Say I have an answer which gets 30 upvotes. After a time I realize the answer is wrong. Should I:

a) delete the answer?


b) edit the answer to be correct?

My concern with the approach in b) is that whatever I write in my new answer will not be vetted by the 30 people who upvoted my original answer. Since they aren't notified of the changes, I don't think it is fair to keep their stamp of approval attached to whatever I write. That would amount to the 30 upvotes being an endorsement of me as a knowledge source rather than an endorsement of the ideas in the original answer.

Additionally, according to the help page on editing, edits are to be used for correcting minor mistakes or clarifications which do not change the meaning of the post. I believe the reasoning behind that is to address the above concerns.

Please treat the scope of this question as general, but the discussion was spurred by Fundamental question about dimensional analysis as suggested by dmckee.

  • $\begingroup$ That post is a community wiki. Go ahead and fix the mistakes $\endgroup$ – Jim Jul 15 '15 at 18:30
  • $\begingroup$ NB: dmckee even said So I'm making the answer Community Wiki and anyone who cares enough can get it done $\endgroup$ – Kyle Kanos Jul 15 '15 at 18:33
  • $\begingroup$ Also, you can't delete accepted answers. I know because I have an answer that is wrong but accepted. I didn't have the drive to do the math to fix it, so I just wrote a note beneath it that it was incorrect $\endgroup$ – Jim Jul 15 '15 at 18:36
  • $\begingroup$ Only problem is that whatever the community wiki comes up with will have 30 upvotes to start with rather than zero. dmckee can delete the answer since he is a mod, but regardless the scope of this question is general. $\endgroup$ – pentane Jul 15 '15 at 18:41
  • $\begingroup$ How radically different would the answer be after your edits? $\endgroup$ – Jim Jul 15 '15 at 18:45
  • $\begingroup$ @Jimself: Methinks that the issue was that it was already edited (cf. the revision history) & pentane wants to know if that's okay. $\endgroup$ – Kyle Kanos Jul 15 '15 at 18:47
  • $\begingroup$ Well, it's not like those votes are frozen-in anymore... $\endgroup$ – Jim Jul 15 '15 at 18:48
  • $\begingroup$ For what it is worth, I believe the post has recieved 1 up and 1 down vote since the time I put in the header pointing to the paper that shows the original argument to be incorrect. It's been there for some time and I don't know if the original voters simply haven't seen it since then or don't care. I'm certainly not worried about the rep. $\endgroup$ – dmckee Jul 15 '15 at 18:58
  • $\begingroup$ Would you post an answer you know to be wrong? No. So why would you not delete an answer you know to be wrong? You might say "I will edit it to make it correct," and that is great if it is early enough that you have no votes. If you did already have votes, then people who already voted don't get a notification that you did so and it is not their obligation to check every question they've voted on to make sure the substance hasn't changed. $\endgroup$ – pentane Jul 15 '15 at 19:10
  • $\begingroup$ Pentane, there are and were many references to that answer elsewhere on the site. Simply deleting the answer would have broken those links and left a bunch of wrong assertions both hanging and not corrected. So I owned up in a prominent way and expected to be able to scrub the answer when a more complete one was available to take it's place. The more complete one never materialized, so the problem with deleting the answer remains, but at the same time it was wrong (as you helpfully pointed out). At some point getting it fixed is better than leaving it wrong. $\endgroup$ – dmckee Jul 15 '15 at 19:49
  • $\begingroup$ @dmckee I'm regretting referencing the specific post because I would like to come to a generally applicable solution. I think in the specific case what you did may be the best we can do because of all the links that would be broken. Please read my post below and give your feedback. $\endgroup$ – pentane Jul 15 '15 at 20:13
  • $\begingroup$ This question is very unclear and inconsistent with several of your later comments. Is "you" the author of the post, or someone else? I.e. are you asking for the course of action of an answerer who realizes their answer is wrong, or for what to do if you see an answer by someone else which is incorrect? $\endgroup$ – Emilio Pisanty Jul 16 '15 at 16:55
  • $\begingroup$ @Jimself A thousand times no. Community Wiki is Stack Exchange's "billion dollar mistake". Not quite as drastic as null in programming, but it is the least understood feature in the entire app, and I honestly am having trouble thinking of a time when it really should be used. I can't think of a single one. $\endgroup$ – corsiKa Jul 27 '15 at 22:45

A similar case was addressed in this Mother Meta post (does not mention "high vote," but I expect that the answer is the same regardless). The accepted answer there says that you should edit it if there is no existing correct answer and delete it if there is an existing correct answer.

