What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 149 Stack Exchange communities.

The exact rules for the automatic question and answer bans are not public, we know some general principles behind them, but not much more. I agree that some automated mechanisms like this are necessary for a site as large as Stackoverflow, but I'm wondering if the exact details really need to be secret for the bans to be effective.

The reason why they are secret is, as far as I know, to prevent users from gaming the bans. But I wonder, from what we know there isn't much a user can do except to provide some good content in the categorie he isn't banned or to improve the existing content enough to gather some upvotes. I might be wrong as I don't know the details, but I don't see how you could game the ban without accidentally posting good content, exactly what should be encouraged. Knowing the details could make it more effective to lift the ban, as you would know exactly which criteria you need to meet. But you still would have to improve your posts or add some new ones.

Is the secrecy necessary for the bans to be effective, or wouldn't they work as well as they do know even if all the details were known? While I think they are necessary, I'm a bit uncomfortable with such drastic measures that cannot be reviewed by the community due to the secrecy.

share|improve this question
I can already hear "Hm, so I need a sockpuppet to do X to free me of my ban...." –  Ben Brocka Jul 21 '12 at 18:53
And, FWIW, we have pretty good details on how they work, just not the exact specifics how how many downvotes, how many deleted questions etc. I think the important details are already publish, so focusing on the exact numbers is the wrong idea. –  Ben Brocka Jul 21 '12 at 18:55
On the other hand, what would be gained by having it fully public? –  Bart Jul 21 '12 at 18:55
I'm actually wondering: How many of these "unwanted" users even care enough to try to game the bans? –  Mysticial Jul 21 '12 at 18:55
@Ben We have tools to detect sock puppets, and if you already went to the trouble of creating a sock it doesn't really matter if you know exactly how much you need to vote, or if you just randomly vote on your posts until you see the ban lifted. –  Mad Scientist Jul 21 '12 at 18:56
Something to keep in mind: the entire purpose of the ban is to handle cases where moderators aren't able to step in. The more that mods need to step in to handle issues surrounding q-bans, the less useful even having that feature becomes. –  Shog9 Jul 21 '12 at 19:08
@Bart Transparency, the ability to review the criteria, argue about them or maybe even improve them. –  Mad Scientist Jul 21 '12 at 19:14
Fair enough. Though we have a reasonable idea of what contributes to a ban and I have yet to see a ban that was completely unjustified. I would rather see a preemptive warning for someone getting close to a ban. –  Bart Jul 21 '12 at 19:33
@MadScientist I also think it's not up to the general public to discuss what constitutes a ban or lifting it. The mods know the rules, have full visibility of all the folks that have been banned (including those that have improved), and I have faith in their ability to adjust them as they see fit. –  Aaron Bertrand Jul 21 '12 at 19:49
@AaronBertrand: This is about the automated question and answer bans, not moderator-dispensed suspensions. The elected moderators do not know the exact rules and do not have any direct power over them; the implementation of the ban is developer access only. –  Josh Caswell Jul 21 '12 at 20:11
@Josh ok, so change my statement from "mods" to "developers"... I still don't feel ordinary users should be able to get into a democratic decision about what those automated rules should be. The whole point is to reduce the human effort that goes into them. –  Aaron Bertrand Jul 21 '12 at 20:20
I think there is quite a bit of value to be had in public discussion, @Aaron, but documenting the exact rules makes it pointless. Most of the important factors are documented (not the thresholds for them though) - if someone complains of being blocked and an examination of their profile suggests that they shouldn't be, that's definitely worth talking about. As you note, the idea is to reduce effort for folks who aren't willing to respond to feedback, or put any effort into correction. –  Shog9 Jul 21 '12 at 21:40
@Shog9 Determining if a user shouldn't be blocked is pretty much impossible for anyone but SO mods as deleted content plays a very important role for the ban. A normal user examining a profile won't be able to tell anything as they don't see the deleted content. –  Mad Scientist Jul 21 '12 at 21:44
@Mad: there are definitely situations where you can't know for sure (and currently there are situations where no one can know for sure short of checking the logs), but as a sanity-check looking over the posts that aren't deleted is a pretty good start: it's rare to see someone q-banned who doesn't have at least a few (and usually quite a few) poor-quality posts publicly visible. –  Shog9 Jul 22 '12 at 17:20

3 Answers 3

I don't think secret is a problem - it makes it an order of magnitude harder to game the system and the exact rules are largely a distraction from the real problem - bad questions/answers.

