5 ms·

You helped creating a bot that punishes users for posting certain words. You used those same words in your message and got a treatment not unlike the one you we

You helped creating a bot that punishes users for posting certain words. You used those same words in your message and got a treatment not unlike the one you were helping to automate. Maybe there is a lesson here somewhere.

freemint 4y ago
Maybe even exactly the treatment. It could be the reddit runs automoderator on DMs to flag people. It would be ironic if by fixing an automoderator config and chatting about that in Reddit DMs this code gets used on oneself.
jasonhansel 4y ago
Looks like OP ran into a version of https://en.wikipedia.org/wiki/Scunthorpe_problem https://en.wikipedia.org/wiki/Scunthorpe_problem
- FaviFake 4y ago
  That was a fun rabbit hole! Thank you!
nousermane 4y ago
Keyword-based moderation has some pretty bad failure modes, yeah. False-positives from context, false-positives from different meaning of the same word... But what alternatives are here? "Report" button, and human reviewing each and every instance of it? We've all heard horror stories about fb moderators burning out and getting PTSD from this kind of work.
- Levitz 4y ago
  I'd say in a site like Reddit where every comment is voted on to change its visibility to begin with, there is no need whatsoever for such a tool. The amount of overmoderation on the site is ridiculous. You have the site admins, then the subreddit mods, then the moderation by the users themselves, then the automatic stuff. You can still block users and filter subjects.
  - canistista 4y ago
perihelions 4y ago
This is uncharitable; one merely filters isolated comments, the other temporarily* bans humans from an entire social media site. If you click through to their offending thread, the AutoMod profile they built has a field value of "action: report". As I understand it, that means it does nothing but forward the flagged comment to a human moderator for review. As benign and judicious a use of automation as you can get. *edit: s/permanently/temporarily/ (sorry!)
- djbebs 4y ago
  They are both censoring
- rossvor 4y ago
  >This is uncharitable: one merely filters isolated comments, the other permanently bans humans from an entire social media site. For three days, not permanently: >"You’ve been banned from Reddit for three days"
- soneil 4y ago
  I'm assuming they were showing an example of a regex that will match a list of terms, and the sheer quantity of terms (rather than the initial presence of) weighted its score heavily. It's a little unfortunate, but 3 days isn't terrible, and it's a good lesson in false positives.
FaviFake 4y ago
> You helped creating a bot that punishes users for posting certain words No, I did not. The bot would not punish the user in any way, it would simply hide the comment from other users without even informing the OP. I'm not sure if you're understanding the situation, the bot would remove comments containing the n-word, r-word and other very insulting slurs
- blueflow 4y ago
- 1983054104 4y ago
  It's either a side effect and you accept it peacefully, or you're in denial and don't want to accept that the automatic moderation on Reddit is awful. Which one do you choose?
  - FaviFake 4y ago
    This wasn't caused by the automatic moderation of Reddit, the official Reddit Help Center says that "site-wide suspensions can only be applied to accounts by employees of Reddit and are done so after review of the actions and the context in which it took place". Maybe a bot flagged my comment, but a human had to review it.
    - ericfrazier 4y ago
      Totally, because every company policy is true.
- TMWNN 4y ago
  >No, I did not. The bot would not punish the user in any way, it would simply hide the comment from other users without even informing the OP. This is shadowbanning, and is a very bad feature of Reddit. If someone's comment or post is removed, that person ought to a) be notified or, at least, b) get the chance to see for himself that it has been removed. Fooling the poster into thinking that all is well is terrible from both UI/UX and general human being perspectives. >I'm not sure if you're understanding the situation, the bot would remove comments containing the n-word, r-word and other very insulting slurs Thinking that shadowbanning is a good thing is really retarded.
jareklupinski 4y ago
"Don't build systems you wouldn't want to be a victim of." https://en.wikipedia.org/wiki/Brazen_bull https://en.wikipedia.org/wiki/Brazen_bull
Ekaros 4y ago
Sounds like system is working as designed.

freemint 4y ago

Maybe even exactly the treatment. It could be the reddit runs automoderator on DMs to flag people. It would be ironic if by fixing an automoderator config and chatting about that in Reddit DMs this code gets used on oneself.

jasonhansel 4y ago

Looks like OP ran into a version of https://en.wikipedia.org/wiki/Scunthorpe_problem https://en.wikipedia.org/wiki/Scunthorpe_problem

FaviFake 4y ago
That was a fun rabbit hole! Thank you!

nousermane 4y ago

Keyword-based moderation has some pretty bad failure modes, yeah. False-positives from context, false-positives from different meaning of the same word... But what alternatives are here? "Report" button, and human reviewing each and every instance of it? We've all heard horror stories about fb moderators burning out and getting PTSD from this kind of work.

Levitz 4y ago
I'd say in a site like Reddit where every comment is voted on to change its visibility to begin with, there is no need whatsoever for such a tool. The amount of overmoderation on the site is ridiculous. You have the site admins, then the subreddit mods, then the moderation by the users themselves, then the automatic stuff. You can still block users and filter subjects.
- canistista 4y ago

perihelions 4y ago

This is uncharitable; one merely filters isolated comments, the other temporarily* bans humans from an entire social media site. If you click through to their offending thread, the AutoMod profile they built has a field value of "action: report". As I understand it, that means it does nothing but forward the flagged comment to a human moderator for review. As benign and judicious a use of automation as you can get. *edit: s/permanently/temporarily/ (sorry!)

djbebs 4y ago
They are both censoring
rossvor 4y ago
>This is uncharitable: one merely filters isolated comments, the other permanently bans humans from an entire social media site. For three days, not permanently: >"You’ve been banned from Reddit for three days"
soneil 4y ago
I'm assuming they were showing an example of a regex that will match a list of terms, and the sheer quantity of terms (rather than the initial presence of) weighted its score heavily. It's a little unfortunate, but 3 days isn't terrible, and it's a good lesson in false positives.

FaviFake 4y ago

> You helped creating a bot that punishes users for posting certain words No, I did not. The bot would not punish the user in any way, it would simply hide the comment from other users without even informing the OP. I'm not sure if you're understanding the situation, the bot would remove comments containing the n-word, r-word and other very insulting slurs

blueflow 4y ago
1983054104 4y ago
It's either a side effect and you accept it peacefully, or you're in denial and don't want to accept that the automatic moderation on Reddit is awful. Which one do you choose?
- FaviFake 4y ago
  This wasn't caused by the automatic moderation of Reddit, the official Reddit Help Center says that "site-wide suspensions can only be applied to accounts by employees of Reddit and are done so after review of the actions and the context in which it took place". Maybe a bot flagged my comment, but a human had to review it.
  - ericfrazier 4y ago
    Totally, because every company policy is true.
TMWNN 4y ago
>No, I did not. The bot would not punish the user in any way, it would simply hide the comment from other users without even informing the OP. This is shadowbanning, and is a very bad feature of Reddit. If someone's comment or post is removed, that person ought to a) be notified or, at least, b) get the chance to see for himself that it has been removed. Fooling the poster into thinking that all is well is terrible from both UI/UX and general human being perspectives. >I'm not sure if you're understanding the situation, the bot would remove comments containing the n-word, r-word and other very insulting slurs Thinking that shadowbanning is a good thing is really retarded.

jareklupinski 4y ago

"Don't build systems you wouldn't want to be a victim of." https://en.wikipedia.org/wiki/Brazen_bull https://en.wikipedia.org/wiki/Brazen_bull

Ekaros 4y ago

Sounds like system is working as designed.