-16

I just took accepting answers for granted on Stack Exchange, but then I came across Quora, a Q&A site that doesn't have accepted answers (Just to be clear, I'm not suggesting that we make Stack Exchange like Quora - there are tons of reasons why SE >> Quora - but we should be open to other ways of doing things). So, I thought about the pros and cons of accepting answers. What I have here is not an exhaustive list, but just what I came up with off the top of my mind just now.

Cons of Accepting Answers

  1. The best answer isn't always accepted. Often, the accepted answer and highest voted answer are not the same. For questions with 20 views, this is okay, but on many popular questions, the accepted answer isn't the highest voted. In my opinion, the highest voted answer is the best answer, since I value the community's opinion above that of the OP. There are cases where this isn't true, but the vast majority of the time, it is.
  2. The Accepted Answer goes to the top. I wouldn't have any problems with point 1, except that the accepted answer is always at the top. I think that the best answer should be at the top, because some people don't scroll down below it (myself included, oftentimes).
  3. Accepting seals the question. What I mean is that to accept an answer to a question is almost the same as closing it. People may still go to it and look at it, but definitely fewer. It reduces the traffic.
  4. It discourages better answers. Accepting an answer doesn't mean it is the best possible answer, it just means that it satisfies the OP. But since accepting an answer seals the question, a person who can provide a better answer is unlikely to stroll along and give that better answer.

Pros of Accepting Answers

You guys already know these, so I'll just go over a few of them briefly:

  1. In many cases, the "right" answer is showcased by sending it to the top of the list.
  2. The answerer of the "right" answer is rewarded with 15 rep.
  3. New users can reward helpful answers, if they don't have upvoting privileges yet.

Possible Solutions

I'm not sure how to solve the problems of accepted answers without abolishing them - but I think the pros, primarily the first two, outweigh the cons - which is why I'm asking this question here. How can we fix the problems with accepting answers without getting rid of all the benefits?

39
  • Maybe show a line below questions with accepted answers that encourages those with better answers to still post them.
    – jmort253
    Jan 8 '14 at 1:28
  • @psubsee2003 They're not duplicates. That question asked if they are useful. This question declares that they do cause problems, and asks how we can solve the problems.
    – anon
    Jan 8 '14 at 1:34
  • 2
    Your #1, #3 and #4 are incorrect. The answer the poster of the question found to be the answer to their question is what's being indicated (not the best - the best is indicated by the number of votes it receives), and accepting does not "seal the question* because answers can still be posted (and voted on) whether there's an accepted answer there or not. I've posted many answers where there was already an accepted one (and also had the accepted answer changed by the poster afterward more than a few times, which refutes your #4).
    – Ken White
    Jan 8 '14 at 1:35
  • @cloudcoder2000 if you look at the answers, a number of users discussed the use and the solving the problem. We don't need multiple discussions on the same subject. Jan 8 '14 at 1:35
  • 4
    "Accepting seals the question." - I'm not sure this is universally true. The acceptance checkmark can and has been moved if a better answer came along.
    – Adam Lear StaffMod
    Jan 8 '14 at 1:37
  • 1
    You're making a lot of assumptions about player behaviour (certainly #1). I can't support that with my own experience, do you have any objective numbers that do? Jan 8 '14 at 1:37
  • @JeroenVannevel I'm not good enough at SQL to use SE Data Explorer to find out. But I'd be glad if someone more qualified than I could do that research.
    – anon
    Jan 8 '14 at 1:39
  • 1
    @KenWhite Often the first answer posted that solves the problem is accepted; it's often not the best. And no, it doesn't technically "seal the question," but that's a valid concern - it probably does reduce traffic. However, as for #4, if someone able to produce a better answer does visit the question, I think they're pretty likely to post it.
    – Trojan
    Jan 8 '14 at 1:40
  • 2
    Consider that Quora has a very different culture of acceptable questions. "What are some uplifting anecdotes?" is a perfectly acceptable Quora question, and one for which having an accepted answer makes no sense. On StackOverflow, that kind of question would never be welcome. Jan 8 '14 at 1:40
  • 3
    I think the most reasonable solution, if any action at all is taken, is to order answers by votes, so that the accepted answer isn't always the first shown (the rest of the model would remain untouched). On the other hand, you have to acknowledge that occasionally the highest-voted answer is not the "best", but that's much less common than an accepted answer not being the best or highest-voted.
    – Trojan
    Jan 8 '14 at 1:40
  • @AnnaLear By sealing the question, I mean that people are less likely to add new answers. If a new answer comes along, then it can be changed, but the chances diminish.
    – anon
    Jan 8 '14 at 1:41
  • @psubsee2003 On that question, discussion of solutions to the pitfalls of accepting answers, was if anything, a mere side note. Here, it is the focus of the question.
    – anon
    Jan 8 '14 at 1:42
  • @trojandestroy: As I said, voting indicates the best answer, and if the poster accepted it must have answered it for them. I'd much rather post an answer that's accepted and upvoted*, but I'll be satisfied with posting an answer that gets 10 + votes where the accepted answer only has the accept checkmark (and I'll gladly use information in the higher voted answer over the info in the accepted, because I'm capable of comprehending that the higher vote count means something). Only the person who posted the question can decide which one to accept, and it's their decision to make.
    – Ken White
    Jan 8 '14 at 1:43
  • 6
    My point is that this attribute of Quora (no accepted answers) is tied in with it's other attributes. Quora encourages open-ended questions, which don't lend themselves to accepted answers. StackOverflow insists on objectively answerable questions. Jan 8 '14 at 1:48
  • 6
    You say "I don't always scroll down". If you were looking for an answer for a real problem you have, you would scroll, right? If I have an actual problem, I'll read at least the first few answers, just to see which approach sounds better suited to my tastes/style. If, on the other hand, you're just browsing, what does it matter?
    – Geobits
    Jan 8 '14 at 1:54
9