Even further, Shawn Chin states,

Better yet, to keep the post clean, replace the wrong answer with an footnote which mentions your previous answer and include a link to the relevant revision of your answer.

Because revision histories are public, anyone (even unregistered users) can see them, so why not give them a direct link to the old wrong answer1. This will let future visitors see the history of what was the original answer, if they so desire, but keep the correct answer present for all to see first (which is more important).

If the answer has been accepted by the OP, you cannot delete it without moderator intervention. What you can do here is try asking the OP to unaccept your answer (comments under their question would probably be best) so that you could then attack it as above (edit if no correct answer is there, delete if a correct answer exists).

If they don't respond then you're in a bit of a pickle: do you (a) edit it for future viewers, OP's non-acceptance be damned or (b) post a new, correct answer, and never get the right answer accepted2? This, of course, is the dilemma we face here, as the OP in the question in question has not visited the site since September of 2011 (and has only asked that one question). I'd probably lean towards (a) myself, but it really is a judgement call best left to the answerer.

1 Like this one for this particular post.
2 Technically speaking, the accepted answer only indicates that that answer helped OP the most. Contrary to the opinion held by many people on this site, it does not signify correctness of the answer. So really this point is kinda a non-issue.


Just deleting a wrong answer is not good, especially if it is far from obvious that the answer is wrong. Silently correcting the answer is not much better either. You could instead leave the original answer basically untouched, add a warning at the begin that the answer is wrong, and possibly add an appendix explaining why the answer is wrong, and how this might be fixed.

Note that this is also what dmckee did initially (https://physics.stackexchange.com/revisions/7673/5):

Addendum: It has been brought to my attention that the argument herein is specious.

See DOI: 10.1021/ed1000476 "Can One Take the Logarithm or the Sine of a Dimensioned Quantity or a Unit? Dimensional Analysis Involving Transcendental Functions"

When later comments asked for more details about why the argument is specious, he made the answer a community wiki, such that somebody else could add these details, if he cares enough.

Ruslan cared enough to provide those details. Funnily enough, he initially also provided an argument why the cited paper is wrong and the original answer is correct. He later weakened this to saying that the cited paper is only physically wrong, but mathematically correct. My guess is that the paper is just wrong, but Ruslan is unsure how to nail down the mathematical mistake, so he tries to stay on the safe side.

  • 1
    $\begingroup$ No, you should not edit in a warning that it's wrong (cf. this Mother meta post) $\endgroup$ – Kyle Kanos Jul 15 '15 at 18:49
  • $\begingroup$ @KyleKanos That post refers to editing "This is incorrect" into someone else's accepted answer. This post is saying you can edit it into your own answer. But I do agree that a comment indicating its incorrectness is better in all cases $\endgroup$ – Jim Jul 15 '15 at 18:53
  • 1
    $\begingroup$ @KyleKanos The topic of this site has to do with science. How could we ever learn from our mistakes in science, if the expected way to proceed would be to hide mistakes and try to remove any traces they left? To avoid any confusion, I explicitly wrote "especially if it is far from obvious that the answer is wrong", which seem to be the case for the referenced answer. $\endgroup$ – Thomas Klimpel Jul 15 '15 at 19:02
  • $\begingroup$ @Jimself: The issue is that pentane didn't write the answer but is asking as if he did. In this case, the Meta post I linked to is correct. $\endgroup$ – Kyle Kanos Jul 15 '15 at 19:04
  • 1
    $\begingroup$ @Thomas: there is this really great thing called "Revision history" in any post that has been edited. No one can remove traces of edits because one can simply look at the revision history. Same thing happens in science, we submit errata for mistakes in our papers. $\endgroup$ – Kyle Kanos Jul 15 '15 at 19:05
  • 1
    $\begingroup$ @KyleKanos You forget that not everybody has access to moderator tools. I have no way to see the revision history of a deleted answer on this site. $\endgroup$ – Thomas Klimpel Jul 15 '15 at 19:10
  • $\begingroup$ IMO, deleting an answer should only be done if the answer doesn't actually answer the question (e.g., Q:"Why a in f=ma" A:"Roses are blue"). I'm currently writing an answer as to what should be done. $\endgroup$ – Kyle Kanos Jul 15 '15 at 19:17
  • $\begingroup$ I also wasn't thinking about the >10k rep viewing deleted answers bit, I was thinking purely about the everyone, even unregistered members, can see edit history of every post. $\endgroup$ – Kyle Kanos Jul 15 '15 at 19:22

