Detecting non-trivial computation in complex dynamics

Abstract

We quantify the local information dynamics at each spatiotemporal point in a complex system in terms of each element of computation: information storage, transfer and modification. Our formulation demonstrates that information modification (or non-trivial information processing) events can be locally identified where "the whole is greater than the sum of the parts". We apply these measures to cellular automata, providing the first quantitative evidence that collisions between particles therein are the dominant information modification events. © Springer-Verlag Berlin Heidelberg 2007.