Managing Multiple Interacting Adaptive Systems Via Game Theory

We consider the problem of a "manager" who controls some variables y that affect a set of interacting, goal-oriented agents. We model such agents as players in a noncooperative game, with y being parameters of that game. So the manager's problem is how to change the game so that the resultant equilibrium behavior of the players optimizes a utility function of the manager. This can be viewed as an extension of the problem of mechanism design, to allow bounded rational players, to exploit knowledge concerning players not directly affected by y, and to allow arbitrary types of y. We introduce a solution to this problem and illustrate it with computer experiments.