What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 131 Stack Exchange communities.

This conjoined twin of a feature request was surgically separated. The other portion of the request is now found at: Fine grained suspension

Please give the existing answers a bit of slack for addressing portions of this request that have been moved.

tl;dr version: Rather than locking people out of the house for increasing periods of time, wouldn't it be better if we had a robot that detected regular patterns of abuse and performed the banning automatically? This would be perceived as "fair", as opposed to some assertions that our current banning process is too opaque and subjective.

Problem

A significant problem that suspended users and their supporters complain of frequently is that the current ban system is not fair (being singled out, rules aren't stated, here are examples of others doing the same thing, etc)

Evidence

While I hate to open old wounds, and there is an unwritten social agreement to avoid mentioning specific people who have been banned (Primarily for privacy reasons), I feel it's important that we consider some concrete examples that did happen, rather than postulate based on what could or might happen.

Threatening emails from Jeff Atwood <-- User asked privately to change their comment posting behavior to avoid suspension.

http://stackoverflow.com/users/5640/geoffrey-chetwood <-- User suspended due to numerous complaints and flags against user's comments and other interactions on the site.

Numerous other users have been banned at various times for various reasons, including extraordinary users (one of the top superusers, for instance, and that is a site that could still use a little more help). Would these cases have been better if we could remove one or another ability, rather than removing them from the site completely?

Solution

Automate the suspension system based on flags and moderator attention

  • While users are told that they are being banned for specific violations, there appears to be a lot of wriggle room for the suspension trigger to occur. It would be better if, over time as problems are identified, the system is designed to detect inordinate numbers of flagging and moderator actions against a specific user for specific violations. Triggers should be set that provide the moderators/team with notices so that they can perform suspension based on a well known set of conditions. As long as the system is automated to at least this degree, then people cannot easily complain about how the system is unfairly targeting them. Further, the moderators then have clear guidelines as to what constitutes suspendable violations for some types of regular infractions.

This feature should also reduce moderator workload as no warnings would need to be given to users who receive so many flags so consistently that they lose the ability to comment, so the whole email communication aspect could be removed, and they could be encouraged instead to come here for support regarding semi-automated suspension. It may reduce the need for moderators to discuss specific cases and determine the appropriate action, and suspensions can be enacted immediately, rather than a day or more after the latest incident.

Discussion Points

  • Are there behavior patterns that we could detect that are currently not detected?
  • Is it reasonable to formalize this process so once a trigger occurs it is acted upon, rather than starting discussion, or deferring the decision to Stack Exchange, Inc?
  • Would this actually resolve the problem, or would banned users still be able to complain about the unfair nature of it due to some other link in the chain?
share|improve this question
3  
Man, I can't believe I spent 90 minutes making a question. Time to turn leechblock back on... –  Adam Davis Mar 3 '11 at 15:08
    
Automate the suspension system based on flags and moderator attention: are you aware that one of the example user you picked IS a moderator? :P –  Andreas Bonini Mar 3 '11 at 15:12
3  
oh, great - now that you've admitted to putting all that time into this, I'm tempted to just reply with some dismissive meta comment instead of actually bothering to read it... –  Shog9 Mar 3 '11 at 15:13
1  
@Shog, come on, we all know you'd do that regardless of how much time Polly put into the question. –  Pops Mar 3 '11 at 15:24
1  
That thread on math.se is hilarious, even my 1.5 years old doesn't act up like that... I think someone needs a hug –  jmfsg Mar 3 '11 at 15:27
2  
@Juan Let's avoid attacking them. That's one of the reasons I generally avoid bringing up examples - there's no need to rehash old issues, nevermind bring them here. –  Adam Davis Mar 3 '11 at 15:37
2  
You brought it up, and now I can say whatever I want... na na na na na na na na na na na na na na na na –  jmfsg Mar 3 '11 at 15:39

3 Answers 3

We don't want to suspend users. That's why Jeff sent that email, of course. The ideal scenario is that in any confrontation, we talk it over with the user and get the problem resolved without needing to resort to that kind of end result. Automated suspension, then, seems like a very terrible idea to introduce to the network.

