47

There's been problems with abusive comments since the site started. What's with all the rude comments recently?

We still haven't managed to conquer the issue. Negative comments drive new users away, how can we fix this?

Users can be issued chat bans and post bans. The API automatically bans IP addresses with validated spam flags. (Would it be difficult to implement comment bans? Would mods have the power to impose comment restrictions?).

Although comments have been classically referred to as second class citizens, rude comments on the site are the one place that is driving people away.

Can we implement temporary automatic comment ban on 'x' number of validated rude comment flags?
It would probably be better if the exact heuristics were not disclosed publicly, much like the serial voting scripts

This question Can we have a "Sorry, we are no longer accepting comments from this account"? was asked 7 years ago. Can we revisit it?

This question is related, but is asking for a more complex approach:
Implement a Selective Comment Ban

44
+800

This suggestion is coming up with increasing frequency in various forms; one of my colleagues expressed amazement this morning that we don't already do this - he'd spent the morning cleaning up after an abusive comment-author, only to find this person had kept right on leaving abusive comments the whole while.

Journeyman and Catija have suggested a couple of ways in which this might be problematic, but in truth we need not speculate: such an automatic-ban system already exists in Chat!

...this system is not exactly well-loved. While I don't think it's as bad as folks make it out to be, it's hard to ignore the fact that it often has the opposite effect from what is desired: abuse-banned folks often come back angrier and more abusive than when they left.

Of course, comments aren't supposed to be chat. But that raises an important question...

Why aren't we rate-limiting comments based on how they're used?

If the intent of comments isn't to chat up other users all day long, then why does the system allow them to be used this way?

  • Some users are in the habit of leaving dozens, even hundreds of comments on dozens or even hundreds of posts... Every day.
  • Some users are in the habit of conducting long, detailed back and forth conversations in comments under their posts.
  • Some users regularly leave multiple comments back-to-back under the same post.
  • Some users do all of the above even when 50-75% or more of their comments are deleted on a regular basis.

...and the system does very little to discourage any of this. There's a tiny bit of rate-limiting in terms of how fast you can post comments, but nothing stopping you from posting just a bit slower than that all day long, every day, on as many posts as you wish to.

For a moderator, the only options for handling persistent comment abuse are:

  1. Suspend the user from the site
  2. Do nothing.
  3. Laugh madly while the world burns (functionally equivalent to #2)

So from a user's perspective, they get absolutely no pushback at all until suddenly all of their privileges are gone... And on larger sites where comments can readily go unnoticed, this may take months or even years.

In essence, the system and UI facilitate - even encourage - abuse, even while flaggers and moderators are expected to stop it.

Proposed rate-limits

Instead of automatic bans, I'd suggest something like this:

  1. Limit everyone to some number of comments per day and some smaller number of posts on which they can leave comments. Let's say... 100 comments / day and 50 commented posts / day by default.

  2. Reduce both of these limits severely if the % of deleted comments for a given user exceeds some nominal threshold over the past week. For the sake of discussion, let's say anything over 20% gets you knocked back to 20/10, and anything over 50% puts you at 2/1. Yeah. If more'n half of your comments get deleted, you only get to post two comments on at most one post per day.

  3. For the purpose of these tests, ignore auto-generated comments (duplicates, canned off-topic reasons, review deletion comments) - those are supposed to be deleted in many cases.

  4. Show the number of remaining comments below the comment form whenever a user starts to compose a comment.

My prediction is that this system would dramatically reduce the number of noisy, snarky, rude comments even for people who never hit the limit, simply by making it obvious that they're using up a limited resource.

But for those determined to push forward, it would add real teeth to flaggers' and moderators' cleanup efforts by drastically reducing authors' ability to generate huge numbers of worthless comments.

