I am frequently told that I must have a relevant, real-world question to address before I pursue research. This is sensible enough from a utilitarian perspective but, on the other hand, historians like Lisa Jardine have argued convincingly that technology, the piece of apparatus or model, frequently precedes “the question” and drives the biggest scientific discoveries historically. Take for example the 2012 publications of a new type of computational neural network called a convolutional neural network that showed beyond-human abilities in image classification. This model led to a flurry of publications and discoveries in a range of fields, most notably in medicine where a range of tools are being developed that are better than medical professionals at diagnosing disease from medical images. Clearly here the practical innovation came first and the medical breakthroughs followed.
So it is with the new scientific paper I and colleagues from the organic agriculture sector have recently published in the journal Agronomy for Sustainable Development. We used a model called a Boolean regulatory network. I first came across this model many years ago when I worked as a postdoc in evolutionary biology at Leeds University. This class of model was developed in the 1960s by a computer scientists called Stuart Kauffmann who presented it as a model of gene regulatory function. Kauffman believes that gross aspects of organic structure are responsible for some of the order we see in nature and that these evolutionary processes work alongside natural selection on an equal footing. He modelled the temporal dynamics of genes interacting with each other in a completely randomised Boolean regulatory network and showed that such systems have inherent order: they end in numerous different structured patterns of gene expressions such as regular cycling or constant expression of a subset of genes. Most provocatively, he showed that a randomised model network with roughly the same number of genes as the human genome displays the same number of stable expression patterns as cell types in the human body…
I suppose Kauffmann is what we might term an evolutionary “pessimist”. He deemphasises the primacy of natural selection in evolution and sees the direction of evolution as highly dependent on traits that have already evolved. It would be great (optimal even) if I could develop green skin, photosynthesise, and produce my own food, but I can’t because I don’t have the right evolutionary history and genetics for that to happen. This contrasts with what I will term “optimists” who tend to see traits of an organism as relatively unconstrained and optimal for their function. Optimists mainly work in the area of human and animal behaviour and their point of view isn’t ridiculous. The nervous system is extremely versatile and there are many examples of behavioural traits that are optimal or close to it, at least in the context of how organisms live in nature.
Optimists and pessimists have traditionally been at logger heads in post 1950s evolutionary biology and likely for reasons that aren’t wholly scientific. Scientists guard their areas of endeavour from subversion (so the following would never be publicly admitted by an evolutionary biologist) but likely there is a political dimension to this debate. Doesn’t evolutionary optimism tend naturally to the view that society is optimal and developed through optimal processes implemented by optimal humans? There are poor and there are rich; that’s just the way it is! Fossil fuel use developed because it is convenient in so many ways and we can’t do without it. And so on. My person interactions with evolutionary biologists from both camps tend to reinforce this idea of a political (or at least personality) dimension to this “optimist” vs “pessimist” debate and I don’t exclude myself from this categorisation.
So I read Kaufmanns magnificent book, Origins of Order, and then the need to answer someone else’s scientific questions (a key feature of postdoctoral research and, increasingly, tenured research) came in the way and time passed but these models stuck in my mind and, perhaps though unconscious processing, I realised one day that they could be used to model systems other than gene networks. Then I fell out with the high-ups and left the university system and applied for a bunch of jobs and was offered one in applied agricultural research (the area I wanted to work in) in the third sector and here I am today.
Agroforestry (putting trees in and around crops and livestock) is increasingly popular across the world, and England is in the process of getting its first agroforestry incentive scheme for farmers. Working in this area I came to realise that agroforestry models don’t include the really interesting and inspiring thing about trees, with the living things (birds, insects, plants) they bring with them completely absent from agroforestry modelling. Kaufmann’s models immediately sprung to mind. What if I represent each member of this ecological network as a node in a Boolean regulatory network. BRNs are very simplified and each node can only be high or low (not in-between), but that’s OK: we tend to think of crop pests as present or absent or crop yields as good or bad anyway. This is dialectical thought, formalised by the ancient Greeks and adopted by the statist communists millennia later. This dialectic structure of the Boolean model made me think of the disappointing lack of field data I have to parameterise the model and whether this could be overcome using expert opinion, as they do in decision analysis, and I was able to persuade several of the top experts in agroforestry to help out. I told them nothing about the study to avoid unconscious bias and got them to analyse the 128 possible unique states of my model ecosystem and predict the state of the ecosystem in the near future. This was plugged into my model and the dynamics of the system iterated over time using code. Miraculously, realistic ecosystem dynamics emerged. Numbers of pest predators rose when pest numbers rose, weeds proliferated when crops became diseased and withered away. This is the best bit of science: when you press the button and all the secrets emerge, and the thrill last no more than a few minutes. This is the rush we share with those that jump from mountain tops in wingsuits.
Then came “the question”.
A big EU project I was working on wanted to know if crop yield was more resilient in agroforestry or crop monoculture, so I ran this scenario through the model. Crop yield is not more stable in agroforestry but yield of arable crops is overall is higher with trees.
I don’t claim that our new paper is a “breakthrough” by any means but its innovative and unusual in its approach and potentially the sort of thing that could influence other to adopt the same methodology for more hard-hitting questions or practical applications. Producing unusual work like this is very much like organic evolution: messy, lengthy, and profoundly dependent on personal history. The UK state bemoans the lack of transformative research in the UK but at the same time values the scientist who can produce two dull papers over the scientist who produces one innovative paper in the same journal.