Antifragile: Things that Gain from Disorder

From The Jolly Contrarian
Jump to navigation Jump to search


So many good things here.

  • Fragile: damaged/destroyed by volatility: a porcelain cup.
  • Robust: Not affected by volatility. a rubber cup.
  • Antifragile: Positively benefits from volatility, and is harmed by a lack of volatility: An immune system. Volatility creates stability. Lack of volatility creates fragility. The robust and antifragile don't need as accurate comprehension of the world as the fragile and do not need forecasting. Ideas are antifragile.

It is better to have lots of people hating your idea but a few very loyal supporters than everyone being lukewarm: Farage versus Lib Dem. (See also Scott Adams)

Fragile: Being cash poor/indebted is fragile Anti fragile: Being cash-rich is antifragile.

Mediocristan versus Extremistan

Mediocristan: Where one event does not increase the likelihood of another. “Bottom-up” volatility which generates randomness that cancels itself out over time. Volatility creates stability. Eliminating speculators and their moderate volatility gives the illusion of stability. In a nonvolatile market, when prices do vary, it will be assumed "someone knows something" and everyone jumps. Some confusion adds stability.


  • Maximally devolved government versus autocracy – Switzerland versus the Soviet Union. NATO a better model than the EU, which is a better model than the Soviet Union.
  • Air crashes each makes the next less likely because we learn from each.


Don’t mistake absence of evidence for evidence of absence


  • Evolution is antifragile. It only works because of its antifragility. But individual organisms are fragile, while the gene pool is not.
  • Evolution is not a competition between ideas but between humans and systems based on those ideas.

Iatrogenics and the agency problem

Causing harm by trying to help. To intervene implies a denial of fragility, removal of shocks to protect the fragile. Just as shocks hurt the fragile, shocks don't hurt (or benefit) the robust, an absence of shocks hurts the antifragile.

Caused/compounded by the agency problem: The agent's personal interest (being paid) and to be seen adding value conflicts with the actual needs of the principal. Much easier to sell "what I did" than "what I did not do". Doctor will want to recommend treatment. A lawyer will recommend changes. Doctors and lawyers get no credit for not taking action.

A performance-based bonus system exacerbates this.


  • Reliance on data causes severe side effects. The proportion of spuriousness in the data increases as you get more data. Data is toxic and in large quantities. It is infinitely narratisable. The more you have the more easily narratisable.
  • The more frequently you look at data – the more immersed in it you get – the lower the signal to noise ratio. Annual stock prices tell you more than weekly, tell you more than daily, tell you more thsn hourly.

Signal-to-noise ratio

“Say you look at information on a yearly basis, or stock prices, or the fertilizer sales of your father-in-law’s factory in Vladivostok. Assume further that for what you are observing, oat a yearly frequency, the ratio of signal to noise is about one to one (half noise, half signal) — this means that about half the changes are real improvements or degradations, the other half come from randomness. This ratio is what you get from yearly observations. But if you look at the very same data on a daily basis, the composition would change to 95 percent noise, 5 percent signal. And if you observe the data on an hourly basis, as people immersed in the news and market price variations tend to, the split becomes 99.5 percent noise to 0.5 percent signal. That is two hundred times more noise than signal — which is why anyone who listens to news (except when very, very significant events take place) is one step below sucker.”

Nassim Nicholas Taleb, Antifragile

74 – he who is without sin is less reliable than he who has sinned once - and someone who has made a lot of (different) errors is more reliable than someone has made none, because he has more experience to go on.

101 – light control works; close control leads to overreaction.
108 – modernity is the spirit of rationalisation. The ideal society is understandable => predictable. Hence the bell curve and the notions of efficiency and optimisation.
109 – you need a name for a narrative, not for a practical action. (lecturing birds on how they fly).

