See Brian Train's Algeria. I think that the simulations and games communities have been so dominated by mathematics-based thinking that language based approaches to the problem have been neglected. Chris Engle's Matrix Games seem promising in this regard. The "End State Problem" is indicative of this. Properly writing desired end states is always a linguistic task. "Victories" and outcomes are ultimately defined in langauge. Perhaps games used to the end of better understanding COIN should be accordingly language-based.