“Wind extinguishes a candle and energizes fire.”
I’m often asked what my favourite book is. I guess it’s all about being the right book at the right time. To date though, Antifragile probably had the most influence on the way I think and see the world.
Unfortunately it’s written in an inaccessible style. Too many unnecessarily fancy words. This is very much intentional. The author doesn’t want skim readers to spread misrepresented ideas.
This is a shame though, as so much wisdom remains buried under pages, without getting a good chance to influence even more people.
So, this is my attempt at articulating some of its key concepts in the simplest way possible. Rather than just regurgitating the book, what I’ll do is augment Taleb’s central ideas and examples, along with my own observations and ancillary concepts I find relevant. There’s 5 sections:
- Defining and recognising antifragility: distinguish fragile vs robust vs antifragile; recognise that antifragile systems are everywhere (horizontally across different disciplines, and vertically across different scales)
- Modernity’s denial of antifragility: how modernity’s obsession with reducing disorder hurts antifragile systems; exploring economic super-efficiency to illustrate this
- Dancing with disorder: a deeper dive into some more technical supporting concepts in order to minimise harm and maximise gain from antifragility
- Potential misuse and abuse of antifragility: raising awareness on the misuse and abuse of this concept, especially by those with no skin in the game
- Wrapping up: summary, why this is my favourite book (personal development and complexity science)
1. Defining and recognising antifragility
1.1. Antifragile: things that gain from disorder
What are some things that are fragile?
What makes these things fragile? What are the properties of fragility?
Okay, then what’s the logical opposite of fragile?
Doesn’t break easily.
So something along the lines of: robust, resilient, strong, tough?
All of the above present a false dichotomy. Saying that robust is the opposite of fragile is like saying the opposite of positive is neutral, when it is in fact negative.
“Some things benefit from shocks; they thrive and grow when exposed to volatility, randomness, disorder, and stressors and love adventure, risk, and uncertainty. Yet, in spite of the ubiquity of the phenomenon, there is no word for the exact opposite of fragile. Let us call it antifragile.”
Let’s define fragile as things that dislike disorder, volatility, randomness, uncertainty, stress, time etc.
Anything from the disorder family:
- dispersion of outcomes
- the unknown
- imperfect, incomplete knowledge
Antifragile, being the exact opposite, would then be: things that gain from disorder, volatility etc.
- Fragile dislikes volatility
- Robust doesn’t care
- Antifragile likes volatility
- Fragile breaks when dropped
- Robust remains intact
Antifragile becomes even stronger than before
- Damocles dines with a sword hanging above his head – does not want change.
- Phoenix gets reborn from its ashes – doesn’t care.
- Hydra grows two heads for every head it loses – wants change.
Antifragile goods would love to be shipped in a box labeled “please mishandle, please be careless”.
1.2. A transdisciplinary lens
Many things don’t have antifragile properties. Especially inanimate objects.
But many other things do. Especially networks and living systems.
Natural selection. Recall high school biology: (i) there is a variation in a population; (ii) when there is a change in the environment, those with favourable characteristics to this new environment survive, and those without perish; (iii) favourable characteristics are passed onto the next generation, (iv) after many generations these favourable characteristics become common in the population. But if there are no major environmental changes for a long time, quirky characteristics that are unfavourable in today’s environment (but would be in a new world) become less and less common. Variation decreases and the population trends towards a monoculture. Then, if there is a big environmental change, the monoculture can easily be wiped out.
Immune system. Your body responds much faster and with more intensity, after every encounter with a pathogen (something foreign to body that triggers immune response, e.g. bacteria or a rusty blade). So if your blood has been previously exposed, it’ll be much more effective when the next invasion occurs. This is how natural immunity and vaccinations work.
Bone. Bone-building cells (osteoblasts), and bone-destroying cells (osteoclasts) are given instructions on where to build new bone / destroy old bone based on where stress is being applied. Add bone to high stress areas, and remove bone from low stress areas. This is why astronauts lose 1-2% of bone mineral density for every month spent in space: no gravity means no stress means less bone.
Start-up ecosystems. Start-up concentrations like Silicon Valley thrive (create and capture value) when there are disruptive changes to society. Some directly commercialise new technologies (personal computing, the internet, smartphones, AI etc). Others do even better by bringing new business models that ride on a tailwind of social megatrends (share economy, platforms, social networks etc). Either way, if the market, or business landscape is not changing, the incumbents will just continue to dominate.
