Simplifying some thoughts from a post I did earlier:

Here are some user-configurable factors a UI could use for spam filtering on nostr. Some of these are possible using the current version of the nostr protocol, with no changes to relays, with extra processing being done client-side. Some others might require extra processing on the relay's end or something.

Signals to trust a post:

- It's only been reported by a user you have flagged as untrustworthy with reports.

- It's only been reported by a user flagged as untrustworthy by someone you have flagged as trustworthy with flagging other users' trustworthiness.

- It's only been reported for a reason you don't care about (e.g. impersonation that might just be for meme purposes and doesn't override cryptographic verification anyway).

- Someone you follow has liked it.

- Someone you follow has reposted it.

- Someone you follow has followed whoever posted it.

- Someone you follow has previously liked a post by the same npub.

- Someone you follow has previously reposted a post by them.

- [2 degrees of separation] Someone followed by someone you follow has done one of these.

- [More degrees of separation] Someone followed by someone followed by someone you follow, etc.

- You've assigned higher trust weights to users who liked posts by this npub than users who reported posts by this npub.

- [2 degrees of separation] You've (followed or) assigned higher trust weights to users who have (followed or) assigned high trust weights to users who have liked posts by this npub etc

- [More degrees of separation] You've assigned whatever, you get the point

Signals to filter a post - basically just the opposites of the above:

- It's been reported by users you have followed or flagged as trustworthy with reports.

- It's been reported by users flagged as trustworthy by someone you trust.

- It's been reported for a reason you care about (e.g. impersonation that you don't feel like dealing with for the occasional good meme).

- The user or post's ratio of people liking vs reporting isn't good among those you follow.

- The user or post's ratio of people reposting vs reporting isn't good among those you follow.

- [2 degrees of separation] Everyone followed by anyone you follow is factored into these calculations.

- [More degrees of separation] Everyone THEY follow, etc.

- You've assigned higher trust weights to users who reported posts by this npub than users who reposted posts by this npub.

- [2 or more degrees of separation] People followed by people you follow have done whatever I was gonna say, you get the point again