In general, I agree. I'm fairly sure the feature tends to discourage new answers. And when new answers come in, accepted answer marks rarely change, often leaving an incorrect or outdated answer at the top. (And when we raise that concern, we're told that we shouldn't pay attention to the accept mark. So why is it there then?)

The accepted answer feature does, however, have one big advantage. It gives the asker the power to get a workable solution. Without the answerers' hope for the accept mark, many questions would receive answers that may be correct, but not helpful for the asker - for example, when the question is about a less-than-optimal practice in a real-world situation where it's the only choice.

I'd be in favour of a compromise. Retain the feature (and the 15-point rep gain that goes with it), but make the accept mark a less prominent signal to the rest of the world:

  • stop displaying accepted questions in a special way in lists.

  • stop giving accepted answers preferred treatment order-wise.

  • stop displaying a giant accept mark; introduce a more toned down way of showing the OP's approval instead.

8
  • 2
    My problem with this is that it hurts people that are actually looking for answers. If I run a search, I want the result list to show me which questions have an accepted answer, so I know which are more likely to have a solution for me. You're looking at this from the angle of an answerer.
    – Geobits
    Jan 8 '14 at 2:06
  • 1
    @Geobits, spot on. We need to think of the people coming in from google, not the user looking for +15 rep for their answer.
    – jmac
    Jan 8 '14 at 2:06
  • @Geobits, jmac - you're assuming that the accept mark is always a quality signal. I strongly disagree with that. When I search, I don't care for the accept mark because I know not to trust it. There's a million reasons why people award it and that it was the most helpful answer is only one of them. I go through each question looking for an answer that actually helps me.
    – Pekka
    Jan 8 '14 at 2:08
  • I'm not assuming it's always a quality signal. I'm assuming it is about 90% of the time, though, because that's what I've experienced. I look at multiple questions/answers, too, but most often, the accepted flag works.
    – Geobits
    Jan 8 '14 at 2:12
  • 1
    FWIW, I do agree that it discourages new answers. When I'm looking for something to answer, I skip accepted questions more often than not. I just don't agree with your proposed remedy.
    – Geobits
    Jan 8 '14 at 2:13
  • 4
    @Geobits OK, point taken about the quality signal - different people will use SO in different ways. I'm still strongly in favour of abolishing the silly sort order preference though. Show what's most upvoted on the top. Don't let the OP's choice (that they made at a random point in time) trump what the community has upvoted since then
    – Pekka
    Jan 8 '14 at 2:13
  • That I could get behind. I believe I've seen that particular issue brought up a couple times and status-declined, though.
    – Geobits
    Jan 8 '14 at 2:15
  • 2
    Why do you think that green checks from users with tinfoil headgear indicate that the question has an answer that you will find useful?
    – Rosinante
    Jan 8 '14 at 2:38
2

No. Definitely not for your reasons. Let me debunk them...

  • The best answer isn't always accepted.

    With this, you may have a valid point; I've seen my fair share of answers that are substantially better, or simply put, not as dangerous, but by no means does acceptance mean that the answer is correct. I've always interpreted it as this being sufficient to answer the OP's original question.

    This is why one must be careful when reading answers - don't just go for the accepted one right off the bat, but see if there is more substance in a higher upvoted answer.

    This is more caveat lectorus than anything else.

  • The Accepted Answer goes to the top.

    You're probably right about this, but this is likely an issue with answer ordering than anything else.

  • Accepting seals the question.

    In a sense, it could, but if a better solution exists, and the community identifies it, then there's absolutely no reason why someone else couldn't come along later and post a better answer. If the OP also agrees, then they could award the "accepted" answer to them, instead. In fact, this past Winterbash, there was a hat exclusively for that (Buccaneer).