Capitalism. The market determines the demand and price of goods and services. This way, the ‘invisible hand’ directs more of society’s collective resources (in the form of capital) towards value-generating activities, and gives less from value-destroying ones. Disorder, volatility, change – continually forces the system to re-arrange itself to maximize value.
Military. Some say it won’t be long until China’s military power surpasses the US. Such predictions are typically based on: (i) China’s GDP having already surpassed the US on a purchasing power parity basis; (ii) US retracts into protectionist policies, while China’s expands along its ambitious Belt and Road. But I say it’ll take a lot longer than most may think. Because, the US has been continually involved in wars post WWII, while China’s last major involvement was in the Korean War in the 1950s. With continual military involvement comes: more highly-trained generals, a repertoire of field-validated strategies, and updates to equipment/internal processes/structures etc based on real battle experience. For now, the Chinese military is like a super-fit UFC figher that’s trained hard for years but never actually been in a ring. I know this military example is more contentious, more on this in Part 4. when I discuss ethical dimension to antifragility.
1.3. A fractal lens
Here’s when things get more interesting. Antifragility in the presence of layers and hierarchies.
In order for a system to be antifragile, some of its components need to be fragile.
This may seem counter-intuitive. Perhaps we’re too familiar with robust things being made up of robust components. Robustness at every layer.
Start-up ecosystem. The fragility of start-ups (having a high failure rate) is necessary for the broaderthe economy to be antifragile. Society has limited resources, and it needs to allocate more to the more value-adding ones. The faster the low-value start-ups fail, the less resources they burn off, and this is better for the system as a whole.
Biology. For a multicellular organism, the trouble-maker cells (old and damaged) absorb resources without adding much value. These cells need to be fragile in order for them to be easily replace with fresh cells. For example, when you’re deprived of food (i.e. low in resources) – your body kills the old cells first. Through this mechanism of sacrificing the fragile, the system (organism) stays healthy.
The above examples considered going to lower levels in a system hierarchy. Systems contain subsystems, and those subsystems are compromised of even more subsystems, and so on.
But you can go the other way too. Any system is often a subsystem of a bigger system, and that higher level system is part of another one above it. So if a subsystem is antifragile, then other subsystems on its same level need to be fragile for the parent system above it to be antifragile.
Back to the biology example to illustrate this recursive antifragility:
- A multicellular organism, say a human, is antifragile if some of its unicelluar components are fragile
- A tribe of humans is antifragile, if some humans are fragile. For our late ancestors, fragility here meant the vulnerability of getting killed. Killed by being eaten by a lion, or being abandoned by the tribe for physical weakness/injury/disease etc. But over time, this also included more mild forms of fragility like being ignored/punished for violating norms, or just being an asshole.
- For a larger human society (where multiple tribes interact) to be antifragile, some tribes need to be fragile. Different tribes have certain norms and lifestyles.
- Basically, recursive Darwinism at different scales.
Perhaps there are ontological implications in this antifragility-fragility hierarchy notion. From subatomic particles scaling up to galactic clusters with alternating degrees of antifragility/fragility at every layer.
2. Modernity’s denial of antifragility
2.1. Modernity is obsessed with reducing disorder
Much of our resources goes into taming and reducing disorder.
- Agriculture to stabilise the food supply.
- Fridges keep food fresh by narrowing the temperature range
- Damping systems keep skyscrapers from wobbling from wind/earthquakes
- Fuses protect electrical appliances from from over-current damage
- Safety rails in high places protect people from falling to their death
- Insurance protects bank accounts from going bust from unexpected accidents/misfortunes
Shielding the fragile from disorder is good and necessary. And modernity has become more and more effective at doing this. Perhaps even too effective.
2.2. Denying antifragile systems of volatility
The problem is we’re shielding disorder not just from fragile systems, but from antifragile systems too. This denial stems from a lack of understanding of antifragility. And also, our institutions have not been designed to take care of antifragile. Maybe this is partly because the rate of social, economic, and technological change has been so fast these past several decades.
