39

In the Stack Exchange voting system, we have two classes of votes: standard up / down votes that anyone can cast, and special accept votes that only the asker can.

Although accepting an answer does not always pin it at the top of the answer list, that green check mark does send a strong signal about the correctness of the answer. This gives a single person much greater weight than the rest of the community. That green check mark can seem confusing to visitors, who see potentially wrong or much lower quality answers marked as being correct. I've heard this from several friends recently who were surprised when they came across factually wrong answers that had been downvoted, yet still had that check mark next to them.

Moderators regularly field flags from even more experienced users who demand that the accepted answer be changed from one that is incorrect or even harmful to another. We decline these, because we're not supposed to be making technical judgments like that and because we have no way of accepting answers on behalf of the asker. However, these flags do show that some people have problems with the presentation of accepted answers vs. the community voting on all the others.

There's also all the fun that came with arguments over accept rate, and people badgering askers to accept their answers.

Questions on the Stack Exchange network are intended to be broadly useful, to help out many others in the future beyond the person asking the question. If that's our goal, why should the person asking the question have any more say than the rest of the world? If it's a broadly applicable question, why should one person be able to highlight something as correct simply because they asked it first?

I guess what I'm asking is: if we did away with accepted answers, converting them to upvotes where appropriate, and only let community voting sort out the best answers, would we lose something important? I thought I'd ask after reading Rosinante's answer to this recent question, because I wasn't sure how I felt about accepted answers anymore.

  • 5
    I'd say most of the time the checkmark marks a correct answer. The exception is the opposite. Why not rather put accepted, downvoted questions to the bottom i.e. – juergen d Dec 30 '13 at 18:29
  • 9
    Well, converting accepted answers to upvotes would cut Jon Skeet's rep in like half, overnight. It would also hurt users in lower volume tags/sites a lot more than higher volume tags/sites, where a higher percentage of rep comes from acceptances than from upvotes. – Servy Dec 30 '13 at 18:29
  • @Servy: It could be rewarded as bounty to get around that. – juergen d Dec 30 '13 at 18:31
  • 21
    I'd be more inclined to just remove the "accepted answers are shown on top" feature, rather than messing around with the rep awarding features. – Servy Dec 30 '13 at 18:31
  • 1
    Oh, and it would make earning rep on brand new sites much harder; it takes 10 rep to upvote after all. Sites get established using acceptances for early rep. – Servy Dec 30 '13 at 18:32
  • 15
    @RyanCarlson Telling people to stop downvoting something is a good way to increase the number of downvotes it gets. For the record though, the question is encouraging removing acceptance entirely, and is thus making a proposal, which is subject to approval/disagreement votes. It's simply in the nature of how the question is asked. – Servy Dec 30 '13 at 18:46
  • 5
    @RyanCarlson - As Servy said, I don't mind the downvotes at all. I do implicitly suggest something, and I take the downvotes as disagreement for that proposal. You can't take downvotes here personally. – Brad Larson Dec 30 '13 at 18:54
  • 1
  • 1
    -1 for whining about votes. Oh, wait, that wasn't you. Nevermind. – John Dvorak Dec 30 '13 at 19:14
  • 1
    What if the accepted answer is outdated and is no longer correct. An example would be an illustrated solution with a much older version of a language. Is there a provision for revising old Q & A? – gsndev Dec 30 '13 at 19:47
  • 1
    @gsndev The answer would still be useful to people using that version of the language so an update to add what version of the language the answer applies to could be useful. – Joe W Dec 30 '13 at 20:10
  • 2
    I think there's a flip side to this. Yes, sometimes the accepted answer is wrong or not as good as the highest-voted answer. But sometimes the accepted answer is better than the highest-voted answer. This seems especially true when some answers say lots of good, wholesome, and correct things (thus earning lots of votes) but that don't actually answer the specific question being asked. I think there's value in having both signals. – Adrian McCarthy Jan 14 '14 at 22:58
  • 1

11 Answers 11

21

Because Stack Overflow Is a Problem/Solution sort of website, the OP of a question is uniquely qualified to select a Solution because he or she is the one who is actually facing the problem.

Most of the early votes on answers come from other potential answerers, who are not actually facing the problem, but are offering a solution.

They cannot verify the solution as truly as the OP can, because they are not actually facing the problem. They could very well be coding in a different environment or they might be missing other details that matter.


One of the things I like about getting technical help from Stack Exchange and sites like it is that you're not reading about theoretical problems, but you're reading about actual problems that people have actually faced. If anyone but the OP is able to select a solution, some of that "real-ness" gets lost


That being said, I have begun to doubt that that's the case on other Stack exchanges that are not comprised of Problem/Solution types of questions. When someone is asking for a ruling for a popular card game at https://boardgames.stackexchange.com/, than the asker is actually not uniquely qualified.

