• 106 Posts
  • 50 Comments
Joined 8 months ago
cake
Cake day: January 26th, 2024

help-circle






















  • I made this system because I, also, was concerned about the macro social implications.

    Right now, the model in most communities is banning people with unpopular political opinions or who are uncivil. Anyone else can come in and do whatever they like, even if a big majority of the community has decided they’re doing more harm than good. Furthermore, when certain things get too unpleasant to deal with on any level anymore, big instances will defederate from each other completely. The macro social implications of that on the community are exactly why I want to try a different model, because that one doesn’t seem very good.

    You seem to be convinced ahead of time that this system is going to censor opposing views, ignoring everything I’ve done to address the concern and indicate that it is a valid concern. Your concern is noted. If you see it censoring any opposing views, please let me know, because I don’t want it to do that either.


  • It’s difficult. A downvote from an account with no history does nothing. Your bot has to post a lot of content first to attract upvotes from genuine accounts. Then once you’ve accumulated some rank, you can start giving upvotes or downvotes in bulk to the accounts you want to manipulate. It’s impossible to completely prevent that, but you have to do it a lot to have an impact.

    I think this model is more resistant to trickery than it would seem, but it’s not completely resistant. I do expect some amount of trickery that will then need counter-trickery. On the other hand, the problem of tricking the system also exists in the current moderation model. You don’t have to outwit the system to get your content posted or ban your enemy if it’s trivial to flood the comment section with your content from alt accounts and drown them out instead. I don’t know for sure that something like that is happening, but it wouldn’t surprise me if that was one reason why there are so many obnoxiously vocal people.


  • You’re not banned or even close to it. The ban list is surprisingly lenient in terms of people’s differing political views. You have to habitually make enemies of a lot of the people in the comments, one way or another, with a big fraction of what you post. Most people don’t do that, wherever on the political spectrum they might fall.

    Whether that’s a good idea or not remains to be seen. I had some surprises today.





  • I looked at the bot’s judgements about your user. The issue isn’t your politics. Anti-center or anti-Western politics are the majority view on Lemmy, and your posts about your political views get ranked positively. The problem is that somehow you wind up in long heated arguments with “centrists” which wander away from the topic and get personal, where you double down on bad behavior because you say that’s the tactic you want to employ to get your point across. That’s the content that’s getting ranked negatively, and often enough to overcome the weight of the positive content.

    If Lemmy split into a silo that was the 98.6% of users that didn’t do that, and a silo of 1.4% of users that wanted to do that, I would be okay with that outcome. I completely agree with your concern in the abstract, but that’s not what’s happening here.




  • I’ve already declined two reports requesting that I take moderator action against content that’s people directly going out into their community and helping get things done, because that is “not politics.” People definitely seem to want their mods to be vigorously engaged in enforcing the boundaries on the stuff people are allowed to say.

    As far as my take on it, we can have overlap between the peasant politics and the pleasant politics. The community was for the latter, but the former sounds great, too.



  • Don’t let the python fool you. It is not simple python. I’ll try to add some comments later on to make it more clear what’s going on.

    For tuning parameters, it was complicated. Mostly, I did spot-checks on random users at different ranking levels, to try to check that the boundary for banning matched up pretty well with what I thought was the boundary of an acceptable level of jerkishness. That, combined with deeper dives into which comments had made what contributions to the user’s overall rankings. And then talking with existing moderators, looking over the banlists, and bringing up users where they thought the bot was getting it wrong. There were a lot of corner cases and fixes to the parameters to fix the corner cases. Sometimes it was increasing SMOOTHING_FACTOR to make users more equal in rank with each other, when we found some user that was banned because of one bad interaction with some high-rank person who downvoted them. Sometimes it was changing parameters to change how easy it is to overcome a few negatively-ranked postings by being generally positive with the rest of your postings. There are always users for which the right answer is a matter for debate or opinion, but as long as the bot isn’t making decisions that are clearly wrong, I think it’s doing pretty well.

    You can look over some places where I talked with people about the bot’s opinion of their user, in this post and this post. I don’t want to publicly do those breakdowns for people who haven’t agreed to have it done to them, but that might give you an idea of how the tuning went. What I did to tune the parameters was the same type of thing as I showed in those comments, just a whole lot more of it.


  • I know exactly what you mean. If I had to pick one type of comment that the bot is designed to ban for, those are them. It turns out to be pretty easy to do, too, because the community usually downvotes those comments very severely, even if the current moderation rules allow them even when someone does them 20 times a day.

    Pick a name of someone you’ve seen do that, search the modlog on slrpnk.net, and I think you will find them banned by Santa. And, if they’re not, DM me their username, because there might be some corner case in the parameter tuning that I have missed.


  • I was kind of like rooting for you, but it just seems like from what you said here that you’re only gonna allow people to be rude to whatever party. It is that the majority people on Lemmy don’t like.

    You’re absolutely right to worry about this. This was one of my biggest concerns when I was setting it up. Lemmy already has a definite community vibe and consensus opinions to go with it, and I think censoring the “opposition” opinion is one of the quickest routes to turning any political community into a useless circle-jerk. Most lemmy.ml communities are like that.