“Modernity corresponds to the systematic extraction of humans from their randomness-laden ecology—physical and social, even epistemological. Modernity is not just the postmedieval, postagrarian, and postfeudal historical period as defined in sociology textbooks. It is rather the spirit of an age marked by rationalization (naive rationalism), the idea that society is understandable, hence must be designed, by humans. With it was born statistical theory, hence the beastly bell curve. So was linear science. So was the notion of “efficiency”—or optimization.”
Why is this artificial suppression a bad thing?
(i) First, antifragile systems don’t just want disorder, they need it to stay alive. Without disorder, the antifragile actually suffers, stagnates, and eventually dies. Absence of stressors leads to atrophy.
“…if antifragility is the property of all those natural (and complex) systems that have survived, depriving these systems of volatility, randomness, and stressors will harm them. They will weaken, die, or blow up. We have been fragilizing the economy, our health, political life, education, almost everything … by suppressing randomness and volatility.”
(ii) Second, from the outside, the system exhibits no visible risks. Silent risks accumulate beneath the surface. So when the fire eventually burns, it’ll really burn.
Fires have gotten larger over the past 100 years. Climate change is one contributing factor, but not the main one. It’s more to do with land management. In the US, for example, forests used to be more patchy. Wildfires were more frequent but smaller.
“These different forest types, the environments that they grew in and fire severity – they all worked together to shape this historical patchwork… It provided a natural mechanism to resist the spread of future fires across the landscape. Once a patch of forest burned it helped to prevent the flow of fire across the landscape. A way to think about it is that the burned patches helped the rest of the forest to be forest.”
Roads and rail became potent fire breaks. Forests that would otherwise burn more regularly were protected. Logging removed the larger trees – survivors of centuries of wildfire. The forest became more dense and filled with small, fire-sensitive trees.
In 1910 there was a huge wildfire near Washington – the “Big Burn”. A large area burned and many died. From then on the newly formed national fire service was mandated to put out all wildfires across the country. But by continually suppressing every small fire, the forests accumulated dry dead matter, making it increasingly vulnerable to inevitable bigger ones. By suppressing disorder, the forest has become more fragile.
This is the point of back burning, (aka prescribed burning): intentionally burning off parts of the forest in a controlled manner. We just don’t do enough of it.
2.3. Relentless pursuit of economic efficiency
Let’s turn to an economics example. For 200 years and running, globalised economies have embraced Ricardian comparative advantage.
David Ricardo, early 1800s, comparative advantage: each country should focus on its relative advantages over other countries, rather than its absolute advantages. The classic example:
- Suppose the entire world economy was just England and Portugal
- Raw material and labour costs are higher in England
- To produce 1 unit of cloth, England requires $100 while Portugal requires just $90
- To produce 1 unit of wine, England requires $110 while Portugal requires just $80
- Each country wants to figure out how it should best allocate its limited resources to maximise economic results
- Portugal wants more cloth. Should it make its own for $90 or buy from England for $100? Comparative advantage says Portugal should pay the extra $10 and buy from England for $100. Counter-intuitive right.
- Why? Opportunity cost. It’s better for Portugal to allocate maximal resources into wine production and selling it to England at a $30 profit/unit, rather than making cloth and selling it at a $10 profit/unit.
- So even though Portugal can produce cloth more competitively than England, England actually has a comparative advantage over Portugal in cloth production
We’ve all benefited from specialization at a national level.
But there is a hidden price to superefficiency: fragility.
We have too many eggs in one basket. If something unexpected happens (i.e. more disorder), the whole system will take more damage.
Industries are becoming increasingly concentrated. Winner-take-all effects are becoming more prevalent. Pareto distributions are becoming super-Pareto.
- In 1978, the top 100 most profitable firms earned 48% of all publicly traded companies’ profits.
- In 2015, the top 100 earned 84%.
The 2019 HBR article: Rethinking Efficiency illustrates this point. In the following charts, the left horizontal axis shows aggregate market share of the top 4 companies in a given industry. The right horizontal axis shows the same for 2012. So an upslope indicates increasing concentration.
850 industries were studied.
- Two thirds increased concentration (shown left in orange)
- Note the emptiness in the top part of the right chart. Most of the highly concentrated industries in 1997 amplified their concentration in 2012.