Not to mention, it'd require a lot of development because any kind of approach towards suspension occurs from human analysis. Dismissals and flag actions don't actually tell the system anything about the user affected. Instead, you need a human to read the moderator flags (most of the seriously problematic stuff is going to be hand-typed, not the auto-messages), and probably review a series of moderator annotations (also hand-typed).

This also gives us better leeway to analyze why people are contacted. We can make notes on what kinds of comments were being troublesome, what kinds of voting patterns they did, and how we tracked or discovered it. And, importantly, by reaching out to the user with the evidence we have, we can try to prevent there being any suspension at all.


There are some kinds of fine-grained blocks already introduced, like the one that prevents users with a history of bad questions from posting more questions. I think that adding fine-grained levels adds a bit of extra work, but I don't think it is entirely meritless. Let me think some more on your discussion points. I do think, as well, that it does add some levels of complexity to really tracking things and knowing what is happening to everyone.

Is this worse than a full suspension (ie, being able to participate, but not fully, might hurt worse than simply being kicked out?)

Depends on the user. In the case of utilities like comments, which a user may use as often for helping explain answers as for whatever punishable offense, the removal of the utility may be so damaging that they might as well consider themselves completely blocked.

Is there validity in the idea that kicking them down the ladder would be better than just removing certain abilities? (for instance, if you can't comment properly, perhaps you shouldn't be allowed to do anything else above commenting that your reputation would otherwise allow)

There is some merit in thought that someone with extremely high privileges might, in retaliation to losing a lower privilege, go on like a crazy abuse spree. But I don't think it's enough merit to make this terribly valid.

Is it fair making a user's ability contingent on something other than reputation, such as how well they use that ability?

This is already the case with normal suspensions and the other blocks we have. But I imagine that users would be more pleased to not lose any reputation in the process.

Is it wise to remove their ability to defend themselves in comments? (Are comments a primary feature of the site?)

It would be exceptionally awkward. But this is generally why we do things like email for this. That way, we can't restrict their ability to defend themselves.

Are there abilities that should or should not be considered for this treatment? (Posting questions and answers is the primary purpose of the site - if we remove either of those, isn't that essentially the same as suspending them completely, even if it's only for a few days?)

As mentioned, posting questions already can be blocked. We offer lenience for the severity of their offenses via warnings. Once they get past that lenience, I'm not sure that certain privileges should be exempt.

Specifically regarding commenting, should comment-suspended users be allowed to comment on their own questions and answers, or if someone explicitly calls them out using the @username feature? (for instance in the case of someone commenting on how the suspended user edited a question)

I guess this scenario highlights in the end why I'm not really a fan of fine-grained privilege loss. People use all of their abilities, and losing a critical component like editing or commenting can be quite torture-some, especially if they're still trying to participate like normal. Making exceptions for them just opens up opportunities to evade the semi-suspension, ultimately. If you're only going half way, why only go half way to half way?

share|improve this answer
    
I purposefully left out the technical aspect of the question. Yes, a discussion would have to take place regarding cost/reward. However, I suspect that in some cases (flags on comments being the primary example) it wouldn't be overly difficult to detect and there exists examples that show commenting is being abused to a degree that may merit this attention. I generalized it without calling out corner cases where detection is hard, but there are some areas where automated detection would be workable, and areas where fine grained suspension would be good, even if not automated. –  Adam Davis Mar 3 '11 at 15:34
1  
@Pollyanna It's hard to introduce a purely mechanical system and ignore the actual technical aspects. Especially when you consider that suspension is currently not mechanical at all, other than the interface. Flags on comments don't give any reason, so all we'd know is that their comments are getting flagged. A reason for that is what is necessary. –  Grace Note Mar 3 '11 at 15:37
    
@Grace The problem with example two is not the reasons why they were flagged, it was that they were consistently and regularly flagged well above and beyond the average user. While knowing why they were flagged is important, one of the many reasons they were suspended was simply because they caused so many flags to occur. Quite frankly, I don't think a user can complain about having their commenting ability removed if they get 10-20 flags a day on their comments, regardless of the content. –  Adam Davis Mar 3 '11 at 15:40
    
