Depends what you study…

by aepxc

Simple, bounded systems can be examined reductively – with controlled experimentation reducing the observed complexity to a smaller number of generative rules. Parts of larger, complex networks should really only be studied descriptively – as in “this is the most detailed available account of this part of the network”.

The former can allow one to deduce the best way to act in the future. The latter is best for finding solutions through trial-and-error.