34

From the help center, auto-delete section (emphasis is mine):

If the question is more than 365 days old, and ...

  • has a score of 0 or less, or a score of 1 and a deleted owner
  • has no answers
  • is not locked
  • has view count <= the age of the question in days times 1.5
  • has 1 or 0 comments
  • isn't on a meta site

... it will be automatically deleted. These are termed "abandoned" questions (RemoveAbandonedQuestions).

I have been reading about the scripts for auto deletion of posts in SE, and reviewing borderline questions to those scripts. By borderline I mean, almost all criteria are met except one.

The 'comments' requirement caught my attention. For example, all assumptions above are true except the question has two or more comments which are as old as the Q (see questions from query 'b').

Wouldn't it be more in line with the Stack Exchange saying "comments are second-class citizens" to instead of setting the requirement based on amount/count of comments, to set it based on the date from the most recent comment (no matter how many comments)? The name of the script refers to 'abandoned questions' and a pile of old comments under an old question does not make it less abandoned, does it?

There are many questions being kept from auto-deletion due to the comment criterium, consequently hindering useful searches (see query 'a').


Data to support proposal:

Take a look at the following data retrieved from two queries in SEDE:

Table1. Number of questions to be deleted based on different scenarios for the comment rule in the RemoveAbandonedQuestions script (first column is the baseline). Results from January, 2019:

                  total           most recent comment
                  n# of comments  older than n months          
              ------------------  -------------------
SE site****     < 2*         0**  12***      6      0(**)
------------  ------------------  -------------------
Super User        0  24750(6.3%^) 24625  24733  24802
Ask Ubuntu        0  12530(4.0%)  12412  12560  12602
Server Fault      0   7905(2.9%)   7839   7899   7922  
GIS               0   3137(2.9%)   3097   3128   3140
------------  ------------------  -------------------    
  • ^ Proportion of questions to be deleted in relation to the total number of questions.
  • * Number of comments < 2 is what is currently implemented (the baseline), i.e., no additional posts are deleted.
  • ** Number of Qs to be deleted if the comment rule was deactivated (no comment threshold required). Data was retrieved from query a).
  • *** Number of Qs to be deleted if date required from the most recent comment was older than 12 months. This was the maximum threshold for period of time used because it would not make sense requiring the comment to be older than the time threshold for the post itself (365 days). Data was retrieved from query b).
  • (**) The number of questions here should have matched the second column from section 'total n# of comments', but they were pretty close.
  • **** Unfortunately, the queries ran out memory for Stack Overflow and Mathematics.

If my queries are correct, it seems 'age of comments' are irrelevant to this matter, i.e., completely removing the comment rule would yield almost the same results than requiring the most recent comment to be at least 1 year old.

In other words, 'abandoned' questions which are currently being prevented from auto deletion due to the comment rule, also do have only 'abandoned' comments.


Meanwhile, if anyone is interested knowing about own comments preventing questions from auto-deletion, see this query and its description.


The following post is a near duplicate in Stack Overflow Meta: Time for roomba to ignore comments (credits to user gnat).

  • 13
    That criteria was probably used for performance reasons, because CommentCount is a denormalized column on the Posts table already. But if joining in the Comments table to get the last comment date isn't too expensive to cause the query to start timing out, I think that's be a good tweak. – animuson Jan 10 '19 at 23:39
  • @animuson if that's put in place I wonder if it'd be worth reducing the last comment date to a month or so (if not removing entirely)? Another year seems like a long delay for a "any luck with this?" comment that goes unanswered. – PeterJ Jan 11 '19 at 0:50
  • 4
    @animuson another possible reason is that in the past it was possible to use unanswered questions as dupe targets, see eg Time for roomba to ignore comments "Back then... it made solid sense to be careful about deleting old unanswered questions... in historical context that requirement for minimal amount of comments looks more like sanity check to ensure that question is not totally useless. Nowadays things are very different - practically opposite..." – gnat Jan 18 '19 at 11:10
  • Andre you've barely asked any question on Stack Exchange, but if you did you'll probably realize it's a pain to have one's perfectly on-topic questions be removed on the grounds that the score was 0. – Franck Dernoncourt Feb 15 at 2:57
  • @FranckDernoncourt I have asked more than 40 questions on main sites and more than 100 considering per meta sites. Every time a question of mine went unnoticed, I edited it; put a bounty on it (check for yourself); and 100% avoided auto-deletion. I realize you already have had bad experiences with roomba (meta.stackexchange.com/questions/269392/…), but if one goes through the question list which results from my queries, will see more than 90% of Qs are indeed crap or abandoned (OP took off) which makes searches more difficult. continues .... – Andre Silva Feb 15 at 10:31
  • However that is kind of beyond the point here. You are questioning if roomba should exist; I am saying that once the “abandoned question” script exists; it does not make sense to keep questions only because comments, when they are as old as the question (at least older than 365 days). This is shown with data/stats in question body. – Andre Silva Feb 15 at 10:35
  • @AndreSilva I've asked ~100 times more questions than you, it's important to remark to some extent because you have likely not the same level of experience when it comes to asking questions and how people vote them. I don't question here whether Roomba should exist, but rather that if we remove the rule you mentioned then we have to change another rule to avoid some very deleterious side effects. – Franck Dernoncourt Feb 15 at 19:19
12
+50