  • 3
    100/50 seems like quite a low number. This is far below the top undeleted amount of 564/348 and there have been 4,214 times where the number of undeleted comments would have broken this limit. Is there any particular reason why this limit? – ben is uǝq backwards Apr 23 '18 at 18:00
  • 12
    It'd be worth considering that if a user removes their own comment (a behavior we want to see), this should not count against them on the path to rate limiting and, (assuming the comment were posted that day) might actually give them back an additional comment to use. Neither of these would be done if the comments were removed by moderators or the other users. – Catija Apr 23 '18 at 18:04
  • 3
    Yeah, that's worth a look @Catija; also, I'd want to simulate this with past data to establish less arbitrary thresholds should we ever decide to implement it. – Shog9 Apr 23 '18 at 18:10
  • 3
    @benisuǝqbackwards 100/50 seems low for a day? I think it will probably depend greatly on the site. And, perhaps, it can be like flagging privileges... users who show they know how to use comments well, can be given more/day to use. Those who fail, lose them. I have many users on IPS who only have ~300 comments total but close to half of their comments are deleted, so 100 in a day would be quite more than needed for us. – Catija Apr 23 '18 at 18:10
  • 5
    100/50 per day is incredibly rare, @ben. – Shog9 Apr 23 '18 at 18:11
  • 7
    IPS has users with over 80% of their comments deleted in recent days, @Catija. IPS is becoming the "pathological example" in comment discussions. – Shog9 Apr 23 '18 at 18:13
  • 4
    "Some users do all of the above even when 50-75% or more of their comments are deleted on a regular basis." You aren't actually told when your comments are deleted, are you? So people probably just never look back and never realize... – Southpaw Hare Apr 23 '18 at 18:47
  • 5
    @SouthpawHare Oh, sure... some don't... And some spend large amounts of time on Meta complaining about it. Either way, though... they should be aware that the comments are problematic. I am somewhat concerned about users who actually do post helpful comments that are deleted because the post has been edited. These are definitely in the minority but I'd want to be certain that these very helpful users weren't penalized in the same bucket as users who are actually abusing comments. – Catija Apr 23 '18 at 18:49
  • 1
    Related meta.stackexchange.com/questions/35501/… – Nobody Apr 23 '18 at 19:13
  • 3
    Two very major things missing here: what about self deleted comments? And as other comment mentioned, if such thing takes place, we must let users see their own deleted comments, same way we can see our deleted posts, so we can know what we did wrong. Sometimes comments we posted years ago are flagged and deleted, we can't be expected to remember what we wrote back then. – Shadow The Princess Wizard Apr 23 '18 at 19:19
  • 5
    I think you're probably overestimating both how much you comment and how many of those comments are deleted (for any reason), @ShadowWizard. That said, I think allowing for self-deleted comments is reasonable for the simple reason that it encourages a useful behavior. – Shog9 Apr 23 '18 at 20:03
  • 4
    yeah... I think it's probably worth considering comments that everyone can make (own post, answers to own post, automatic dup-comments, etc) and comments that only privileged users can make as two mostly-different categories, with vastly different motivations and abuse scenarios. – Shog9 Apr 23 '18 at 21:24
  • 2
    What @ShadowWizard is talking about was actually one of the first things I came to MSE to look for a solution to... and I was pretty much told that the comment being gone was all that mattered, not my being abused. – Catija Apr 24 '18 at 3:50
  • 3
    Numbers above are fairly arbitrary, @Lilienthal (should be in the right ballpark, but based on a spot-check not detailed analysis). That said, a great many limits here are set higher than the vast, vast majority of users will ever hit: the primary purpose is to stop unusual forms of abuse, not cause daily hardship. There is a potential secondary benefit however... Simply setting a visible limit on something can change how folks perceive it. Right now, comments are often seen as essentially free - but this hides the cost to others of gobs of comments clogging up a question or answer. – Shog9 Apr 25 '18 at 18:04
  • 2
    I did, @Fabby. It's an important badge and an important behavior; I think we should all try to earn it. – Shog9 May 10 '18 at 0:53
14

This doesn't sit well with me for reasons I outline in a similar post on Meta Stack Overflow.

That said, I strongly reject the idea of an automated warning being sent to a user for this. If the likelihood of a comment being incorrectly validated as rude/abusive was low - or if there were a way to change the flag validation after the fact, it'd be one thing but the default moderator UI - particularly when viewing comments on posts - makes it very easy to accidentally validate a rude/abusive comment flag just by deleting the comment - if they're viewing the post directly, the moderator may not know the comment was flagged at all!

This is about a warning and that's not even nearly as problematic as an actual ban. The automatic ban, as I see it, has several problems:

  1. It's really easy for moderators to inadvertently validate a rude or abusive comment flag on a comment that they don't think is rude or abusive and there's no way to rectify the situation after the flag is handled.
    Without any sort of improvement to the moderator UI to emphasize the fact that a comment is flagged as rude or abusive, the chance for error is very high. This is particularly problematic on comments that have both rude or abusive flags and no longer needed flags. You'd have to decline all of the flags, since you can't pick which flags to decline on comments.
  2. Three users can validate rude or abusive flags on comments if the comment has no upvotes.
    I'm not a huge fan of using "users could abuse it" as a reason to not do something in general but (right now) we have zero information as moderators that comments were deleted by the community unless we happen to be reviewing deleted comments on a post and notice it and, in the case of rude or abusive comment flags, we get an automatic flag letting us know that three comments were deleted as rude or abusive in seven days.
    At the very least, such an implementation would require this to trigger the auto flag first and consider whether the moderators have acted on it or not (with a mod message or suspension)...

And at this point, I have to wonder... why should this be automatic if it's easily accidentally triggered and we're already telling the moderators "hey, this may be a problem, you should look at it"... just let the moderators take away the comment everywhere privilege themselves for X days instead of a full suspension...

  1. It shouldn't be a comment ban... it should be removal of the Comment Everywhere privilege.
    This may sound like splitting hairs but if we're not taking away their ability to post, they need to be able to comment on their own posts or on answers to their questions. If they can't even be civil there, they should be getting a suspension, not a comment timeout.

Plus, if they're going to make it possible to remove the comment everywhere privilege... it should be something that a moderator can do and it should be possible to see all rude or abusive flagged comments in one page. Right now, you can see all flagged comments but can not sort by flag type.

