Abstract. Many complex systems can be characterized at two well-differentiated levels: the micro-scale (e.g., growth rates of organisms, incomes of individuals) and the macro-scale (e.g., ecosystem productivity, GDP). In steady state, probability distributions over micro-scale variables can sometimes be predicted from knowledge of the static macro-scale state variables using the maximum information entropy inference procedure (MaxEnt). If the state variables are changing in time, however, as in a disturbed ecosystem, the constraints imposed by instantaneous values of the dynamic state variables fail, empirically, to accurately predict instantaneous micro-scale distributions. The dynamics of such two-tiered systems are especially complex if the macro-scale variables both influence the dynamics of the micro-scale variables and are also sums or averages over them. I present a hybrid, mechanism-plus-MaxEnt, theory to describe the dynamics of disturbed systems in which reciprocal cross-scale influences occur. A toy model illustrates general concepts and generates some of the rich behaviors emerging from the theory.
Noyce Conference Room
US Mountain Time
John Harte (University of California, Berkeley)
John HarteExternal Professor