We are allowed to edit without changing the meaning of our posts.

This is because if I espouse Theory A and you upvote my post and then I realize Theory A is wrong and change my post to say Theory B is right, then it appears that the community supports Theory B. Regardless of whether Theory B is actually right or not, the community never got to vote on Theory B, and that is bad.

This is why I believe we should recommend that if an OP realizes their answer is wrong that they delete it and post the correct answer separately (if they know it).

  • $\begingroup$ And if they aren't sure how to fix the answer, no other answers posted are correct, and/or they can't delete the answer (without mod intervention, obviously)? $\endgroup$ – Jim Jul 15 '15 at 20:29
  • $\begingroup$ @Jimself good point, updated $\endgroup$ – pentane Jul 15 '15 at 20:57
  • $\begingroup$ This is based on misunderstanding the edit how-to. Those are guidelines for editing someone else's post. You can edit your own post all you please, right to wrong or wrong to right. $\endgroup$ – Kyle Kanos Jul 15 '15 at 22:37
  • $\begingroup$ @KyleKanos - So, the moot question here is - do the same guidelines also hold for community-wiki posts? I thought, those are collectively-ours type of posts. $\endgroup$ – 299792458 Jul 16 '15 at 15:33
  • 1
    $\begingroup$ @TheDarkSide: So long as the answer is improved, editing a CW is free (cf. this Mother Meta post) $\endgroup$ – Kyle Kanos Jul 16 '15 at 15:40
  • $\begingroup$ @KyleKanos Say I post an answer that gets 10 upvotes. Then I change it to be the wrong answer (i.e. if I had posted this originally, it would have gotten 10 downvotes). Now it looks like the wrong answer is supported by the community. Do you not see the problem here? Maybe this deserves its own meta post. $\endgroup$ – pentane Jul 16 '15 at 19:04
  • $\begingroup$ If an edit is made to your post, those that voted on it are allowed to change their vote. This renders your "problem" moot. There's also the timeline page that shows the number of answers & upvotes per day and when they were edited. $\endgroup$ – Kyle Kanos Jul 16 '15 at 19:13
  • $\begingroup$ @KyleKanos There's no notification. Is everyone expected to continuously cycle through every post they've ever voted on to make sure the OP didn't change the meaning of their answer? AFAIK, upvotes on this site are for a specific idea not a specific user. They are not endorsements that say "I agree with whatever this person says in the future on this topic" $\endgroup$ – pentane Jul 16 '15 at 19:18
  • $\begingroup$ (a) Edits bump the post back to the front page, so you can find it that way. (b) no one expects you to scroll through all 50k+ posts to ensure there were no changes, but the option to change the vote is always there (which is the important thing). Upvotes mean that that person found the post to be useful (hover your mouse over the upvote icon). $\endgroup$ – Kyle Kanos Jul 16 '15 at 19:24
  • $\begingroup$ @KyleKanos right but they found the post to be useful in the form it was in when they upvoted it. The don't-edit-to-change-the-meaning-of-your-own-post rule is good because the votes are for a specific meaning, not for a specific person. Does that part of my argument make sense? $\endgroup$ – pentane Jul 16 '15 at 19:30
  • $\begingroup$ You'll need to provide some SE source that says you can't edit your own post to change the meaning. The link you're reading from is for editing someone else's post. And, again, when you make your edit, I am free to change my vote to reflect the current status of your answer. $\endgroup$ – Kyle Kanos Jul 16 '15 at 19:34

You must log in to answer this question.

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