Should users be able to vote on questions too? In both directions?

If questions have votes, what effects if any should they have beyond displaying the vote counts as we do for answers?

6 Likes

Yes. We need a way of protecting ourselves from bad questions, and from users that repeatedly post bad questions. That can’t be done without somehow judging between good and bad. The obvious way to achieve that is by voting.

Good questions don’t need much comment, and I don’t see any point in handling questions differently as long as they aren’t bad. This means we need downvotes to tell bad. However, since users disagree, we need upvotes too so that the combination of up/down votes results in a consensus.

6 Likes

Absolutely questions should have votes. In some contexts/communities it really doesn’t matter. But in others it really helps a lot. Up & Down. (Otherwise it isn’t a Vote, just a Like).

5 Likes

If I’m new to a Meta-site I’ll search it for [meta-tag:faq] tags.

If I’m new to a site I might read the list the list of tags, select the topics for a tag, and sort by Votes – to try to see the more popular/important topics.

It’s very imperfect (!) but it’s all there is apart from searching for keywords – but if I’m really new I don’t know what keywords to search for.

And doing that depends on topics having votes – as a proxy for their being popular, welcome, or important, “good questions”.

If you’ve a better way (than voting on question) to organise/present a communities best to newcomers then I’d like to hear/see it.

1 Like

Downvoting questions makes a lot of intuitive sense to me, but I think it might be worthwhile to play the devil’s advocate a bit to make sure it’s really essential. In particular, the single most obvious use for question score is to allow answerers to pick questions they can answer effectively and usefully, but, after many long hours of digging through question lists, analyzing patterns, considering my habits, and so forth on multiple sites, I never found question score anywhere near as useful on any SE site as I would have expected. It simply doesn’t help much for its number one use case.

This makes a lot of sense, but it does occur to me that there are two other signals that SE already uses to good effect in its automated Q-ban system: question closure and deletion. It’s not very clear yet how aggressively we’ll be deleting questions, manually or automatically, and the deletion signal apparently isn’t weighted very strongly in SE’s implementation, but closure is still worth considering. If we can do a better job of recognizing specific show-stopper flaws in questions that make them unsuitable for answering (temporarily or permanently) then a Q-ban system could reasonably use that data to at least approach SE’s speed and accuracy of automated bans.

What’s more, if there were no up- or down-votes on questions, any potential reasons to close would be more thoroughly considered by anyone who disliked the question at all, making closure a more reliable signal and keeping the site tidier. Admittedly, this could result in overreacting, especially early on before a mature site culture handles voting norms, so leave-open and reopen mechanisms would need to be robust. And the difference between a temporary hold to get some needed editing done and a mistaken close that was reversed might also need to be made explicit in the system.

Have an easy-to-access sort by questions’ answer quality. “Questions By Best Answer”, or something. A canonical question will inevitably have a highly-scored answer. In fact, usually the answer score is at least a slightly better indicator of the question’s true usefulness than its own score. A +100 question with a +80 answer may be very well asked, but most people can’t benefit from it as much as a +80 question with a +100 answer, never mind a +80 question with a +200 answer.

Strictly speaking this kind of sort is actually also useful for answerers on the main site. I have often wished to have an easy way to find questions with inadequate answers so as to contribute where there was a need, but searching for these is extremely non-trivial on SE. (In fact, I think you have to use either SEDE, on a week-long lag, or write an API client yourself.) Being able to sort by answer score in either direction lets you see either the best- or the worst-answered questions, and thus either learn from or teach about some of the most fertile questions respectively. (Being able to further sort by views, subscribers, etc would help a lot as well.)

4 Likes

Yes the principal immediate effect of a downvote is to send a message, “I don’t like this” – which gets that off your chest, for what that’s worth, and is of course received as unwelcoming or hostile.

I’ll upvote a question for two reasons:

  • I couldn’t answer this but I hope someone else will
  • I answered this and given that I found it worth answering it was worth upvoting too

Downvoting is much rarer:

  • I think the OP (probably a known user, not new) is misbehaving somewhat in posting this question – e.g. trolling – behaviour that I want to discourage and to warn off other users from participating in

Downvoting isn’t exactly “constructive criticism”.

That’s not entirely true, but it be mostly true and very good idea anyway.

A potential counter-example, my own top-voted question – an outlier, abnormal – has a score of +62/-1 and 16 answers of which the top score and accepted answer plus +38 (the next was +19). It was actually the first question I asked on the site, so self-evidently of interest to a newcomer or novice … or somewhere in a sweet spot between ignorant and knowledgeable (and it’s the highest-voted question on the site).

It is very interresting to follow votes on questions and answers as function of their age.

  • On the on hand: Questions receive much less votes than answers (possibly partly the reason why SE/SO has decided to change the rep per vote).

  • However, on the other hand, for the questions and answers from the first years the questions have a stronger increase in the number of (up)votes that they receive in comparison to the answers, and as a result the questions have acquired more (up)votes than the answers although initially it were the answers that attracted more votes.


I will dig up the data some day to see if the ratio of question score and max answer score is dependent on the age of the question. I suspect that it plays an important role.

1 Like

I don’t remember how the voting in this question increased over time.

