Tuesday, August 07, 2007

The Self-Made Tapestry - Philip Ball

I'm currently in Melbourne, working with the CEMA group at Monash Uni. Among other things, we've been talking about artificial ecosystems, growth, morphogenesis and self-organisation - and I've been working on a generative art piece that has had me casting around for models and mechanisms. Jon McCormack passed me Philip Ball's 2001 book, The Self-Made Tapestry: Pattern Formation in Nature.

The book looks at morphogenesis - the self-creation of form - in physical and biological systems, and computational models. It updates the work of D'Arcy Thompson, whose 1917 book On Growth and Form showed that natural forms could often be explained as products of dynamic physical interactions as much as adaptation or evolution; for example, a seashell spiral emerges as a result of the growth rate of the organism living inside it. Ball takes a similar approach to morphogenesis, in that he treats physical systems and living systems as fundamentally interlinked, often examining the material mechanisms in biological form. Unlike Thompson though, he has a modern reservoir of complex-systems science, biology and physics to draw on. In a great piece of pop-science writing, Ball knits together a wide range of work under a useful set of headings, and the text is full of enticing illustrations. The image below is by Eshel Ben-Jacob, whose bacterial growth work is featured extensively in the book.


It's a treasure trove for the generative artist/designer; flick through until you find an illustration that catches your eye - maybe a bacterial growth form, a reaction-diffusion system, or fracture patterns - and then read up on the morphogenetic models involved. Generative clip art? Not quite; Ball's text explains the principles and processes clearly, but links them organically to each other through systemic properties: symmetry breaking, bifurcation, fractal dimension and so on. While there are verbal descriptions of plenty of generative algorithms, understanding them really requires coming to grips with the underlying models and their shared characteristics.

Ball also talks explicitly about the use of computational models, which play an important role in the book. This is especially important for anyone using the models as (generative) ends in themselves, rather than empirical devices. Ball clarifies the scientific sense of "model" as something selective and partial, rather than representational or exhaustive: there are plenty of things that such models omit, either because it's too hard to include them, or they don't seem to influence the outcome. Along the same lines, some (maybe all) phenomena can be modelled effectively in several different ways, using different assumptions and techniques. Conversely, the interrelations between morphogenetic systems often come down to shared models, cases where pattern formation in different domains (say, fracture patterns and plant growth) can be modelled using similar, often very simple, techniques.

I've been critical in the past of the simplistic models used by generative artists - but also argued that generative art's ability to play with models (creatively and intelligently) is what makes it interesting. So my recommendation here is half creative and half critical; in other words it's got eye-candy (bacterial eye candy even) as well as substantial model-y goodness.

No comments: