Emergence of Observational Hierarchies in Natural Evolution

Hauhs, Michael & Lange, Holger

The starting point for our work are (models for) forest ecosystems. We seek to integrate two sources of knowledge about such ecosystems: the scientific approach from the viewpoint of an exo-observer (one that is detached from the system and has theoretically unlimited observational capacity) and the management approach for utilization (one that has a common history with the managed system and includes the possibility of an endo-observer). Within the scientific approach, these ecosystems are often regarded as being among the most "complex" systems that can be abstracted as objects whereas their practical management for human utilization (including regular interferences of an entangled endo-observer) sometimes has allowed economically reliable predictions over time scales from years up to a century.

Each approach alone seems to be insufficient to solve the theoretical and applied problems of contemporary ecology. Practical experiences on one hand cannot be extrapolated and becomes invalid under new environmental conditions, whereas ecosystem research on the other hand has not come up with a single example where such experiences could be derived from an theoretical understanding of ecosystems. We conjecture that this failure has theoretical rather than technical origins. Its solution, however, will not only require a broader observational basis including the perspective of endo-observers, also the abstractions from this basis may change. Our metaphor for modeling ecosystems is shifted from energy and matter turnover to a computational problem.

To this end we suppose that information and energy are the primary irreducible entities driving self-organization and structure formation. The history of natural evolution, from the creation of matter until now, may be understood as steady change of phases where the symbolic aspects, and phases where the matter aspects dominate. We argue that these phases are related to classification and construction aspects, respectively, and hypothesize that exploration of unknown structures by "agents" always implies this alternation. However, the structures are also partially created by the agents, bringing self-reference into play.

The ("phenotypic") appearance of agents can differ drastically depending on the phase (or the age of the universe) considered. Each agent type is "genotypically" characterized by a certain maximal information processing capacity, eventually leading to an information crisis when no more structures are exploitable. These crisis induce a hierarchy of ("genetic") codes, with increasing capabilities for hypothesis building (the agent as endo-observer). These hypotheses are constituted as matter (phenotype construction) and tested; successful hypotheses are characterized by persistence (the agent as interactive constructor). We try to clarify the hierarchy building or emergence of new levels within it by analyzing the relationships among and roles of a number of dualisms: genotype/phenotype, classification/construction, endo/exo point of view, local observer/global context. In a self-referential universe their respective explanatory power depends on the "situation" and on observer perspective. However, it becomes possible to find and test analogies between the corresponding computational crises that must occur in the evolutionary history of such an universe.