4 ms·

This precisely points out the problem. Moderation is human, algorithms can't do it and humans are inconsistent. I've been a moderator before, within a team of

This precisely points out the problem. Moderation is human, algorithms can't do it and humans are inconsistent. I've been a moderator before, within a team of moderators. There were varying views within that group from 'ban the slightest aggression or just disliked person' to 'give people as many chances as can be afforded, ban is last resort (my view).' Members got 'banned for nothing!' or regarded as 'why is that person still here?' Depending who you asked both statements were wrong... It just doesn't scale massively without something like Reddit's subsite moderation structure. FB, Twitter, etc all can't do it at scale across one public community and algorithms can never be nuanced enough and are extremely vulnerable to political/ideological bias ('misinformation' labels are a simple example of this, regardless of what you think of their accuracy or usefulness).

IgorPartola 5y ago
Something I have learned from moderating: you can’t just moderate statelessly or with localized state. Moderation is not just about deleting someone’s offensive post or banning a troll. It’s also about setting community norms by performing moderation as a public act. It’s about acknowledging when moderation happens. It’s why sometimes you give someone multiple chances: you hope they get better but also if not at least they’ll dig their own reputation a deeper grave. And sometimes you ban at the first sign of trouble. It’s quite possible that an AI could learn to do all this but I suspect that a big part of modernism that participants in a community know and respect the moderators. Not many humans respect AI in the same way.
- headsoup 5y ago
  Yes, not to mention the 'behind the scenes' where the moderators discuss disagreements in approach, what actions to take in some situations, need on occasion for PSA's to reiterate the guidelines or explain recent actions, etc. An algorithm can't ask itself for feedback. I hope Musk does make the Twitter algorithms open source, that would be fascinating and put pressure on other platforms if it turns out to be eye opening or useful
  - IgorPartola 5y ago
    To be fair, isn’t the whole GAN methodology all about asking itself for feedback? But I think a more refined version of my initial comment is that if the users respect moderators, then moderation will be much more cohesive and well received.
- nradov 5y ago
  There are no community norms at the scale of Twitter. It's helpful to have community norms for a small forum to discuss model trains or mountain biking or whatever. But Twitter users are a cross section of the whole world. There's nothing to unify them.
  - IgorPartola 5y ago
    Yup, that’s my point. This is why dang doesn’t scale. Twitter clearly has utility. But it will never be a nice place to hang out except in some very localized parts of it.
  - cbozeman 4y ago
    The norms are, and should be, what the United States considers norms. It was built in America, by Americans, and now, is about to be owned by an American. And like America, everyone is (should be) welcome... provided you honor our Western traditions of free speech. I like what Musk said about free speech within the limits of the law. I think that's a good place to settle things.
    - IgorPartola 4y ago
      Does the US have a uniform set of norms? Last I checked it was pretty divided.
      - nradov 4y ago
        The US has a uniform set of federal laws and Supreme Court opinions, which is what matters here.
        IgorPartola 4y ago
        Wow that is pretty naive. Let’s see: A lot of stalking and harassment laws are state level and not the same in every state. Since online harassment is a common problem on Twitter, which state’s laws would you like to apply? Spreading misinformation that results in violence towards an individual is not governed by federal laws unless it falls under the definition of a hate crime. Spreading misinformation about results of elections isn’t illegal but is very obviously quite damaging and has real world consequences. How do you tackle this issue? That’s just a very very small sample. The problem you are going to face with your approach is that the US laws were not designed with the idea of governing stuff on Twitter.

IgorPartola 5y ago

Something I have learned from moderating: you can’t just moderate statelessly or with localized state. Moderation is not just about deleting someone’s offensive post or banning a troll. It’s also about setting community norms by performing moderation as a public act. It’s about acknowledging when moderation happens. It’s why sometimes you give someone multiple chances: you hope they get better but also if not at least they’ll dig their own reputation a deeper grave. And sometimes you ban at the first sign of trouble. It’s quite possible that an AI could learn to do all this but I suspect that a big part of modernism that participants in a community know and respect the moderators. Not many humans respect AI in the same way.

headsoup 5y ago
Yes, not to mention the 'behind the scenes' where the moderators discuss disagreements in approach, what actions to take in some situations, need on occasion for PSA's to reiterate the guidelines or explain recent actions, etc. An algorithm can't ask itself for feedback. I hope Musk does make the Twitter algorithms open source, that would be fascinating and put pressure on other platforms if it turns out to be eye opening or useful
- IgorPartola 5y ago
  To be fair, isn’t the whole GAN methodology all about asking itself for feedback? But I think a more refined version of my initial comment is that if the users respect moderators, then moderation will be much more cohesive and well received.
nradov 5y ago
There are no community norms at the scale of Twitter. It's helpful to have community norms for a small forum to discuss model trains or mountain biking or whatever. But Twitter users are a cross section of the whole world. There's nothing to unify them.
- IgorPartola 5y ago
  Yup, that’s my point. This is why dang doesn’t scale. Twitter clearly has utility. But it will never be a nice place to hang out except in some very localized parts of it.
- cbozeman 4y ago
  The norms are, and should be, what the United States considers norms. It was built in America, by Americans, and now, is about to be owned by an American. And like America, everyone is (should be) welcome... provided you honor our Western traditions of free speech. I like what Musk said about free speech within the limits of the law. I think that's a good place to settle things.
  - IgorPartola 4y ago
    Does the US have a uniform set of norms? Last I checked it was pretty divided.
    - nradov 4y ago
      The US has a uniform set of federal laws and Supreme Court opinions, which is what matters here.
      - IgorPartola 4y ago
        Wow that is pretty naive. Let’s see: A lot of stalking and harassment laws are state level and not the same in every state. Since online harassment is a common problem on Twitter, which state’s laws would you like to apply? Spreading misinformation that results in violence towards an individual is not governed by federal laws unless it falls under the definition of a hate crime. Spreading misinformation about results of elections isn’t illegal but is very obviously quite damaging and has real world consequences. How do you tackle this issue? That’s just a very very small sample. The problem you are going to face with your approach is that the US laws were not designed with the idea of governing stuff on Twitter.