What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 129 Stack Exchange communities.

I keep on seeing a fair number of questions (such as this one) which get marked as answered within 5-10 minutes of being posted. Sometimes the questions like this are insanely popular and actually see some proper discussion, however often I see questions have few views and only 1 answer. The trouble with these questions is that very few people have scrutinised the answer for correctness, or possibly a better solution.

I always make a point of not accepting my answers until a day or so has passed — would it not be a good idea to put a short (1 hour or so) time limit on accepting answers on new questions with few views or votes, or perhaps just an additional confirmation dialog, something along the lines of:

Your question is new and not many people have had a chance to look at it yet. Are you sure you don't want to wait to see if anyone else has a better suggestion?

share|improve this question
9  
+1 If someone accepts one of my answers very quickly, I normally add a comment saying they should really wait a bit. It seems a bit churlish to do it for other people's answers, so a confirmation dialog would be a win from my perspective. –  nb69307 Feb 3 '10 at 13:16
7  
-1, I don't agree, sorry. Accepted answers can easily be changed if a new one is posted. As soon as the asker gets an answer that satisfy him I think he should accept it. Anyways most people would just forget to come back to accept it later (but if they accept answer A right away, and then answer B gets posted, they'll be notified with the orange bar..) –  Andreas Bonini Feb 3 '10 at 13:18
1  
+1 - though I wonder if rather than simply a time-based thing it could be a combination of time and views (views > x || age > 1 hour) where x is to be determined. –  Dominic Rodger Feb 3 '10 at 13:18
28  
I think having an accepted answer discourages some other people from answering, if not all of the time then some of the time. I think it might also discourage voting to a certain extent to, but I've no evidence of either. –  ChrisF Feb 3 '10 at 13:27
4  
@Koper As usual, your ideas on how SO should be used are ... strange. –  nb69307 Feb 3 '10 at 13:31
5  
I've noticed a couple of things occur when a response to a question is marked 'the answer': 1) if the question is getting up-voted quickly, as soon as there is an answer, the up-voting stops/slows tremendously, and 2) answers stop flowing. It is important to be able to get as much feedback as possible. –  IAbstract Feb 4 '10 at 7:11
    
@ChrisF - I feel exactly the same –  Matt Feb 4 '10 at 15:36
4  
+1 Earth takes 24 to make a tour. The other half of the planet is sleeping when you do a question, so letting them the time to answer your question before you accept another one is a sign that you want the best answer. –  Vicente Botet Escriba May 4 '10 at 22:00
    
I like the added feature but I think the time period should be longer. I often come in as the 2nd or 3rd answer because I'm not sitting on the questions page hitting refresh all day - I have other things to do besides help people, sorry. When that happens I'll typically be too late to be the chosen answer because the OP has already decided (this just happened to me today, where my answer is - I feel - better). And they very rarely change their accepted answer, even though they can - some don't know this is possible, and some think it's rude to the person whose answer was originally accepted. –  Aaron Bertrand Aug 17 '11 at 14:07

12 Answers 12

up vote 35 down vote accepted

I think I can support a limit, but much less than an hour.

Anyone who accepts an answer within, say, 15 minutes of posting their question is almost by definition being lazy in a bad way. That is, they're accepting the first answer that comes along without considering all their options.

Greg Hewgill's graph shows that very quick answers are accepted less often than the first probably more complete answers, but gives no clue about how soon the winning answer was accepted:

edit: this is implemented. I think 15 minutes (the current setting) is completely reasonable; if you're accepting earlier than that, it's irresponsible IMO as others haven't even had a chance to even LOOK at it.

select extract(epoch from a.creationdate-q.creationdate) from post q join post a on (a.id = q.acceptedanswerid)