This trend is even more evident if we just look at the industries where cumulative top-4 market share moved by more than 10 percent points. Of 285 industries, ~75% (216 of 285) became more concentrated. When we look at the 92 industries that moved by more than 20 percentage points, an even higher proportion, ~90%, (82 of 92) increased concentration.
Not surprisingly, the trend persists when we look at the largest 4 sectors.
Why? Efficiency gains from economies of scale and winner-take-all effects. For example, some of the advantages one large trucking company has over many smaller ones include:
- shared costs, instead of each company having their own: fleet maintenance and repair; back-office functions
- increase buyer bargaining power to acquire fleet
- increase supplier bargaining power to customers they serve
Usually, this is great for most people – we get better, cheaper, faster stuff. Efficiency creates value for society.
But too much efficiency can actually destroy value in the long-run.
- Big firms generally respond slower to change, and are less adaptable
- Highly concentrated oligopolies can turn into quasi-cartels. Each benefiting from lucrative margins at the end user’s ultimate cost.
- This is positively reinforcing. More profits, more power, more lobbying, more control, more profits, and so on. Big firms get bigger. The rich get richer.
- Increased market power means the big company can more easily get away without responding to bottom-up trends/changes in the market. This accumulates hidden fragility.
- When they fall, they fall hard: many lose their jobs, many customers are suddenly left without a supplier. And this creates much bigger second-order effects than if a couple of small ones had failed
A superefficient dominant model elevates the risk of catastrophic failure. Bigger fire.
In short, modernity’s suppression of disorder comes with good intentions. Modernity is a neurotic parent. This has been a good outcome for fragile systems (e.g. the helmet the child was forced to wear by mum prevented permanent brain injury). But a bad one for antifragile systems (e.g. preventing the child from ever going on a bicycle out of fear they’ll get hurt).
We need to develop an eye to recognise such fragility, and minimise its effects…. Right?
Yes, but that’s just saying let’s move from a fragile design to robust one.
We can do better. Let’s move from a fragile design to an antifragile one. Gain from disorder, not just be immune to it.
3. Dancing with disorder
How to gain from disorder? How can I benefit from this concept? What are the practical applications?
We first need to dive deeper into some more technical concepts. Specifically, the mathematical properties of fragility, as well as a few closely-related ancillary concepts. Here we’ll cover:
- Convexity effects
- Black Swans
- Event vs implication of event
- Limitations of antifragility
- Jump from a 0.1 m height 1,000 times. Your knees would hurt a bit but otherwise you’re fine.
- Jump from a 100 m height once and you’re dead. 1 x 100 m jump has a much bigger effect than 1000 x 0.1 m jumps.
This is non-linearity. A single large shock has a disproportionately greater effect than the equivalent cumulative effect of many small shocks.
- Drive into a wall at 5km/hr 20 times versus drive into a wall at 100km/hr once.
- Drink 1 bottle of wine a day for 7 days versus drink 7 bottles of wine in 1 day.
Fragility is generally non-linear. This is due to the structure of survival probabilities. While the fragile is still alive, it’s harmed a lot more by single large shocks than lots of small ones. Once it’s already dead, more shocks no longer affect it. The fragile is subject to accelerating harm.
Commute time is also subject to non-linearity.
- Adding more cars into the road network when there’s very few cars doesn’t change your commute time very much. eg. going from 10k cars to 20k cars.
- But if there’s already quite a lot of cars (e.g. 100k) on the road, adding another 10k won’t increase commute time by 10%, but something much more, say 50%.
- So travel time is fragile to the volatility of the number of cars on the road. Travel time is subject to accelerating harm.
- Having 90k cars for one hour and 110k cars for another would be much slower than two hours of 100k cars.
- Since travel time is a negative, we can represent it as a cost. So graph going down is higher cost (longer commute time).
Liquidation cost is another non-linear function. The cost of selling off $70b worth of assets is significantly more than 7x the cost of selling off $10b worth.
Antifragile is also non-linear. Although in contrast to the fragile, big shocks bring more benefits than lots of small shocks. Up to a certain point, of course. The antifragile is subject to accelerating benefit.
3.2. Convexity effects
There are two kinds of non-linearity: convex (smile) and concave (frown).
Antifragile is convex. Fragile is concave. How so?
