I've torn apart some of the code for Government, and am re-implementing it in a kind of neural network. This is definitely the way to go, for both the policy, and the back end simulation side. The simulation code isnt in yet, but the basic relationship stuff that controls voters reactions top policies is sketched out.
I got the current network to plot itself as a map of connecting neurons. Its a bit bizzare just how dense the map is at the moment. (see below)
I might include a mode of the game where you can fly around the simulations network and examine the insides ;)