What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 130 Stack Exchange communities.

I was discussing this question on team chat:

How do you perform an IF...THEN in an SQL SELECT?

This question is a high value asset on Stack Overflow, it already has 150k views, 120+ total votes between questions and answers, and a very high ratio of positive anonymous feedback.

I approached improving this question aggressively, too aggressively for the teams taste:

  1. I made massive edits to the accepted answer

  2. I deleted 8+ "duplicate" answers in that question, including a 17 upvote and 13 upvote answer in that question.

enter image description here

(Since then I have undone my actions, except for a comment conversion and the big edit)

This raises the question: what should be done to keep this high value question in great shape so it effectively helps lots of people down the line?

Specifically:

  • Should there be any answer deletion? No answer deletion? Some answer deletion?
  • Should we play favorites improving the top voted answer / first answer?
  • Will deleting answers with +17 and +13 score turn away high rep users?
  • Has this asset become a public resource, whatever that means?
  • What should we do when new answers trickle in and are dupes?
share|improve this question
19  
They're too soft on duplicates –  random Aug 19 '11 at 0:52
37  
They're too soft on duplicates. –  jcolebrand Aug 19 '11 at 1:24
4  
They're too soft on triplicates(?) –  davidsleeps Aug 19 '11 at 1:33
1  
@davidsleeps +1 for effort. But -1 for getting it wrong. So .. A null vote for you. –  jcolebrand Aug 19 '11 at 2:41
2  
@jcolebrand, much appreciated :D –  davidsleeps Aug 19 '11 at 2:45
3  
A very good test of an answer, before even beginning to judge its correctness is "can I figure out the question that this is answering?". If you can't, then the answer should be (at the least) edited to have some context. I'm surprised that several of those answers haven't been flagged as non-answers. –  Tim Post Aug 19 '11 at 5:11

8 Answers 8

When I was a teacher's assistant at the university I learned some important lessons about teaching people effectively.

One of the most important things I learned was that sometimes it took multiple tries of telling a student the exact same thing in slightly different ways before they would grasp the concept. I don't think duplicate information gets in the way or harms a question.

As long as there is no misinformation in the answer, I don't think you should be deleting answers.

I think one of the main reasons I like duplicates in particular is because they help clarify each other. Someone looking for help can examine all of the answers and discover the commonalities in all of the examples. Now, they know the base concept instead of one or two individual's twist on it.

share|improve this answer
1  
@waffles, I'm saying that people shouldn't be so quick to deem duplicates useless. Unless an answer isn't providing the Exact same solution, it adds some value to the question. (It definitely adds more value than it takes away.) –  jjnguy Aug 19 '11 at 2:00
6  
that is what I am struggling to understand ... there are 8 answers here that use CASE WHEN ELSE END with slightly different column aliasing, casing and formatting. essentially they are dupes, are you saying even if 50 new answers are added, as long as the binary diff is there - do not delete? –  waffles Aug 19 '11 at 2:04
    
@waffles, "with slightly different column aliasing, casing and formatting." - In this case, someone reading all of these answers that initially had no idea what-so-ever what they were doing would find value in all of those. They can examine all of the answers and discover the commonalities in all of the examples. Now, they know the base concept instead of one or two individual's twist on it. –  jjnguy Aug 19 '11 at 2:17
    
@waffles, I added a clarification in the last paragraph of my answer. –  jjnguy Aug 19 '11 at 2:23
8  
A better answer explains what is going on and not just throw code at the screen and walk away. Lazy is no excuse for having variants on a theme because no one bothered to say why they were viable. –  random Aug 19 '11 at 2:38
    
@random, if we have thousands of people willing to answer questions using their own tone and style, why should we discourage them from putting any effort into it? Beggars can't be choosers. –  jjnguy Aug 19 '11 at 2:44
    
@random, I obviously won't argue that people shouldn't put effort into answers. –  jjnguy Aug 19 '11 at 2:46
4  
Tone and style would be phrasing and explanation. Not Y over Yes or putting an opening parenthesis on the same line as opposed to the next. That's just diff for diff's sake. –  random Aug 19 '11 at 2:47
3  
@random, if those were the only differences between two posts. (And it was clear that Y vs. Yes is only a stylistic choice) the I'd say delete one of them. However, one great way to show someone which parts of a concept are necessary and which parts are stylistic is to show them 4 different answers that solve the problem the same way with slightly different style. The bits of style that each author had added will no longer be an important part of the solution. Instead the general structure of the solution will become clear. –  jjnguy Aug 19 '11 at 2:51

I think there are a general set of guidelines for question maintenance that should apply to any extremely popular question:

  1. There should be multiple "correct" answers; all of them deserve editing and improvement.

    Favoring only the accepted answer in editing over all the other answers is deeply wrong; that is like saying "of all my children, this one is my most favorite". While the accepted answer is indeed important, it's never the entire story. Anyone with a genuine interest in this question wants to hear multiple perspectives and decide the pros and cons of multiple answers, and will read at least 3 or 4 different answers.

