14

Accepted answers are oft misunderstood.

As developers we often think that every answer should have the one accepted correct answer, if this is not the case ... well it is time for duel.

The trouble is that the concept of accepted answers is a convention between the question answerers and the asker. The rest of the world is only indirectly involved, and can not control this bit of information.

However, accepted answers get special treatments in the answer stream. They always show up as the first answer (even if better ones exist).

This bump is handy when a question starts its life, it quickly tells us: "This has been taken care of ... move along". However, as questions age this special bump is counter productive. There is no way to displace the incomplete or incorrect accepted answer from the #1 spot. So people, edit it with warning, downvote it more than it deserves, and so on.

Should answers revert to standard ordering, as G-D intended, 2 weeks after an answer is accepted?

Data points:

  • There are approx 119 thousand questions where the accepted answer has less votes than the top voted answer.
  • You can browse through 20k of them here.
  • Browse through 30k of the answers that out-scored the accepted AND were posted after the accepted answer. (total 42k)
Improve this question
9
  • 3
    Yuck, yet another way for a popular answer to out-shout an expert answer.
    – Uphill Luge
    Apr 20, 2011 at 1:42
  • 3
    @Hans ... you are assuming that the people asking are experts in the question domain, which they are usually not
    – waffles
    Apr 20, 2011 at 2:38
  • @Hans see this for an example: stackoverflow.com/questions/250874/… displacing Jeff's now incomplete answer is going to be impossible ... even if I get the votes ... which is unlikely cause this has dropped off the radar
    – waffles
    Apr 20, 2011 at 2:41
  • What does Paul McDermott have to do with anything?
    – random
    Apr 20, 2011 at 3:14
  • Why didn't you improve the answer? I took care of it.
    – Uphill Luge
    Apr 20, 2011 at 3:19
  • And how many of those popular answers are just "use a library" instead of whatever the OP asked about?
    – random
    Apr 20, 2011 at 3:27
  • @Hans I am not following how "Edit an answer into correctness" is the correct action... we may have to do that 119 thousand times ... You are re-enforcing the stigma that "accepted" = "correct" ... "accepted" is simply a convention between the asker and answerer ... nothing more
    – waffles
    Apr 20, 2011 at 3:39
  • Of course, that was intentional. Accepted == correct in the vast majority of cases, 119,000 times is only 8% of all questions. Contemplating editing your question to add that data point...
    – Uphill Luge
    Apr 20, 2011 at 3:59
  • @Hans sure this actually needs a couple of edits, I also want a query that shows places where the community disagreed with the acceptor after the accept ... so for example an answer was posted after the accept and got more votes that the accept. I suspect that in all those cases, the new answer deserves the #1 spot. the query is kind of nasty though.
    – waffles
    Apr 20, 2011 at 4:05

8 Answers 8

Reset to default
10

After thinking about this, I think the accepted answer should always be at the top, as it is easy to see when the answers below it has a lot more votes.

I don’t think the benefit in changing the system is great enough to make the change worthwhile.

Improve this answer
4
  • 5
    It's not always clear what the numbers and glyphs mean to someone who is not an SO user, coming from a search result perhaps. Typically, you see the big green check mark and assume "This is the correct answer, the others are probably not worth reading."
    – user159834
    Apr 19, 2011 at 8:58
  • I fail to see how this answer deserves the #1 spot stackoverflow.com/questions/2521188 ... the trouble is that as people hit this page from google ... confusion will ensue
    – waffles
    Apr 20, 2011 at 7:18
  • 1
    @madmartigan, anyone that does not know to look at most of the answers when they see a question on a internet site, should not be using the internet. Using the internet is all about deciding what to trust...
    – Ian Ringrose
    Apr 20, 2011 at 12:59
  • That's a grim assessment, but OK.
    – user159834
    Apr 20, 2011 at 13:00
6

I always understood that the idea behind an accepted answer is that it's the answer that helped you -- the question asker -- the most.

Thus, if we follow that logic, upvotes by other users should have little effect on the actual accepted answer, as the purpose of accepting an answer is to say "this answer helped me the most."

In summary, just because 10 other people liked a particular answer more than another, that does not necessarily mean the most upvoted answer is the one that will solve my unique situation.

Improve this answer
3
  • 3
    But waffles doesn't want to change the accepted answer - he just wants to change the display order in which the answers are shown. Arguably, the highest voted answer is very likely to be more useful to future generations than what helped the OP in their specific situation (and may have been wrong, or a bad practice)
    – Pekka
    Apr 19, 2011 at 7:11
  • 2
    And, these sites not being some personal help forum, the you is not that important for future visitors who have the same problem.
    – Arjan
    Apr 19, 2011 at 7:26
  • 1
    Makes sense. I can understand the logic behind that. Is there a way to perhaps beta test the change to see if it makes any difference?
    – jmort253
    Apr 20, 2011 at 4:42
5

You are assuming that the voters are experts, they are usually not. They didn't post an answer, likely never once did. There are two users that usually really understand the question. The one that asked it and the one that answered it. The myth that a questioner cannot understand an answer is hogwash, that completely defeats the point of asking a question.

