Rube Goldberg invention image www.rubegoldberg.com via Wikimedia.
The more I learn about it, the more Nature looks like a Rube Goldberg machine: energy and materials flow through convoluted paths to accomplish something seemingly simple. Take the process by which a leaf decomposes. The sheer number of biogeochemical processes involved in the use of that leaf during decomposition is amazing: consumption by fungi and bacteria, use as shelter / habitat by soil nematodes and other microscopic organisms, consumption at a macro scale by larger soil organisms like earthworms, beetles, and crustaceans, use by natural geochemical processes in soil formation or water storage, and far more that I don’t know about (and that I’m sure even researchers in the field are still discovering).
A wild ecosystem involves countless intricate relationships, some essential to the functioning of the broader ecosystem (e.g. those between keystone species and others). Many of the relationships in such ecosystems produce yields of energy (and thus life) for creatures that have no immediate use to humans, and may even be pests on some level.
In many of our human systems, however, in the name of efficiency we engineer out the middle steps. Obvious examples of this abound in the industrial food system. An industrial CAFO is a bastion of efficiency, of a narrow sort. Food arrives (from the outside, typically) in a highly energy-dense and processed form, ready for conversion by the machine-animals from carbohydrates into meat and other products. Any waste products that can be used by humans are siphoned off during this process for sale; any that can’t be used are sent to waste lagoons and the like. The process is linear and of low complexity — food comes in one end, meat and waste on the other.
Consider how it’s often found that eating a fruit delivers vitamins more efficiently than taking a multivitamin. What in the mimicry of the natural vitamins is the multivitamin tablet missing? Instead of opting for the complexity of the fruit, we get orange tablets that can be churned out much faster and cheaper than growing oranges.
The problems inherent in such a way of thinking about and working with Nature is well known. What’s a better way of approaching it? It’s often been argued that the solution is to mimic nature. (This is indeed the approach taken by several agroecology systems, like permaculture.)
But there’s something missing here — and maybe the problem is inherent to some (but hopefully not all) attempts at biomimicry. Suppose I were to build a computer system that tries to learn from Nature. We spend some time analyzing how the natural system works, but to simplify matters we use the usual techniques of scientific reductionism and make the natural system much simpler than usual. Then we take a subset of the natural system and simulate or mimic it. Have we captured the right parts?
More than just capturing the relationships between the parts correctly, such a reductionist approach is at risk of opting for efficiency over resilience or even sacrificing both efficiency and resilience. That is, such an approach could confuse a Rube Goldberg machine that mimics Nature for the real thing. One of the hallmarks of a Rube Goldberg machine is not only its complexity but also its fragility — its lack of resilience. If even one step along the way fails, the whole system fails to achieve its objective, and is not self-repairing.
Thus there might be an important distinction between inherent complexity and apparent complexity. That is, if biomimicry is an important approach to solving problems to meet human needs in a more sane way, we need to be able to differentiate between the Rube Goldberg machine and a bona fide web of life. While Nature might look like a Rube Goldberg machine, it has inherent complexity.
I think this distinction between inherent and apparent complexity arises is many contexts. Consider the subfield of mathematical topology known as knot theory, which is about the study of, well, knots. A circular piece of string might be of low apparent and inherent complexity — the “unknot”, which is just an open loop. Take that string, jumble it in your pocket, and take it out and lay it flat on a piece of paper. You can write labels for crossings that you see using dowker codes, and may arrive at the conclusion that the knot is complex — that it has many crossings (that, presumably, could be hard to untangle). However if you haven’t actually changed the loop of string in any way, and if you were to hold it in just the right position, it’d be clear that all you have is the unknot — that the inherent complexity is low despite high apparent complexity. In this context inherent complexity is captured in the concept of crossing numbers.
When we build a complex system using biomimicry — say the construction of a watercourse, selection of plants in a fruit-tree guild, design of a composting system intended to prevent phosphate loss, or any number of others — is the inherent complexity high or just the apparent complexity? How can we tell? Perhaps one easy way to tell is to break the system. If I identify, say, 10 places I could break the system, and break it in those spots, what happens? Does the system route around the failure, or does it fail catastrophically due to my actions? If a leaf is decomposing and there are no appropriate fungi present, bacteria will get the job done, and vice versa; if neither are present, something else will take over. Only in degraded ecosystems — ones that have low inherent complexity — will decomposition not take place at all.
Perhaps then we should evaluate our systems not only by whether they mimic Nature well in the ways that they function but also how well they mimic Nature in the ways that they don’t.
