As the Fediverse grows, the amount and frequency of harmful spam also increases. The feudalistic structure of the network creates notable moderation overhead as every instance admin has to ban offending remote accounts and remove their content themselves, if the originating instance does not act quick enough or such delete actions are not federated properly.
An example (and my motivation for this post) are reports from the last days about accounts which post CSAM content. This is especially problematic, as media content gets duplicated to every receiving instance.
The following idea came to my mind: A software which sits on top of Activity Pub and creates a kind of court. Instances of this software are independent courts. All users can elect admins / mods of their home instance as judge. Everyone can post a case via normal Fedi posts to the court and the elected judges give a verdict via majority vote (votes are also normal Fedi comments with special formatted content). Once a case has a verdict, all instances should immediately apply the decided action against the offending content / account. This can be done via a bot which is set as an instance moderator and follows the published verdicts of a court instance the local admin trusts.
This way swift action can be applied through all instances which follow the court at once. By having periodic elections it is kept democratic.
And the task of the court is to defend universal rules all following instances agree on. Not to handle politics between instances.
What are you thoughts on this matter? Or do other approaches to solve this issue already exist?
Update:
From this came the idea that instances can follow mod actions of other instances and apply them if x% of other instances did the same.
I think this is an interesting approach which might be better than my proposal. So I would like to hear your thoughts about this too.
Posts hitting the fediverse over the past 24 hours or so are essentially creating a new account for each post. The accounts are across many instances and posts across many communities.
I'm not sure how this proposal would work in the current environment.
Good point. If they have a shared identifier (IP or something) than I could see the block list working. If not, and it's probably not, then yeah, I don't think the block list would be able to handle that spam.
I wonder if there is any approach that could. Different origins, different entrances, and different targets, that's a difficult one. It's not my knowledge base, so maybe something exists that could handle it? I'd be curious to hear about it, probably pretty cool.
Maybe machine learning on interaction rates (sign in to post timeline)? I'm assuming it's automated though...