Why ecosystems ?

This project draws inspiration from natural systems to develop creative computer software. In this research project we are focusing on the concept of the ecosystem to motivate alternative ways of conceptualising the design and realization of creative systems.

Mimicking natural processes isn’t new. The field of evolutionary computing successfully uses the metaphor of Darwinian evolution to solve problems in optimisation, search and learning. Other biologically inspired techniques such as swarm optimization, stigmergy, Lindenmayer systems and artificial immunology have similarly been applied in engineering and artistic domains. This project develops the metaphor of the biological ecosystem in a similar way that expands the core ideas of evolutionary computation, integrating evolution with other self-organising processes. The ecosystem as a conceptual model of the workings of biological systems is particularly interesting, as it focuses on processes that drive adaptation and diversity, characteristsics that are of prime interest to generative artists.

Some properties of ecosystems covered by this research include: niche construction; organisms as ecosystem engineers; resource recycling; trophic levels; homeostasis and bistability. These topics are well studied in biology and in recent years have been the subject of study as systems in Artificial Life. Some features found in ecosystems (co-evolution and island models, for example) have already demonstrated themselves to be useful in evolutionary computing by solving problems difficult for simpler algorithms such as the Simple Genetic Algorithm (SGA).

Why would ecosystem metaphors be a useful way to think about creative processes? Our reasoning is that evolution is a process capable of discovering interesting and appropriate novelty, i.e. it is creative. The diversity and complexity of life on Earth serves as inspiration and evidence for this. A key research question that biologists face is how this novelty and diversity arises. Our challenge is to understand how a process that produces novelty in one domain (biological life) can be successfully applied in another (generative art and design).

Organisms adapt to and modify their environments. An environment consists of biotic and a-biotic elements for our purposes. In order for appropriate novelty to arise, we need to design our environments and the interaction of their elements successfully in a creative context.

The use of evolutionary methods in relation to art, design and music has thus far almost exclusively focused on two methods:

1. trying to craft a machine-representable fitness function for human aesthetics, or
2. the Interactive Genetic Algorithm,  (aesthetic selection), whereby a user performs the fitness evaluation according to aesthetic (or other subjective) criteria.

While these schemes have had their successes, we argue for a completely different approach, based on our intuition that (i) is too difficult a problem and (ii) is too limiting as a creative technique. The ecosystem methods we are researching do not solve these problems, rather they re-conceptualise the creative process to become one of designing ecosystem components and their interactions as opposed to fitness functions or aesthetic selection systems.

Our methods are not general purpose solutions to arbitrary creative problems in the way evolutionary computing algorithms can be applied to solving a wide variety of engineering problems. Rather they promote a way of conceptualising creative systems and processes that will hopefully lead to rich and diverse creative outcomes.

Links to:
Foundation papers