That doesn't have to be mutually exclusive to warning user before they get banned. For instance a simple thing that would be possible would be for Community ♦ to explicitly warn users that question bans are possible and their questions are unacceptable when it raises the automatic consecutive closed questions flag on users which are not currently banned.

A similar thing could happen for consecutive deleted answer (currently no flag exists for that, but it wouldn't be hard to add one too).

The problem at the moment seems to be that everyone who gets banned is surprised when they get it. They didn't know it existed, they (usually) seem to fail to realise their questions aren't being well received. More feedback along the way seems to be the solution, not showing everyone/anyone some super but largely arbitrary secret score.

I say focus on the bigger issue - giving users better feedback before they get to the point of a ban.

share|improve this answer
I have a feeling that the answer to your last sentence can be summed by this answer: meta.stackexchange.com/a/113179/169611 (Not saying that I agree with it though...) –  Mysticial Jul 21 '12 at 19:03

Actually, there are two methods of "gaming" that I've already observed, even without published rules:

  • Sockpuppet up-voting

  • Repeatedly creating new accounts (in some cases with proxies to avoid IP blocks / limits / detection).

These are fairly rare, as they take a fair bit of effort to implement and the folks being blocked as a rule tend to not like putting much effort into... anything. But they do happen.

There are potentially much easier ways of avoiding the bans without actually producing something of value though. I hesitate to even refer to these as "gaming" - they're closer to the effects you might see if we published the exact algorithm used for the quality block - folks doing just enough to work around the specific checks, without actually putting serious effort into improving their posts overall.

Combating this would likely involve making the checks more strict... Which would also increase the false-positive rate.

share|improve this answer
Not sure what you mean by "false positives". Do you mean people who (by some subjective criteria) didn't deserve to be banned? –  Asad Jun 5 '13 at 19:49
Yes. For instance, at one time it was possible to ask several bad questions without hitting the ban, and then end up banned by deleting them. You weren't making the site worse by deleting bad questions - that would be a false-positive. OTOH, routinely deleting questions as soon as an answer comes in (so you get help but no one else benefits) is classic help-vampire behavior - that should result in a ban. –  Shog9 Jun 5 '13 at 20:23
But can't that problem be addressed by working a distinction between answered questions and bad, unanswerable questions into the algorithm? –  Asad Jun 5 '13 at 20:25
It can be, and it was. Other tweaks have been made over time to handle similar problems. My point is, we don't want to tell someone, "you can only disrespect the folks helping you 9 times here - do it 10, and we'll ban you". Because then they'll do it 9 times. We want to tell them "don't disrespect others. Ever." - and if they go out of their way to do it anyway, then block them. –  Shog9 Jun 5 '13 at 20:31

Psychologically, unpredictable rewards are more motivating than predictable ones. A user who receives a ban and doesn't know exactly how to fix it will put much more effort into recovering an account than a user who knows all he or she has to do is have x amount of upvotes on x number of questions. I doubt the algorithm for deciding when to lift a ban is variable, but keeping it a secret is much more motivating than providing an obvious solution.

The best solution to helping new users recover from bans is to provide constructive criticism and resources they can use to improve their posts. It is important to impress the need for patience, hard work, and careful study when posting questions and answers on any SE site.

share|improve this answer
Unpredictable rewards, or unpredictable punishments? Seems to me there's some of both here. –  Nathan Tuggy 6 hours ago
That's true, but I'd argue that the uncertainty of receiving a ban is also helpful in keeping people on their toes. I've actually been banned on a site so I have experience. –  brandaemon 4 hours ago

You must log in to answer this question.

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