The following pattern is common on math.SE:

1. Question Q is posted: is statement $S_1$ correct?

2. Answer A shows that the statement is wrong, by displaying a counterexample. As far as Q&A of specific questions is concerned, this is a correct and conclusive answer to the question as stated.

3. Question Q is modified in light of counterexample A: is statement $S_2$ correct?

4. Correct answer A is now subject to downvotes because it is inapplicable, wrong, or irrelevant to the revision $S_2$, and later readers of the exchange do not see that it refuted an earlier version of the question, except in the rare case when they review the edit history of the question.

More generally, editing questions can improve them but substantial changes to the content can turn the Q & A into a blog which is confusing unless answerers "post defensively" with lots of redundant material in case the Q changes.

The FAQs I have seen do not address the following points:

-- focused Q&A is not a "blog about the topic of the question"

-- advice to questioners on how to operate when editing the question (e.g., explain the change and its connection to any existing answers)

-- advice to answerers on how to avoid confusion when posting (e.g., include material that makes the answer more self-contained). This is not necessarily desirable because placing more demands on the respondents would make it marginally more time-consuming to post answers in some circumstances --- if a stand-alone paragraph is needed where a more context-dependent one-liner might have sufficed, some answers might never get posted.

-- should a refuted (or fully answered in its original form) question be edited, or frozen and with modifications of the question posted as separate threads?

• Is it unwanted behavior to simply ask a new question and leave the original up, as well? We can simply accept both resulting answers as valid correct answers to two different questions. – Justin L. Sep 20 '10 at 17:14

Question Q should retain statement $S_1$, possibly struck out like this, and explain that because of someone's useful answer the OP was led to statement $S_2$. In other words, the goal should be maximal transparency. This isn't an issue with the reputation system as much as it is a misuse of editing.

• Is there an FAQ or other easily cited statement of this or similar principle? – T.. Sep 2 '10 at 17:45
• I don't know. It evolved by consensus on MO. – Qiaochu Yuan Sep 2 '10 at 18:34
• Diagnosis of any given case as misuse of editing is clear, but something like an FAQ would be useful for reducing the incidence in post-beta phase. Stackoverflow has a large number of FAQs so it's not always clear whether something is already written or not. – T.. Sep 2 '10 at 18:42
• Sample FAQ-able text: "Repairing questions is fine, but silent editing of refuted questions is discouraged, because it turns correct answers into "wrong" ones, leading to (1) downvotes of the initially correct answers, and (2) confusing inconsistency between answers, where some say the (revised) question is correct and others that the (original) question is wrong, without any answer being mistaken. Correct editing practice is to retain the original question content, adding (and displaying) modifications in a way that makes the evolution of the question transparent." – T.. Sep 2 '10 at 20:55
• I have tried to strike the LaTeX code, but apparently it does not work. Test <strike>$\frac{a}{b}$</strike> – Américo Tavares Sep 2 '10 at 23:55
• @Américo Tavares: I noticed that too. It's a bug I'd say. However if one strikes a larger piece of text surrounding the formula too, it then becomes a bit more workable, even though the LaTeX parts still aren't struck. For fixing only formulas, one would have to add something explicit like "<new formula> (was: <old formula>, fixed thanks to comments)". – Fizz Feb 10 '15 at 11:38
• This works: $\color{#ff8888}{\frac a b}$ $\frac{a^2} b$ – wizzwizz4 Dec 10 '17 at 11:48

Let me concatenate my three longish comments to Qiaochu's answer into an answer of mine, since I think it will be easier to read and respond to. I have added a fourth paragraph.

As Qiaochu says, the general principle here is that edits should only add text and not subtract it. It is much better to strike out, rather than delete, a wrong statement, especially if it has already been responded to. I agree that it would be helpful to add this to the FAQ.

What I said above is pretty clearcut and easy to get in the habit of doing. What is less clearcut is how many times it is kosher to modify a question in response to answers which correctly answer the question as it was asked but not in the way the OP meant. Here is one idea on this: if you are passing through and you (i) immediately know the answer to the question as stated and (ii) suspect that the OP intended to ask a different, more interesting question, then leave this as a comment rather than an answer.

On the other hand, it is arguable that if someone puts real time into answering the question as asked (it is hard to know this directly, but one can make reasonable guesses based upon the length, intricacy and quality of the answer), then s/he deserves to have this answer accepted and the OP should try again with a new question. Once the question has been changed, if people follow the editing practices described above then downvotes for the "old answer" should be avoidable but it is unlikely that this answer will get the upvotes it may deserve.

Finally, although I expect not everyone will agree with me, I do encourage answerers to parse the question literally and attempt to answer the question that has actually been asked rather than the question that one thinks the questioner might have meant or wishes the questioner had asked. (Whatever else you want to say in your answer is fine with me as long as it includes a good faith attempt to answer the OP's precise question.)

• I'd agree with the latter half (" if someone puts real time into answering the question as asked (it is hard to know this directly, but one can make reasonable guesses based upon the length, intricacy and quality of the answer), then s/he deserves to have this answer accepted and the OP should try again with a new question") and provide as concrete example: math.stackexchange.com/questions/1133083/… – Fizz Feb 10 '15 at 11:41
• As for deleting vs. fixing a wrong statement if it hasn't been noticed by someone else (as wrong)... it depends if it substantially affects the post or not. My only question here insofar was alas motivated by an example which turned out to be wrong in the sense that it didn't have the property I thought it had. In that case, it made sense to struck out the example and add an explanation, even though nobody had called me on it, although I'm sure someone would have had I left it unchanged for a longer time period. – Fizz Feb 10 '15 at 11:47
• On the other hand, I've removed a conjectured improvement (to the method/algorithm used) from then end of an answer of mine within minutes of posting, after realizing it was actually incorrect, and I don't think I've done wrong doing that for it would not have helped much to get into details (in an already long post) on stuff that doesn't work. – Fizz Feb 10 '15 at 11:52