One could use Darwinian methods, where "hate speech" carries a higher risk of 'predation' ie = terminal modding. There may be a way to use recursive bots who look back at assorted offender threads and detect a pattern. A person who reserves his venom for a few select threads, and on all the other threads he comments on he is, as Mae West said, 'pure as the driven snow...' The bot would need to be able to detect places, where, as Mae West also said - 'but I drifted' The unpaid mods should be honored for their work, and work with these bots to increase the level of persistent detection and harassment of the hate speakers. 'Eternal vigilence = price of freedom' - someone said.