I have all of the policy data for the game converted to the new flexible function format for evaluating effects. I still need to convert some of the event data, and need to take a fresh look at the dilemma stuff.
I just started skimming an article on some AI website. It was explaining a very simple concept, basically a single neuron in a neural network, but it threw around lots of equations and phrasess like "perceptron convergence fomula" which makes it all sound more complex than it is.
A lot of game AI is filled with buzzwords and people being very academic about it. My approach has always been far simpler and more practical.
The way I see it, what works.... works, and it really doesn't matter whether you are using 'zyminskys inverse consciousness paradigm' or just making it up as you go along. After all, I'm writing games, not academic research. (there is a big difference in priorities between the two approaches, something some game studios often forget).