This are more 'just curious' sorts of questions rather than problem/solution questions.

It might make sense for individual Stack exchanges to be able to configure how answers are selected, but for a Problem/Solution website like Stack Overflow, it wouldn't be appropriate for anyone but the OP to select an answer.

  • 20
    At the same time, they are also demonstrably uninformed about the subject, given that they needed to ask a question about it. – Servy Dec 30 '13 at 18:33
  • 18
    they didn't know when they asked. But they are in a great position to test the answer and confirm it worked. Others are often just thinking "yup, that looks like it should work." The asker usually tries what is suggested in their own live code. – Kate Gregory Dec 30 '13 at 18:35
  • 1
    "Stack Overflow Is a Problem/Solution sort of website" -- or is it? (I know many questions have little or no value for future visitors, but that is also exactly why I find it's becoming harder and harder to quickly find an answer.) – Arjan Dec 30 '13 at 18:56
  • 6
    In 99% of cases, the OP shouldn't be uniquely qualified to select a solution. If the OP has information about the problem that others reading it don't, such that others can't assess whether a given solution will help the OP but the OP can, then that almost certainly means that the OP has asked a bad (and quite possibly close-worthy) question. Now, it may be that the OP typically cares about and understands the problem more than most voters, and so will apply better judgement in selecting their favoured answer... but does the evidence really bear out that theory? – Mark Amery Sep 27 '15 at 15:01
  • 1
    Yes, it is a QA website, but at the same time we continually stress the importance of the idea that the QA isn't so much (at least primarily) about OP getting an answer to their question as it is about farming good questions, and related answers, that are broadly applicable (and helpful) to other people. In my mind, that makes the green checkmark almost entirely irrelevant. – J... Sep 28 '15 at 3:39
  • 2
    @J... The OP is a prototype for all future consumers of that question. The OP is the person who is most like the people who need an answer to the question. The other potential answerers, who provide the most early of up votes, are not. – Sam I am Sep 28 '15 at 14:05
  • @J... also, this is not a QA website , despite the number of people that use it as one. It's a Q&A website. – Sam I am Sep 28 '15 at 14:06
15

Yes

Accepted answers are still very helpful to the community. It helps increase the signal/noise ratio by showing that one answer is better than the rest, or was more helpful to the OP. There are edge cases, of course, with non-accepted answers having scores of 2x or 3x more than the accepted answers, but for the vast majority of questions, the green checkmark next to the answer, coupled with a high score, tells me that that answer is the solution that I should use.

There's also the past reputation to think about. People who regularly hit the rep cap rely on bounties and accepted answers for a significant part of their reputation. Removing accepted answers would have a massive impact on the rep of the community, people with high rep as well as low rep.

I know that SO questions are meant not only for the OP, but also for other people coming to the page, but it is up to the OP to tell what answer helped him the most, and therefore is most likely to help others the most. If there's another answer with a significantly higher score, perhaps we should have some indication of it (although there usually is, in the question, there are often comments saying "you should accept @betteranswerer's answer", and the like

11

I think the problem is the outward appearance of the accepted answer, and not the fact that it exists. To knowledgeable users (users knowledgeable in the Stack Exchange format), we understand what the check mark actually means and how we should use it to evaluate the answers ourselves.

However, to someone who is not familiar with how the site works, it lends the appearance of a community supported correct answer given the presence at the top (in most cases) and the fancy green check mark next to the answer, so it's nature can be misleading.

So maybe it could use a face lift for appearances sake without changing the underlying functionality.


More often than not, I think the accepted answer is a good answer, and I'm sure there are badly accepted answers caused by the OP not understanding the meaning of the check mark, or even worse, was badgered into accepting by another user, or users being spiteful and refusing to accept an answer from a specific user or other childish nonsense.

But I think there are 2 larger problems:

  1. The OP is not always the most qualified person to decide which is best. Especially in many questions from low rep users, the accepted answer is the first answer that solved their specific problem. It might be a bad solution in the long term, or it might be bad practice, but it solved the problem first so the OP quickly accepts it and runs off to create their next issue.
  2. The longevity of the site becomes a problem. 2 or 3 year old answers may no longer be the best answer due to changes in the technology but no one revisists these questions, or because the OP has disappeared (or doesn't care), there is no one around to change the checkmark when a newer (and better answer comes along).

This comes back to appearances. Someone coming in from Google might not pay attention to dates or votes. They just see a solved question with the same problem they have and they just go ahead and use the answer that the original asker said solved their problems.


