Black Swan - Chapter IV
Series of Summaries on Taleb's work
In this series, we are looking at Taleb’s The Black Swan. Earlier chapters were discussed here and here.
In Chapter III - ‘One Thousand and One Days or how not to be a sucker’, Taleb discusses the problems of inductive knowledge. Induction is the process of going from specific indvidual instances to reach general conclusions. However, for a turkey, inductive knowledge teaches it that it will be fed everyday with grains and treated properly. Until Thanksgiving that is, when this belief will have to be seriously revised.
The main problem with inductive reasoning is we try to know the future based on given knowledge and the infinite unknown from the finite known. This does not work. Merely because something worked or was true in the past does not mean that it has to work or will be true in the future. Contrarily, such belief makes us complacent and can have negative consequences, by lulling us into a false sense of security.
Taleb argues that the Black Swan is relative to knowledge (and the belief that we possess such knowledge). For the turkey, thanksgiving - the first day in a thousand days when it is not fed and is killed is a Black Swan. But for the butcher, that was the plan all along. He also argues that Negative Black Swans happen quickly and Positive Black Swans take time to show effect. In other words, Black Swans which destroy are quicker, Black Swans which build are slower to impact. The Turkey Problem is Hume’s Problem in philosophy - on what basis do we use past inferences to posit future possibilities?
There are a couple of themes from our blindness to the Black Swan -
First, we focus on preselected portions of what we know and generalise from it to the unseen. This is the error of confirmation. For example, if we toss five coins in succession, and it always lands heads up, we may posit that that coin always gives us heads. While all individual instances were heads, it is insufficient to inform whether subsequent outcomes (or a generalised theory) would be heads or not. Sidenote: It is worthwhile to learn how the law of large numbers operate to understand coin tosses better. Merely because a coin toss results in 100 heads in consequence does not mean the 101st time would be a tail. Each coin toss is an independent event. A 50:50 probability of being heads only means that on a large enough number of tosses, the number of heads and tails would roughly be equal.
Second, we fool ourselves with stories to cater to our Platonic thirst. We want to categorise things, build a story, and make sense of the world. This is the narrative fallacy (which we will discuss in the next edition).
Third, what we see is not necessarily all there is. History hides Black Swans from us and gives us a partial or mistaken idea of the odds of event. This is the distortion of silent evidence.
Fourth, we tend to tunnel or focus on well-defined sources of uncertainty. This is at the cost of other uncertainties which do not come to our mind. When we have deadlines to submit exams, we keep a buffer of a couple of hours to avoid instances of power cuts. But we do not factor in other unlikely events such as sickness, failure of our computer systems, etc… A 2-hour buffer might suffice for some uncertainties, but not for all uncertainties. We could convince ourselves that we have proofed ourselves from all adverse events and will be able to make our deadlines.
I hope you liked this edition. Looking forward to sending the next edition.