Because the systems under study are not isolated external systems that follow well-order rules like a laboratory science. Prediction of human systems is like prediction of weather systems one week ahead: there are ways to model and improve methods for such a task, but our predictive success will continue to be just a little bit better than playing slots.

It would be folly and vain to misunderstand the difficulty of the problem at hand in an effort to signal that one is confident and optimistic. Social science from a century+ ago suffered a lot from folks with that frame of mind, sometimes ending in utter disaster.