share|improve this answer
3  
"if you're accepting earlier than that, it's irresponsible IMO" -- not necessarily true, but 15 minutes seems like a reasonable minimum amount of time before accepting. –  Jon Seigel Mar 27 '10 at 22:31
9  
Does this feature block accepts before 15 minutes, or just warn about them? If it's the former, I hope that it does not encourage those same lazy users to simply leave and never bother accepting an answer at all. Some questions are very "hot" and it may be completely reasonable to accept an answer after 10 minutes if there are already 5 answers; on the other hand, those bars in the 54s and 2m range are a little ridiculous. –  Aarobot Mar 27 '10 at 22:41
4  
@Aar: I think it's fair to say anyone that posts a question, leaves within 15 mins, and never comes back (to that question) doesn't care enough to put much thought into accepting in the first place. We're better off with only the community deciding (in the form of votes) for that question. –  Gnome Mar 27 '10 at 23:25
2  
@Gnome: That is all well and good for popular/semi-popular questions, but there are many completely reasonable technical questions that are not common enough (or maybe just not interesting enough) to garner more than a tiny handful of views and 1 or 2 votes or maybe none. If these pop up in somebody's Google search results, I would prefer for them to see an answer which was tested and verified first. –  Aarobot Mar 28 '10 at 0:18
2  
@aaron it's a block. And if you think someone accepting an answer within 15 minutes of posting the question means they have tested and verified that answer, well.. you're quite the optimist! :) –  Jeff Atwood Mar 28 '10 at 1:23
2  
I'm curious about the 20-second answers: were those people posting their own canned and predetermined answer to their own question? Where would I find the links for those questions? –  Mark Rushakoff Mar 28 '10 at 2:04
7  
This graph looks wrong. Jon Skeet answers questions before they get posted, and yet there's nothing below 7 seconds. meta.stackexchange.com/questions/36591/question-asked-in-future Is this because you're trying to take the logarithm of negative numbers? –  Andrew Grimm Mar 28 '10 at 7:22
    
@Jeff though it worth pointing out one case (from my own question) where I tested and verified the answer within 7 minutes of it being provided. stackoverflow.com/questions/248989/… Not that this invalidates the change in any way - at the time I wished I hadn't accepted so quickly, since I wanted the answerer to get as many upvotes as possible. –  David Hall Mar 28 '10 at 20:52
    
@Jeff: stackoverflow.com/questions/2535516/… –  Jon Seigel Mar 29 '10 at 3:01
3  
Note that that graph indicates the time until the (eventually) accepted answer was posted. I had originally titled it "Time to best answer" to indicate it wasn't the time until the OP clicked "accept", but changed the wording because the best answer could be considered to be the highest voted answer or something else. So that histogram indicates how long it takes the community to provide the answer that was eventually accepted. –  Greg Hewgill Apr 7 '10 at 5:57
    
@Greg, any chance you could easily create an updated version of that (nice!) graph? Might be interesting to see how things have changed. (And if the overall accept-rate has dropped, or not.) –  Arjan Apr 25 '10 at 12:10
1  
@Arjan: I update it every month with the new data dump: hewgill.com/~greg/stackoverflow/stats-SO.html –  Greg Hewgill Apr 26 '10 at 0:33
1  
Jeff, I removed your "I was shocked to see the number of accepts that happen even under 15 minutes" as Greg's graph gives no clue about how soon answers are accepted, but only that early answers are accepted more often than late answers. (Like Greg already explained above.) –  Arjan May 5 '10 at 16:54
    
@Arjan: Thanks, that's an old one. I keep finding and fixing them. –  Greg Hewgill Oct 7 '11 at 18:58

I definitely agree. There have been several situations where I arrive on a question and have something new to add to the discussion or solution, but simply hit the back button because an answer has already been accepted. Maybe that's wrong of me, but I only have a limited number of hours in the day, and I'd rather spend them productively.

Perhaps this is another feature that could only be turned on for very new users (measured either by time or reputation) -- either disable the accepted answer button entirely, or leave it up with some extra nag text like "Please consider waiting a little longer before accepting an answer, so as to encourage more answers to be posted!"

This suggestion would be suitable to go along with the text suggested here: When a question is posted, can we advise the poster to hang around and respond to comments?

share|improve this answer
    
+1 Good point about new users. –  nb69307 Feb 3 '10 at 16:40
    
I think new users should be strongly encouraged to select an answer (as it is, many don't) and if they get their solution in 5 minutes and accept it, that simply means the system is working as designed! –  Adam Davis Feb 3 '10 at 16:46
    
@Pollyanna: agreed, but they are probably not aware that they may get more answers if they held off on accepting for a bit. Obviously, if they are a hit and run user, I'd rather they accept now than not at all, but they should be aware of the other options too. –  Ether Feb 3 '10 at 17:57
    
But if the answer meets their needs, solves their problem, then does it really matter in the long run if they don't get more answers? Why not let others say, "Ah, solved. Moving on to the next question." This actually increases the number of questions answered, whereas a question with no accepted answer (even though the first one is correct and fine) might eat a lot of user bandwidth with other answers that are not needed. Yes, they may be better than the existing answer, but the existing answer is good enough, and they might be able to squeeze in another answer elsewhere. –  Adam Davis Feb 3 '10 at 18:20

I don't agree with that idea.

If I found the exact answer within 5 minutes, why wouldn't I be able to accept it ? Often see questions like "this is not working, help" with an answer like "do this, it'll fix it".

