10
\$\begingroup\$

Currently our consensus is that incorrect solutions should be deleted.

However it's not clear if this applies to questions. questions are sometimes a bit like regular code-golf challenges, but usually it's more like Q&A. SE sites that are Q&A (to my knowledge) don't have moderators delete incorrect answers. They just let users downvote them. Deletion by moderator is reserved for spam, harassment or just pure nonsense.

So the question is: What policy to answers fall under? Are they required to be correct like solutions? Or should work like Q&A on Stack Exchange and wrong answers can stay?

As an example here is an answer which is not correct.

\$\endgroup\$

2 Answers 2

6
\$\begingroup\$

Yes if the answer is unfixable, because deleting unfixable, objectively incorrect, answers to questions would be beneficial even on Stack Overflow

(Note: this post is only about answers that are demonstrably, objectively, incorrect, and that inherently can't be fixed to be correct. With fixable answers, fixing them is obviously preferable to deleting them, unless the fix would cause them to become a duplicate; and with answers whose correctness is subjective, the correct response is often to close the question.)

Stack Exchange, in general, has a policy of not deleting incorrect answers to questions. I think that, assuming that the site has enough moderation power to delete answers discovered to be incorrect, this is a mistake for Q&A sites in general. (It's possible that the reason that Stack Overflow doesn't delete incorrect answers is that there isn't enough moderation power to determine which answers are incorrect and delete them, but our community shouldn't have that problem.)

The normal assumption made by the Stack Exchange software is that "The best answers are voted up and rise to the top". But the meaning of "best" is a bit fluid here, and in particular, it's rare for people to use their votes as a referendum on whether an answer is correct or not. An upvote does not, in practice, mean "I'm sure that this answer is correct"; rather, people upvote because they find an answer funny, or thought-provoking, or because it made them feel like it taught them something, or because it's well-written, or because it looks like a lot of effort went into it. None of these are necessarily an indicator that the information in a post is correct – in particular, a post can be incorrect and still get upvoted by the majority of voters.

Additionally, on Code Golf in particular, the majority of our questions hit HNQ, which means that if an answer is posted early, the majority of people who see it will be people from outside the site, who are able to upvote but not downvote. Additionally, the visitors from outside will generally be less equipped to determine whether the post is correct or not than the site regulars would be – they are quite likely to upvote an incorrect post due to not realising that it's incorrect.

As an example, when I posted this answer, I was confident that it would gain a score of 100+, and my prediction was correct. The answer is correct, but how many of those 137 voters do you think a) know Brachylog and/or b) checked enough of the Brachylog documentation to determine that it was a valid answer to the question? Probably not very many of them; so if I had hypothetically posted an incorrect answer instead, the answer would have received upvotes at pretty much the same rate until someone who knew the language turned up and commented that the answer was wrong (at which point it would likely stabilise at a fairly high positive score if not deleted). Although that answer isn't , I don't see any particular reason why answers would be different in this regard.

As another example (using a Stack Exchange site with a more normal Q&A format), I found an HNQ on Chess Stack Exchange for which the top answer was wrong, and voted to +8 or so (because it was interesting). I was able to persuade the author to delete it by posting a comment demonstrating the mistake in the comments, but I suspect it had a positive score at the time it was deleted. If the author hadn't been around to delete the post, this could have developed into quite an issue, because nobody else could delete it either (and arguably, by Stack Exchange deletion guidelines, the author was wrong to delete the post!).

This means that in practice, Stack Exchange doesn't have a reliable "this is wrong" signal – about the closest you can get is upvotes on a comment stating that the post is incorrect. In particular, that means that the software can't identify answers as incorrect; if the answer is upvoted, it will be treated as correct by the software, meaning that it will give reputation to its author, be sorted high when sorting by votes, and the like.

Deleting incorrect answers fixes all these issues, and has no obvious downside (other than the difficulty of determining that the answer is incorrect and is unfixable, but we already go to that trouble with non- questions). An unfixably incorrect answer is kind-of embarrassing for everyone involved; it makes the author look bad, and it makes the site look bad, and it can mislead readers if they miss the comment stating that it's incorrect (or if they believe the post author over the commentor – "this is incorrect" comments are sometimes themselves incorrect). Deleting such an answer basically sets things back to the state they'd have been in if the answer had never existed, which is probably best for everyone.

Such a deletion policy would therefore probably be helpful for all Stack Exchange sites (even the Q&A-oriented ones, which is basically what is). Code Golf isn't an exception – it's atypical in some ways (high community moderation power, very high HNQ/non-HNQ ratio, lack of review audits), but those differences imply that deleting incorrect answers should be an even better policy here than it would be on Stack Exchange globally, and it would probably be a beneficial policy even on Stack Exchange globally. (I agree that this isn't the policy on Stack Exchange globally, but think that that's a mistake.)

\$\endgroup\$
4
\$\begingroup\$

Yes, demonstrably invalid tips should be deleted

Note the distinction between a "valid, but not very good" tip and a "demonstrably invalid" tip. The first is what downvotes are for, the second should be deleted.


I believe we should have a consistent rules set for all answers on the site, and one of those rules should be "If an answer can be shown to fail to satisfy the conditions set in the question, it should be deleted". For questions, those conditions are generally

  • Your answer should be shorter than the original provided sample, and
  • Your answer should perform identically to the provided sample.

Therefore, if an answer to a challenge doesn't shorten the provided code, and if it doesn't perform the same task as asked for, it's invalid, and should be deleted, just the same as any other invalid answer on the site.

\$\endgroup\$
3
  • \$\begingroup\$ I feel like this doesn't really address the why? It proposes something but doesn't really provide an reasons to support that. \$\endgroup\$
    – Wheat Wizard Mod
    Jun 25 at 9:32
  • \$\begingroup\$ @WheatWizard The "why" is the first sentence of the second paragraph: "I believe we should have a consistent rules set for all answers on the site". If an invalid answer to a [code-golf] challenge is deleted, so should an invalid answer to a [tips] question. \$\endgroup\$ Jun 25 at 9:36
  • \$\begingroup\$ I guess I just don't see a reason why other alternatives are not consistent. It seems to me like a very weak reason. \$\endgroup\$
    – Wheat Wizard Mod
    Jun 25 at 9:38

You must log in to answer this question.

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