Just let individuals ban whoever they want from THEIR view.
If you want to be super-fancy, you could then see if some account X is banned by many of individual users from appearing on their feed, and give individuals an option to have those automatically banned from their own feeds after some threshold percentage.
So, if X is a jerk/spammer and individual many discussion group users have banned them (from their own view), give users the option to automagically have X banned from their own feed too once they hit say 10% of other members banning them.
This off-loads bannin a little, and as long as individual users have the option to check which those "auto-banned" are and e.g. except them from being auto-banned for them, it still maintains freedom.
In HN with showdead etc, I've never seen any "dead" comments that I couldn't just have as regular comments and just ignore on my own...
1. Blocking people is reactive. It means that everybody still sees the first time somebody DMs them calling them a slur. If you instead take the approach of "block everybody that the ML system thinks is alt-right" or "block every post that the ML system thinks is spam" then you are right back at the fun problems of false positives and defaults.
2. People aren't just concerned about their personal experiences on these services. Advertisers are concerned about their ads showing right next to posts calling jewish people evil. Citizens are concerned about the radicalization effect such that even if I don't see conspiracy posts about liberals eating babies, those vortexes still lead to social harm.