There has been some good debate in a Finnish GMO-Awareness facebook group recently regarding both the pros and cons of GMOs. This has prompted me to participate as well, because I feel I have something specific to contribute to the discussion: the explanation of systemic risk.
All living organisms are complex adaptive systems individually and also on an environmental and a species level. In other words, complex adaptive systems can be nested within larger systems, and also contain smaller systems within them. A cell in itself is such a system, but so is an organ consisting of multiple cells, and an individual body consisting of multiple organs and other tissues. A collection of bodies might be called a tribe, a pack, a clan, or an organisation. In an even larger scale we can talk about species and the entire biosphere of Earth.
What all these systems have in common, among other things, is non-linearity. This means that a small parameter change in one part of the system can have cascading effects and result in vastly different outcomes in other parts of the system, or in other connected systems. A classic example of this is the flap of a butterfly in Brazil causing a tornado in Texas, as shown by Lorentz [PDF] using computational weather simulations: a minuscule change in parameters resulted in a difference between sunshine and a tornado due to non-linearity. Another example would be the way genes work. The genetic instructions for building a human body instead of that of a chimpanzee has about 4% difference, yet the outcomes are very much unlike one another, and when searching for ultimate causes that 4% difference is also underlying everything man-made that has ever come to existence and will come in the future. After all, chimpanzees are not programming computers, designing airplanes, or composing symphonies.
So what’s the problem? Thanks to their complexity, interrelations and links to other systems, and non-linearity, we are utterly incapable of accurately predicting system-level risks. I will use GMOs as an example in this article, but this issue is not limited to them. What I am talking about is a property of complex adaptive systems, and therefore the same reasoning applies to all other contexts where these systems are concerned.
Changing a parameter within a system – say for example, changing a gene in a tomato to make it more resistant to a pesticide – can have effects elsewhere in the same system (a tomato), or even in other related systems (a human being eating the tomato). In fact, we have evidence of this: Dwarf wheat is a result of selective breeding and has become the dominant species globally since 1970s, as it provides a better yield than traditional wheat strains. However, what made dwarf wheat dwarfish also had an effect on the genetic structure of the grains we eat, and dwarf wheat has been linked to the growing numbers of coeliac disease and various other autoimmune conditions.
These non-linear effects can also go a long way: a worm may eat the tomato, get eaten by a bird, the bird poops on the ground, and soil bacteria eats the poop and starts dying en masse. Sounds pretty far fetched? Indeed the probability for such a chain of events to take place is extremely small – but not impossible – and that is the entire point. To test for such an effect simply cannot be done using the scientific tools we have available today. Nor would it be a feasible investment of resources to test for every conceivable small probability outcome. However, if we calculate the impact of such an event, making a further assumption that the death of soil bacteria would result in severe nutrient depletion in the soil, we would be soon facing a global food crisis.
These kinds of high-impact system-level risks would not be so scary if they were contained locally, but unfortunately the trend seems to be going in the opposite direction; towards “optimised” laboratory-grown strains with a global reach. If we discover in, say, 40 years of time that a particular GMO crop causes birth defects in human babies, it would be a tragedy if the effect was local (e.g. within a country or a state), but on a global scale it would be a catastrophe. This brings us to why GMOs cannot be compared to naturally occurring mutations within a species: The natural ones are always locally contained, and it takes generations for a mutation (or an adaptation) to spread to a population, which gives other species and the environment (i.e. other systems) time to adapt.
In the end, due to the nature of complex adaptive systems the risks we are facing are huge and we have no means to predict them. What we are doing with GMOs is effectively playing a lottery: We are staking calculable short-term benefits against the incalculable probability of a devastating negative outcome.
Where this whole GMO debate goes to a morally shady ground can be explained with the agency effect: GMOs benefit first and foremost the corporations that develop, patent, and sell them for profit. Yet they do not carry any risk for these system-level negative effects. Instead, this risk is carried by us who buy their products, and largely by the environment (which should be of everyone’s responsibility). If a serious system-level issue becomes visible after a few decades, and we can confidently say that a particular GMO is the underlying cause, the developer of the GMO (who also made the most profit out of it) will not be held liable, nor do they have responsibility to compensate for the damages (if such would even be possible). This kind of environment vastly rewards risk-taking at the expense of others, and we the people end up paying the price.
Update on May 27th, 2014:
I have been asked here in the comments and on facebook about why put GMOs to a pedestal? Why is it different when compared to e.g. traditional methods of guiding the development of a species, such as selective breeding? Also, mutations taking place in nature are much more imprecise and potentially more significant than the ones done with GM techniques, so why should they be considered any less risky?
First, let me address why the risk imposed by GM is statistically and categorically different from the risks inherent in e.g. selective breeding. Naturally occurring mutations have been around for millions of years, yet life continues and as far as we can tell, these mutations in various species have not had sudden dramatic effect in the environment. If there is a statistical possibility for a mutation that has such a radical impact, the time frame of natural evolution suggests that these kinds of events have likely already happened multiple times. However, so far the only drastic change we know of regarding life on Earth has been the death of dinosaurs, and to my knowledge a natural catastrophe was the cause, not evolutionary change in some species.
As for selective breeding, it has also been around at least for tens of thousands of years, and in that time no drastic changes to the environment caused by selective breeding have happened either. In other words, we have tens of thousands of years of empirical evidence stating that this method is safe. With GM our amount of evidence can only be measured in years or decades. This creates the statistical difference between GM and the naturally occurring methods of genetic change, which affect more genes (shotgun approach) instead of the selected few as in the case of GM (scalpel approach, as one commenter eloquently put it). There is much more historical evidence suggesting that selective breeding is safe than there is for GM.
However, for all fairness I must say that selective breeding may have also become more risky than what our tens of thousands of years of empirical evidence suggests. This is because for the majority of history selective breeding has had only local impact. Now with globalisation however, the scale of impact can be much larger and happen much faster than ever before.
For understanding systemic risk I recommend the book Antifragile by Nassim Taleb. For a short version explaining risk and fragility there is this letter, which also explains how systemic risk is fundamentally different from the classical understanding of risk (classifiable outcomes and probabilities).
As for understanding “scientific” research and possible reported beneficial effects from GMOs (this also applies to all research on medical, nutritional, and public health issues), I recommend reading Richard Feinman here, here, and here.