Again I think I erred in including "reputation system" in the claim. I was trying to draw a distinction between voting up or down on a claim/comment/post and between providing evidence for or against.
While I'd like to think it could be evidence all the way down, in reality you would need voting at least at the leaf nodes. And the voting is probably useful at each level.
I wasn't trying to make it exclusive, rather that conversational structures to support things like evidence, cruxes, and tests is an unexplored area in the field, while voting systems are both well known and hard to perfect. So the ROI feels higher for the first.