Sauce for the Black Swans

I am reading The Black Swan (second edition) by Nassim Nicholas Taleb, Lebanese in exile, New Yorker at heart, financial trader, philosopher. A book that has caused a modest stir in some circles. Its author would probably agree that most books that cause a stir in the circles of financial managers are a waste of good ink and paper, but The Black Swan is worth reading.

The thesis of the book is that large areas of human activity are dominated by the Black Swan events: Unpredicted, seemingly random occurrences with a high impact. Unpredicted, rather than unpredictable, as Taleb concedes that what constitutes a Black Swan is in the eye of the beholder. Therefore the book focuses not on the nature of such events, but on the human psychology that causes people to overlook their possibility, indeed (given a sufficiently long period of time) their probability. And why Taleb adds a third characteristic of a Black Swan that people are always, with hindsight, able to find a explanation for them, which enables them to forget about the lesson for the future. As a potential cure for these ills, Taleb advocates skeptical empiricism: Proceeding on a cautious observation of the world, while resisting theories and models. The underlying philosophical problem is how skeptical empiricism (supposedly taking into account the Black Swans) can be separate from inductive reasoning (supposedly not). As one cannot prove a negative, one can never exclude the possibility that a Black Swan may occur – Taleb's concept is not falsifiable! – yet it is obvious that inductive reasoning is often useful, despite the risk.

Taleb presents his concepts from the background of a financial trader and analyst, which provides him with experience and examples. However, very much to his credit, he does not roll out endless tables and graphs to try to prove his case using anecdotal evidence, nor does he confuse the reader with financial jargon. This is a very accessible book. At times the perspective does appear a bit too narrow and biased: In a footnote Taleb argues in apparent seriousness that the outbreak of the Great War in 1914 came as a surprise, because "bond prices did not reflect the anticipation of war." Maybe bond prices didn't, but military preparations did! Yet Taleb is probably wise to draw his arguments from the financial arena, because this has the benefit of offering a lot of quantitative data -- and besides, judging from some errors in the book, his knowledge of general history is scant.

That is not to say that Nassim Taleb is a one-dimensional figure which understands only market numbers. He clearly possesses erudition and makes an effort to demonstrate it. His erudition, evidently, is that of a writer, scholar and soft scientist. It is also that of an egocentric full of an idea, who can at times be unthinkingly dismissive towards activities in other areas. Nike, Dell and Boeing, he writes, make money through "thinking, organizing and leveraging their know-how" while outsourcing "the grunt work" and "the noncreative technical grind." That is managobabble. I doubt that Nike is leveraging much beyond the American dominance of the media markets and its established brand position. But Boeing? Modern airliners are enormously complex machines, and in the endless pursuit of savings in fuel and cost of operations, every little detail counts. Creative design work cannot easily be separated from the "grind" of technical skill, and knowledge of how the "grunt work" is done is vital, because it determines the limits of the feasible. Yes, Boeing outsources a lot, but that actually includes the outsourcing of a lot of creative work, while a lot of the technical grind and construction work is still done in Seattle. Boeing splits up the work by components – fuselage, tail, wings, flaps, engines, instruments, seats – and not by easy categories such as "creative" and "non-creative".

The author, then, is a erudite man with an idea that he proposes and defends energetically, forcefully, and sometimes bitterly. His strong rhetoric can be confusing as he repeatedly slices and dices the universe into categories, subtly changing the dividing lines by introducing new terminologies: Creative versus noncreative, scalable versus nonscalable, techne versus episteme, Mediocristan versus Extremistan, domains in which useful experts can be found and domains were the supposed experts do a very poor job of predicting the future, Gaussian versus Mandelbrotian. The proposed division of reality is a ragged, frayed one, but attentive readers will nevertheless get the point: We should be more cautious in the use of our mental predictive "machinery" with its reliance on the law of averages. We usually depend on it in everyday life and it serves us well for many  purposes. But we should be aware that there are also areas in which it breaks down because the unpredicted, the outlier, is the dominant force there, instead of a detail that can be glossed over: A natural disaster, a major invention, a very rich man, a financial collapse. And, Taleb argues, these areas are growing more important as society gets more complex and interconnected, and may now involve the majority of the decisions that we will be required to make. We need to be aware of this, because we are not mentally equipped to take Black Swans into account: It takes a conscious effort to do so. 