    So while 12 competing answers is clearly too many, 1 is not nearly enough. The answer lies somewhere in between, and we shouldn't improve one special blessed answer at the expense of utterly ignoring all the others.

  2. Stack Overflow citizens in good standing deserve credit for their contributions.

    If an answer is correct, but weak, I'd be willing to extend the benefit of the doubt and leave a comment asking for clarification if the user

    • has a significant amount of reputation

    • is still actively participating on the site

    Whereas if it is a weak answer by a user with 1 rep, or a user with 50 rep who hasn't been seen in a year, I'd be much more inclined to delete it outright.

  3. It's OK to combine answers, but not into one giant single katamari ball answer.

    If you accept the premise that "there's more than one way to do it", the best way to combine answers is to pick 3 ~ 5 approaches and identify the best starting answers that cover that approach. Then roll up related, lower-quality answers into those compatible answers, and extend the same editing TLC to each of them.

    Good answers should not be limited to the unreadable, encylcopedic, giant bullet point list from hell, they should tell a coherent, succinct story of "well, here are the pros and cons of this particular approach..." and the reader can compare and contrast this with the other answers.

  4. If answers are to be outright deleted and not combined, they should be provably bad by concrete metrics.

    We know that quality equates strongly with length (TWSS). Therefore, answers that are strong candidates for deletion:

    • belong to low rep or anon users with no real commitment to the community

    • are provably duplicate, that is, were added well after (30+ mins later) other answers that contained the same exact information

    • are short in length

    • do not explain much of anything

So in short, I don't think the way you handled that question was correct. I support some deletion, but only under the quality terms outlined above -- and I strongly object to the "favorite child syndrome" where only the accepted answer gets first class treatment and all the other answers are either deleted or ignored. That's not fair to the community, and it sets a bad example for everyone.

share|improve this answer
2  
But Jeff, only my firstborn child has any merit. All the ones I killed before my firstborn had no merit, of course. –  jcolebrand Aug 19 '11 at 16:19
  1. I made massive edits to the accepted answer

No you didn't. You tweaked the code formatting and added a couple more examples. The spirit and meaning of the answer is alive and well.

II. I deleted 8+ "duplicate" answers in that question, including a 17 upvote and 13 upvote answer in that question.

This is tricky... I hate duplicate answers. I've been known to down-vote every answer to a question save one, purely because they added nothing to that one. Even if that one answer preceded them by seconds... Because, really, who wants that noise?

As The jjnguy is fond of pointing out, different wording for the same thing can be useful. Some folks just need a little different wording to understand what's being shown them. So the fact that there are 18 answers to that question isn't necessarily a bad thing...

Except that a pile of them are duplicates. Exact duplicates. And in two years, no one down-voted them, no one commented on them, no one improved them... They just sat there, repeating what the accepted answer already said, taking up space. I'm not sure why you deleted the answer that opted to describe the syntax for CASE instead of giving concrete examples; it may not have been all that great of an answer for most folks, but it did give the answer in a different way. But that still leaves at least 5 answers that add nothing.

Anyway... The specifics:

  • Should there be any answer deletion? No answer deletion? Some answer deletion?

My personal preference is that authors delete their own answers when they fail to add anything new. Commenting and down-voting is a reasonable way to encourage this...

Failing that, I advocate a conservative approach: leave them alone until they're doing actual harm. What harm can a duplicate answer do? In this case, obscuring the new answers, specifically: one that was inapplicable (but not down-voted) and one that correctly noted the existence of a new method (previously suggested, but inapplicable at the time).

  • Should we play favorites improving the top voted answer / first answer?

Yes! If you do that, who cares how many other answers there are? Sucks to be the guy not getting any rep for noticing that IIF now exists, but it's great to be the next reader who doesn't have to troll through 18 answers to find it!

  • Will deleting answers with +17 and +13 score turn away high rep users?

Impossible to predict. Shame no one bothered pointing out that they were dups two years ago though...

  • Has this asset become a public resource, whatever that means?

Well, yes. The asker got his answer two, almost **three* years ago. The 160K readers since are the only reason it's worth keeping around, and the core reason for having this site in the first place!

  • What should we do when new answers trickle in and are dupes?