There's a strong tendency at SO to vote for answers that are easy to understand. Many, many complaints about this at meta. This is entirely natural and inevitable as long as posts are rated by votes. The liability however is to get an answer that's merely related to the question, is easy to understand, draws a lot of votes because of that, but doesn't actually answer the question. It takes an expert to read between the lines and understand to true problem. Such an answer is rewarded with the answer mark, rarely with votes.

Maybe it is important for Stack Exchange Inc. to get all the popular answers moved to the top. Looks good, few complaints. Actually useful? Hmm, doubtful. Actively harmful if that makes experts find a better way to spend their free time.

Improve this answer
4
  • Hans, look through the list here: data.stackexchange.com/stackoverflow/q/98359/… and you be the judge. From much browsing in the vast majority of cases here the #1 spot is not deserved, when a new answer outscores the accepted, after it was accepted usually it means the asker is not an expert, or just abandoned the question. When its outscored by a lot like this puppy - stackoverflow.com/questions/3216046 or stackoverflow.com/questions/2521188 it is dangerous advice
    – waffles
    Apr 20, 2011 at 7:15
  • granted, this is a great way of finding questions that do not belong on the site, however for ones that do, at a certain threshold the community is more of an expert than the asker.
    – waffles
    Apr 20, 2011 at 7:16
  • 2
    Yes, granted. You found a clever way of finding crappy answers. What you don't see from that query are the many more cases where the accepted answer is the best answer. The absolute number is tempting, the relative underwhelming, this is still only 8% of all questions. My personal record for accepted answers with 0 upvotes hovers at a nearly constant 14%. Self-inflicted no doubt by methodically working through a tag. But having those answers disappear into a hole certainly adds insult to injury. It makes it really pointless to add a late answer.
    – Uphill Luge
    Apr 20, 2011 at 13:01
  • if we required a threshold this should be fine, I doubt any of you 0 score accepted answers were out-scored after the fact by more than a vote.
    – waffles
    Apr 22, 2011 at 0:22
4

I like this idea.

Future visitors are likely to benefit more from the highest voted answer than the accepted one.

In my experience, when the accepted answer is not the highest voted, one of these mostly applies:

  • the accepted answer is incorrect
  • the accepted answer has become incorrect over time
  • the accepted answer recommends something that works for the OP, but is bad practice

there also are cases where the community is wrong, or where the highest-voted answer turned out to be incorrect later, so the OP changed the mark to a new, more correct one. But I'm willing to bet money that that occurs more rarely.

Improve this answer
3
  • 5
    There are also plenty of cases where the asker just checked the first good answer within 5 minutes of asking and moved on, while the much better answers came in later (as a retort in some cases to the accepted answer).
    – user159834
    Apr 19, 2011 at 7:22
  • may be one more fact - another answer is better and well-written than accepted answer?
    – YOU
    Apr 19, 2011 at 8:47
  • 1
    To your second point: It is not always the case that the accepted answer has become incorrect, but very often some months later, a better solution is available. I personally often see that in Android questions, when a new release is available.
    – RoflcoptrException
    Apr 19, 2011 at 10:14
4

I like the idea but there should be some minimum threshold somewhere or additional requirements before it goes into effect. We can't apply this to just any ol' answered question. Something like:

  • This only applies if this question has generated x upvotes across all answers. That way, a 0 vote accepted answer will not get trumped by a 1 vote answer.
  • This only applies if the net difference between scores of the accepted answer and the highest is greater than n points.

It should only affect questions that has at least a certain amount of activity and where the difference between the accepted and highest voted answer is overwhelming at the very least.

Improve this answer
1

When you accept an answer you are saying "this worked". When you search for an answer, you're looking for an answer that works. A better answer may exist (and often does), and we do still show those. However, I think showing the accepted answer first is the right way to do it.

Improve this answer
1

Certainly not in the general way you propose. What about the cases where the community got it wrong, but the asker tested answers and marked the one that actually worked as accepted? I've certainly seen that happen. Now I haven't dived into the database to make statistics, so this is anecdote only, but even if the accepted answer is the right one in a minority of cases, I'd prefer to privilege attentive askers, because they're the ones who tend to ask the most interesting questions.

If you change the ordering, that should only happen if a newer answer (posted after the acceptance) beats the accepted answer.

Thus:

  • Answer A accepted, then answer B posted, then answer B overtakes A (by some margin): maybe it's ok to change the order.
  • Answers A and B posted, B has more votes than A, asker accepts A: don't change the order.
Improve this answer
0

To me the accepted answer is not only what the user thinks is correct but it is THE verified and tested answer in the situation that was brought up in the question. While other people may think that another answer is correct - nobody but the asker can tell us "yes, this answer is correct, and I have tested it in my situation as lined out above as being absolutely correct and working"

IMHO that should always be given top billing.

Improve this answer
1
  • That sounds like a very optimistic thing, but it just doesn't pan out with reality. How can the OP be a creditable verifier of the answer when they don't know the answer themselves?
    – Mitch
    May 1, 2011 at 23:52

You must log in to answer this question.

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