I stumbled upon “Black Swans” as I browsed through non-fiction section of Audible audiobooks, looking for something that could help me make sense of the economy and investing. It seemed as though the author meant it to be more than a book about the stock market however, more something of a reflection on the tools we have to predict and prepare for the future. Although Nicholas Nassim Taleb starts of by introducing valuable ideas about the nature of uncertainty and our relationship with it, the chapters quickly degenerate into endless ramblings about the incompetence of economists and statisticians, and what was passing for elaborate skeptic thinking quickly turns out to be a simple grudge against academics and practitioners of social sciences.

Black swans : a summary

The author coins the concept of Black Swans, rare events that have a disproportionate impact on our lives and the world around us. They can occur in many domains and have either negative or positive effects. The essence of black swans according to Taleb is that they catch us off-guard. They can impact us negatively, like in the case of war or a market crash, or we could be eagerly awaiting one such exceptional event, like a scientific or technological breakthrough, success for an artist or an innovative business.

They are therefore of an unpredictable nature. Much has been done in the way of making predictions in various domains but these are most useful in “Mediocristan”, the realm where uncertainty can be modeled using the Gaussian distribution, which will assume observations of a random variable will cluster around the mean, or other tame distributions. The author mentions that things such as the distribution of people’s height, weight, or the outcome of games of chance; dice, slot machines,… all belong to Medicristan. “Extremistan” on the other hand is where complex and unpredictable social, economic or political events unfold. Taleb argues that winner-takes-all mechanics favor the emergence of events with large impact, and the apparition of a “fat-tail”, where observations far from the mean are more common. One could model book sales in such a way, because a few successful books will tend to be more successful than all of the other books combined. This sort of randomness, along with the complexity of the world and the fact that many phenomena are interrelated, conspire to create events that are unpredictable. Technology is deemed to increase inter-dependence in the world and make it better at predict normal events but more vulnerable to black swans as consequences of small variations are more far-reaching and unexpected.

We easily fall for the “Narrative Fallacy”, blame our failures on unpredictable events and explain our successes by our skill, which can be somewhat exaggerated, many of our lives being subjected to untamed uncertainty. History produces many unique events, and our knowledge of the past is simply too sparse to completely understand it, and make accurate predictions from it. This also highlights too the problem of correlation not imply causation. Causation sometimes is not reexamined after it has been inferred from past events, a scientific approach and the use of falsifiable statements in historical theories is needed in order to keep history honest.

The unbridled skeptic

NNT portrays much of the academic world as “suckers” that fall for the trap of thinking they are living in Mediocristan. The truth is, in the face of enormous uncertainty, forecasters are often asked to produce a prediction, which is nothing better than the mean of all the eventualities considered by a model, and can literally extrapolate as far as we want into the future based on simplistic assumptions. Even the most sophisticated models can be absolutely useless if not used properly, knowing their limitations and understanding their assumptions. Their application in business, or even academia, can be clumsy or inappropriate. The antidote is however more erudition and better knowledge of the models used, not moving away from formal studies, so that one can be critical of the tools they are using. The supposedly technical chapter is little more than a diss of the statistics and economics establishment, where one would hope for a dive into the techniques used to cope with uncertainty.

Throughout the book, the author takes a skeptic stance, and I agree with the point that people exit doubt too easily and fall prey to dogmatism even in scientific fields, but it seems to me as though the book falls into the trap of discounting all predicting and planning. We can make plans and give predictions that we all know will be inaccurate, or plain wrong, in many human endeavors we cannot rely on our forecasts and the key is to be adaptable and understanding that the assumptions made are just as important as the prediction itself.

The failings of platonicity

In order to make predictions, we have to use well-defined forms and formal systems and we reduce the world and its mechanics to symbols we can manipulate. This consists of all the modelling we do today, and randomness is reduced to a distribution, which is a description of how events happen over time. This is best illustrated by the uniform distribution of the outcome of dice, the distribution of that event telling us that the probability of each number being rolled is $\frac{1}{6}$. In the real world however, uncertainty appears where we don’t even model it, in a fortuitous manner. NNT boasts endlessly of being able to deal with these, but in my opinion the book fails to really address this point, offers little in the way of solutions, much is said about Fractal distributions, which falls under the umbrella of the platonic models so criticized, though they arguably work better in some cases.

Nicholas Nassim Taleb seems to be extremely polarized when it comes to the choices made to model uncertainty, really too much for a skeptic. With bold statements such as “the Nobel prize for economy should be cancelled” and a taste for provocation, Nicholas Nassim Taleb seems desperate to become the black sheep of the world of statistics and economics.