Comment on Bluesky has started honoring takedown requests from Turkish government
sugar_in_your_tea@sh.itjust.works 2 days agoYeah, I’m working on something that I think should improve on things, but I keep bringing it up in the hopes that someone beats me to it. Here are some notes:
- P2P network based on something like IPFS or Iroh (I picked Iroh)
- a “community” is a distributed hash table, with posts, comments, etc as structured keys
- everything is cryptographically signed by the author, so you can check for tampering (built-in feature of Iroh)
- moderation is also distributed, based on “trust”; everyone is a moderator, and you “trust” others’ moderation either explicitly or by happening to moderate similarly; options are “like,” “dislike,” “relevant,” “report” (spam, CSAM, etc)
- everyone contributes a little storage to the network, and you can adjust your storage quota
Some interesting side effects of this design:
- single namespace - no “instances” since hosting is distributed (so just “Technology” instead of “Technology@lemmy.world”)
- everyone will see a different feed due to differences in moderation choices
- no concept of “all” since you wouldn’t sync communities you don’t care about - I would add a discovery mechanism to help here
- could be “sneakernetted” if countries block this service, provided you have a way to discover other users in each closed region
- nobody can censor you since moderation is opt-in, so I literally cannot respond to takedown requests by governments
- there’s a very real risk of echo chambers, but that’s on the user not centralized mods
When launching, I’d have a default set of mods that automatically “block” things like CSAM, but users can choose to remove those and/or adjust weights. The idea is for moderation to be transparent, but also something users aren’t expected to change.
The only hosting needs would be:
- relay servers to connect people - relay servers would be federated and incredibly lightweight
- storage instances - only needed in the early days until enough people join the network
- website for documentation and whatnot
It’s very early days (still working on the P2P part, but have a POC for the moderation algorithm). I’ll probably post once I feel like it’s actually useful, which won’t be for a while.
Fredthefishlord@lemmy.blahaj.zone 2 days ago
Sounds like you’re mistaking filters for moderation.
sugar_in_your_tea@sh.itjust.works 2 days ago
They’re essentially the same thing no? The main difference is in how they’re applied:
With Reddit/Lemmy, moderators are chosen by other moderators/admins, or are the people who create the community. It’s arbitrary and frequently leads to people mass-leaving the community if the moderation is poor. Other social media sites are moderated by algorithms or employees, which can also lead to people mass-leaving if the moderation is poor.
This approach preserves the distinction, but leaves the control in the hands of the user. If moderation is poor, it’s something you can fix using features like:
Hopefully that’s an improvement. Maybe it’s not, idk, but I like the idea of removing centralized moderation.
Fredthefishlord@lemmy.blahaj.zone 2 days ago
I don’t think your distinction between moderation and filters is correct.
I would say it’s closer to filters are something to curate what you say, and moderation is to curate the community.
Trust based systems are absolutely an amazing idea, but it is a very good idea to make it so once you have a level of ‘trust’ you don’t just apply filters, but actual moderation instead. Having a hybrid system, where you can apply filters from a certain set of users, but also allow purely trolling or immoral(ie cp and gore) to be fully removed so that new users or visitors to the community do not have issues.
It also serves as a stronger prevention measure against racists and nazis.
It also serves to preserve the correction function of communities, without allowing popular will to reject reasonable expectations just because they dislike them. That has happened a lot in redditor communities.
I strongly agree that purely centralized moderation is bad, but some level of centralization of moderation is beneficial.
sugar_in_your_tea@sh.itjust.works 2 days ago
That’s the traditional definition, sure. Traditional moderation essentially forces others to not see certain content based on the moderator’s opinion, and that’s incompatible with a properly P2P application where there is no central authority.
Perhaps here’s a more satisfactory definition:
So an individual client wouldn’t have the CP, gore, etc content for a given community because it has been moderated out. However, another user with different moderation settings might still have that content on their machine. If most people in the network remove the content, then the content is effectively gone since it won’t be shared, but there’s no guarantee that nobody has that content. Content nobody sees value in will disappear, since things are only kept when someone wants it.
Make sense? The only exception here is for that moderation queue, so you might have that content depending on your settings, but it wouldn’t be shared with others (client feature).
Perhaps. I just have trouble figuring out how to decide who the moderators are in a way that doesn’t lead to the problem of new users flooding a community and kicking out existing users, so elections are out. There are no admins to step in to mediate disputes or recover from a hostile takeover.
So my solution leans on the users to generally prefer to not associate w/ scammers, spammers, pedophiles, etc, and that disassociation would help them benefit from the moderation efforts of their peers that think similarly to them. However, this also means that Nazis, pedophiles, etc can use the platform to find like-minded people. But the only people impacted by their nonsense should be those who believe similarly to them, since other users wouldn’t see their content.
So we’ll still end up with silos, but they’ll be silos that users choose. If they don’t like what they’re seeing, they have the tools to fix that. Hopefully this is good enough that most people will get what they want, and in a way with user-driven censorship instead of platform-driven censorship.
The nice thing about this setup is that we can add centralized moderation if we choose in the form of public filter lists. It would be completely opt-in, and clients could be tuned for that use-case. But because of its distributed nature, there’s no protection at the protocol level to prevent undesirable people from forking the client and removing those types of filters, in much the same way that Lemmy doesn’t prevent someone from ignoring all moderation.
I’m open to suggestions. I also don’t like the idea of Nazis and child abusers using my platform, but the distributed nature means nobody has any form of top-down control. Either we elect a moderator (which is subject to bots and whatnot), or we remove the concept of moderator entirely.
I think we’ll end up with accounts that people can trust completely, such as bots that identify CP, gore, extremism, etc, and then you can just explicitly add them to your trusted moderator list. And I’ll probably add something like that to the codebase once it’s created. But yeah, it’s a tricky problem to solve, and I’m trying to lean on reduced centralization when I have to make a choice.