The latter precludes the former in my way of modelling internet contributions.

Ah, I see. Unfortunate that "trolling" is so ambiguous as to whether it's about results or motivations (i.e.(?), immediate results or expected future results (potentially conditional on feeding/anti-feeding)). Results in e.g. Eliezer calling XiXiDu a troll even when XiXiDu clearly isn't trolling in the conative sense. Steve suggested ghost netting for the non-conative case.