Federated bayesian filters or local machine learning ? You could have a personal (stored on device) Bayesian filter that learns from your preferences, and a sh

Federated bayesian filters or local machine learning ? You could have a personal (stored on device) Bayesian filter that learns from your preferences, and a shared one to block spam spotted by your contacts As other people suggested, I don't think it should be at the node operator level, for freedom of speech concerns