What I had in mind was a system of interacting agents in which the A.I. agents automated certain aspects of their response with other agents but in which the controller for each agent could intervene on whatever basis seemed appropriate to that controller. The goal would be to define sub-sets of assumptions that would tend towards stability under a wide range of perturbations, including the desire of individual agent controllers, or players, to win.
Well I want a sound reproduction system with 100% accuracy. So ... GET TO WORK, dammit!
(LOL)
Your requirement are easy to 'spec' but - alas - a wee tad harder to accomplish.
;-)
