Manipulating artificial ecosystems

Artificial ecosystems extend traditional evolutionary approaches in generative art in several unique and attractive ways. However some of these traits also make them difficult to work with in a creative context. This paper addresses the issue by adapting predictive modelling tools from theoretical ecology. Inspired by the ecological concept of specialism, we construct a parameterised fitness curve that controls the relative efficacy of generalist and specialist strategies. We use this to influence the population’s trajectory through phenotype space. We also demonstrate the influence of environmental structure in biasing evolutionary outcomes. These ideas are applied in a creative ecosystem, ColourCycling which generates abstract images.