@jwz @lwndow

Hmm interesting. Given the structure of the Fediverse, with thousands of instances, and the fact that brand-new troll instances pop up regularly, it'd be nice to have some sort of shared-block capability.

@timbray @jwz @lwndow Hence my proposal for curation.

People subscribe to one or more curators who promise to find them content they are likely to 'like' and unlikely to 'dislike'. They are held accountable by the client suppressing results from poorly performing curators.

The curators take the like/dislike responses from a large number of users (a million say) and perform AI analysis to identify a much smaller number of orthogonal traits (hundreds). Each user is classified according to the extent to which their responses have a positive/negative correlation with the traits. This is then used to provide individualized recommendations for each user.

In this model, the people producing the spam will of course create a citation cartel liking the garbage as a sibyl attack on the system. But these responses will be identified as a separate trait which most users correlate negatively. So the fact that the spammer citation cartel is liking the post makes it less likely I will see it. Contrawise, the fact that Putin's people are disliking posts critical of the diminutive dictator will actually make it more likely I see them.

In this model, the instance you post from is irrelevant. Only the interactions of people whose responses closely correlate with your own are going to affect what you see. So the NAZIs will get feeds containing lots of the pathetic content they enjoy and I will see none of it.