Is there need to patrol for deleted questions?

In a recent discussion Mariano Suárez-Alvarez told us, that if we are aware of users, who delete their questions after receiving an answer, we should bring them to the attention of mods. (IIRC the OP can only delete a question if no of the answers have been upvoted.)

In an older discussion about the same problem Willie Wong said that:

While 10K users can see all deleted posts, the posts are only presented in a nice summary form in the 10K tools panel if they are "deleted by committee". Deleted by owner posts don't show up there. This makes it hard for users to "patrol" for such behaviour. Also, one can always use the free-form flag field.

and

nothing easy that I know of (and there's also a difference between what diamond mods and 10K can see/search). Hence my previous comment.

The fact that self-deleted questions are not listed anywhere was mentioned in this discussion, too.

So I want to ask you this:

• Should we try to find self-deleted questions which could be a sign of cheating? Does this problem appear often enough to warrant additional work for users/moderators?
• If yes, how this could be done?

I was not able to come up with anything more intelligent than downloading links in the form https://math.stackexchange.com/questions/XXXXXX using wget and then checking which of them could not have been downloaded. I've done a test where I found out that in 1000 posts I've tried I have found 14 deleted questions. (The reasons why I was doing this experiment were twofold: I wanted to see whether this method is feasible. And I also wanted to have some data about the proportion of deleted questions among all questions.)

Maybe someone who is more familiar with SE data explorer can come up with a better method how to find such questions.

Feature request mentioned in Jonas Meyer's answer would be another possibility.

EDIT: Since this post was made, there were some changes which now make possible for users to find their own answers which have been deleted (at least in some cases).

Namely:

For details see here: Is there any way to see my deleted questions or answers?

• MathOverflow users used to do it with help of the administrator (which we would not have here). Apparently they ran out of steam about a year ago. – user53153 Feb 21 '13 at 14:04
• +1 for your effort... – draks ... Feb 21 '13 at 14:23
• I tried tinkering with the data explorer, and I have not found any way to locate deleted posts. There is a "Post History" table that records deletions, edits, and other actions, and I can find posts that have been deleted, but only those that have been subsequently undeleted again. – MJD Feb 22 '13 at 5:20
• @MJD A few queries (although they show somewhat incomplete information about the post) can be found here: Are there some tools to follow deletions on this site? (And, of course, further additions and improvements are welcome, it is a CW answer.) – Martin Sleziak May 10 '18 at 4:33
• math.meta.stackexchange.com/users/recently-deleted-answers/… with the users id instead of mine. – Roddy MacPhee Sep 4 at 11:25
• @RoddyMacPhee Yes, that's the first bullet point in the part after "EDIT". However, this is only way how a user can get to their own posts that were deleted. – Martin Sleziak Sep 4 at 14:34

Changes on the SE part do not look likely. One thing an ordinary user can do is to develop a habit to save his/her answers with a service such as Evernote. (Chrome extension makes it straightforward to clip the entire page.) If it later turns out that an answer has disappeared, a search through the notes will recover the answer and provide a point of reference for the moderators, who can undelete the question.

Of course, a simple notification ("your post ... was deleted") would be nice, and if it came with a copy of the deleted post (by email) that would be even better.

Added: I found and adapted a query which retrieves the last 50000 posts ordered by id. (Apparently, 50000 rows is the most DataExplorer can return). Downloaded the data and ran a script to generate the list of the missing ids. There were 7192 of them, which gives the proportion $7192/57192\approx 14\%$ of deleted posts. Despite the apparently large number of processed posts, they only cover the period back to mid-December 2012. Note that posts include both answers and questions, and presumably most of the deleted posts are answers. There seems to be no way to tell which ones are questions (let alone questions with substantive answers) without navigating to each by the URL, if anyone is so inclined.

• By extension, it is easy to get the id numbers of all non-deleted posts by using "between 50000k and 50000(k+1)". Then take the complement of this set. Still no way to separate answers from questions, though... – user53153 Feb 27 '13 at 6:50
• Thanks for providing the list with links. I went through some questions from your list. The list seems to be deleted by now, but it should be easy to generate a similar list. Perhaps if I have some spare time in a future, I'll check a few of such questions again. – Martin Sleziak Mar 10 '13 at 11:17

Personally I wouldn't want the burden the community by requesting that we hunt down cheaters. I think it is enough that people know that this sort of stuff happens and that they can do something (i.e. flag) if they see it happening.

In general, if you spot a contribution to the site that you think is unnecessarily deleted, you can flag for moderator attention and we will do our best to sort it out.

Lastly, I am more in favour of preventative measures (which to me feel more "positive"):

