10

I want to talk about comments today.

You may or may not be aware, but we get a few comments here. Even if you think we get a lot of comments, you may not be aware of how many... and I'm actually not, either, because the system has only a partial count of comments because, as we all know, comments are temporary and sort of second-class citizens in the SE world.

So, what I want to see is some data so that we can decide if we need to do something about the sheer volume of comments we get. I have been able to use a SEDE inquiry to get info about the number of comments that currently exist on each post... or at least existed the last time the SEDE query ran - Sunday the 27th - as of right now the site reports that we have 394 questions and 1822 answers but these numbers are a bit wonky. This has been updated since the first posting to reflect a more recent SEDE query.


The majority of answers have no comments - 931 vs 844 for a total of 1775 answers but they add up to 2863 total comments on answers. So, this represents about 1.6 comments per answer, though if you exclude answers with zero comments, you get a little over 3 comments per answer.

Most of the questions have comments, though - 45 vs 337 for a total of 482. This is probably expected. We want to help people improve their questions, which requires comments. In total, there were 2045 comments on questions the last time SEDE ran. This means that, on average questions get a little over 4 comments each or a little over six per question if you exclude those with zero.

While this doesn't sound too bad, there are 58 questions and 38 answers with ten or more comments each - up to 29 comments on a single post.

That's a total of 4908 comments on 2208 posts (2.22/post) since this site went into public beta. I don't know if this is completely out of line but it seems somewhat high. Computer Science Educators, a site that is at 95 days in beta has 3985 total comments on 1970 posts (2.02/post). Bioinformatics has 2158 comments on 1223 posts (1.76/post) after 105 days in beta.


Also remember what I said earlier - this data is incomplete. This represents only non-deleted comments. Any comment deleted by either the moderators or by the user who posted the comment or through flags aren't included in this count.

I can add a little bit of light to that. As a moderator, I can see a count of total moderator-deleted comments (includes CM deletions). That number is 1544 (as of right now, not when SEDE ran). You read that right. In two months, we have deleted over 1500 comments - or about 24.5 per day. Compare that to the total from the SEDE query, that's about 1/3 of the existing comments, so about one in four comments ever posted on this site have been deleted by a moderator.

A - smaller, I assume - number have been deleted by the person who wrote them, through abusive comment flags, or have a keyword that makes them auto-delete with one "no longer needed" flag. I don't have numbers for that.


So, what I'd like from a CM is the rest of the data I'm missing - or as much as the system is able to serve up.

  • total current comments on all posts (I understand it may not be possible to include deleted posts in this).
  • total current deleted comments
    • by type (self-deleted, mod-deleted, flag-deleted but no diamond involvement)

If it's possible, I'd also like to see the correlation between deleted comments and posts on the HNQ list. I know there's no "tag" for HNQ posts but I think we can safely say that if views > 1000 it was on the HNQ, since the site is new enough that it's uncommon for non-HNQ posts to get that many views. Giles made me a query for existing comments in SEDE here. It's pretty nifty but doesn't take into account deleted ones.

I'd also like to know how many auto comment flags have been raised by Community - these flags are raised on a post that gets more than 20 comments. We get a lot of them but they're not tracked. There's an MSE post that seems to indicate that you can give us that data, too.

Some thoughts regarding the comment volume would be useful, too.

| |
  • 2
    Do we really need all this data? It's pretty clear that comments are a problem here, in the same way that they are a problem on other sites. I'm not sure how more data about this site will increase our understanding of te problem. – user288 Aug 26 '17 at 15:23
  • @Hamlet It can be valuable to know whether the HNQ is a major source of the comments, for example... There are several solutions recommended on MSE and which one might be the better solution depends on where the comments are coming from and whether they're actually a problem. Our comments numbers aren't that far out of line from the other two recently-started beta sites but I can't tell how many have been deleted over there. – Catija Aug 26 '17 at 17:43
  • To be fair, a lot of the questions we get are lacking in relevant details. Often this is cultural and the OP doesn't realize initially what information might be relevant. I think there may be some skewing because there is so much potential for clarifying questions in comments. Sure, there are also a problematic number of "answers" in comments; but the nature of the questions can explain some of the discrepancies for comments on questions. – JMac Aug 28 '17 at 12:01
10

If I'm reading this correctly, your goal is to get something of a high-level picture of how comments are used (and abused) on this site.

The following data covers all comments posted on Interpersonal Skills for its entire history, unless otherwise noted. It includes deleted comments and comments on deleted posts; the latter are treated identically to comments on visible posts.

  • There've been a total of 8349 comments posted: 3872 on questions, 4477 on answers. The distributions look like this:

    distribution of comments by post type

  • There've been a total of 2373 comments deleted, 1412 on questions and 961 on answers. The distribution for those looks like this:

    distribution of deleted comments by post type

  • The reasons for deletion break down as follows:

    • 670 were deleted by their authors
    • 758 were deleted in response to a flag on the comment
    • 945 were deleted by moderators (possibly in response to a flag on the post but not when a flag was pending on the comment itself).
  • There've been a total of 881 deleted comments on questions with > 1000 views, 748 deleted comments on their answers, and the distribution looks like this:

    distribution of deleted comments on high-view posts

Oh... And there have been 46 "Too many comments" auto-flags raised during the life of the site. Since that's boring, here's a chart of those flags by week:

too many comments flags by week

| |
  • Thanks! Did you have any of the info on the number of automatic 20+ comments flags? – Catija Aug 28 '17 at 18:09
  • Thank you. A question I had is: are these data accessible only to mods? Or is it pulled from publicly available SEDE thingies? – NVZ Aug 28 '17 at 18:11
  • Added that info, @Catija. – Shog9 Aug 28 '17 at 22:15
  • None of this information is public, @NVZ; I created this report specifically for this question. – Shog9 Aug 28 '17 at 22:15
0

Some thoughts regarding the comment volume would be useful, too.

What I found unusual, coming from English.SE myself, is that as you observed in your question,

The majority of answers have no comments

This is unusual based on my experience at ELU, though I am aware that comments pattern can differ widely from site to site, and ofcourse IPS.SE is still in beta.

Based on your wide experience with multiple Stack Exchange sites, you might be much better placed to decide whether this is in any way unusual, @Catija.

Have other members noticed anything else significant about comments?

| |
  • 1
    Well, I think that ELU has more users who aren't afraid to call out a poor-quality answer as such. There's a sad reality here that (probably partially because we're still figuring things out) we are not asking people to support their answers with experience or sources. Many questions get dozens of answers and many of those answers are lacking and yet they are not flagged, or commented on or downvoted. Unsupported answers on ELU get comments and downvotes and flags. – Catija Aug 28 '17 at 19:00
  • That probably explains it, @Catija. However Interpersonal.SE has made rapid progress and this is the 'expansion stage' where volume of user participation needs to be encouraged. Ironically many high rep members (not me) who routinely feel the stress of ELU actually 'relax' by writing some good posts at IPS.SE! This is a very positive and supportive atmosphere and consistent quality control will surely follow. – English Student Aug 28 '17 at 19:38

You must log in to answer this question.

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