Consider the graph below. Horizontal axis is some variable (e.g. height you jump from). Vertical axis is gain/loss from that variable (e.g. your physical condition after you jump). Your health is a function (f(x)) of the jump height (x). Given this shape, you prefer less volatility in x. Because on average, you’re going to lose more than gain.
The convex is the the same shape but flipped over.
For the convex, more volatility brings more gains than losses on average. So the convex likes volatility and disorder.
If you earn more than you lose from fluctuations, you want a lot of fluctuations.
A more technical way of interepreting convexity is: the average of the function is greater than the function of the average. Especially more so as x is more volatile and dispersed. This is illustrated for two x-values below, but can be generalised for more.
Another implication: the convex does not need to be right that often. Suppose you were wrong 5 times, and right just once. If the gains you make from that one right time is greater than the total losses you make from all 5 wrongs combined, then you’re still up.
Someone with linear payoff needs to be right more than 50 percent of the time. Someone with convex can still be ahead even if they’re frequently wrong.
3.3. Black Swans
The metaphor originates from Europeans being shocked from seeing black swans for the first time when they first came to Australia. All swans they ever saw back home were white. And so, they presumed all swans were white.
Black Swan: an event that is: (i) an outlier, with (ii) extreme impact, that (iii) people make ex post facto explanations for (“I told you so”).
- Natural disasters, especially those that affect humans: tsunamis, earthquakes
- Non-natural disasters: terror attacks, underestimated extremist political candidates winning elections, Titanic sinking, Aztecs encountering the Spanish
And of course, the current COVID-19 pandemic is a Black Swan event…
(On financial markets, I’d actually argue COVID-19 is more like the final straw that broke the camel’s back, but let’s keep this post on topic.)
How effectively governments respond will drastically affect the next election results. And for non-democracies, it’ll affect how people view the current ruling party. All of these factors may reshape global order and leadership dynamics in ways we can’t foresee right now.
Need I say more?
No one saw it coming.
What makes Black Swans unpredictable is we don’t know what we don’t know.
The turkey problem. Imagine you’re a turkey. Every day that gets closer to Thanksgiving, a turkey is continued to be fed very well. As each day passes, a turkey can conclude with more and more statistical confidence that Thanksgiving is good for the turkey. Then, right before Thanksgiving we all know what happens.
This is a classic example of conflating absence of evidence for evidence of absence. Here, evidence is the turkey not being harmed by the farmer.
Many forecasts extrapolate historical data into the future.
“The fool believes that the tallest mountain in the world will be equal to the tallest one he has observed.”
But the danger of this is that the risks remain hidden. And these hidden risks is what makes catastrophes catastrophes.
Take the Fukushima nuclear reactor, as another example. It was built to withstand the worst earthquake …that has been recorded in the past. It had no precedent for the magnitude of the 2011 earthquake. Catastrophe.
For the fragile, wins and losses over time may look something like this.
Small frequent visible variations with large occasional hidden losses. Well, hidden, until it finally happens.
For the robust, the variations could small and insignificant. Black Swans don’t really affect them. Fluctuations are predictable.
For the antifragile, variations could also be small most of the time. But every now and then an unexpected event brings a monster win.
So the fragile is exposed to negative Black Swans, while the antifragile is exposed to positive ones.
In a financial markets context, an option is a right, but not an obligation, to buy/sell something at a fixed price within an agreed time frame. They’re a type of derivative – a contract, a financial instrument, that has its value determined by some other underlying asset. There’s two types of options.
- Call option: right to buy something at a fixed price within agreed time frame
- Put option: right to sell something at a fixed price within agreed time frame
For example, a certain stock is at $100 per share. You think it’ll go up. You have $500.
- You could buy 5 shares with your $500 ($100 x 5).
- If stock goes up to $125, you make $125 profit ($25 x 5). Not bad.
- But you’re greedy. So you borrow money from a friend to leverage up, say another $2k.
- You buy 25 shares with $2.5k ($500 your money, $2k friend’s money)
- If stock goes up to $125, you now make $625 profit ($25 x 25). Better.
- But leverage can hurt you much in the same ratio. If the stock goes to $75, instead of losing only $125, you now lose $625.
- And if the stock goes much lower, you lose so much money you can’t even pay your friend back.
Alternatively, you could buy call options.
- Think of each option contract as a coupon that says “for a limited time only, with this coupon, 1 share of this stock can be bought for $100.”