In the end, I do not think the accepted answer feature a bad feature, but maybe it could use some tweaking to ensure the best answers are seen first.

Some thoughts (in no particular order):

  • Institute a community solution feature. This would be marked differently than the accepted answer and would not come with a rep bonus, but would be a signpost to post incoming users to an answer that the community thought was best. It could be awarded simply by the answer with the most upvotes after a period of time and would change constantly based on the highest scored answer, and it could be the same as the accepted answer.
  • No longer let the accepted answer bubble to the top or allow it to be placed below an answer that had significant more upvotes (maybe 2x the votes)
  • 3
    Yeah, I think it's the perception that the big green check mark provides that's the core issue here. Comments like "This should be the best answer." seem to conflate acceptance with which one is the true best answer to a question. Those who understand the difference can tell them apart, but your average Google visitor has their eyes drawn right to the answer with the very visible mark next to it. – Brad Larson Dec 30 '13 at 19:15
  • I don't know about other people, but I groked the difference between the check mark and the upvotes right away. But that's probably because the first time I've seen that paradigm, it was a verbose "accepted answer", instead of a symbol – Sam I am Dec 30 '13 at 19:22
  • 2
    Maybe instead of a check mark change it to text that says "Answer chosen by User" or something similar. This way you still keep the accepted answer but make it more clear about who accepted it. – Joe W Dec 30 '13 at 20:13
  • I don't see how "Answer chosen by User" would have any value at all to readers. If a unique signal on the page (the green tick) doesn't identify the "best answer" then it should be removed because then it is just noise. Even if you use it like a "support ticket closure" that would be bad because that would suggest that better answers were not welcome. When there is a "best answer" award to strive for, people will work harder and harder to post better and better answers. Without that award, people will just post quicker / sloppier answers to earn the early green tick. – mickmackusa May 27 '17 at 15:51
10

The accepted answer mark should have one purpose, to show which answer solved the problem the user had. But that doesn't mean that this answer is actually the most useful to all the later visitors that view the question.

I don't think eliminating the accepted answer concept entirely is a good idea. We would lose some information about which answer worked, and it would cause some drama because it would reduce the amount of reputation answerers could earn (especially the interaction with the daily reputation limit).

But what we should do is eliminate the effect the accepted answer mark has on sorting. The default sorting should be entirely determined by voting, and not put a lower voted accepted answer above a highly voted, not accepted one. The accepted answer mark should still be visible on the answer, it should just not affect sorting by votes. I believe voting to be more accurate in identifying the best answer than the accepted answer mark.

  • 1
    How much of a bonus does being the accepted answer give for the sort order? – DanTilkin Dec 30 '13 at 20:36
9

In ordinary cases, accepting answers is neither useful nor harmful, since the answer that gets accepted is likely to either be the highest-upvoted answer anyway, or be one of the obvious potential best answers to the question.

There is one situation nobody has mentioned in which the ability to accept answers is important and useful, which is when somebody comes along to an old question that has an ancient, highly upvoted answer, and posts an even better answer. In this situation, a diligent question owner can accept the new, superior answer and make it the most visible answer on the page, like it deserves to be. Without this ability, new answers to old questions with many highly upvoted answers have little chance of being noticed. This can be especially important when the correct answer to a question changes over time due to technology evolving, and the highest-upvoted answer becomes obsolete. For an example, here is an answer of mine that has a fraction of the upvotes of the highest-upvoted answer, but is rightly and usefully accepted.

There is also the situation that many people have mentioned in which accepted answers are harmful: when a question asker accepts an inferior or outright wrong answer.

In order to preserve the benefits while mitigating the harms, perhaps what we need is a heuristic that detects when a bad answer has been accepted and doesn't sort it to the top. For instance, if the accepted answer has negative score or the rate at which it has gained upvotes over time is less than half that of the highest upvoted answer, the accepted answer could be sorted according to score like self-accepted answers are.

7

The community often does a great job of pushing the better answers to the top, and pulling the worse/dangerous answers to the bottom. I believe that is working as intended.

However, the OP has the option to say, "this answer helped me the most."

It could be that the answer is dangerous, or against best practices, or introduces some security hole, but what the OP is saying is that this answer helped them out.

I realize that this may be dangerous to those that skim answers, looking for a green checkmark to try and implement it, but the really dangerous thing there is that those looking for answers are willing to take the first solution they see as "correct" without much background research.

So to be frank, I'm not worried about a checkmark providing a bad answer unless the person searching for it doesn't continue to do their research.

Why not let accepted answers that are downvoted sink to the bottom, and answers that are more favorably upvoted float to the top? Give it a ratio of 2:1 - if an accepted answer has two downvotes, then it should appear lower than a question whose total accept score is equivalent.

