"This person is evil/crazy/retarded/etc, don't listen to them!" in an automated way is still just more speech. If you don't agree with it, don't listen to it. When the architecture of the protocol (rather than applications built using it) prevents speech, I'll be right there beside you fighting.