118 – ingenuous interventionism: give copy to a sub-editor, he will propose a certain number of edits. Give the corrected text to another editor and he will propose the same amount of fresh edits, sometimes reversing the first set of edits.
121 – 126 – 127 – easier to mistake a stone for a bear than a bear for a stone. Significant signals have a way to reach you, as long as you are not drowning in noise.
131 – when constrained, fragile systems collapse, failure is seen not as a consequence of fragility, but as the product of poor forecasting. Misses the point to see the straw that broke the camel's back as the cause. Subprime was a symptom of the crash, not its cause.
135 - . .
136 – after the occurrence of an event switch the blame from an inability to predict that event to the failure to build the system to withstand those types of event.
138 – randomness down the tail is intractable. It is nonmeasurable and nonpredictable by definition. If it were predictable it wouldn't be down the tail. Don't try to predict events in the fourth quadrant: get out of the fourth quadrant.
150 – you can't predict in general, but you can predict that someone who relies on predictions will take more risks, will be fragile to prediction errors. If prediction leads you to take more risk.
158 – fragile implies more to lose than gain – unfavourable asymmetry. Antifragile implies more upside thsn downside – favourable asymmetry.
161 – barbell – extremes at each end, with avoidance in the middle. Play it safe in some areas (be robust to negative black swans): take small risks where exposed to positive black swans. So buy a lottery ticket?
170 – teleological fallacy: that you know where you are going, that you knew where you were going in the past; that other people have succeeded by knowing where they were going. Steve jobs /Henry Ford: people don't know what they want until you give them it. Don't ask people what they want.
179 – 180 – societal growth comes not from raising the average across the board but by large increases at the tails – crazy risk takers.
186 – education is the result, not the cause of wealth. (lecturing birds how to fly)
190 – simplified definition of fragility: the more obvious the discovery, the less likely we are to figure it out by complicated methods.
190 – The significant can only be revealed through practice.
194 – two types of knowledge. 1: apophatic - a way of doing things we do well but can't articulate in clear language 2: academic – learned in school: theorisable; formalisable. We overestimate 2 and underestimate 1.
198 – greed is not new and it is not curable. Don't try to change humans: build a system that is not vulnerable to greed.
198 – the mafia don's lesson: lack of vigilance was not the cause of death: making enemies was the cause of death.
199 – difference between practice and theory : detection of the sequence and retaining the sequence in memory. Life is lived forward but remembered backwards – Kierkegaard.
200 – tourist brochures look better than reality. Bias measurement: difference beyween brochure and actual experience.
210 – people with too many complicated tricks miss elementary things.
213 – optionality is Promethean. Narratives are Epimethean.
215 – trial and error is not devoid of narrative; just not overly committed to it. They don't depend on the narrative being true.
215 – 215 – when you are fragile you need to know a lot more than when you are antifragile.
220 – practitioners don't write. They do. Birds fly.
222 – heuristics precede mathematical theorems. Pyramids, wheels, etc. Theory follows practice.
225 – development of Internet: science of use but at no point did academic science lead the way in setting the direction.
232 – in complex domains established knowledge inhibits research [for justifiable reasons from the perspective of thr prevailing paradigm]
235 – absence of evidence is not evidence of absence. For antifragile phenomena good news tends to be absent from the data; for the fragile bad news is absent from the data. (If your coffee cup is still intact, it cannot have smashed).
239 – just because some non-academic practitioners are quacks, it doesn't follow that all are.
246 – avoidance of boredom is the only worthy mode of action.
260 – an event and how it affects you are not the same thing. A predictor's reaction to prediction failure: we need better computation. A better reaction: modify your exposure. Be able to get out of trouble if the unpredictable happens.
276 – the central problem of efficiency: non-linear errors compound in one direction. You will be later, never earlier.
279 – squeezes are exacerbated by size. In a nonlinear way. Small players can't get squeezed.
279 – economies of scale hurt at times of stress. It is a bad idea to be large at times of stress. Mergers rarely play out.
282 – do not use the calculus of benefits feom scale without including the probability of failure.
284 – add uncertainty to a project and it will last longer and cost more. Never shorter, or less. The distribution goes only one way.
285 – One problem can halt the entire project. Projects tend to be as weak was the weakest link in the chain. Acute negative convexity.
