The Mathiness of Nassim Nicholas Taleb
inexactchange.org
inexactchange.org
Namely, you claim (necessarily without proof) that a real-world phenomenon is analogous to a certain mathematical model. Then you prove some things about the model. Finally, you claim that the things you've proved about the model are also true about the real world. It doesn't matter how rigorous the mathematical steps in the middle are, the strength of this kind of argument is determined by the first step: how strong is the analogy between the mathematical system and real life?
In physics, we can build models where the analogy is very strong: when we try out new situations in both the model and in real life, we get the same answer in both. In (macro-)economics, we typically do something weaker, where we start with a set of models determined by intuition, and then choose the one that gives results most similar to real life in the situations we've already seen. Prediction of future events is really difficult.
It seems like the approach in Taleb's paper is even weaker. We choose a model based on pure analogy, explicitly disclaim the usefulness of evidence ("While evidentiary approaches are often considered to reflect adherence to the scientific method in its purest form, it is apparent that these approaches do not apply to ruin problems"), and then attack anyone who says that we've done something other than prove a mathematical theorem about the real world.
It's too bad, because underneath all of the "mathiness" is one of the strongest arguments against GMOs that I've heard (and I say that as a GMO-eater). It just should have been an opinion piece in the NYT rather than a research article on the arxiv.
However, he has failed to grasp, or educate himself sufficiently, that gene transfer between organisms is in fact a common occurrence in natural ecosystems [0].
Further, at this moment at least, the genes that are being transferred artificially are in general not genes that would cause havoc if they somehow became mobilized from the species into which they were introduced. By way of example, the protein Bt confers resistance to insect predation,it is originally derived from a bacterium that killed insects in a parasitic cycle [1]. Plants, generally, have hundreds or thousands of ways of resisting insect predation, and adding one more to the mix is not going to devastate any ecosystem, at least not in the long term.
Nature has a way of eliminating failures; humans have a way to offload failures off to someone else.
The effective aspect of the scientific method that separates it from scientism is falsifiability. The comparison for falsification here is "GMO farming" versus "status quo farming"; nowhere is nature involved.
Of course, GM-technology is a huge area of tools, tools which are being used in all kinds of situations and outcomes: modifying bacteria to create insulin or washing powder ingredients (both of these have been done on an industrial scale for decades now); modifying plants to have a better defense against various pests (being done with Bt cotton and soy IIRC), or just to glow in the dark (there's a kickstarter), etc.
These are all more or less "good" outcomes of GM, but you could also use GM for "bad" - you could make maize express a human neurotoxin and spread that plant in order to attack a country (a not very directed attack). The only thing that's stopping you is that it's still very hard, expensive and time-consuming to do so.
You are assuming that we have sufficient comprehension of these genes and their effects to make such claims definitively. And while we may have a very high degree of confidence in that, say enough confidence to entrust my life, your life, or hundreds of individual lives - entrusting the wellbeing of the entire planet requires a different scale of rigor.
And that, I think, is the essence of Taleb's argument. 99.99% sure might be enough to get a drug approved, and it well should be, because only some fraction of the population is at risk. But because DNA is inherently self-propagating, a mistake can have extremely far-reaching consequences, and as such the tolerable level of risk must be orders of magnitude lower.
This has been Taleb's essential, consistent thesis throughout his career. That humans are bad at assessing the probability of relatively unlikely events, and that when those events are extremely catastrophic, that cognitive deficiency can be disastrous.
"Even if we can agree that some things are natural and some are not, what follows from this? The answer is: nothing. There is no factual reason to suppose that what is natural is good (or at least better) and what is unnatural is bad (or at least worse)."
https://en.wikipedia.org/wiki/Appeal_to_nature
Gene transfer may have already screwed up the planet once:
http://www.smithsonianmag.com/smart-news/how-single-act-evol...
In the case of the LHC, we know that particles collide at the speeds planned all the time, yet this does not lead to black holes erupting everywhere.
In the case of the gene transfers currently done, we know precisely what they are and they are already part of the environment, having been derived from other natural organisms. The risk of the current crop of engineered plants becoming super-weeds that choke the planet is not zero, but also not plausible from a biological perspective.
That someone could engineer a super weed is of course possible, as would be engineering the flu virus to produce cholera toxin. So, precautions should indeed be taken when releasing engineered organisms.