    My goal was to set the parameters broadly enough that people who disagree with the community are allowed to say whatever they want, but still strict enough that people who are outright jerks in any big fraction of their comments get removed. The current tuning bans about 1.4% of the community. You’re still not banned. I don’t think limiting it to 98.6% of the community will create too much of a circlejerk. There’s only one user that I’m aware of that is banned, for which I disagree with the ban, and I talked to them for a while, and sent them some detailed examples of what the bot concluded about their posting. I concluded by saying that while I disagree with silencing them, I think amending the way they present their posts will help the bot’s conclusions about them, and also for the same reason get their point across more effectively to any person who’s reading them. The huge amount of downvotes they’re getting doesn’t necessarily mean they are wrong, but it does mean most people are putting them in that bottom 1.4%, which is a problem if they want to convince anyone or accomplish anything.

    It helps that I sympathize with some viewpoints that are unpopular, so I get it if someone wants to have the right to speak their mind without some person looking over their shoulder deciding if they’re allowed to, or if they’re being civil enough about each individual comment. You’re right. That’s ridiculous.

    You’re just going to take their word for it, as if they’re some certified expert and shit? “I don’t like what you said, therefore i deem you a this or a that”

    Absolutely not. Part of what came through over and over again while I was tuning the bot, and looking over mod decisions to contrast with it, was that a lot of times the moderators are coming in and making snap judgements that are far less complete and accurate than can be gotten from looking at what the whole community consensus thinks is a problem.

    You’re doing exactly what Lemmy is already doing.

    Why is it that some of you moderators and admins can’t just be equal without letting your feelings dictate who is right and who is wrong?

    Assess both sides under the SAME scrutiny, even if you don’t like something. I mean, really who, even wants to be a part of a discussion like this?

    This is the algorithm. It’s not going to be clear what it’s doing, since it’s not commented well and it would be complicated to understand even if it were, but surely you can see that there is no “if my_llm_thinks_is_fascist:” block in it or anything.

    Like I said, you’re not banned, as of the current parameters. Part of the idea is to give people the freedom to come in and say what they want, instead of having an overworked mod decide by hand on the spot what is disinformation, what is incivility, what sources are reliable and not, important and not trivial decisions like that. I don’t know how to duplicate for you the time I spent looking over what the conversations really look like, how to draw the line so that the people everyone thinks are clearly bad actors are removed, but the people who are simply unpopular or have a minority opinion are welcome, but that’s what I tried to do.

    One way to cut to the chase: Just try it. Come in, say some political opinions, see if it works. The bans are mostly static based on past behavior, so as long as you’re not posting porn or KKK flyers or something, I think you’ll be fine.

    If it’s something outside the realm of politics I will probably moderate it by hand. I’m not trying to offer a blanket “free speech safe space” for racism or anything else that anyone feels like posting. Sorry. If you want that, you can go to Twitter. It’s up to you of course, but I think that this is a step closer to what you’re saying here that you want, not a step away from it.



  • The code for the bot is open source. It’s not an AI model. It’s based on a classical technique for analyzing networks of relative trust and turning them into a master list of community trust, combined with a lot of studying its output and tweaking parameters. The documentation is sparse, but if someone is skilled in these things they can probably take a few hours to study it and its conclusions and see what’s going on.

    If you’re interested in looking at it for real, I can write some better documentation for the algorithm parts, which will probably be necessary to make sense of it beyond the surface level.


  • I completely agree with you on that. “Pleasant” might have been a misleading way for me to frame the community. As far as the bot is concerned, you’re free to be as unfriendly to fascists as you want.

    As a matter of fact, part of what I think is wrong with the current moderation model is the emphasis on “civility.” I think you should be allowed to be unfriendly.

    I’ll give an example: I spent some time talking with existing moderators as I was tweaking and testing the bot, and we got in a discussion about two specific users. One of them, the bot was banning, and the other it wasn’t. The moderator I was talking with pointed it out and said that my bot was getting it backwards, because the one user was fine, and the other user was getting in arguments and drawing a lot of user reports. I looked at what was going on, and pointed out that the first user was posting some disingenuous claims that were drawing tons of hate and disagreement from almost the entire rest of the community, that would start big arguments that didn’t go anywhere. The second user was being rude sometimes, but it was a small issue from the point of view of the rest of the community, and usually I think the people they were being rude to were in the wrong anyway.

    The current moderation model leaves the first user alone, even if they want to post their disingenuous stuff ten times a day, and dings the second user because they are “uncivil.” I think that’s backwards. Of course if someone’s being hostile to everyone, that’s a problem, but I think a lot of bad behavior that makes politics communities bad doesn’t fit the existing categories for moderation very well, and relying on volunteer moderators who are short on time to make snap judgements about individual users and comments is not a good approach to applying the rules even as they are.

    So come in and be impolite to the fascists. Go nuts. You don’t have to be pleasant in that sense. In fact, I think you’ll probably have more freedom to do that here than in other communities.


  • I would suggest that what is to you “correcting misinformation” can easily be received as just being cantankerous or offensive.

    If you accept that the other person has a choice whether or not to agree with what you are saying, and show respect for both their ability to make up their own mind about it and the possibility that you might be the wrong one, I think you will be more successful at correcting the misinformation. As it is, I think you’re gathering a lot of downvotes because you’re airing deliberately combative opinions in places they aren’t welcome, and often not much more than that.

    I think a better solution would be to find a way to present your opinion in a way that still preserves the health of the community as you say, and stay, rather than to either hold on to your current way or else go. I didn’t read your entire profile, just that parts of it that the bot took issue with, but even those, I agreed with your unpopular opinion a lot of the time. But I do think the bot has a point that you’re creating your own unwelcome reception by the way that you are presenting them.