24

Today on a site I moderate we got a disputed low quality review flag on a post with the reason "Post was undeleted by the author". By the standards of our site, this deletion was entirely valid and the post had a negative score as required for this kind of deletion. The post hadn't been edited since being deleted, so no improvement to it had been attempted. It was deleted again by a moderator for these reasons.

Some background:

From the Deletion FAQ:

Normally, if your deleted post is not self-deleted, you can't undelete it yourself (though you may be able to vote to undelete). An exception is if an answer is deleted from Low Quality review queue by "Recommend Deletion" reviews, without three trusted users voting to delete, then it can be undeleted by the author.

Part of an early explanation for this implementation from 2014 seems to be largely technical:

  1. It doesn't happen that often
  2. Fixing it in the way you describe requires stuffing reviewers names into the deletion record - that is, we can't just attribute the deletion to Community, or undeletion is locked. It might actually make more sense to just unlock Community-deleted posts (that is, allow 3 trusted users to undelete them), but we never followed up on that previously because there were such a small number of these in play (see #1).

Part two is no longer an concern - Community deletions no longer carry the weight of a diamond and can be undeleted by three trusted users voting to undelete the post. From the FAQ:

If your post was deleted by trusted users or by the Community user, it will require three undelete votes to be undeleted - politely asking for this on the per-site meta may attract the necessary votes if you make a good case for why the question should be restored.

So, this leaves only the "It doesn't happen that often" explanation for it. Now, as I noted (and as the response to the Feature Request asking for this to be changed states) the solution was to raise an automatic moderator flag on the few occasions it does happen. A bug report was also marked as a duplicate of that Feature Request and marked as by-design with the explanation

You're be surprised how many of these auto-flags I dismiss because the post should have never been deleted in the first place, or the post has been improved since deletion and is now a valid post. If I had to guesstimate, I'd say well over half.

So, now we've got two explanations:

  1. It doesn't happen that often
  2. The deletions were invalid

So that makes me wonder, both of the linked posts were written several years ago

  • are these still that rare?
  • what percentage of the mod flags result in deletion/no action?

A bit about my concerns:

Our site has been struggling to get users to recognize certain requirements for post quality and get them to act so that these low-quality answers can be deleted by the users rather than the moderators. Many of our users are frustrated and confused about what qualifies as a sufficiently supported answer and when they should recommend deletion. It's an uphill battle that we're working through on meta and expect to continue to work through for a while still.

That being said, having the work of these six users single-handedly invalidated by a user with no reputation on the site seems unkind to the users who do. There is nothing in place to even cause a speed bump to this or make it visible to the non-moderator users of the site:

  • No edit is required (yes, we can argue about how valuable such an edit would be were it required - but it's something)
  • Without an edit, undeletion doesn't mark a question as "active", so it doesn't show up on the front page to draw any attention.
  • Any delete votes from trusted users are removed upon undeletion, requiring those trusted users to cast the vote a second time if they still think the post should be deleted.
  • When a moderator reviews that post and deletes it again, such a deletion requires a moderator to undelete it if it's ever fixed.

So, why do we do it this way? Could we do something better? Different?

Thoughts I've had about it:

  • Could the moderator flag be raised before the actual undeletion goes through?
    • the user could be given the option to "request undeletion".
    • This could also be useful in cases of mod-deleted answers that require moderator undeletion.
  • Do moderator deletions still need to prevent trusted users from undeleting the post?
    • Moderators could lock deleted posts that should stay deleted rather than forcing this locked state.
  • There is no delete/undelete queue, trusted users have to visit the "10K tools" page to see if posts have delete/undelete votes
    • would having one help?
    • what would cause something to go into this queue?
  • How often are moderators being forced to re-delete items undeleted from review deletions? If it's a lot, then sure, we could potentially look into a change. If it's not, and the mods aren't being burdened with more undeleted review deletions than they can handle, then why would we need to make a change? – Servy Jun 20 '18 at 21:48
  • I was shortcutting to refer to the undeletion of posts deleted in review. The posts that your whole question is about. Sorry if it didn't come out clearly. – Servy Jun 20 '18 at 21:51
  • 1
    That is part of my question - The very last two bullet points. Are they rare, are they a problem... – Catija Jun 20 '18 at 21:54
  • please take into account that at SO four reviewers can delete, not six – gnat Jun 21 '18 at 5:38
  • 2
    Also note that on SO, the LQ review queue only consists of answers; questions are reviewed in Triage instead. – Sonic the Reinstate Monica-hog Jun 21 '18 at 17:06
5

So, here's the short answer: it almost never makes sense for anyone to invest time into an answer that's been deleted via LQ review.

  • Moderators have enough other stuff to do. They have brand-new low-quality non-answers to delete. It could take hours, even days, to get to a deleted answer.

  • 20K users can undelete answers, but it takes three of 'em. That's three people's time, assuming they can even find the answer, which they probably can't.

  • The author could just post another answer. If it was deleted via LQ review, it scored 0 or less - so there's no reputation to retain. Unless they're answer-banned (which is extremely unlikely just as a general rule), posting a new answer takes very little effort.

If that last bullet - authors undeleting their own posts - seems kinda borderline, that's because it is: copy-pasting text is still slightly more work than just clicking a button that says "undelete" on it. So this is where the utility of this weird behavior comes into play: it gives the one person who might have a good reason to spend time undeleting an answer the ability to do so.

Now... To be clear, this was entirely accidental. We sorta made this up as we went along after it became apparent that relying entirely on 20K users to delete all the low-quality answers in LQ Review wasn't gonna be effective.

But... I really like how it worked out. It's one of the few parts of the moderation system that doesn't add a ton of overhead in the name of preventing abuse; instead, it just raises a flag that can be handled at the leisure of the moderators.

The best part about it? It's rarely abused. Mostly because it's barely ever used: only about 2-5% of answers deleted via review get undeleted by their authors (anywhere on the network); usually less. Of those, 2/3rds tend to get re-deleted... But remember, the alternative here is that the author re-posts the answer, so there's not much lost even then.

Alternatives

Most of your suggestions add overhead - either for the author, for someone else, or for the author and someone else. That's... kinda defeating the purpose. In particular, custom mod flags are gonna be a lot more work to handle than these; they're not likely to be consistent or easy to identify compared to a standard system flag, and are thus even more likely to languish.

We could require an edit here, if we wanted to emphasize the necessity of improving posts that several folks found problematic. I think that might be useful, but I'm not entirely convinced it's warranted right now given the low volume of answers affected - it might just end up tipping the balance toward more re-posted answers. If we ever wanted to implement something similar for questions though, I think requiring an edit for undeletion would be a sensible enhancement here.

  • 1
    Don't those flags get automatically cleared if a user undeletes (raising the flag), deletes again (clearing the autoflag), then undeletes it again? – Sonic the Reinstate Monica-hog Jul 16 '18 at 5:26
  • 2
    Lots of flags do, @Sonic. That's rarely important. – Shog9 Jul 16 '18 at 16:48
  • 3
    although I agree in general, this answer contains particular factual inaccuracy which looks worth correcting, namely this statement is incorrect: "it scored 0 or less - so there's no reputation to retain" - this happens only when answer has no votes up or 5x more votes down than up. But if, say, negative score answer has vote split like -10/+9 there are 80 rep points to retain. Consider editing to account for that – gnat Jul 17 '18 at 4:36

You must log in to answer this question.

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