@Pollyanna They can complain about it if we can't really explain why we figured the flags to be valid. Not every flag is valid, and behind a flag being valid is a reason for it to be valid. I think it's much nicer that if we have to contact a user, we can give them something substantial so that they know what they have to fix. –  Grace Note Mar 3 '11 at 15:41
2  
... wow, did you ever expand your answer while I was writing mine. –  Pops Mar 3 '11 at 16:03
    
@Grace I've separated the two requests, Would you please consider moving the parts of your answer that address fine grained suspension to the new request here: meta.stackexchange.com/questions/84655/fine-grained-suspension –  Adam Davis Mar 24 '11 at 20:26
    
@Adam You'll have to give me some time to slice this up. –  Grace Note Mar 24 '11 at 20:34

I support automation, but not automated suspension. One or more human moderators — sorry, Rovio — should always review every potential suspension. Suspension...

  • is infrequent, so manual review won't take up much mod time
  • is restricted, so it should have priority on the mod queue
  • is severe, so it shouldn't be left up to pure standard deviation counts, where the cutoff would be somewhat arbitrary anyways

This answer was edited after the question was split into two pieces.

share|improve this answer
    
I've separated the two requests, Would you please consider moving the parts of your answer that address fine grained suspension to the new request here: meta.stackexchange.com/questions/84655/fine-grained-suspension –  Adam Davis Mar 24 '11 at 20:26
    
Eh, I'm pretty much behind Jeff's and Rosinante's answers on the new post, so I'll just pare this answer down to its gist and call it a day. Thank you for the notification comment, though! –  Pops Mar 25 '11 at 15:26

Automate the suspension system based on flags and moderator attention

The single best use of suspension (IMHO) is the temporary removal of users who, for whatever reason, have taken a sudden nose-dive into negative / abusive behavior: started deleting/vandalizing all their posts, posting nonsense, stalking other users, etc. It gives them a chance to cool down and a chance for the community to repair the damage.

And it needs to happen quickly; waiting around for sufficient flags to trip an automated system doesn't work.

So that just leaves users with a constant, long-term pattern of flag-attracting posts/comments. Frankly, I don't care for the use of suspension here at all - we have plenty of ways for the community to remove content it doesn't care for, and moderators can and do obliterate accounts that have no redeeming value. I can appreciate that it might be necessary on a new, low-traffic SE site simply to keep the peace when there simply aren't enough active users to flag-delete offensive comments... But that also implies that there's not enough participation for this to be properly automated.

Worst of all, automated flag-based suspension promotes the idea that a vigilante mob running around removing the wrong sort of user is an acceptable part of the site. As distasteful as moderator-led suspensions might be in some cases, its understood that this is their job, one they were elected to perform because it is not something we want to put in the hands of the community at-large.

share|improve this answer
    
"automated flag-based suspension promotes the idea that a vigilante mob running around removing the wrong sort of user is an acceptable part of the site." Quoted for emphasis - this is probably the biggest reason to avoid fully automated suspension. Putting the moderators in the middle (ie, the system alerts the moderator that the trigger conditions have been met, and the user should be considered for suspension, but the moderator is still responsible for performing the suspension) would resolve it. –  Adam Davis Mar 3 '11 at 17:43
    
I do now agree with you that I should have split these suggestions into two separate feature requests. I'll consider splitting them up later when I have time to re-approach the issue. –  Adam Davis Mar 3 '11 at 17:45
    
I've separated the two requests, Would you please consider moving the parts of your answer that address fine grained suspension to the new request here: meta.stackexchange.com/questions/84655/fine-grained-suspension –  Adam Davis Mar 24 '11 at 20:27
    
@Adam: I've removed the now-irrelevant portions of my answer. Response to your new suggestion will have to wait... –  Shog9 Mar 25 '11 at 15:33

You must log in to answer this question.

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