Down-vote them, perhaps leave a comment noting that this answer already exists and that they should try to add something new if they really must bother. Protect the question if useful. Delete once noise-level injures the ability of future readers to evaluate new and novel answers.

share|improve this answer
7  
"My personal preference is that authors delete their own answers when they fail to add anything new." - I agree with this 100%. I wish all authors did that. –  jjnguy Aug 19 '11 at 2:36
    
Your answer to "What should we do when new answers trickle in and are dupes?" I disagree with though. If they are just sitting at the bottom of a list of answers, then what hard are they doing? What information are they obscuring? Why can't we just ignore those answers. Someone may find them helpful. –  jjnguy Aug 19 '11 at 2:39
5  
@jjnguy: if they're near-exact duplicates of existing answers, then unless they get down-voted future new answers will sort below them - even if those future answers actually bring something new to the table. As I mentioned, this specific example had a new answer posted about a month ago, in that time it got one up-vote - still sorting it below 9 other answers, most of which were superfluous. Donno 'bout you, but I'd probably not read through ten answers that all seemed to be saying more or less the same thing in order to find that new one that said something different... –  Shog9 Aug 19 '11 at 2:48
    
I guess I was assuming that all of the new answers were not adding any more value. However, the same problem happens with relatively new answers too. When new answers get added, if it is below the fold, it has a far worse chance of getting read. –  jjnguy Aug 19 '11 at 2:54
3  
The system breaks down as the number of answers increases, yes. All the more reason to make each answer COUNT! @jjn –  Shog9 Aug 19 '11 at 2:57
    
"All the more reason to make each answer COUNT!" - My point is that each answer does count for something. It is far to subjective for someone who completely understands a topic to delete answers that they deem to be duplicates. A person reading a question for a solution can deem an answer duplicate and move past it if they want. (If someone really needs to solve a problem, they will read through as many answers as it takes until they fix the issue.) –  jjnguy Aug 19 '11 at 3:03

I delete answers if:

  • They're 100% a duplicate of another answer -- they literally don't include a single new fact, however trivial
  • They're newer than the other answer
  • They have a lower score than the other answer

So my actual deleted answer count is pretty low; on SO I would expect it mostly on trivial questions that get 10 duplicate answers in the first 30 seconds. It sounds like super popular questions like this also fit the criteria, since new people show up a year later and submit a duplicate answer without reading the existing ones. In any case, if people submit an answer and realize somebody beat them to it they should be deleting it on their own, so they certainly shouldn't get bent out of shape if somebody else does it for them.

I definitely wouldn't have made that edit, and I probably would've rejected it if it were a suggestion; you more than doubled the amount of information in it -- it should've been a separate answer.

share|improve this answer
1  
define 100% duplicate ... if you alias a column with AS vs = vs nothing qualify a differing answers? –  waffles Aug 19 '11 at 1:14
    
also, do you consider the example question to have ANY duplicate answers? if so which? –  waffles Aug 19 '11 at 1:19

I think dupe answers aren't a problem, as jjnguy said, IF it contains some explanation rather than just a chunk of code.

For me there is one thing wrong, shallow answers.

I think people who post just a chunk of code without any explanation shouldn't get upvotes at all. The answers might help someone, but the effort to guide the OP is so shallow I think it it is not worth an upvote.

The +17 answer is a good answer, it is showing the syntax of the CASE statement and it mention that you can use to order the result as well. It has interesting and useful information.

But if you look at the +13 answer it is just a chuck of code, that misses some points mentioned by others. The text entered on it is obsolete. Although the answer is awful, it is correct, I am not sure what deleting it would mean, but I don't think people would like to have a 'correct' answer deleted. Instead of deleting this answer there should be a way to "lock" votes on it, people shouldn't vote on answers like that.

I remember a post on the blog about creating pearls not sand, that's what is missing, pearls. People should incentive complete answers rather than chunks of code without explanation (aka sand).

share|improve this answer
    
I think it is very insightful answer, the issue that I had was exactly that, shallow - chunk of code - answers. There was no explanation of the subtle differences of how you can alias a column etc... just use "chunk of code" ... here is your "chunk of code" ... –  waffles Aug 19 '11 at 2:33
1  
also I agree with you on the +17 answer ... my finger was way to heavy on the trigger there, its a good answer I should have left it untouched during my rampage ... if anything I probably should have improved it –  waffles Aug 19 '11 at 3:20
    
I personally welcome all code answers, as long as it solves my problem. While more explanation is nice, it's not usually necessary, in fact it can be insulting sometimes when they want to explain the most basic of concepts. Just show me the syntax that I should be using, and I'm a happy man. –  Lance Roberts Aug 19 '11 at 4:10

I'll bite:

Should there be any answer deletion? No answer deletion? Some answer deletion?

That depends.