    The mentality of acceptance meaning resolved answer also breaks down in the face of bounties on questions with accepted answers; perhaps the answer that is accepted doesn't sufficiently cover the question as well as originally thought, or more detail is preferred.

  • It discourages better answers.

    This is patently false. If you could provide some scenarios in which this is the case, I'd be happy to retract this statement, but by and large, I've seen that, if there is a better answer out there, then people tend to post it.

6
  • 1
    I've seen that, if there is a better answer out there, then people tend to post it..... and if there is a worse answer out there, people tend to post that too. Jan 8 '14 at 1:54
  • 3
    I don't think the point about discouraging answers is completely incorrect. I'm not versed enough in statistics to do the analysis, but I bet if you'd look at the likelihood of a question receiving an answer before and after it is accepted, you'd find a big change.
    – Pekka
    Jan 8 '14 at 1:54
  • @Pëkka: I'd like to see some statistics about it too. That's why I left it mostly open-ended. I don't believe that this occurs, but I can be convinced.
    – Makoto
    Jan 8 '14 at 1:55
  • 1
    @psubsee2003: You've got a good point in "worse" answers, and I won't disagree that they're posted either. But, typically, the worst of the answers is downvoted into oblivion, and the answers which are actually better are indeed voted up.
    – Makoto
    Jan 8 '14 at 1:57
  • @Mako to not always, some are worse, but not awful. They just get posted and forgotten so they sit there with no votes and few views. Jan 8 '14 at 2:00
  • @psubsee2003: That feels more like question/answer rot, more than anything else. If the answer isn't seen by a lot of people, then yeah, it kind of gets forgotten about.
    – Makoto
    Jan 8 '14 at 2:18
1

Executive Summary

You seem to have three faulty premises:

  1. The goal is to provide the 'best' answer to each question
  2. Accepting an answer reduces visibility of other answers
  3. Accepting an answer prevents future improvement

'Best' Answer

When I come across a question through Google, my primary goal is to solve my problem. The accepted answer checkmark tells me that, for the person who asked the question, the accepted answer works. That is a hugely valuable signal to me I would hate to lose.

Upvoters may not have tested the code. The person asking probably has. They have tried the solution and it worked. No amount of upvotes can tell me if that is the case, and for many questions the gain of testing multiple answers is less than implementing the accepted answer solution.

Visibility

If I see an accepted 5 upvote answer and read to the end, I will notice that the answer below it has 60 upvotes. When there is a large gap, I will likely read through the second answer as well, and then I have two resources to look over to solve the problem. Browsing answers that got the populist badge, many are alternative methods, slightly more optimized, but not dramatically better than the accepted answer for someone looking for a quick fix.

The point is that even though the accepted answer goes first, the other answers are still there, and visible, and the bigger the contrast in score, the more of a signal toward the potential worth I have to make a decision on how to spend my time. If I need a quick fix, the accepted answer will probably suit my needs 99 times out of 100 or more.

Future Improvement

Looking back at those populist badges, you will see that there are several where the answer is one that provides an updated answer to the question. For instance, this question provides an answer for a later version of the software:

Change UITextField and UITextView Cursor / Caret Color

The accepted answer is pre-iOS7, and the most upvoted is for iOS7+. Both are good answers. And having an accepted answer hasn't prevented people from adding answers or from finding the answer they were looking for an upvoting it.

5
  • Re "Best" answer - it's often enough the wrong signal, and no one except the OP has the power to change it. That sucks, big time. You have a point with the second paragraph though
    – Pekka
    Jan 8 '14 at 2:01
  • @Pëkka, I don't know about that. Of the 6.4 million questions on SO, 1.7 million have only one answer which is accepted. That means 25% of our questions are giving the right signal off the bat. Someone can dig in to the data explorer, but I would wager that there are more accepted answers that have the most upvotes than non-accepted answers with the most upvotes, further increasing the usefulness of that signal.
    – jmac
    Jan 8 '14 at 2:05
  • But look at your example question. There is going to be an iOS6 answer at the top, signaling "this was the solution", for all eternity. That happens often enough, and no one but the OP (who may have long gone) has the power to ever change it. How is that not broken? And it'll become brokener and brokener the older the data pool gets.
    – Pekka
    Jan 8 '14 at 2:06
  • @Pëkka, if you are looking for the answer to that question, does the current format not allow you to still find it easily? Even the first answer has been edited to point out what it applies to. Nothing is preventing someone from creating a new question explaining how to do it in iOS7+ as a separate question either. The info is there, accessible, and clearly marked. How is this a strike against having an accepted answer exactly?
    – jmac
    Jan 8 '14 at 2:08
  • Nothing is preventing someone from creating a new question that may very easily end up being a duplicate though. The accepted answer system simply isn't a good fit for questions that change in time. If you ask me, we either need a way for the community to change the accepted answer on abandoned questions, or to get rid of it altogether.
    – Pekka
    Jan 8 '14 at 2:11