2. Vote early, vote often. By voting up answers to a question, you are affirming that both the question and its answer has some value on the site, which will help prevent their being removed.
• Concerning 2: I find it appropriate to quickly upvote any answer that has no votes at all if, upon a quick scan, it looks like a serious attempt at the given question. I will not necessarily check the correctness at this stage. This is by analogy with an established process (at some better journals): quick pre-refereeing message to journal's editor merely confirms that the paper is something worth looking at, not that it works. If the question was already voted on, then that stage is over, and I have to actually read it to decide how to vote. – user53153 Feb 21 '13 at 14:23
• @5pm If your prior comment proposes that one should employ lower upvoting standards on unvoted answers in order to help prevent question deletion, then I don't think that is a good idea. This could cause problems much worse than the problem of question deletion, e.g. incorrect positive valuation of wrong or misguided answers, and. consequently, possible encouragement of further such problematic posts by the same author. – Math Gems Feb 21 '13 at 15:00
• @5pm Further, some students post answers which they are not sure of, hoping to receive feedback on their understanding, correctness of proof, etc. Lowering the standard of that feedback due to orthogonal (lower-value?) concerns is not fair to those seeking thoughtful, correct feedback on their answers. – Math Gems Feb 21 '13 at 15:11
• @MathGems The upvote arrow says: "This answer is useful". An answer does not have to be entirely correct in order to be useful. – user53153 Feb 21 '13 at 15:59
• @5pm: Some people feel that only if an answer surpasses their own it merits an upvote, otherwise it merits a downvote. Some people don't even vote on threads they don't answer, because they don't visit threads they don't intend to answer. – Asaf Karagila Feb 21 '13 at 16:26
• @5pm Eh? If you believe that incorrect or misguided answers are "useful" then it seems we have a very different conception of the meaning of "useful". I think that, generally, one should exercise more care in voting - not less care. Doing so would help alleviate some of the problems, e.g. that good or later answers often receive fewer votes than poorer but quicker answers. – Math Gems Feb 21 '13 at 16:27
• @AsafKaragila This entire chain of comments can be summarized as "people do things"... Willie may want to clean it up, as it seems my comment sidetracked the discussion of deleted answers... And yes, Math Gems, incorrect answers can be useful to me. Your idea of usefulness may be different. – user53153 Feb 21 '13 at 16:29
• @5pm I think the comments are quite useful in that they help to show they different ways that various users think about votes - so help to promote better understanding of each other. Further, they are quite on topic, having to do with how one should follow the "vote early, vote often" SE advice, i.e. whether or not one should compromise voting standards to do so. – Math Gems Feb 21 '13 at 17:09

There is no reasonable way to find self-deleted questions. They don't appear in the Data Explorer, and crawling the site like you did could cause SE to block you from accessing the site, if you are not careful.

But such automatic measures are not necessary anyway in my opinion, there is always one user that likely notices such behaviour, and that is the user that answered the question. If that user notices such behaviour, he can flag for moderator attention and the moderators can easily see if it happened more often for that user.

An old feature request of mine on Meta.SO might also be of interest in this situation: Preventing misuse of question self-deletion

• SE could block you from accessing the site, if you are not careful. I've tried to be careful as far as this is concerned. I waited 5 seconds between any two downloads, I hope that it is sufficient. Anyway I agree with what you wrote here: My main point is that the moment you're scripting the SE interface to achieve your goal, it indicates that the tools that are available are insufficient. I'd rather convince SE to improve those than hack around these deficiencies. – Martin Sleziak Feb 21 '13 at 14:34
• I don't have a high expectation that the average MSE user playing the role of the answerer would simultaneously: realize we want things to be flagged; be motivated enough to do it; know how to go about doing so. Also, those who answers in comments are much less likely to notice this happening. – user14972 Feb 21 '13 at 15:37
• "there is always one user that likely notices such behaviour, and that is the user that answered the question" How will they notice? I don't know how I would notice if one of 361 questions I answered so far got self-deleted (in many of these questions the answer has 0 score). – user53153 Feb 23 '13 at 3:58
• @MartinSleziak Good luck convincing SE... The 3.5-year-old request Show all of my question/answers to me even if they are deleted has the score of $\mathbf{354}$ and status-declined. Atwood responded to it with "Will not get implemented as long as I am still alive to prevent it from happening." -- and got downvoted to $\mathbf{-107}$ as of now. – user53153 Feb 24 '13 at 0:51
• @5pm I don't know how I would notice if one of 361 questions I answered so far got self-deleted. I know of two possibilities where it is possible to notice that a question has been removed. One case is when the question is in your favorites. Another possibility is to save your reputation breakdown to a text file from time to time. If you then compare two versions, you can see which answers lost rep. (But this only work if your answer has been upvoted. IIRC robjohn mentioned in chat that he saves this file.) – Martin Sleziak Mar 7 '13 at 6:53
• @Martin Thanks but unfortunately neither would help me notice. I use favorites to collect questions I intend/hope to answer, and consequently remove them from favorites once they are answered. And since the unupvoted answers are the ones easiest to remove, reputation count does not help as much. // It's no big deal, of course: answers are not academic papers. – user53153 Mar 7 '13 at 12:59
• @5pm Of course, you are right about upvoted answers. Such questions cannot be deleted by OP. I knew that - I'm not sure how I have overlooked this problem with my suggestion. – Martin Sleziak Mar 7 '13 at 13:07

I will add that in SEDE there is the table PostsWithDeleted which contains data with deleted posts. This gives possibility to find links to deleted questions and if the users with 10k+ reputation can also view them.

For example, it is possible to find all recently deleted question, such query is listed in this answer: Are there some tools to follow deletions on this site?

In particular, we could also try to look for question which could have been deleted by the OP after receiving an answer: query1, query2.

These queries search for questions which have only one answer (a question with two answers cannot be deleted by the OP) and where the answer does not have positive score (if the answer has an upvote, the OP can no longer self-delete).

Still there are som issues with these queries:

• There will be some false positives - since we can use the query to check whether the question satisfies the criteria for the self-deletion to be allowed, but it's still possible that it was deleted in some other way.
• Some questions can be missed by the first query - since we count all deleted answers; if some of the answers were deleted by the answerer and only one remained, the OP is still allowed to delete. (In the second query I only take into account the answers which have the same DeletionDate as the question.)
• The data in SEDE are only updated once a week - in this way we do not find the most recent deletions.