- For this fixed price (aka. strike price) of $100, let’s say each option contract is $5 (aka. the premium).
- So each coupon costs $5. You buy 100 of these coupons for $500.
- If the stock goes up to $125 at any point within the agreed time frame, you have the right to use (aka. execute) your option, to buy stock worth $125 for just $100.
- This advantageous price difference, makes each one of your ‘coupons’ have an intrinsic value of $25. Since you paid $5 for each one of these coupons, the profit you make for each coupon you own is $20.
- 100 coupons each profiting $20 is a $2k profit (see below). $2k profit is better than the $625 profit in the previous example.
- Now imagine the same nightmare scenario of the stock plummeting to $30. At this price, you can just choose to not execute your options. There’s no reason to pay $100 for something that’s worth $30. You simply let your coupons’s time frame expire.
- Your 100 coupons are now worth $0. So you lose the $500 you paid for these. $500 loss is much better than the $625 loss in the previous example.
- If the stock went even lower, say to $10 a share, you still only lose $500.
There are 3 points to note here about options.
First, note the favourable asymmetry. You have a capped downside and uncapped upside.
Second, this means that if you’re holding such options, you want to see more price volatility. You don’t care about the average outcome, you only care about the most favourable one. Options like a dispersion of outcomes. Options like volatility.
The luxury goods industry, for example, cares less about the average income/wealth in its target market. It only cares about having enough people that are super-rich.
- Market A: 10m people with $50k income, and 100k people with $1m income
- Market B: 5m people with $10k income, 1m people with $1m income
Market B is more attractive to luxury goods. Inequality doesn’t matter. Dispersion does.
Third, you don’t have to be right that often. So long as the cost of each time you’re wrong is low. You can lose small often, but win very big every now and then.
“If you “have optionality,” you don’t have much need for what is commonly called intelligence, knowledge, insight, skills, and these complicated things that take place in our brain cells. For you don’t have to be right that often. All you need is the wisdom to not do unintelligent things to hurt yourself (some acts of omission) and recognize favorable outcomes when they occur. (The key is that your assessment doesn’t need to be made beforehand, only after the outcome.)”
With this, we also define optionality as: optionality = asymmetry + rationality.
By rationality, we mean keeping the good and ditching the bad. That is, it does take some intelligence to recognise favourable outcomes and knowing what to discard. This also implies that action needs to be taken to realise the value of the options.
We actually pay for optionality all the time. One example is purchasing insurance. Pay some fixed amount per year. If something goes wrong (Black Swan), we have the option to neutralize the effect.
But explicit, well-defined options, are expensive to purchase. Especially when the value of the optionality is identified and mapped in a contract such as an insurance policy. An even more expensive example of the additional premium paid for an option is flight tickets with flexible cancellation/date change policies.
However, there are other places where options remain underpriced, or not priced at all. We just strrugle to recognise due to the domain dependence of our minds. It certainly helps to be transdisciplinary.
The antifragile has options, and the fragile does not. An option is the weapon of antifragility.
3.5. Event vs implication of event
Now let’s bring all these concepts together.
- Non-linear means accelerating harm or loss. Increasing x has a disproportionately larger effect on f(x). Many systems are non-linear. And many people mistake non-linear to be linear.
- Convex is the good type of linear where fluctuations lead to more gains than losses overall. The convex wants volatility. The antifragile is convex.
- No one can predict Black Swans. We don’t know when they will come, where they will come from, and what exact impact it would have. All we know is that they’re always a possibility.
- Asymmetry plus rationality means optionality. Options don’t care about average outcomes, it only cares about the most favourable ones. Options like a dispersion of outcomes.
All of these more technical concepts, illustrated why the antifragile wants more volatility, dispersion, and randomness.
Now when it comes to the applying of our understanding of antifragility, we actually care more about the consequence of an event rather than the event itself.
- We care more about the payoff than the thing that causes the payoff
- We care more about the value of f(x), rather than the value of x.
- What happens to something as a consequence of some event, is not the same as the event itself.
“There is something (here, perception, ideas, theories) and a function of something (here, a price or reality, or something real). The conflation problem is to mistake one for the other, forgetting that there is a “function” and that such function has different properties.
Now, the more asymmetries there are between the something and the function of something, then the more difference there is between the two. They may end up having nothing to do with each other.”
