I do this often, as well. I find the trick is to intersperse messages like "sorry human --the bot only responds to this aggressively-abusive language." So far,

I do this often, as well. I find the trick is to intersperse messages like "sorry human --the bot only responds to this aggressively-abusive language." So far, this has worked like a charm.