Tempted to write a "What is Fedi" blog post explaining how wanting to have builtin automated blocklists and "trust and safety" features won't fly with almost every Fedi server dev out there. So I can then send it to every post I see about users wanting this and get yelled at for being XYZ.
Timeline
Post
Remote status
Context
1
@phnt Aren't those features already in the process of being added to almost every fedi server? GoToSocial and a number of others already have block list subscriptions
Replies
14
@silverpill GTS can do it by I think importing remote CSV files, Mastodon has their FIRES grant. Those are the only two implementations I'm aware of doings this and they are nothing compared to tens of others. If you don't account for install base that is.
@silverpill My point is that you can't enforce stricter moderation on this network no matter how you try. There will always be ways to circumvent it. Like respawning under a new domain for $2. Not to mention that the mentality of I think most Fedi devs is that users should moderate their timelines themselves, however they want. Instead of some Big Brother doing that for them. And one of the reasons why I think user-level moderation is way better, is that it lacks the possibility of a turnkey change in moderation. Elon bought Xitter, instead of removing blatant censorship, only who gets censored and how changed.
@phnt There is more
The first that comes to mind: https://github.com/Automattic/wordpress-activitypub/blob/trunk/includes/class-blocklist-subscriptions.php
I remember seeing other devs promising to implement automated import
@silverpill Did you know that this apparently exists?
https://github.com/swicg/activitypub-trust-and-safety/
https://github.com/swicg/activitypub-trust-and-safety/
@phnt Yes, this is probably the most active SWICG task force, they even got funding from NLNet recently: https://github.com/swicg/activitypub-trust-and-safety/discussions/144
@phnt The work plan seems to be relatively benign, though:
GtS interaction controls, flagging, blocking, best practices.
Automated filtering was voted down.
@silverpill It's not like much can be done anyway, decentralization and some "funny" admins running malicious instances turn that effort upside down. Similarly how BitTorrent still wasn't moderated and regulated into oblivion with how much pirated content is on it.
At least semi-standardized Flag/Block will be a nice thing unless they go "implementations MUST prevent blocked users from viewing blockee's posts". Interaction controls are meh. And automated filtering will probably end up there someday after the frog is boiled for a little longer.
At least semi-standardized Flag/Block will be a nice thing unless they go "implementations MUST prevent blocked users from viewing blockee's posts". Interaction controls are meh. And automated filtering will probably end up there someday after the frog is boiled for a little longer.
@phnt By the way, I found a fun way to sabotage those task forces. You can publish a FEP describing some mechanism (e.g. interaction controls), and say that it is patented. No one in fedi would give a shit, obviously, but this seems to be very important for those working at W3C
@silverpill @phnt >Automated filtering was voted down.
Hmm, I see that one of the recommendations from the person who opened this issue includes a proprietary enterprise service. I think that's a major point against it.
https://github.com/swicg/activitypub-trust-and-safety/issues/28
But what I'm more interested in is whether they wants to filter text? If so, wouldn't it be easier to just crudely add a lexicon containing tens of thousands of sensitive words to block users from posting, like some social media censorship? 
@caohuak @silverpill The person that created that issue is someone that more or less created Fediverse long before ActivityPub, you can see his name as an author of ActivityPub, OStatus and more.
Things have changed since the creation of that issue, he now pushes IFTAS which is a nontransparent organization that built an automated attachment filtering service that died with USAID drying up and that operates an automated blocklist subscription tool you have to give full admin rights to. Sources where they get data for that blocklist aren't disclosed, reasons behind additions/retractions to the blocklist aren't disclosed either. Members of the org are hidden on a page on their site that is hard to find. But one of the prominent members is a guy that keeps pushing his "trust and safety ideas" into ActivityPub related dev forums and pushes the view that Bluesky/ATProto is part of the Fediverse.
Things have changed since the creation of that issue, he now pushes IFTAS which is a nontransparent organization that built an automated attachment filtering service that died with USAID drying up and that operates an automated blocklist subscription tool you have to give full admin rights to. Sources where they get data for that blocklist aren't disclosed, reasons behind additions/retractions to the blocklist aren't disclosed either. Members of the org are hidden on a page on their site that is hard to find. But one of the prominent members is a guy that keeps pushing his "trust and safety ideas" into ActivityPub related dev forums and pushes the view that Bluesky/ATProto is part of the Fediverse.
@caohuak @silverpill Safe to say, he's not trustworthy anymore.
@caohuak @phnt I think they are interested in CSAM detection, but once the infrastructure for large-scale filtering exists it can be re-used for whatever authoritarian governments come up with.