Taleb has an axe to grind with economists who use formal mathematical models to predict the markets: He is not opposed to the concept in principle (he suspects that fractal statistics may work) but dismissive about approaches based on Gaussian statistics, which in his view have been proven a failure again and again, because they are too reductionist and start from flawed assumptions on uncertainty and risk. The fierceness of his attack on the idea may be justified, but in defense of the economists I would say that they would not be the first ones to formulate a model under a given set of assumptions and approximations, then see it abused to make predictions under entirely different circumstances. Some of them may be guilty of hubris, but there may be some innocent victims.

I am not very impressed by the forays in evolutionary psychology, and there are other weak points. For example, there is a section with the bold title Information is bad for knowledge, but in this he only demonstrates that a gradual flow of information is bad for knowledge – and that a less gradual flow may be better. If subtle, the difference is very important. Even worse is the technical gaffe in the chapter on the Gaussian distribution, where he claims that the distribution will become narrower as sample size increases. This is a common misunderstanding among people with a tenuous grasp of statistics, but that does not make it excusable in a book that pretends to challenge conventional statistics. Nor is it accurate to assume that the  fundamental uncertainty is the quantum world, as defined by Heisenberg's principle, obeys a Gaussian distribution. There are some other strange statements in the extension to the second edition, which lead to the conclusion that for the more academically and mathematically schooled collaborators of his scientific publications, Taleb must be a difficult man to work with.

Yet these problems do not undermine his central thesis, and he does make some strong points. If you think that the inventions we see around us came from someone sitting in a cubicle and concocting them according to a timetable, think again. Having spent most of my career in research, I can observe that although it is many ways the art of generating (falsifiable) predictions, the activity itself is inherently unpredictable, because the study of the unknown produces inherently unpredictable results. Nevertheless I have seen a lot of people sitting in cubicles who were expected to deliver new inventions according to a timetable. That approach does not work, but so far that has not stopped anybody from trying it – quite the contrary. The human (managerial) desire for predictability, control and guaranteed outcomes is so overwhelming, that we continue to rely on predictive frameworks even after they have failed time and time again, and seek to impose them on others.

Taleb's strong recommendation is to try to incorporate room for the unknown and unpredicted in your (business) strategy. The Black Swans cannot be predicted, but by a smart strategy you can mitigate the impact of negative ones and maximize the effect of positive ones. He highlights the foolishness of acceptiong the risk of rare, but very damaging events on the assumption that they will never happen – especially if they actually have a history of happening from time to time, like financial collapses. On the other hand, you should be ready to grasp the once-in-a-lifetime opportunity: A serendipitous discovery should not be ignored, merely because it doesn't fit into your strategy.

Having finished reading the book and the lengthy extension added to the second edition, I cannot escape the observation that Taleb has something of the crank about him. There are the dazzling intellectual maneuvers, the scattering of small factual and logical errors, the solipsistic ranting of a self-declared genius against the establishment: You can also find these in a book dedicated the the mysteries of the pyramids, the Bermuda triangle and Atlantis, all explained at one stroke. The redeeming quality that makes the book worth reading is that the central points are nevertheless entirely valid. The more straightforward point is that in a domain where the outcome can be determined by rare events with a high impact, statistical approximation breaks down; and the abuse of Gaussian statistics in areas where they are not applicable is inherently misleading. The less obvious point, which really makes the work significant, is the argument that this condition of statistical breakdown is not the exception but commonplace, and dominates financial markets.