I do a bunch of things, don't care if any of them work
the ones that do, in combination, tell me something about what may be going on
to save steps, I get my intuition involved alongside my scientist
I scry for patterns that I can test using novel tools - tools that would create results that either prove the fit (if not the fact) of the pattern
I build the tools, and try to break them
(I am always trying to break them)
new tools in hand, I do a bunch of new things I specifically couldn't do before, don't care if any of them work
etc
the result of this loop is a toolset that works pretty well. the result of this loop does *not* include knowledge of what's actually going on.