I agree with you on this. In particular, when you say this it resonates with me:

Wouldn't it be more in line with the Stack Exchange saying "comments are second-class citizens" to instead of setting the requirement based on a low threshold for number of comments keeping the question, to set it based on the date from the most recent comment (no matter how many comments)? The name of the script refers to 'abandoned questions' and a pile of old comments under an old question does not make it less abandoned, does it?

I review many old unanswered questions and I think that if a question has had no answer posted, and it has long since ceased attracting comments to try and make it more answerable, then it can be considered abandoned.

I propose that:

  • has 1 or 0 comments

be taken out of the algorithm, and in its place, this be used:

  • has had no new comments for 365 days
|improve this answer|||||
7

I just went to propose something similar myself.. In my mind, the logic is backwards.

More comments => worse question.

More comments likely implies the question needs more work, e.g., they're unclear so people ask for clarification. More specifically, comments require that

  1. a user has found the post,
  2. they chose to refrain from upvoting it [putting aside upvote/downvotes cancelling out], and
  3. they were willing to interact with the post.

If the question is attracting comments because it's good, it'd also get upvotes or answers.

Roomba is being held up by comments like the following on Stack Overflow:

Actual comments preventing roomba. (Obtained by this Data Explorer query.)

|improve this answer|||||
  • 1
    Nah. The logic (fear) is that the comments contain the answer. – Catija Feb 14 at 2:36
  • I'd guess that questions that attract answers in comments also tend to be low-quality (with the answerer expecting an edit). Plus the answer-in-comment itself is also usually low-quality. I don't feel like we'd lose any pearls. Maybe I'll investigate this more carefully in the future. – Rebecca J. Stones Feb 14 at 2:45
  • 6
    @Catija if occasionally that is the case and a question clear enough to get a “comment answer” but not a curated answer gets Roomba-ed then it just leaves opportunity to for a clear question that attracts a clear answer to be formed next time it’s important to someone. – PolyGeo Feb 14 at 2:55
  • @Catija the logic is more likely an (outdated) fear of losing duplicate target, see comments above – gnat Feb 14 at 9:27
  • How about we flag these comments, like: "any solution found ?!?!?!?" as no longer needed. Then the comment count drops and they can be roomba'd right? Or do deleted comments also count towards the threshold? – Luuklag Feb 14 at 10:37
  • It's more efficient to downvote the question to trigger roomba. – Rebecca J. Stones Feb 14 at 10:37
  • @RebeccaJ.Stones For starters I'm not going to downvote a question just in order to have it Roomba'd. My sense of ethics objects there. Flagging a non-useful comment in order to have something Roomba'd sits much better with me. If it really is a poor question I'd obviously would downvote, but then because of the question quality, not to trigger Roomba. – Luuklag Feb 14 at 10:39
  • Fair enough. (Related Data Explorer query and StackApp.) – Rebecca J. Stones Feb 14 at 11:13
  • 1
    Just to be clear... that's the logic... I didn't say I agree with it. But if you're interested in scale, I think there's around 800k-1mm questions on SO that haven't been deleted. – Catija Feb 14 at 13:56
-1

If one day the "has 1 or 0 comments" criterion is removed, then I would recommend to be changed with the "has a score of 0 or less" rule to "has a score of -1 or less", because plenty of perfectly on-topic questions have a score of 0, especially on Stack Exchange websites with low frequentation and for advanced questions. I've had hundreds of my questions removed by the auto-deletion script, and the vast majority of them had a score of 0.


Here are some examples of questions that got automatically removed because they had a score of 0 (but which I later reposted):

