Sporns is providing a measure of functional complexity of cognitive networks. This involves looking at statistical measures of integration and segregation. He suggests that brain-like networks provide an optimal mixture of these to achieve really good complexity.
We were talking about this in ALife class, in particular how if complexity is in some sense a measure of covariance, then using complexity as a fitness function migh push an ALife system towards intelligence, in the Pavlovian sense. I've been trying to figure out how you might create a physics for an ALife system that drives the system towards complexity in the way that Swenson and Turvey suggest the 2nd law drives the earth towards increasing complexity.
Are these approaches different? Is "using complexity as a fitness function" the same as "devising a physics with complexity as an non-epiphenomenological* law?"
* yes, I know this is a stupid word, but I don't know the proper one.
http://www.indiana.edu/~cortex/publications.html