I non-verbally recite the word "epicycle" when confronted with theoreticians who refuse to entertain the possibility that all or nearly all their assumptions are false, and who continue to tweak their "models" as if these models were asymptotically just "that" close to full and universal understanding of the global economy. This is in reference to the epicycles that our predecessors used to desribe the troublesome orbits of Mercury and their attempts to fit facts into prevailing theory (at the time a Ptolemic circular architecture presumably based in part on the soundness of Platonic forms...in this case the circle).
So the below remarks from a respected macro-economist come as no surprise...if only we had more experimentation and data to truly understand...we are so close. Ha. There is no experimentation that can be valid in the real world of animal spirits and reflexivity. There can be no theory of economics without a theory of power...even Marx understood this...and the institutional structures in place at any one time in history are non-repeating events given the various reflexive actions of participants because of them.
That said, the good news is that we recognize that in finance and parts of macroeconomics, the models or frameworks are incomplete. That represents a challenge to the academic community. But it also means that, in the short run, participants and regulators will be operating with incomplete models. This will require judgments (which will be uncomfortable in contrast to the earlier sense of certainty). There will be mistakes. And, as Olivier Blanchard said in his excellent summary, we will proceed step-by-step, evaluating the impacts of policy choices and sometimes reversing course.
This is relatively familiar territory in developing countries, where changes in the institutional depth of the economy mean that models (especially advanced country models) are not very precise in predicting market responses to policy choices. Nevertheless, theory is still useful when used judiciously. In that context, you can think of this as analogous to navigating with charts that are incomplete.
Having said that, we do in principle have the option of returning to old patterns while waiting for different or more complete models to be developed and tested; I think there is a widespread recognition that this would be a risky mistake