If you added the equivalent of a spam button ("I don't like this"), took the user's history into account, and relations with other users, would there be a way to pre-flag posts much like a spam detector?

Although I suppose even if it were an interesting and feasible AI project, and even if it were useful to some users, it's probably something Twitter just wouldn't want to get into. They would be setting themselves up for endless political flak when it did or didn't work correctly.