2. We recognize that there is no one-size-fits-all approach to moderation—no single company can get online safety right for every culture and community in the world. We’ve also been building an ecosystem that gives communities power to create their own spaces, with their own norms and preferences.
Still, using Bluesky feels familiar and intuitive. It's a straightforward app on the surface, but under the hood, we have enabled real innovation and competition in social media by building a new kind of open network. Read more: bsky.social/about/blog/0...
Will moderation services be commutative - that is, does the order of layering matter?
the ordering doesn't matter, they all combine based on your preferences however the app can hardcode a labeler (which ours does) and give that labeler the ability to apply full takedowns (which again ours does with our moderation)
Thanks Paul. Love your work.
Is it possible to roll your own app with a different hard-coded labeler, and if so, would posts subject to a "full takedown" by the hard-coded BlueSky labeler still be visible in that alternate app?
yes and yes
well "full takedown" is a bit ambiguous, we talk about app takedowns and infra takedowns an app takedown, it would still show. an infra takedown would take it out of our relay and appview
Thanks for humoring me Paul. How does something get taken down on an infra level? Do the people manning the hard coded labeler have some sort of "kill it with fire" button for illegal content that third party Ozone users don't?
pretty much
very cool
currently, no, the order does not matter. "stronger" interventions take precedence. (the logic is actually pretty complex. we did a lot of prep, but I might be slightly wrong, and it is possible we'll need to tweak the logic as corner-cases emerge)
Thanks! Sorry for the somewhat nerdy question.
It's a good question. Has a effects on the behavior of mods for smaller vs. larger groups.
I would like to see other labels added, like: Divination, horoscope, tarot and the like Conspiration theories and disinformation Suspected bot
Will the label API be opened up for third-party labellers?
They said that's the goal, and then users can subscribe to third party labelers (and clients can add support for recognizing and processing arbitrary labels)
This is huge. Community notes will draw a tonne of people. Once you have private messages done, I think working in multi-user accessible profile for brands and businesses would bring a lot more over. Then businesses could link staff or mods to help run them. It'd be good for accessibility too 🤘
You know @safety.bsky.app a simple tried & true method for adding safety to social network accounts that has been in use for over a decade is Private Accounts. Every other social platform has this feature with newer services implementing at launch. It would be a really nice feature to have here.
I found these posts from Paul helpful at explaining the situation:
Thanks. I will patiently and eagerly await a private account feature.
It's simple in a closed platform (or even in a federation of closed small ones like Mastodon). But in an open platform like atproto/Bluesky, it's a lot more complicated (to do it well without any possibility to bypass the privacy of the accounts).
Thanks for that insight. I hope they get it figured out soon. I think more people would feel more comfortable about posting on the app if they had control over who sees their posts.
Is there a way to mute all sexually explicit content?
yes there is :) Go to settings, then moderation
Thank u
overall excited about this development, but I've had a question continuously raised in my mind: if users are able to opt out of moderation defaults, build their own servers, etc. what happens with content that should definitively not be ANYwhere (CSAM, explicit hate speech, etc.)?
illegal content like CSAM gets handled at the infra layer
And hate speech? He asked about that too
figured it’d be that way but good to get that confirmed
I've seen some pretty explicit hate speech just labelled with a warning.
Remember when Bluesky launched and it just had a Hate Speech toggle in the moderation options? I'd have to see how all this can be implemented, but I wonder how far they've improved since then...
which part? the relay?
yep
Could someone theoretically bypass the bluesky moderation if they were using a third party relay?
yes, but they'd be somewhat segmented from the network somebody banned from Bluesky wouldnt be able to interact with the Bluesky users
Is there a particular jurisdiction that governs your definition of "illegal" or is this policy exposed to legal fuckery from countries with suppressive speech laws (ie the UK where saying something negative but factually correct about a rich person can be construed as libel)?
Obviously illegal stuff (any kind that could put Bluesky in trouble if not removed) will be filtered out on the Relay and/or AppView level, as I understand, and also removed from the PDSes in case of Bluesky-run ones
is it theoretically possible for 3rd parties to build Relays or AppViews that leave these filters out? or is there something inherent about atproto that doesn't allow for that?
Absolutely possible! No one apart from Bluesky runs those *yet*, but it's important for the decentralization of the network to have more options, so hopefully someone will
hmmm considering I was wondering about a hypothetical where someone could build infrastructure that circumvented CSAM filters I don't know that I'd use the word "hopefully" lol. I know what you mean, but this is a major worry I have about decentralization
Well, uh, yeah, you know what I meant ;)
They already do with private websites, email servers, chat rooms, etc, so atproto doesn't uniquely enable anything illegal in that regard. It's just like the email protocol itself, the protections are at the interfaces between servers and between users (spam filters, etc).
What's the point of having reporting options for spam, trolling, and harassment when you don't moderate it? There are accounts that violate these 24/7, all have been reported and no action is ever taken! #Twitter2.0
Sounds like you're shaping up to be an even worse place for public discourse. bsky.app/profile/vael...
Just, please, don't go full Reddit.
It's federated so it kind of is by default, with less external moderation. This is a suped up Mastodon
Of this goes that way I will not stay. There is most value to me in seeing random things, not just being in an echo chamber.
The architecture with relays and 3rd party feeds does a good job in avoiding the echo chamberism of Mastodon
I just don't want to see Walled Gardens everywhere. It's fine to curate your experience with the block and mute tools and even to use community mute or block lists if you are so inclined, but letting groups police their own places makes the whole site poorer, in my view. Variety is the spice.
There are people who love the echo chamber and others who don't. Wouldn't this end up serving everyone?
It would take away from me if people closed themselves off in Reddit-like sub forums. I like to see many, many different interactions and if there were echo chambers that I eschew then this service would be the lesser for me.
From my understanding of this, unless you're blocked, you could still see them regardless of how many so-called sub forums they join. Bluesky will have a default feed where everyone shows up. I don't think people can post directly to a sub forum and not have the post show up on the default feed.
Seems like a lot of hoops, to me.
Yeah but basically, you'd still use Bluesky as usual without noticing anything different. It's just that some people will not use the default feed, hence they'd only see your post if you mention a key word that matches that of the custom feeds they follow. I could be wrong about this of course.