This is the conflation problem. It’s very important not to conflate x with f(x).
Especially if f(x) is non-linear and asymmetrical. This would just amplify the difference between x and f(x) – especially for large x values.
We all do this all the time. Our brains are optimised to save energy. This is why it’s so important to put additional effort into re-training our neural networks, to sharpen our critical thinking continuously.
Oddly though, appreciating this distinction between x and f(x) is comforting.
Because x is hard to predict.
- Causes are complex
- Systems are interdependent
- Second order effects are extremely difficult to foresee
- In fact, for complex systems, there’s not enough atoms in the universe to compute all of the possible outcomes
- There’s hidden chaos in order, and hidden order in chaos
- Hence, the perpetual possibility of Black Swans
Even without fully understanding x, so long as we understand f(x), we can position ourselves to win more than lose.
We want favourable non-linearity: convexity. This exposes us to positive Black Swans rather than negative ones.
We want favourable asymmetry: where a few big, uncapped wins are much greater than the frequent small, capped losses.
We want optionality to ditch the bad and keep the good.
This way, it’s less about how often you’re right, and more about how much you gain every time you are right.
This way, it’s possible to benefit from a world without fully understanding it.
Embrace uncertainty. Welcome volatility.
Focus on the payoffs of an event rather than the event itself.
3.6. Limitations of antifragility
There’s limitations and nuances to antfiragility though.
“What doesn’t kill you makes you stronger.”
Yeah, but you can get permanently disabled.
Exposing your immune system to trace amounts (often deactivated/dead) of a pathogen makes you stronger. But injecting excessive amounts will just kill you. Your immune system is only antifragile up to a certain point.
Bone exposed to too little stress is weak. Bone exposed to the right amount is strong. But bone exposed to excessive stress, like jumping off a building, will break. Bone is only antifragile up to a certain point.
Children should not be over-protected, but neglecting them is even worse.
We can visualise this limitation as function with some convex sections, and some concave sections.
As the dose increases beyond a certain point, benefits reverse.
Convex exposure likes volatility (up to some point) over some range.
Balance matters. Context matters. It’s naive to take a mental model and apply it wholesale to everything. Many misinterpret a summary of Antifragile (and the summary itself likely a misinterpretation to begin with), and go around calling everything antifragile. “If you’re holding a hammer everything looks like a nail.”
I consider the learning-process-application cycle of any mental model in stages:
- (i) Never heard of it
- (ii) Heard of it
- (iii) Read a summary / someone told me about it
- (iv) Read the book. Recognise antifragility when it is present. “Wow I see antifragility everywhere”
- (v) Thought more about it. Oh wait, I’m confused
- (vi) Thought even more about it. Distinguish false antifragility from true antifragility, and recognise the nuances.
- (vii) Consciously apply it in life and work
- (viii) Unconsciously apply it in life and work. You don’t even realise you’re doing it.
- (ix) Go back to step (v) and repeat…
Stage (iii) and (iv) – “I think I know” – is most dangerous.
4. Potential misuse and abuse of antifragility
Furthering this warning on misintepreting antifragility is misusing and abusing it.
4.1. Stealing an option from society
Unfortunately in modern society, fragile disadvantages are often transferred from some onto others. Without also transferring the antifragile advantages. Reaping the benefits while someone else take all the risk. It’s like eating the top part of a cupcake and giving someone the rest.
“Antifragility for one is fragility for someone else.”
Let’s take the 2008 GFC for example. A select few at a select few financial institutions were benefiting from frequent, large wins without consideration of the hidden risks. Or rather, knowing that they won’t be affected by the risks. When things finally did blow up, it was the taxpayers that bailed them out. Too big to fail. Taxpayers ended up paying for the fragile consequences while a select few took home the bacon.
Essentially, these select few were stealing an option from society. Benefiting from the value of optionality, while the rest of society ended up paying for the option ‘premiums.’
This is part of a broader problem of misaligned incentives. Or more specifically, incentives without disincentives. All carrots, no sticks. Action without consequences. Not just at an individual, or group level. But at an institutional / system-design level. The commercial world works by addition rather than subtraction. Big pharma doesn’t gain by people avoiding sugar. There are hidden dangers of laissez faire capitalism.
