Oddbean

Relay-based filtering. Only load comments, replies and mentions from relays that are deemed (by you) to be safe. That should be the default in all clients. What relays are safe? Relays that will require some form of friction for accepting events, and that will actively ban people that have been flagged with kind 1984.

We’re working on a moderation bot that will accept encrypted report notes, then the bot can review the content via AI and a human, and then issue a report on that content. Basically it’s not always a good idea to be public about your report. Also clearly we need the ability to vet reports, because often the report is not accurate or isn’t enough. So i an say, sure bob reported jane as being as spammer, but I don’t agree, or this moderation group looked at Jane’s content and decided Bob’s report didn’t have merrit. Sure others might agree with Bob’s report, but the point is that we need a second level, and the ability to have some reviews.

I think that approach is valid, but I don't know if it's worth trying to arrive at the "best" judgement of whether someone should be banned or not. It's better to foster an ecosystem in which relays can ban freely according to whatever absurd criteria they decide and users and clients can pick relays wisely according to their preferences (most clients are not ready for this as far as I know, but it wouldn't take much). On https://pyramid.fiatjaf.com/ my plan was to somehow let hierarchy of the relay somehow police itself. If people higher in the tree don't like someone down they can just ban that person -- and if they are not in the same subtree their vote still counts more than someone that is lower. Or something like that. Of course I didn't implement any of that yet, I was even blocking kind 1984 by mistake until a minute ago.