292 – fragility comes from nonlinearity and convexity. Detecting acceleration of harm applies to all decision making under uncertainty.
293 – if you can make a simple point in a complicated way, even if the complexity adds no rigour, people will take the idea more seriously. Emperor's new clothes
295 – effect of nonlinearity:conditions under which the average does not matter.
297 – the number of cars is the something. Traffic times mm is the function of something. The thing andvuts function diverge under nonlinearities. In a convex situation the average of the function will be higher then the average of the thing. And vice versa for the fragile.
305 – subtractive knowledge – innovation is saying no to 1000 other things. – Steve Jobs.
305 – subtractive knowledge is barbell shaped and convex. What you know is wrong is robust. What you don't know is fragile.
309 – antifragile implies the old is superior to the new. Time will know more about its fragilities than any narrative.
312 – futurologist (a) predicted things that did not happen (b) failed to predict the things that did.
315 – technology is at its best when it is invisible. When it displaces alienating, inherently fragile, technology.
317 – domains of life expectancy:
Perishable (organism) : young outlive the old. Every additional day means shorter life expectancy.
Non perishable (species) : young and old have equivalent life expectancy [it is ahistorical: how long you have lasted to get here is not relevant. What matters is that you are here.] additional days have no effect on life expectancy
Non perishable (genera) : informational – old has longer kife expectancy thsn young. [because historical matters… The old has been subjected to more shocks and if concave will have been found out). Every additional day means longer life expectancy.
319 – note expectancy is a probabilistically-derived average. Can ve supplanted by actual knowledge
319 – it is not that technologies do not age, but that most technology that is prone to aging is already dead.
321 – information hides failures. Selection bias.
321 – another bias : we notice change not stasis.
324 – top down effect: top down changes are hard to change so mistakes tend to stick. Bottom up changes are incremental, with creation and destruction along the way. So mistakes don't stick.
334 – risk managers focus on "evidence" so on things that have hurt them in the past. But before they happened these events were without precedent.
341 – in a mature market shortcuts to performance are more likely have a hidden cost. Examples: steroids. Transfats. [But counter examples: the Internet. Maybe it is the duration point: is it a paradigm shift? Too early to tell. See genetically modified shorter wheat stems: You mess with an evolutionary process - which has weeded out poor design – at your peril .
357 – longevity is thanks to medical benefits at the extremes – when the outcome would otherwise be lethal. An error to infer that we live longer because of all medicines.
364 – we ignore religion's role in limiting intervention bias and iatrogenics.
370 – a modern illusion: that "we" are the end product. Before the enlightenment we were part of the present collective and future progeny.
371 – the antifragility of a system comes from the mortality of its components. My information – my genes – is the antifragile me who seeks mortality. Not the flesh sack.
376 – power goes to those – bankers, politicians, executives – who steal a free option from society.
389 – heuristic: don't ask a doctor what you should do. Ask her what she would do in your circumstances.
395 – don't listen to those who don't live the exact lifestyle they want others to follow.
398 – volatility benefits those who have optionality. Bankers get bonuses in a volatile sideways market. They get nothing in a flat market. Clients do worse in a volatile sideways marker.
400 – corporations sell junk food and sugar drinks: artisans sell cheese and wine.
401 – executives are not entrepreneurs. They aren't bright – evidence their appetite for internal meetings and bad memos.
402 – artisans sell us healthy products. Large ones sell iatrogenics. *anything that requires marketing appears to carry such side effects *. Marketing beyond conveying information is insecurity.
402 – "cheapest to deliver" pervades the corporate mien.
403 – "it is hardest to be a great man to your chambermaid".
409 – it is a fact that one can rapidly, after I doctrination, become enslaved to your profession.
413 – the more complicated the regulation, the more prone to arbitrage by insiders.
414 – the difference between the letter and the spirit of the regulation is harder to detect in a complex system. Complex environments with nonlinearities are easier to game than linear ones with few variables. Ii 87

415 – via negativa. To make specific forecasts about financial performance you should have skin I. The game. To make general statements about overall welfare you should not be invested.
421 – everything gains or loses from volatility. A crystal glass is short volatility.
423 – distributed randomness is a necessity not an option. Everything big and everything fast is short volatility. Modern times don't like volatility. {everything new is short volatility? Or impossible to gauge?