The exchange indeed seems like a convenient choice as a heuristic, but in reality I feel this question is more analogous to how you would model ecological systems, chemical reactions, metabolism or a computational collective intelligence... the work of Tadeusz Szuba or Doyne Farmer seem to be on the right track.
That's very interesting. I modelled
wicked systems numerically for several organisations and businesses a couple of years back, but I never thought about (or felt the need for) analysing their nature mathematically - strictly mathematically. It still seems nobody is trying to do it. We did solve quite a lot in physics for similar in some aspects, but obviously simpler, open quantum (not relevant here) systems (that's the microscale picture, and in macroscale they give non-equilibrium thermodynamics). I never thought of it as a problem mathematically similar to the optimisation if a cost function when training a deep neural network (a non-convex function) using gradient descent methods.