Methods for open-box analysis in artificial development
Adrian Grajdeanu · 2007
Even a developmental system with very simple building blocks can evolve significantly complex artifacts. Understandingsuch complexity poses a significant challenge. This paper shows how various investigative methods that are typically used in biology, can be transferred and used in an artificial development context. As an instance of evolved complexity, a self-repairing artifact is analyzed using the following methods: ablation of environmental features, chemical concentrations monitors, in silico subsystem simulations, gene knock-outs, and modeling of the gene regulatory map. A number of mechanisms governing size-regulation and self repair are uncovered, such as: subtle timing of gene activations, stable regulation based on attractor points, opportunistic use of the environment and information content replication.