Without meaning it as a personal attack, do you have any reason to think that 6 years of trying to detect meanness has made you any more accurate at it than you

Without meaning it as a personal attack, do you have any reason to think that 6 years of trying to detect meanness has made you any more accurate at it than you were before, or than anyone else is? When you say "trained to pattern match it", is there any feedback to change your models? As moderator-by-fiat it's your decision what comments are mean, which is potentially a self reinforcing loop.