Snowball or butterfly - the underlying assumption is that history (or rather the societies that create history) are so chaotic that small changes lead to large deviations. Problem is that a change that does cause a reaction in one direction is very likely to result in a counter acting butterfly flapping it wings the other way. So the overall effect is dampened considerably. Even it its original home the butterfly effect is being challenged as some believe the weather systems are nowhere near as chaotic as Lorenz believed due to viscous dampening effects.
Thank you for calling out this assumption for what it is. I do firmly believe that history is a chaotic system, but i recognize that it's not necessarily irrational to believe that history is more like Statistical Mechanics when an underlying chaos leads to predictable macroscopic behaviour. In fact, Isaac Asimov, in his Foundation series opeeated on the premise that history on a macroscopic level could be predicted by a sufficiently complicated model, and thus, on that level, was not chaotic.
I think my reason for my belief in a chaotic model making its way up to even the most macroscopic trends in history has more to do in my belief in a sort of free will than anything else. If history is not chaotic, then there is nothing I as an individual can do to prevent future wars, etc. (trying to tie this back to whether WWI would still happen). I firmly believe that it is possible tbat I could be the straw that breaks the camels back of a future war, and that, despite being an altogether ordinary individual, my actions do matter.....
At the same time I recognize that chaos theory would hold that while my actions do matter, there is no way I could ever predict HOW they matter, so I might as well act randomly..... I'm going to stop before I get all existentialist..m