I'd much rather have a comment suspension that requires moderators to explain to a user why they've lost the comment everywhere privilege along with removing it. This should include a link to the user's flagged and deleted comments since the users can't generally see their deleted comments. This allows moderators to prevent comments from both users who are consistently rude in comments (other than outright suspending them) and users who are just constantly using comments inappropriately, which is a problem I have on one of the sites I moderate.

If this were to be implemented, a lot of other changes would need to be made to reduce the accidental triggering and allow moderators to undo the ban if it were triggered accidentally.

  • ... that said, the answer from Tim Post here seems to imply that they'd rather have an automatic ban than a manual one... but I don't really understand the reasoning there. – Catija Apr 23 '18 at 1:07
  • 2
    Honestly, this reads like a post more concerned with moderator error, than addressing the issue. Mods can make errors on anything and to persistently validate rude flags incorrectly to a point to trip such a bam would mean there's a problem with that moderator's ability at that point in time. A comment suspension is different to the comment everywhere privilege. There's no reason to allow the person to comment under their own posts if they don't know how to behave under any post. Currently the community is not gathering and abusing the comment flag system,.../ – Nobody Apr 23 '18 at 5:22
  • ... as with all else, abuse of the system needs to be addressed. As it stands a certain number of rude flags will automatically raise a mod flag, if the flags are being incorrectly validated a mod has an opportunity to review this. There doesn't need to be a custom mod message for every ban, there isn't one for chat. – Nobody Apr 23 '18 at 5:23
  • 9
    It seems like y'all don't actually disagree that much; if the UI were cleared up and there were enough transparency and the thresholds were high enough to avoid accidents, the concerns here would be addressed? It's not that the fundamental idea of automatic bans is bad? – Cascabel Apr 23 '18 at 5:30
  • 1
    If a moderator can accidentally validate enough RUDABUS flags to trigger a threshold, that's a problem with the moderation system or the individual usage. We're not talking three-and-out, but surely something more like "greater than X per month or greater than Y in the last Z months, whichever is the lower limit" with XYZ all being chosen to avoid this exact problem. – Nij Apr 23 '18 at 6:06
  • 1
    @Nij Yes, it's a problem with the UI/system. It still means significant false positives. The common scenario is, people argue, both sides get flagged by the other side as some mix of rude and no longer need, mods clean up the whole thing, flags get validated. So the signal to noise ratio could be really really bad here, with really active users not necessarily looking significantly better than the problem users. Again doesn't mean the fundamental idea is bad, just something to address beforehand. – Cascabel Apr 23 '18 at 14:40
9

I'm against it being purely automatic but I think a good many of the reasons Yvette has brought up is valid. I'd rather this work like a regular suspension (with mods involved) than a 'silent' question or answer ban since in many cases human intervention might be needed. In a sense, comments are very much a people problem.

Although comments have been classically referred to as second class citizens, rude comments on the site are the one place that is driving people away.

This is kind of true. Between the fact that the only option we have is to mod message and/or suspend and that there's no real downside until some moderator goes "Man, this guy has gone too far", I think we're failing here. Policing comments is also pretty darned hard, especially since there's so many of them. Especially where users change their names often, and with the lack of easy visibility of deleted posts, I for one can't keep track of them all. An automatic comment suspension would probably take care of many of them, but I suspect what I'd really want to see are broader trends.

Can we implement temporary automatic comment ban on 'x' number of validated rude comment flags?

Well, this can be an effect. That said, being automatic temporarily puts off the problem, not solves it. I'm not too sure a constantly argumentative user will stop commenting in such a manner. If automatic - suspensions likely need to be along a sliding scale of time.

Also, sometimes you might want to get someone to stop digging themselves into a hole too quickly and if its automatic, clearly the hole is large enough.

If a user is consistently getting RA flags warranted or otherwise, there's probably issues with interactions between people. A comment ban for R/A comments should also have some mechanism to inform mods for follow up.

Much like chat suspensions though, an additional option should be the option to remove the chat ban if a more personal approach is felt to work.

And well, some folks just need to stop commenting, and if slapping them with a suspension helps, that's great.

I'd also add that, well its not just new users. There's a few users who've been around my site a while, who get in trouble in comments. They're too easily baited and being 'encouraged' to step out of an argument would be nice.

So, I like the general idea. I just see that its part of a broader solution.

  • As with any bad behaviour - the automatic suspensions could be annotated into the user's history, if the person keeps offending, then account suspensions can be implemented. Currently there's next to no consequences for rude comments - unless it's a blown up fight - which has allowed users to be rude over and over under other people's posts without pause. A comment suspension provides a pause. AS to whether people learn from suspensions, that's up to each individual and we don't have a lot of control over that. – Nobody Apr 23 '18 at 5:50
  • er ^ pretty much those things - other than that "We don't have a lot of control over that". I'm an optimist and I figure we can at least win a few over ;p – Journeyman Geek Apr 23 '18 at 8:04
  • Annotations aren't necessarily seen. To get mods to validate, whether it's "this was unjustified" or "this should've been more severe", we need autoflags or similar. And if we're going to validate anyway, seems nice to validate before the ban. I think we all agree there should be more consequences for rude comments, no one is questioning that. This is just about exactly what consequences to implement and how. – Cascabel Apr 23 '18 at 14:46

You must log in to answer this question.

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