I would think that we have to consider the entire parameter space or model space of a reductionist view of biology as a space of models, then search for the smaller subspace of models with fewer, stiffer parameters. http://www.lassp.cornell.edu/sethna/Sloppy/WhatAreSloppyMode...