This case is not analogous because in this planet's history there have been 6 different extinction events. So in order for it to be analogous, they would have to say that we know a few times in the past this has caused a black hole, but this is totally different.
Passing genes from one thing into another can very easily have unintended consequences as they do in nature. The link above is an example of what can happen. By the time you figure it out, perhaps it will be too late. The effects might be local or they might be global, there is no way to predict these things and we can not say with certainty that they don't happen, because we have more than one example of where they do. In fact human intelligence may be an accident of horizontal gene transfer. These are not games that are worth playing in my opinion. You are what you eat. If that stuff it in your food supply, it just might end up in you and the bacteria and other life that lives in you.
Another really big problem is dependence. Once you come to depend on these things, just like we are now with carbon producing technology, you can not pull the plug. Even if you know that it's causing harm, the social mechanism needed to pull the plug don't exist in free society and if they do they are very slow moving. That's another big reason to not go down this road, because it's impossible to turn back because of all the people that will die because you can't produce the expected crop yields.
There are two mass extinction events that have a clear biological origin. Minor extinction events are a permanent feature of evolution.
The first is the evolution of photosynthesis causing the Great Oxygenation Event[0]. In this, cyanobacteria converted CO2 into water and oxygen on a massive scale. This lead to a cooling of the planet due to reduced CO2 levels and the precipitation of iron oxides into massive deposits worldwide, today visible as "banded iron formations" [1].
The second mass extinction with a clear biological origin is the current one, where Homo sapiens is converting a massive quantity of carbon previously sequestered during the carboniferous back into CO2 and warming the planet as a result. In addition, this species destroys the habitats of many other species.
Aren't all such (big-)data models (e.g. predicting natural disasters, stock-markets, global-warming, effects on GMO, ...) an attempt on measuring "real life"? And so can not be measured accurately because as soon as you take your snapshot data points and events inside the model will have moved on to a new state? Not that we shouldn't try to build these theories but accept them as theories or "soft-science" and not facts.
As soon as we have simplified a complex hard to understand system into a chart we tend to ignore that these "facts" represent at best a snapshot of reality (one version of it) from a specific point in time but certainly not the future.
For example, Taleb is already measuring earthquakes by the energy released. This is how "magnitude" is defined for earthquakes. Thus statements like "For instance, in the coffee cup example, Taleb measures the intensity of an earthquake by its magnitude. What happens if we measure an earthquake instead by the amount of energy it releases?" make me conclude that the author of the critique understands neither earthquakes nor Taleb.
As for the overall claim of fuzziness, it's probably best to let readers make their own judgement. Appendix C here is Taleb's "Mathematical Derivations of Fragility". While one can dispute where it should be applied, it's far from devoid of detail: http://www.fooledbyrandomness.com/pp2.pdf
No, as the article notes, the measurement of magnitude (in the Richter scale) is proportional to the base-32 logarithm of energy released (its defined as proportional the base-10 logarithm of seismic wave amplitude.)
Taleb's use of the coffee cup example to suggest a disproportionate impact of large events vs. small events -- suggesting that the real relative impact of a magnitude 6 quake compared to a magnitude 1 quake is disproportionately greater than their size because one magnitude six quake does more damage than six magnitude one quakes -- is vacuous from its structure because you can arbitrarily choose scales of measurement to make the impact proportional or disproportional in either direction, and ludicrous in its particular choice of measurement scales since, well, I mean, who would expect an earthquake releasing 1 billion times as much energy to do only 6 times the damage?
Taleb makes a comparison that sounds significant because most people don't know what the numbers involved mean.
> No, as the article notes, the measurement of magnitude (in
> the Richter scale) is proportional to the base-32 logarithm
> of energy released (its defined as proportional the base-10
> logarithm of seismic wave amplitude.)
There are close to 8,000 micro-earthquakes daily
on planet earth, that is, those below 2 on the Richter
scale —about 3 million a year. These are totally harmless,
and, with 3 million per year, you would need them to
be so. But shocks of intensity 6 and higher on the scale
make the newspapers. Accordingly, we are necessarily
immune to the cumulative effect of small deviations, or
shocks of very small magnitude, which implies that these
affect us disproportionally less (that is, nonlinearly
less) than larger ones.