It’s kind of a good question because it’s one which everyone will have at some time, but which it’s difficult to answer like really well.

Such a question with an easy answer – or which, it is possible to answer “canonically” – often doesn’t need asking, or is a trivial to answer or a really novice question, because there’s already some canonical answer which everyone might know already.

I said that’s an outlier; but that might be a reason why the answer didn’t score so high (and there are more than dozen answers) – it’s like “a genuine question”.

I’m inclined to agree with all of your reasons against impementing a voting system on questions. I think it is easier and fairer to deal with question removal using tools such as voting for closure, etc.

Votes on questions help sort good/bad and interesting vs not. Downvotes vs closevotes are milder way of steering on topic questions.

If we have some sort of rep system then I think that whatever imaginary internet points people get for writing questions should be the same as for answers.

After all good questions,

  • Are not duplicates
  • Are on topic
  • Title is clear and matches the body
  • Tagged properly
  • Some level of research has been done before asking

Giving points encourages people to stick around and ask more good questions, so I don’t see a reason to value good questions less than good answers if we do end up having a rep system.

1 Like

Do they, though? Do they really? In my experience, if I see a highly-upvoted question on SO, it’s already been answered so well that there’s no point in my attempting to contribute. (And, as I’ve alluded to up-thread, searchers with the same problem can more easily find the question either by web search or by sorting by answer score, either of which will get them better answers.)

For that matter, even a question with a downvote or two is often worth editing and answering. So in practice I’ve always found that question score seems to save me very little time trying to find a question to answer. I have to actually look at the title, the summary, and then open the full question if it seems promising anyway. It’s quite possible I’m an outlier, though, so perhaps you could explain in better detail how you visualize question sorting to work, at least on the kind of low-traffic sites we’ll have to start with.

1 Like

It depends,

  • As a question asker, scores can be used to guess at the type of questions that will be well received by the community.
  • As an answerer, a question that gets upvoted quickly has a better chance of the answers being upvoted as well.
  • If you are coming from Google it doesn’t really matter.

On a site with a low question rate, you end up reading all the questions anyways, so the score really doesn’t matter if you are looking for questions to answer.

About this, I want to remind of my earlier suggestion that upvotes on answers also give points to the author of the question, on the base that a good question is one that triggers good answers.

Counterpoint, all of the highly downvoted questions on meta.se that SE has been pumping out lately where the answers gets tons of upvotes and the question tons of downvotes.

I don’t see why the question author should get rep for that.

I do find it unfair when 5 people answer a question and none of them upvote, but I am not sure that what you are proposing solves that problem.

1 Like

If the reason is gamification wherever possible then should we also apply a votes and rep system to the contributions to github and the codidact code?

Have all those other free open source projects and volunteer communities without the stackexchange/quora/popular voting been doing something wrong?

For many years people on the internet have been providing/generating intelectual property for free on the internet. They didn’t need money, and certainly they didn’t need virtual kudos.

Is such votes-based gamification really needed to create a healthy community with useful (semi)professionals that wish to spend their free time to make something pretty?

Is such votes-based gamification really helping to shape the community in the way that we want? (or will it lead to heavy micromanaging of processes that will happen naturally anyway and might even work better when they are not placed in a simple systematic framework like plain voting?) (To imagine how voting could ruin something: Imagine how posts will change on Facebook, or other social media, when likes would be recorded in a rep-system - In a certain way this exist already and it are those competitions where people have to gather the most likes, I have even seen this happen in a ‘scientist of the year competition’ where people were mobilising their friends and acquaintances to vote for them, no matter how good scientist they are)

1 Like

Those are a side-effect of SE posting announcements of what they intend to do or have just done on Meta. Meta rep, and quite possibly scoring, should probably work differently anyway, and without a for-profit business running things, those specific examples of the Meta community expressing extreme disagreement and disapproval wouldn’t appear.

That does raise an interesting challenge, though, to getting rid of question scores: if there’s any commonality between meta and main, as has elsewhere been suggested, then meta will still likely share the general pattern seen on SE where the initial question is seldom just a question, but an implicit statement, proposition, value, or suggestion that is then voted on almost as though it were an answer. It might actually be a good thing to make that impossible and force anyone who wants to see voting on such implicit recommendations to break them out in actual answers per se.

1 Like

Meta works differently, and therefor cannot be a counterexample. Note that on SE site metas, votes don’t earn you reputation at all; I’d argue that on the global meta, this should be the same.

1 Like

We have discussed to have different posting types on meta, questions and discussions. If there’s an explicit discussion type, it is less likely that discussions will be shoehorned into the question form.

Also, there’s the suggestion to have blog posts as a separate on-site post type (that is, integrated to the site, not an external extra like on SE). Announcements (such as the heavily downvoted one in the example) would then be posted as blog posts.

Whether blog posts should get voted on is a different question (it is well possible to have votes on blog posts but not questions).

Okay, here are some highly downvoted questions with upvoted answers.

I agree - question score can be especially useful for finding good questions that don’t yet have answers. On several sites there’s a long tail of zero/one score questions that are clear enough, on-topic, &c., & therefore haven’t been closed, but which no-one’s thought worth the effort of answering.

2 Likes