Also new users that come by not often will never come back 1 hour later to accept the answer...

share|improve this answer
    
+1 for new users that come by not often will never come back 1 hour later to accept the answer Seen it happen very often. –  Neolisk Apr 18 at 1:23

I disagree:

  1. There's no reason a question can't be answered faster than that.
  2. There's no reason a question will be answered correctly in that time.
  3. One-shot users will simply not accept the answer and never return to the site.

However, a warning message might be appropriate for when ink is still drying on the question.

This question is only [n] minutes old, and other users may wish to add more answers. Are you sure you want to accept an answer now?

share|improve this answer

I'm not sure that it supports the proposal, but Greg Hewgill's SO stats graphs (scroll to the last graph) show that the majority of questions have accepted answers within an hour of posting.

See also: Meta Stack Overflow statistics graphs.

Edit: Greg's graph doesn't show the time to OP acceptance of the answer - instead it shows the time between question and answer posting - the acceptance might well have come some time later.

share|improve this answer
    
stats are always helpful –  Jeff Atwood Mar 26 '10 at 2:45

Since we have community involvement in voting, the best answer can be voted up by the community.

Granted, this doesn't help questions that are asked and answered very quickly -- but if the answer fixed the problem for the asker, why mess with it?

If you have a better answer, posting it will cause the post to be 'bumped'. Doing this at a high use time of day will almost guarantee upvotes if your answer is better.

share|improve this answer
    
Exactly. The system already takes care of the "wrong" selected answer by placing immediately below it the community's highest voted answer. –  Adam Davis Feb 3 '10 at 16:47

I don't agree... if I ask a question and found my answer, I accept it.

Perhaps you are thinking about subjective answers, but if you have a specific problem and found your solution, why should you wait?

Also, this question is related

share|improve this answer
2  
How do you know the answer is the right one, the best one or the correct one? By definition (as you are asking the question in the first place), your knowledge of the problem domain is severely limited. The only way to find out is to survey a bunch of answers, and the interactions between the answerers. –  nb69307 Feb 3 '10 at 16:43
2  
@Neil - because it either solves the problem, or it doesn't. For questions where you can't test a solution, the question itself is faulty and shouldn't have been posted. But the majority of programming answers, by definition, can immediately be tested for correctness. –  Adam Davis Feb 3 '10 at 16:49
2  
@Pollyanna Well take the perennial - "how do I stop my window closing in a C program?" question. An answer of 'system("PAUSE")' may seem to "work" but it is not correct by any reasonable standard of correctness. And in my experience, the same goes for the majority of programming questions - just because it "works" doesn't mean it is correct and doesn't mean there won't be horrible problems further on down the line. –  nb69307 Feb 3 '10 at 17:00
    
@Pol, exactly what I meant. @Neil, you can accept one, the question still accepts more answers (I have switched accepted answers more than once) –  jmfsg Feb 3 '10 at 17:10

I keep on seeing questions marked as answered within 5-10 minutes of being posted - The trouble with these questions is that very few people have scrutinized the answer for correctness, or possibly a better solution.

[We should] put a short time limit on accepting answers on new questions with few views or votes, or perhaps just an additional confirmation dialog.

We have notifications, so if a new answer is posted, the OP is alerted and they can change their mind, especially if the new answer points out flaws in the accepted answer.

Further, the OP knows best whether the answer worked for them or not. Yes, a better answer may come along later, but that doesn't negate the fact that not only was the accepted answer "good enough" but the person that submitted it delivered it more quickly so the OP could get on with their work.

The effect of preventing or even discouraging people from accepting an answer immediately is going to be overall negative. We've already had to put measures in place to encourage people to accept answers more, and many, many, many comments posted are devoted to asking the OP to accept an answer.

Keep in mind that the community's highest voted answer appears immediately below the accepted answer, and anyone who comes to this question in the future will start at the top, and if the first answer doesn't work, they'll try the second, and so forth until they've resolved their problem.

The system has many measures already to account for poorly chosen accepted answers, and is designed to work despite them.

Lastly, if such a measure is implemented, it might encourage some users to take longer to post their answers. Yes, they may be more finely crafted than a quick and dirty (but correct!) answer, but the OP is left waiting for their answers because some people choose to use this grace period to inflate their chances at getting accepted, rather than posting a good, quick answer.

I do NOT want to, in any way, discourage the quick and dirty answer.

