Abstract: Whereas information-theoretic top-down inferential methods can often successfully predict static probability distributions over microvariables when state variables are constant in time, such methods generally fail in dynamic systems. Building on the observation that complex systems often exhibit bi-directional cross-scale causation, we combine information-theoretic top-down statistical inference with agent-based (bottom-up) process modeling to construct a new theory of complex systems dynamics. The theory predicts the time evolution of both state variables and probability distributions over microvariables. Unexpectedly, analytic expressions for the time evolution of Lagrange multipliers result, thereby allowing for rapid solution even in high-dimensional systems. A worked-out low-dimension ecological example illustrates theory structure. Examples of possible applications of the theory to complex systems with bi-directional causation arise in non-equilibrium chemical thermodynamics, epidemiology, economics and ecology and an exothermic, gas-phase, oxidation reaction in a calorimeter could provide a test of the theory.
Speaker
