What can be done to improve moderation of self-answered questions?

Question

About a month ago, we rolled out a new feature: instant self-answered questions.

So far, the reception has been mostly positive, and I've definitely seen some really good stuff posted in this form.

But I've also seen something else, a side-effect you may have seen discussed here on MSO:

• Is Stack Overflow a central store for tutorials?
• Why was my answer converted to a comment?

Ignore for a moment the specific details of the Q&A pairs discussed in those two posts, and note the following: three poorly-asked questions, questions that ordinarily would have been closed, down-voted, and/or edited... But instead sparked a small outrage among the members who came across them.

Now, we've known for years that self-answered questions can be problematic. Asking is hard. Answering is hard. Asking a question when you already know the answer can be even harder to do well. But that is - or rather, that should be - beside the point: plenty of people ask really terrible questions without bothering to answer them, and we've certainly managed to handle them over the years. If normal community-moderation is failing when it comes to self-answered questions, that's a problem.

These are edge-cases at the moment. But if we're trying to encourage more use of this form to share knowledge on Stack Exchange (and we are), it's something that needs to be faced. I'm presenting one possible solution below (in true gratuitous self-answer fashion), but what do you think? Would some other option work better? Is this even a real problem?

Answer 1

Two thoughts.

1. Part of the problem I see with self-answered questions is that people put so much effort into their answer that they forget the question has to be of acceptable quality as well. If the question is so bare-bones that it wouldn't be allowed as a standalone incoming question from the broader community, why does adding a giant detailed answer magically make it suddenly acceptable? It doesn't. If the question is unclear or barely more than a stub, it'll never get other answers, and any effort to improve the existing answer, whatever its quality, will be hampered by the fact that people have to basically read your mind because they can't understand what the underlying question is.

2. People see that you are getting rep both for the question and for the answer, and getting rep for both feels a lot like intentional double-dipping. Particularly when the question is barely more than a stub (see #1, above). If your answer is so great and so useful, surely you could have found one of the other 3+ million questions that already exists out there and slapped your amazing answer on that question? Synthesizing a question just for the sake of adding your answer feels kinda slimy and/or lazy, unless you do it right and present it as an actual problem you really had and could legitimately find no good solution to, not just "here's an excuse to post this thing I want to post".

One possible solution is to "unify" reputation in these cases so intentional self-answers (that is, answers that come in within 1 second of the question being posted) don't get rep for both the question and the answer, but just the answer. This would at least unify the incentives a bit and prevent the perception, correct or not, of double-dipping.

edit: I also realized that forcing the question part of self-answered questions to community wiki at the time of ask is an easy existing solution. Downside is, that would also force all new incoming answers community wiki, too, at least if the code is still the same (I can't recall).

Answer 2

The answer to this question seems very clear to me:

From Jeff Atwood's comment:

If the question wouldn't be allowed on SO as it stands -- and these clearly wouldn't -- there's no way a self-answer should make it immune to the regular standards we have for all incoming questions...

Let's treat these questions just like any other question. As a community, we need to train ourselves to hold these questions to the same standard we would any other question.

In fact, the community can -- and has -- stepped up to help fix these problems when they're encountered.

For instance, recently someone asked How can I style a part of a single character with overlays using a dynamic width? because he found the answer on his own and thought it would be great to document the solution on StackOverflow. Understandably, the question was closed as Not a Real Question. Here is the original revision before 5 or 6 people all collabratively edited the post and reopened it:

The Question:

I wanted to ask the following question:

Can i style just a part of a single HTML character? For example an "X" which is half-way red and then black?

While playing around with a demo fiddle, i figured it out myself and wanted to share my solution. It's really simple.

By the way, my intention was styling the Font Awesome [icon-star][1], so that i have an exact visualization of scores.

In the question, the op states his intention to answer his question. He broke the 4th wall!

When Jeff says "pretend you're on Jeopardy" what he really means to say is that we must ask our question as if we were the person experiencing the real, actual problem being faced. Then answer it separately, as if we were another person. In other words, pretend you're an actor on a stage playing the role of someone with a problem. Just like a professional actor, you must stay in character.

Self Answerer's Must Role Play:

For example, when self-answering, it's helpful to think of yourself as two people: jmort253(1) and jmort253(2). jmort253(1) has the question and posts it, including what problem he faced, what research he did, what alternatives he considered, and example code, if applicable. jmort253(1) should write the question from the perspective of someone who is having a problem and who simply cannot find the solution. In fact, in his role, he knows of no answer.

jmort253(2) on the other hand, sees the predicament jmort253(1) is in and answers the question, following all of the guidelines of "How to Answer", including real code and explanations and maybe supplementing the answer with a link.

Additionally, a third party coming to this page should not know that the question was self-answered. Imagine if usernames were hidden on all posts and everything was posted by Anonymous. There should be nothing in either question or answer that suggests it's self-answered or that something is amiss.

This is the true meaning of "pretend you're on Jeopardy".

Answer 3

Random thoughts (mine, not random's):

• Get rid of insta-answers

  Insta-answers are pointless (imho). If I have a problem that I happened to solve myself, I might post a question, only to verify my solution or get a better one. However, even if I get a far better solution than my own, I don't care, I've already solved my problem and moved on. I'm might check back months later on a lazy Sunday afternoon and do some refactoring, but other than that, why bother?

  Time to write a good answer (to a problem I don't have) > imaginary unicorn points

  And at the end of the day, it's just a UI trick, you can always self answer your questions if you really want to. Normal self-answers I get, you did some further research, had a light bulb moment and you want to share it with the world. But insta-answers are completely artificial.

• Delaying the answer

  I'm guessing we aren't getting rid of them. Ok, let's keep them, and just delay the answer for a while, 15-30m, whatever, to give the community time to curate the question, close it, or even answer it. The delay could go away after you have a few successful insta-pairs (open question, both question and answer positively scored).

