from an exchange on cliodynamics
(or: The most unintentionally Luhmannian pronouncement I’ve found myself making)
At the hypothetical moment when ‘they actually start accurately predicting future events better than other means we already have’, the nature and concept of an event bifurcates. Now “event” will signify two qualitatively different things – those (henceforth relatively uninteresting) events which, taking place without access to the newly-forged means of prediction (wordplay intended), simply fall under them as their objects, and those events (now of a qualitatively different order), into which access to the calculations reenters as a principle of selection. What happens in the case of the second kind of event – those with respect to which the means of prediction have to operate self-referentially? Either – which seems likely – contingency envelops the new system as a whole (even though it includes, as elements of itself, some accurate predictions about its subsystems) or else someone would have to discover some sort of really stunning fixed-point theorem that prevents access to the prediction from altering it.