Imagine for a moment that I'm on hachyderm, but I prefer how tooters (to pick a random server) does moderation.
With atproto-style labels and a little bit of elbow grease I could make it so that their classifications would propagate to my account and I could decide what to do
Without me having to move servers
Like… just sit with that for a moment
Yes, there are complications in doing that well, yes, I would prefer private feeds for it, and also… just consider what it would look like here.
@jenniferplusplus Sure, and moderation is holistic in that it has multiple aspects there, it doesn't work in isolation.
There is just power in being able to say "this is my community that is separate from my server who I trust to label things" combined with the personal override of being able to dictate how you respond to the label, irrespective of how others might respond.
Right now our current architectures are just servers deciding what their users see as opposed to your selected group.
@jenniferplusplus For example: More than about what I see or don't, it would allow, e.g., communal labeling of content warnings, or processing of nudity, etc and for me to configure a client with how I want it to handle those things.
@hrefna@jenniferplusplus
I am very much in favor of communal moderation, but I have come to believe (despair?) that only a small minority will ever change their server/app default configurations.
It seems to me that as social media grows the common defaults will become more important and filter only the very worst content (that will get you kicked off Cloudflare for example). A few (maybe 5%?) will enjoy much better customized experiences, and be used to defeat efforts to change defaults.
@hrefna Yes, the existence of some kind of content labeling facility for the fediverse could open up a lot of possibilities. There's something about the way it comes up that gives me pause. It has to do with that notion that individuals would choose the labels and labelers they want from a menu, and this would supplant the discretion that instances currently provide.
@hrefna
It's worrisome how individualistic that view is. I think the most radical aspect of the fediverse as it exists is that the service providers are in dialog with their members. There's an active two-way relationship, and its (tenuously) anchored to the instance. It creates a kind of place, which allows those places to establish some behavioral norms. There's behaviors that do and don't fit in. That's really powerful in terms of fostering trust and safety, to use the corporatism.
@hrefna
So. 3rd party, and possibly client-selected labeling is an interesting idea. But I don't want it to weaken the little bit of place-making that we have. I'm much more interested in how it could reinforce that, instead. But that doesn't seem to be how other people imagine content labels would work. So I get anxious about the topic.
For example, I have TW requirements that are simply not reasonable to ask people to adequately flag. What's worse, they tend to CW them differently even when they think to do them
But a small server could have such a rule in place.
It's also not reasonable for me to need to negotiate for my mental health with a bunch of other people, or to do it piecemeal, when it comes to what I see. But doing that via limit/block is reactive. 1/
Sometimes norms exist that would improve someone's mental health, but how do you push that norm out?
People harass others, even on different servers, today to make sure that they do things like apply CWs or mark media as sensitive, even against servers where those are not rules.
Their need can be real, but asking every individual is not sustainable.
But that can be handled via an automated system and a labeling policy, or by sharing such between servers. 2/
To me as well, I see the ability to run a server technically and the ability to manage a community as essentially orthogonal skills and in my case—larp communities—I have a large number of partially overlapping groups that are essentially transitory in nature and where there are goals other than "to socialize"
Do you set up a server for each one?
Do you have a larger server (or set of servers) and encourage forming subcommunities that may have different norms?
I don't know that there are good answers here. There's a lot of nuance and complexity in how these things interact, and real people are endlessly messy. Even concepts like "blocks" are difficult to manage
Ultimately I want "force short of lethal" and not to always find myself negotiating in every space
But as you note, that comes with tradeoffs, and I don't think that those are unreasonable desires either. That radical connection to community is also important 4/
Delegating mod responsibilities also has a long history of harm, and that ability to build and define community around norms is very important.
Basically, the tl;dr: I have no answers. Only complex needs and use cases and a lot of need to find tools that I can beat to fit, paint to match to get them to work, if that makes sense?
@hrefna
These are all very reasonable things. A lot of this puts me in mind of the groups feature(s) that's been requested for years. What you're asking for feels like it lives in the combination of client and group selected labelers and label policies. Which sounds great, and I think also supports instance moderators. It makes it easy to tell the difference between a metaphorical boxing club and a fight club, because now there are gloves and referees.
Yeah, great thread all around -- thanks @jenniferplusplus for tagging me. We're all groping towards good answers. I agree that the ATProto architecture is extremely individualistic and is framed as an alternative to fedi's instances, which can provide some amount of place and community. And I also agree that fedi's current mechanisms for place and community fall far short of what's currently needed -- for example private cross-instance groups are a key feature but don't yet exist.
I look at the ATProto stuff both in terms of potentially-useful building blocks on its own and also as a lens to look at today's fedi. For example I agree with @hrefna that labelers and custom feeds are potentially very good useful for the kinds of things that CWs try to address but don't. And I see it as a good thing that the ATProto architecture makes Relays explicity; they're also important in today's fedi but there's not of awareness of that.
And it's interesting to think about how to build federated communities on ATProto. Rudy Fraser's moderation article (which is great in multiple ways) has a quick description of what he thinks would be needed for a Blacksky community. For federations of communities (caracoles. / fedifam) you'd probably also want a Relay (as you would for fedi) and AppView.
This thread and the attached PRs give me no hope that it will actually solve the use cases we're talking about when it eventually lands, however: https://github.com/mastodon/mastodon/issues/139
Moderation that's about what you can see, or who can use your platform / be in your replies is one thing and should be very doable from the user side. Who can see your public posts at all is another.
And that's a really hard problem. I don't know if there is a good solution other than instance level defederation -- iirc even individual domain blocks don't provide great guarantees.
I do question which threat models require that level of moderation though + find it effective though, since the posts are still public.
I would also argue that "who can see your public posts" is something that even instance level defederation does not and cannot fix—and can only barely improve—without significant changes to how we work with the protocol
I agree with you that it is a hard—albeit important—problem and that the threat model needs to considered carefully
In a way it is a different problem for a different part of the stack. Labeling won't fix it, but it'd help in other ways