Even though the sum of the magnitudes of the smaller
earthquakes is equal to the magnitude of the larger one,
the coffee cup is damaged much more by the big quake than
by the small ones. This is what Taleb means when he
describes the coffee cup as fragile.
> suggesting that the real relative impact of a magnitude 6
> quake compared to a magnitude 1 quake is disproportionately
> greater than their size because one magnitude six quake
> does more damage than six magnitude one quakes
Now, I will grant you that Taleb may well have mistakenly assumed that the measure of earthquakes went up 10^{1} per number of magnitude rather than 10^{1.5}, as this would explain why he chose to use the specific numbers he did. It would be a fair critique to point out this out to him, and I'd be interested to see his response. As you say, 10^{9} would require hundreds of years of earthquakes, and thus is not a particularly compelling example.
> Taleb makes a comparison that sounds significant because
> most people don't know what the numbers involved mean.
(Separately, this is my opportunity to ask: Who are you? What's your background? You consistently impress me with the quality of your comments on HN.)
On the risk side, GMO proponents like to argue that the technology is safe because we haven't found any ill effects. But what they are really saying is that we haven't found any so far. I remember studying genetics and being told about junk DNA, even Francis Crick, co-discoverer of DNA’s double-helical structure, suspected it was “little better than junk.” Since then this opinion has changed. We find out new things about DNA all the time, it is now thought that previous generations experiences can change decedents. Reading the recent articles about Chinese scientists using CRISPR to modifying embryos they started with 86, 71 survived, and 28 were successfully spliced, not really a precise operation. With this in mind I find anyone who says there is no risk is not being intellectually honest. The question that comes up is what are the chances that we are Marie Curie working with something that ultimately harms us?
In terms of assessing benefits I like to take the example of X-ray technology. From the 1920s to the 1970s a machine known as a shoe-fitting fluoroscope was used to see if shoes fit feet (especially children's feet). With the growing understanding of the harmful effects of radiation these machines were phased out, although never banned. That doesn't mean we don't x-ray people, we do, if you break a bone or go to the dentist the benefit of diagnosing the injury or cavity is considered to outweigh the damage from the radiation. For people in Western nations with plentiful food supplies the macro scale benefits of GMO seem limited, we weren't starving to death or going blind before GMOs. However if my family and friends were going blind due to Vitamin A deficiency the benefits would weigh more heavily on my opinion.
I don't claim to be an expert in this area but am interested in being provided with a more neutral analysis to help me form an opinion.
Deconstructing the Precautionary Principle is not "misinterpreting" PP. The entire notion of "mathiness" put forth by the author is that PP is too wiggly, and thus open to whatever vague interpretation is convenient. That is the meat of what Merberg is saying, and handwaving it away isn't an argument.
I put my foot in my mouth all the time, but I put my name to it. Taleb is interesting. Are you saying that the other account is literally Taleb, and that he's known for astroturfing campaigns? I wouldn't discard such a proposition, but I'd prefer hearing it from someone non-anonymous.
(For all I know, you're just an account created to extend the thread.)
Hardly.
Having read his books, he's a bright guy often with compelling ideas. If he would leave them as thought experiments and appeals to common sense honestly I think his arguments would be more effective. His need to formalize everything mathematically, for a reader with a moderately sophisticated mathematical background, ends up undermining his argument.
Any help?
>>We present a non-naive version of the Precautionary (PP) that allows us to avoid paranoia and paralysis by confining precaution to specific domains and problems. PP is intended to deal with uncertainty and risk in cases where the absence of evidence and the incompleteness of scientific knowledge carries profound implications and in the presence of risks of "black swans", unforeseen and unforeseable events of extreme consequence. We formalize PP, placing it within the statistical and probabilistic structure of ruin problems, in which a system is at risk of total failure, and in place of risk we use a formal fragility based approach. We make a central distinction between 1) thin and fat tails, 2) Local and systemic risks and place PP in the joint Fat Tails and systemic cases. We discuss the implications for GMOs (compared to Nuclear energy) and show that GMOs represent a public risk of global harm (while harm from nuclear energy is comparatively limited and better characterized). PP should be used to prescribe severe limits on GMOs.
Leaves little doubt. You can't call him a truth seeker - he is a quack with some talent for wooing people.