I was a witness to a small discussion on twitter today. It started with Biocon MD Kiran Mazumdar Shaw tweeting this article in the Economist which talks about a massive recent study that indicates that genetically modified crops have “widespread benefits” (disclosure: I’m yet to read that article).
— Kiran Mazumdar Shaw (@kiranshaw) November 15, 2014
Next, an anonymous tweeter called Sleuth Stock drew the attention of Nassim Nicholas Taleb, former banker, risk scientist and author of Fooled by randomness, The black swan and Antifragile, and a prominent opponent of GM crops to the tweet. This indicated a link to a paper by Taleb that indicates that GM crops have an embedded systemic risk that has not been exposed yet.
— Sleuth Stock (@sleuthstock) November 15, 2014
Taleb himself then responded to the tweet, with a snapshot of a paper that he is in the process of writing regarding why GM crops are dangerous. The paper is basically a lot of math, with some Greeks, and beyond the reach of this Resident Quant. It was this tweet, by Taleb, mentioning Kiran Shaw, Shekhar Gupta, Devinder Sharma and S Gurumurthy that drew my attention to this discussion.
— Nassim Nichol?Taleb (@nntaleb) November 15, 2014
Then Sleuth Stock responded, saying that people had ignored Taleb when he had talked about systemic risk in banking, and then people started worshipping him when Lehman happened. And now that he is talking about systemic risk in GM crops, we should listen. Bizarrely, Taleb retweeted this tweet (but then some people have the habit of retweeting any faint praise they get on twitter) .
— Sleuth Stock (@sleuthstock) November 15, 2014
First of all, by Taleb’s own explanations in his first “popular” book Fooled by randomness, that you’ve got a forecast right once in one domain has no bearing on your getting your forecasts right in another situation in another domain. While Taleb had very good reasons to call out the systemic risk in international finance, that the risks he cautioned against were actually borne out has an element of randomness to it. Which is why Taleb’s retweet of Sleuth Stock’s tweet is rather bizarre.
Now that that is out of the way, let us get to GM. Cutting through and ignoring all the math in Taleb’s paper, this is my perception of what the matter is about. The basic idea is genetically modified crops (modified as they now are, rather than by means such as grafting, cross-breeding etc. which has been practiced for millennia) carry an element of risk – we do not know what the probability that they are going to be harmful is, but it is yet to be (and never likely to be, according to RQ) proved to be zero. So there is a non-zero probability (which is known to be small, but whose quantum is unknown) that GM crops might be harmful.
Related to this, it is not known what the potential damage caused by GM crops could be if the risk embedded in them bears out. Given the experiments and trials so far, scientists have not been able to quantify the extent of the damage (however small the probability) that might be caused by GM crops in case of the risks embedded in them being borne out.
The expected value of the “trouble” caused by GM crops can be defined as the probability of trouble multiplied by the extent of the damage caused by such trouble. The trouble (pun not intended) is that neither of these two numbers are well defined. We know that the probability of trouble is small, but only have a very loose upper bound on it, and we will never be able to narrow down the precise probability. We know that the damage caused by the trouble can be potentially large, but we don’t know how large, and we don’t know with what probability the damage can be that large.
When you multiply a small unknown number with a large unknown number, where the order of magnitude of either number is itself unknown, the range that the product of these two numbers can take is rather large. This means that there is a small (and again unknown, but not zero) probability that the costs of introducing GM Crops might be higher than the perceived benefits (as measured by the studies referred to by The Economist). Taleb’s argument is that since such a possibility exists, and the extent of expected costs of GM crops is not known, we should not embrace GM crops.
While it is true that there is a small chance the costs might embrace the benefits, the simple truth is that we don’t know, and we never will know. The probability of the “trouble” is so small, and the possible damage so large that even with a large number of studies and spending long years on it, it is going to be impossible for us to get a handle on either of these two numbers, and consequently the expected value. So the expected value of the “trouble” from GM crops is a classic “unknown unknown”.
Given that this unknown unknown will never even become a known unknown, adoption of GM crops is going to necessarily have, albeit however small, a leap of faith. The question is if we should take this leap. If we are extremely conservative, as Taleb normally is, we would be wary of taking even this small a leap of faith. Given the amount of scientific testing we have gone through already, and the lack of information value in any further testing, though, and the facts of growing world population and the need for food and commodity supply to keep up, it would be well worth it to make the leap of faith. We should also keep in mind that such a leap of faith when it comes to GM crops is not so much longer than the leaps of faith we have taken while adopting many other scientific advancements. So we should go ahead with GM crops, but know that there is an unknown unknown in the “tail risk”.
Finally where does Pascal’s wager come in? The basic philosophy there is that even if the probability of God’s existence is infinitesimal, His power is so great that if you were not a believer, you shall be damned. The greatness of God’s power, according to the wager, implies that even if the probability of existence of God is infinitesimal, the expected value of being a believer is positive, and hence you better believe!
It is similar to the arguments of the likes of Taleb in the GM case in that potential harmful effects of GM can be so large (and yet unmeasured so far) that even if there is an infinitesimal probability that GM crops might be harmful we should not be adopting them.