Such a change would have a stronger negative impact on the system than positive.

share|improve this answer
    
+1, however my point is not that someone wont get notified when new answers arrive so much as that they wont get any new answers, due to Accepted questions seeing significantly reduced activity. I also don't believe this will have a significant impact on the quick and dirty answer - personally the main incentive for the "quick and dirty answer" is initial upvotes so my answer is pushed to the top. I don't expect someone to mark my answer as correct until some time later, and so the fact that they cant (or are discouraged from doing so) won't really affect me. –  Justin Feb 4 '10 at 12:11
    
Also, although I agree with the quick and dirty answer from a Q&A "Help me quick I'm stuck" perspective, from a wiki / programming resource perspective it's also important that the right answer also be in there for when other people find the question, even if its posted after the original user has moved on. –  Justin Feb 4 '10 at 12:17
    
@Kragen - Perhaps you can mine the data dump and find out how often the first answer is wrong, and how often the OP chooses a wrong answer in the first hour vs choosing a wrong answer in later hours. If you can't prove that this is happening in a significant number of cases, then all you solution does is assume that, more often than not, the first answer is wrong, and the OP chooses the wrong answer. I believe these are bad assumptions to make based on no evidence. –  Adam Davis Feb 4 '10 at 15:58
    
I've been trying to support my proposition with evidence from the DataDump, but the dump only tells you the day that a question was accepted, not the time. When we are specifically dealing with questions answered within hours of being asked that kind of limits what I can show. –  Justin Feb 5 '10 at 17:33
    
@Kragen - I posted a feature request to include this information, if it exists. We'll see if it gets any attention, but I wouldn't hold my breath... meta.stackexchange.com/questions/38451/… –  Adam Davis Feb 5 '10 at 18:43

I think a minimum of one day should be required for acceptance. A question with an accepted answer discourages other answers -- including better ones. I have answered questions that already had an accepted answer, but it's been a frustrating experience -- spending time on something that is barely noticed, if at all. People move on after an accepted answer.

So why does a question need another answer if one was already good enough for the asker? It depends on what you see as the value of a question. Is it to show that a problem was "solved" in record time, or is it to record a thoughtful answer with long term value to searchers? A carefully thought out answer gives context and raises issues that may not be addressed in the accepted answer.

I've also seen accepted answers that are wrong or misleading or incomplete, and the asker doesn't know any better. One down vote or comment on an accepted answer with 12 votes doesn't get much notice.

People are concerned that this will mean some questions will end up without accepted answers. That doesn't bother me. Doesn't the answer with the most votes count for something?

(I'm glad to see there is no accepted answer to this question yet -- at least no "offical" answer.)

share|improve this answer

I actually think this is a good idea as it could also help cut down on situations where someone provides an answer that may work and gets to be accepted, but someone else later on comes along and provides the best solution for the problem and ultimately gets the accepted answer status. I believe you lost the points if your accepted answer is unaccepted so having a waiting period before an answer could be accepted might cut down on hard feelings because of changing accepted answers.

However, I might go as far as to argue that there should be about 24 hours between when a question is opened and when an answer can be accepted just to give a good opportunity for people to view the question and possibly answer it.

share|improve this answer
2  
My original thoughts was also for a day, but I think if you make it too long you would end up more strongly discourage people accepting answers. –  Justin Feb 3 '10 at 13:54

Answering an already answered question needs to be encouraged by a chance of earning additional reputation, then no delay is necessary. For example, you answer a question, which was already answered, and the OP changes their accepted answer to yours, you get an additional +15 to reputation. Can be a 15 minute timeout, in which the additional reputation bonus is not awarded (=OP changed their mind).

So you get +30 reputation for this accepted answer + any upvotes you would normally get. The number of additional reputation can be adjusted, depending on question age, i.e. +15 (fixed modifier) +1 for every day after a question was posted, to a maximum of +30 (variable modifier), so in total you may get up to 60 reputation for one accepted answer. Is it not a good stimulus for trying to beat an accepted answer with a better one?

All numbers here are subject to discussion and analysis.

share|improve this answer
    
+1, but doubling the bonus might be a bit much. +20 seems reasonable; with an upvote on top, that's still +30. –  Nick Stauner Apr 18 at 2:18
    
@NickStauner: Thanks. Like I said, all numbers are negotiable. The idea is to encourage people visiting already answered questions. A) more upvotes for answerer, if the answer is good already. B) better quality answers overall. –  Neolisk Apr 18 at 11:24

Maybe the right thing to do is to make the time to when an answer become accepted depend on community votes? My rough idea is that if a lot of people are saying that answers are good, then a genuinely good one can be nominated rapidly. On the other hand, if hardly anybody takes any interest in a question then the threshold drops over time.

share|improve this answer

You must log in to answer this question.

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