  • If members delete their own answers because they're wrong/not relevant - absolutely.
  • If answers get flagged as very low quality or get delete votes by privileged users - certainly.
  • If low-voted answers exist which are only superficially different from high-voted ones - usually.
  • If the existing answers are very similar (even if not identical) and the answer count is high enough to push new, original content off the first page - probably.
  • If some answers are trolling, farming, or otherwise notably unhelpful - sometimes.
  • If there is one clear, accepted winner, and others are different but much lower quality - maybe.
  • If all of the answers still fit on one page - probably not.
  • If the top-voted/accepted answer is lower quality than newcomers - no.
  • If all of the answers are offering different solutions/perspectives - definitely not.

Should we play favorites improving the top voted answer / first answer?

If the answer is accepted and needs improvement, yes, because it's the first one people will see. However, unless that answer was genuinely wrong or incomplete on its own, I would not add new information from other/newer answers unless the new answers have extremely poor visibility, for example due to a very high answer count and none of the conditions above being applicable.

Will deleting answers with +17 and +13 score turn away high rep users?

If these users are earning their high rep by posting obvious duplicate and/or low quality answers, who cares? Spend some time on Programmers.SE and you'll be as numb to this as I am. Honestly, once you reach 25k and beyond (and there are a ton of Stack Overlow members above that now), you probably won't even notice the disappearance of 100-200, much less care.

Has this asset become a public resource, whatever that means?

Posts become public resources as soon as they're posted. That's what the CC-wiki license and highly-discoverable edit functions on these sites are for. Public does not mean read-only.

What should we do when new answers trickle in and are dupes?

If they truly do not add anything new, delete them, preferably after leaving a comment to the poster asking them to upvote the post they've duplicated if they agree with that answer.

share|improve this answer

I for one vote for trimming the fat and getting rid of shitty answers. I think that once a question has passed a tipping point (I consider three primary criteria)

  1. Older than six months
  2. Older than 12 months
  3. More than 10k views
  4. Has more than 15 answers (spills over to the next page)

So given those three criteria, if we take any two of those, I think that we can start trimming the fat. We need to get rid of weak answers, for sure, and we need to consolidate similar answers.

I think this is going to annoy high-rep users, but it'll probably piss off low-rep users more.

HOWEVER

I think that it's what the community needs even if it's not what the community wants. We need to focus on canonical and right, not popularity.

Besides, if you spend enough time trying to make sure yours isn't the one deleted, you're going to get more rep for it. If you didn't care about it in the first place, you're going to find that others will care for it even less. And I think this is the bottom line. You have to care about the answers you give more than for the immediate tiny bit of rep. The ones of us who do and find our answer was deleted because it was crap will ask why it was crap, and then we're likely to improve our answers in the future.

Now, we need a system for doing this, so I propose that the rule is you have to give the OPs time from today till the end of the year to start getting serious about updating their own posts, but after Jan 1, 2012 it's open season according to the criteria above. Not everyone logs on everyday, and not everyone logs onto meta, and not everyone reads the blog. But getting the word out there to the high-reps now will get them to decide if this is a big deal for them.

share|improve this answer
    
I will support that. I just want to give a groundrule that there comes a point where your answers may just disappear and sorry charlie. –  jcolebrand Aug 19 '11 at 2:39
    
You listed four criteria in your tree criteria list. –  Paŭlo Ebermann Aug 19 '11 at 16:17
    
What, you want consistency? :p –  jcolebrand Aug 19 '11 at 16:18

As someone who was pointed here having created one of the duplicate answers in question I have the following observations

  1. The duplicates were all close together, and if memory serves before SO had the warning that someone had answered whilst you were typing yours.
  2. None of the duplicates are wrong.
  3. I've seen accepted and popular answers on security questions where the popular answer is wrong and lots of people have ended up pointing it out. Giving favouritism to the accepted answer means you must be 100% sure it's correct, and letting one person with delete privileges decide that is dangerous and strikes me as contrary to the SO ethos.
  4. If you have some sort of sensibility where a popular easy question garners duplicate answers in a short amount of time fine - but putting a comment to this thread on meta under each answer strikes me as much more unhelpful than the duplicate answers and offensive to my sensibilities.
share|improve this answer
    
the comment was placed there so you can join the discussion ... removed it from your post –  waffles Aug 19 '11 at 2:58
    
also ... that answer should at least be fixed, should be Obsolete = 'N' AND InStock = 'Y' ... I am failing to see what value it provides that non of the other answers do. do you care to expand your answer on SO? –  waffles Aug 19 '11 at 3:00
    
Yes, duplicate WRONG answers are even worse than duplicate right ones... –  Shog9 Aug 19 '11 at 3:02
    
@shog: I think blowdart meant duplicate answers pointing out that the accepted one is wrong. –  Paŭlo Ebermann Aug 19 '11 at 16:16
    
Hah, it is wrong, so I deleted it. If it was a correct dupe, then I would not have deleted it as it would have done no harm. –  blowdart Aug 19 '11 at 18:19

You must log in to answer this question.

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