When I read Will Saletan’s big story about anti-GMO movement, I became convinced that the anti-GMO case is pretty unscientific. So it was a surprise for me to see N.N. Taleb, the “Black Swan” and “Anti-Fragile” author, assert in his Taleb-ian way that quite the opposite was true, and everyone who was not against GMOs was either a dupe or a shill. I became curious to see what his case against GMOs was. And so I came to a paper on the precautionary principle that Taleb and co-workers have put up on the ArXiv. Stylistically, the paper is vintage Taleb: frequently coining catchy terms (albeit sometimes for already described phenomena), the talent for saying something that somehow is both obvious and counter-intuitive, blunt disparagement of people who disagree, эпатаж and безапеляционность (since these Russian words are clearly borrowings, it might be that there are analogues for them in English, but I don’t know them). Apparently the paper is being prepared for publication. When it gets published, I’d like to know where, because the journal that publishes this genre of paper probably publishes a lot of interesting stuff that is accessible for a lay reader like me.
The gist of the precautionary principle is this: cost-benefit calculations aren’t a good way to assess risk if one of the potential effects is total ruin. Or rather, total ruin is an absorbing barrier – that is, you can’t recover from it, it’s forever – and thus the “cost” is effectively infinite. Thus we should not do things that have total ruin as a potential effect, even in cases where most of the time the action is clearly beneficial. A crucial point is that uncertainty about outcome only strengthens this intuition. Imagine a graph of distributed outcomes that is your best guess for what might happen, with ruin off to one side somewhere. Uncertainty about outcome just spreads that guess function, and thus makes ruin more likely by your estimation. This applies, for example, if we think a certain amount of global temperature rise would cause total ruin, but are not very confident in current predictive climate modelling. People who are less certain about their predictions should be doing more to avoid ruin, rather than being more in favour of doing nothing.
But all of that’s the easy part – the hard part is figuring out what has total ruin as a potential effect, given that predictions (especially about the future) are famously hard. The answer of course, is many things do, but for most things, it’s vanishingly unlikely. What separates ruin that we should care about from ruin that we should ignore? Taleb et al. point out that global ruin has to do with the risk of spreading, and the risk of spreading hinges on the independence of events. Thus (to use the Taleb et al. example), a bunch of nuclear power plants that could melt down and destroy the environment in their vicinity do not lead to risk of total ruin, because the meltdowns would be independent events, and having them all happen at once is exceedingly unlikely. However, a nuclear war could have total ruin as an effect, because in this case the nuclear bombs are not independent of one another – the first bomb makes the second bomb much more likely, etc. The central limit theorem protects us from ruinous effects of many independent events that could all go wrong, but not from dependent ones. Taleb’s big contribution to the popular discussion as I understand it is to notice that financial transactions are not independent, but people treat them like they are – and that is why they are poorly prepared when market crashes, etc. occur.
But all of that is the easy part of the hard part – the hard part of the hard part is figuring out whether the risk of spreading is high for phenomena where a mechanism of ruin is not well established, such as GMOs. Taleb et al. point out a distinction between thin-tailed distributions (which mean that the underlying phenomenon doesn’t have a high risk of spread) and fat-tailed distributions (which mean it does). While this seems correct, it doesn’t really answer the question – because what are these distributions distributions of? Unless you actually have some idea for what the risk is, graphing a distribution of outcomes is not a trivial step. Another distinction they make is between changes that occur “bottom up” (which do not have high risk of spreading) and changes that occur “top down” (which do). The distinction in this context confuses me somewhat, because it’s clearly not meant to be literal (otherwise, all government actions lead to ruin and should not be made). They do place GMOs in the “top down” camp and artificial selection in the “bottom up” camp, but this seems more naturalistic fallacy than anything that’s well substantiated. Certainly we know that catastrophic risk of spread can exist in organisms derived from artificial selection (Panama disease) or even in the absence of any change in organism geno- or phenotype (pine beetle infestation, Dutch Elm disease).
GMOs can lead to monoculture, and in a monoculture, the risk of spread of any negative outcome is very high. However, monocultures present a risk to a given species. To what extent they present a risk to the biome as a whole has to do with the extent a given species is connected with others, thus the extent to which the destruction of different species is actually dependent rather than independent. This is a subject that requires further study, and I join with the authors of Taleb et al. in saying that I think this is urgent. Furthermore, that the most important studies for GMO safety are not toxicological or molecular genetic but actually ecological – is I think a somewhat surprising and important conclusion. However, because I strongly suspect the interdependence between at least some species to be quite weak, I am not convinced that GMOs present risk of ruin.
Thinking about (and trying to prevent) potentially civilization-ending catastrophes is incredibly important and also very interesting. Taleb et al.‘s precautionary principle paper gives a way to focus this thinking. But to me, they are not sufficiently convincing to turn me against GMOs.