Debugging conference anti-harassment policies

Even though it is is subjective, most people will get it "right" most of the time. Aren't people amazing!
And when people don't get it "right", someone will complain (politely I hope as I think was the case at LCA), an apology can be offered, and the incident becomes information that future speakers can use to guide their subjective decisions.

Thus workable guidelines get defined largely by a set of examples.

Sure: the mathematician in me would rather precise guidelines, but the community member in me knows that it not possible and tends to create as many problems as it fixes (and hence is only valuable if the problems it fixes are much more severe that the problems it creates).