• Don't award Enlightened and Reversal on insta-answers

  Enlightened would be unfair, if you insta-answer no one else has a chance of getting it. Reversal would be just silly, you'd be awarded with a gold badge for a horrible question (and a good answer, but it'd still be cheap gold).

• Put a notice on the insta-answer

  Most people seem to have little idea of what insta-answers are, how they work, what's acceptable and what not. Let's advertise this beast a little bit more, with a simple notice on the answer saying "Hey, this is an insta-answer, check out this blog post to find out more...". The notice could go away after a couple of days, but I think it could help with the initial "you filthy rep whore" backlash on the Q&A insta-pair.

Answer 4

As a frequent self-answerer across several SE sites, I feel like any solution will need to include a community-awareness element. Technical solutions probably won't work in isolation. Perhaps the biggest problem is the perception of "double-dipping". It's absolutely true that I often get a great deal mileage out of self-answered questions in terms of reputation. Since self-answers are often clear and generally useful, they tend to attract upvotes from people who stumble across them looking for solutions to their problems. And since questions tend to be evaluated in part by the quality of their answers, they tend to be upvoted too.

But, at least for me, self-answered questions are actually more work than one regular question plus one regular answer. The answer part is especially labor-intensive since I know many of the pitfalls the asker will run across (since I'm the asker). My answers try to cover all the angles. However, in my zeal to provide the best and most complete answer, I've found that my questions tend to suffer neglect. Quite often, the question has just one sensible answer: the one I provided.

---

Therefore, I suggest defining a set of "best-practices" or community-enforced requirements for posting self-answered questions:

1. Search diligently for existing questions that can be answered with the knowledge you wish to share. (As an aside, many of my self-answered questions begin life as true questions that I discover a solution to in the process of researching the question. Searching for a previously-asked question is one of my early steps.)

   Recourse: if the community discovers that a duplicate question exists, the self-answered question should be closed. Closed questions gather fewer upvotes than open questions.

2. Write the question from the perspective of someone truly struggling with an answer. Be as clear as possible about the difficulties and list some dead-ends that might be chased down. Don't save all your firepower for the answer.

   Recourse: the usual suite of tools should be used with special attention to self-answered questions. Close undeveloped questions as "Not a Real Question". I can't think of a better case for using that reason than a poorly-worded self-answered question—it really wasn't an open question.

3. Provide an encyclopedic answer. Don't just answer the immediate question, but fix it two (or even three) ways. Give the reader everything they might need to answer related questions themselves.

   Recourse: write a better answer or an answer that fills in the gaps. Nothing makes me happier than the ask a question I already have an answer for and have someone else answer it better than I did. Since self-answers are designed to provide long-term, generalized help, better answers will garnish more reputation than mediocre answers.

4. Edit questions and answers diligently to respond to comments and criticism. Tend your blog-like posts as if they were, well, your own blog.

   Recourse: comment on and edit posts that are not as good as they can be. Let the author know that the quality of their work matters to you and that improvements are appreciated.

Conclusion

As a rule, people are good at detecting "cheating" and try to punish it. Since self-answered questions bare some marks of illegitimate reputation grabs, folks who write them must be especially careful to go about asking and answering with integrity. The community should be aware that such posts are often designed to make the internet a better place rather than building reputation on the site. The later, however, is a welcome side-effect.

Answer 5

I see these posts as a way for people to put up mini blog posts with a comprehensive write-up of specific topics or areas of concern. I'd love to be able to do this personally, since I feel I have the ability to contribute blog-quality content, but don't necessarily want to stand up my own blog to do it.

But there has to be some understanding about how these posts fit into the larger body of questions and answers on Stack Overflow, especially with respect to duplication.

A canonical post is supposed to replace any number of more localized posts. It is not a duplicate of those posts, and should not be closed as such. When someone asks the "headers already sent" php question, they post their specific code. The canonical post is supposed to replace those posts with a more general and comprehensive post.

There needs to be some sort of "post notice" letting the community know that this is a canonical post, and is intended to replace any number of "frequently asked questions" on the subject. The post (and its answers) can still be vetted in the usual way with votes, but there needs to be some leeway with respect to closing as a duplicate.

I'm on the fence with regards to CW. While I always CW and protect FAQ posts on Meta (which is our closest model to these canonical questions), if a person is going to put in the effort to write up a canonical post, they should be able to earn reputation for it.

I don't think the answer needs to be delayed. That kind of defeats the purpose. Nor do I think double-dipping is a problem. If people think the OP is getting too much rep from the post, they can just refrain from voting.

On Meta, there is a special tag for canonical posts: [faq]. I would support a similar tag on Stack Overflow.

Answer 6

How about forcing self-answers posted immediately to be Community Wiki? This would accomplish four things:

1. Emphasize the question. Good questions are important; bad ones should be dealt with quickly (edited, voted on, closed).

2. De-emphasize the author. If folks are getting hung up on self-answering for its own sake, this makes that less obvious (without actually hiding anything). Which in turn should aid reviewers in focusing on the quality of the content instead of...

3. End charges of "rep-whoring". IMHO, this is a bogus allegation anyway, twin to FGITW. But it's certainly a common enough charge.

4. Reward finding an existing question to answer vs. posting a new one and self-answering. Since votes on answers are worth more than votes on questions, and answers tend to attract more votes anyway.

The down-sides?

• Potentially seen as a revival of CW as a "Get out of jail free card" for otherwise-undesirable behavior.

• Can be "worked around" by simply manually posting an answer seconds after asking the question, thereby turning what was intended to be a convenience function into a handy way to shoot yourself in the foot.

• Confusing (since CW is always confusing).