2

Lots of good points being made for and against, but at the end of the day, the accepted answer is necessary. It's a critical part of the question and answer model; there is a real need for askers to be able to clearly indicate which answer helped the most. As a concept, it needs to stay.

This discussion seems to stem from the following:

  • The checkmark sends an extremely powerful signal that this is the answer to the question.
  • Random visitors from search engines probably won't understand the finer points of acceptance versus voting.
  • It's difficult, and often impossible, to get acceptance removed from a completely incorrect answer.
  • As Sam I Am points out, some sites don't fit so well with the top-answer-selected-by-the-asker model.

I can't come up with any way to substantially address items two through four that's both feasible and reasonable. But tweaking the strength of the checkmark's signal is certainly doable.

Since it seems like the SE team wants to utilize the user preferences section more, how about having a user-set option for whether or not to pin accepted answers at the top of the list? Allow the default value for the new preference to be set on a per-site basis, by the site's moderation team.

Many users - and all anonymous ones - would never experience anything other than the default setting. This means the site's community decides (at least somewhat) on the level of importance of the checkmark, but experienced users can make an informed choice about data presentation.

This doesn't address the incorrect-but-accepted issue on a site like SO (where I assume accepted answers would remain on the top by default). However, I'm not convinced that the scope of this problem warrants anything approaching drastic, sweeping changes. As juergen d notes, this is the exception rather than the rule. And an incorrect accepted answer is a great opportunity for education on why it's incorrect.

1

Why not give the asker the ability to add an extra upvote (like a bounty) to the "accepted" answer?

It would look like the blue bounty bubbly but only show [+10] in the bubble.

This "bounty" won't take any rep from the asker, but would act otherwise in the same fashion.


So in conclusion -- The accepted answer checkmark could be removed, and this fake pseudo-bounty would replace it.

  • @Krampus You don't use Linux I assume? – qwertynl Dec 30 '13 at 18:54
  • Oh? Were we making a pun? – ale Dec 30 '13 at 18:55
  • ;-) Mayyyybe :-P – qwertynl Dec 30 '13 at 18:55
  • Punnish him for making that edit! – Amal Murali Dec 30 '13 at 18:57
  • Ummmm... How do you suggest I do that @AmalMurali ? – qwertynl Dec 30 '13 at 19:00
  • @qwertynl I am sure if you looked far enough you could find a fortune trove of bad jokes to echo in the general direction of Krampus. One can certainly find material in unix to make puns (No rule to make target 'pun'. Stop.)... If worst comes to worst, you could always send him to read this in its entirety. – user213963 Dec 30 '13 at 19:48
1

Part of the problem seems to be the green check mark seems to mark the solution as the correct and best answer. Now if this was changed from a check mark to text that said "Answer chosen by user" or similar text it should help solve the problem. This would allow a user to select the answer that best helped them without making it seem like it was the best answer to a user unfamiliar with the Stack Exchange format and also keep the current reward to the user who provided the answer.

I know this won't solve all problems that people have with the problems that some accepted answers have but it looks like a solution that can get movement in the right direction and keep the accepted answer system.

0

I would say that accept mark should count as 1.5 upvote when sorting. It would make it an equivalent of "upvote + tiebreaker" and keep accepted answer on top in uninteresting, lowly-visited questions where OP is probably only one who really really examined said answer and not just upvote when passing by because it "looks legit". And on the questions with strong community involvement it would not afefct things too much, making it possible to push down answers generally found to be bad.

Maybe this number should really be 2.5 or 3.5, depending on how much community votes SE team will think is statistically more significant than OP examining an answer. But that's just fine-tuning it.

At the same time it should be clearly pointed out that OP should not accept an answer until he has tried it and actually can testify he found it working. Not "the closest thing to solution", not "best of advices even if not quite right", no. Just "yes, it worked and I can show it to you if you want" kind.

Last but not least, Community Wiki questions should never have accepted answers. Rep from acceptance that happened before CW status should be kept, but checkmark and it's sorting effect should be gone. With possible exception for a moderator to move checkmark around, but CW acceptance should not give rep.

-1

Marking a question as one which has an ‘accepted’ answer also serves to help the review system — if done correctly for other reasons, of course: when a new user posts an answer to a question which already has an accepted answer, it is a good signal that their answer needs notably attention than other ‘new’ or ‘late’ answers to unresolved questions.
Similar to what another user described: either

  • the new answer could be better than the previously accepted one,
  • or it could be a sign that a user doesn't recognize the purpose of accepting an answer — and so posted one of theirs regardless.

You must log in to answer this question.

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