Antifragile

Antifragile is an immensely dense tome spanning a wide array of domains, and attempts to frame various systems and phenomena along a spectrum of fragility - things that are susceptible to, robust and resilient against, or stand to gain from chaos, disorder, and time.

The book is (perhaps overly) ambitious in its application of the idea of antifragility across vast domains, and at times the message seems to get lost amidst the author’s various diatribes against people, ideas, and systems he dislikes, ostensibly due to their fragility. I didn’t find these distractions to be particularly offensive, but found them to be humorous and made the book more approachable.

I had previous read and enjoyed The Black Swan, also by Taleb, and did not recall the same level of criticism against academics, public figures, and non-practitioners as found in Antifragile. Nevertheless, I must say the author knows how to use the controversial, and at times antagonizing narrative style to sell books and to make impressionable points. In fact, after recounting the ideas of The Black Swan, its prescience and timeliness, I can’t help but want to attempt to use the author’s predictions for the next set of large scale, surprise events to enact anticipatory actions to benefit in antifragile ways. However, that would probably be the exact type of hindsight, cherry-picking analysis that the author rails against.

The density of ideas in the book progression follows mostly a skewed bell curve, with a skew of denser ideas towards the front of the book where the initial chapters do a good job of motivating and the middle chapters contain the most convincing narratives and practical fallout. The tail of the book is thinner on ideas and mostly reads as a rant against the phenomena and practices found in modern education and medicine. It’s hard to disagree with the observations about traditional education and its focus on theory over practice, and the fragility of theories that fail to fit large scale events and end up falling apart; additionally, it’s clear that one often does not need to understand theory to be successful in practice in many domains. To me, the overarching theme of the book is less about antifragility and more about asymmetry. Find the asymmetries in life and you’ll find the things that are antifragile.

Some of the novel concepts that stood out to me were:

  • Naive interventionism and its pervasiveness across domains
  • Iatrogenics that arises as a result of interventions
  • Importance of optionality and the practical advice of employing a barbell strategy
  • Favoring trial and error, tinkering, and evidence based approaches
  • Exposing onself to convexity effects through approaches and optionality
  • Looking for asymmetries in upside and downside

This book provides a novel framework under which to evaluate systems, processes, and decisions. As I read the book, I couldn’t help but ponder over the repercussions of taking an antifragile approach when writing software. Some obvious examples of software systems that benefit from chaos and disorder are things like random number generators that require entropy to become better, such as using randomness from a computer mouse movements in order to generate strong cryptographic keys. The practice of introducing chaos into distributed systems by injecting random failures in order to encourage better software design also falls squarely into the antifragile domain.

However, it’s more difficult to imagine general software systems that benefit and get better with additional chaos and disorder. Adaptive systems that are robust or resilient to failure fall in the middle of the spectrum, and don’t improve over time or get better with increasing randomness; you’re typically hard pressed to find systems that get better as uptime increases, as stability is the name of the game. Machine learning systems that benefit from larger and larger amounts of data are markedly fragile, and are typically sensitive to and perform poorly against outliers, due to the effects of underfitting or overfitting.

Finally, I found the section on Seneca and stoicism to be a nice, succinct introduction to the philosophy and surrounding ideas, while providing a philosophical framework for becoming antifragile to randomness. By mentally writing off possessions, and, for example, assuming some amount of wealth that can be lost will be lost and coming to terms with that from the get go, one then has little to lose and much more to gain - the asymmetric, positive convexity effect that an antifragilista always strives for.

Comments