|improve this answer|||||
  • Would it be possible to give examples here please? – Rebecca J. Stones Feb 15 at 3:00
  • This would actually be no change at all - because if they're negative they get deleted after 30 days regardless of comments. – PeterJ Feb 15 at 3:01
  • @PeterJ good point. Then just not remove questions of score 0, unless if closed. – Franck Dernoncourt Feb 15 at 3:06
  • @RebeccaJ.Stones thanks for the link, it points to a question asking for "examples of questions which would be roomba-ed if it weren't for the 'has 1 or 0 comments' -> I don't have that many in mind, my the main issue is that plenty of questions of score = 0 are perfectly fine and removing them is extremely inconsiderate to the OPs. – Franck Dernoncourt Feb 15 at 3:11
  • All examples received answers or upvotes. If you are there to repost them; this is good. You will probably post Qs again with improvements; and a refresh Q might get more attention the second time (many users do this but on a non legit way; keep reposting same question in short period of time). Now, if you had “hundreds” of questions in this situation, probably something is missing on your part as well (not putting bounties; asking too many questions; asking questions which are in the scientific frontier; etc.). continued ... – Andre Silva Feb 15 at 10:43
  • Nonetheless, you are most certainly an exception at most. Many users who have gone over the list of questions not being removed only by the comments appended to such Qs (the result from query “a” in my question) will agree the majority of Qs are indeed crap and/or abandoned (OP took off). If was not for such scripts the mess would be much bigger. – Andre Silva Feb 15 at 10:47
  • @AndreSilva that's my point: in the example I gave, the questions were first removed because they had the score of 0, but then after I reposted them, then some of them got upvoted and/or answered, which means they had some value and shouldn't have been removed. I'm happy to directly link to removed questions with score of 0 but people would need to have over 10,000 of reputations to be able to view them. If the question is crap, then it should be downvoted. If you see a crap question with score 0 then downvote it. – Franck Dernoncourt Feb 15 at 18:49
  • @FranckDernoncourt; but there are also the abandoned questions. From 0 score to 1 score there isn’t much difference, really. If I had hundreds of questions being eligible for the deletion script in past years; I probably would rethink my strategy for asking. I have reviewed thousands of questions in the above situation (being kept by comments) before coming up with this question in meta (in the process, downvoted many from 1 to 0); and to me, the script makes 100% sense. I wish it did not consider number of comments; so to have a more positive experience when searching useful content. – Andre Silva Feb 15 at 19:05
  • @AndreSilva if the issue is just with abandoned questions, then the rules should specifically target them (e.g., if the OP doesn't answer the comments). If you had asked thousands of questions on over 50 different Stack Exchange websites as I have done so far instead of the only 50 questions you posted, you'll realize that some of them get a score of 0 even though they're perfectly fine. I've receive around 150,000 reputations on my questions so I believe my "strategy" to ask questions isn't too bad. – Franck Dernoncourt Feb 15 at 19:10
  • @FrankDernoncourt, that is it. If you have asked thousands of questions and had only a hundred in this situation; you are fine (nothing to complain). Your asking pattern (thousands of questions) is the exception. One needs to consider the overall picture of questions which get deleted (poor or abandoned). This brings benefit for maximizing user experience when searching. If was not for the script; to every question of yours which did not get attention being kept around, there would also be kept other hundreds of questions of poor quality or abandoned ... continues – Andre Silva Feb 15 at 19:39
  • @AndreSilva losing hundreds of questions would represent a significant time loss. Roomba ought to distinguish low quality questions deserving to be removed from decent/good questions with score 0. Especially on the stack exchange websites with low frequentation/voting activity where it's quite easy to post questions and not receive any vote. – Franck Dernoncourt Feb 15 at 19:43
  • One can always repost a deleted question with improvements, and which still continues relevant after a year (which worked for you). One thing we can agree is that if there was a criterium which consider user’s activity (for example, not logged in past 3 or 6 or “x” last months) maybe it could filter your cases; and still bring the benefit of cleaning abandoned questions. So that is, when reviewing borderline questions for the abandoned script I always start for questions where users have not logged in for a while; and they are the most common cases which apply under such conditions. – Andre Silva Feb 15 at 19:45
  • @AndreSilva "One can always repost a deleted question with improvements" -> I wish you were correct but the issue is questions get deleted silently, so OPs don't realize when their 0-score questions get removed. If Roomba was sending some kind of notifications when it deletes questions, then I'd be more ok with it (I still would prefer the 0-score questions not being removed in the first place though.). – Franck Dernoncourt Feb 15 at 20:33
-8

I often comment on questions asking for clarification or giving other feedback to the user on how to get a better answer. I would hate to think that I'm keeping bad questions alive while trying to be helpful.

On the other hand I see situations where the answer is in the comments and nobody ever pasted the answer as an actual answer.

In conclusion; I think that the questions that have comments that are flagged by your new rule as "abandoned" need to go in to a new "Review queues" section where people with review power can choose to close them or paste the answer in to the answer section and close out the question.

|improve this answer|||||
  • 6
    It would be too much work for little return in my opinion. Only a small portion of Qs are answered in comments and usually, such comments are one-liner and/or lack details to escape from downvotes and low quality flags. – Andre Silva Feb 4 '19 at 23:11
  • 6
    A queue like this would probably gain a significant amount of backlog. Answering in comments is already discouraged and doesn't happen enough for it to be worthy of another review queue. – Pika the Wizard of the Whales Feb 4 '19 at 23:22

You must log in to answer this question.

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