So addressing such dangers goes beyond changing the behaviours of a few. The underlying design of the system needs to be updated. Perhaps, into a model that better distributes the benefits of antifragility more proportionally to the risks that are taken on. Not advocating socialism. Far from it. Rather, those that take risks should be paid their due rewards, instead of having it stolen by someone else.
4.2. Inconsistent utilitarianism
To be clear, this section is not drawn from the book. It’s just where my own thinking has taken me. I find exploring inconsistent utilitarianism to be relevant to the ethics of antifragility.
For those unfamiliar with ethical theories in philosophy:
- Deontology: rule-based; intention matters more than results; the means justifies the ends.
- Consequentialism: outcome-based; results matter more than intention; the ends justifies the means.
Utilitarianism is a subset of consequentialism where what’s morally right is determined by whatever gives the greatest number of people the greatest benefit.
In the interest of simplicity, I’ll walk through this point with the classic, and overly-used trolley problem.
- A trolley is fast-approaching 5 people. They’re all about to die.
- Frame 1. There is an option to pull a lever that diverts the trolley onto a different track. But this will kill 1 other bystander that would not have otherwise died. Would you pull the lever?
- Utilitarian ethics says yes. Divert the trolley, save the 5 and kill 1. This is a better outcome for the collective whole.
- Deontological ethics says no. Leave the trolley as is, let it kill 5 people. Because who gives you the right to determine who lives and who dies? Diverting the trolley would kill a person that would not have otherwise been killed.
- 70% say yes (i.e. choose the utilitarian solution). “Killing 1 person is an unfortunate side effect of saving 5 people. This is morally right.”
- But strangely enough this number changes when you frame the same options in a different way.
- Frame 2. Instead of pulling the lever, you now have to the push that 1 bystander onto the tracks with your own hands to save the 5. Would you push the bystander to their death?
- Only 30% say yes. The visceral reality of actually touching the person you’re about to assign a death note to is deeply discomforting. “5 people were saved because 1 person was killed. This is wrong.”
- Frame 3. This time, in order to pull the lever to divert the trolley, you have to quickly push someone standing in front of the lever out of the way, and as that person falls, they die. No other bystanders this time. 5 live saved, 1 killed. Would you do it?
- 80% say yes. “Pushing someone out of way is one thing. Them dying from the fall is another. It’s just an unfortunate side effect. They weren’t killed as a means to save 5.”
“Killing someone intentionally as a means to save five feels intuitively wrong, but the intuition is strongest when the killing would occur right here, right now; doing it in more complicated sequences of intentionality doesn’t feel as bad. This is not because of a cognitive limit – it’s not that subjects don’t realize the necessity of killing the person in the loop scenario. It just doesn’t feel the same. In other words, intuitions discount heavily over space and time.” – Behave, Robert Sapolsky
Bringing this back to antifragility and ethics. Earlier we said that antifragile systems require some of its components to be fragile. This is the mechanism of how it gains from disorder.
Sacrifice a few to save the many.
The first ethical warning here is recognising when antifragility is used as a philosophical justification for extreme utilitarianism. There’s political precedents for this: the rise of Nazi Germany.
The second, more prevalent, but more subtle warning is recognising different frames under which antifragility is articulated around utilitarian positions. Some could argue that efforts to protect the weak denies antifragile systems of their much needed disorder (let the weak die), and this ends up hurting the whole even more.
Because much like the trolley problem, people register very different opinions on what’s morally right based on how things are framed.
Biology. “Weak and damaged cells need to die so that the organism can survive and grow.” Fair enough. That’s the reality of nature. I don’t have a moral position on the live-or-kill decisions over my damaged cells. Actually, I want them to die.
Start-ups. “Weak and damaged businesses need to die so that the broader economy can survive and grow.” Fair enough. That’s the mechanism of capitalism. Efficiently allocate society’s finite resources into more value-adding ventures. In fact, I actually think shit start-ups (much more about leadership and people, and much less about the problem they’re solving, market they’re in etc) should die fast. This is better for the system.
Nation states. “Weak and damaged states need to die so that the future world can be better off.” Well history tells us that states come and go. But the part about whether a weaker state falling is better for the world is questionable – both logically and morally. Let’s explore this case in more depth to illustrate the challenges with using antifragility as a moral justification.
4.3. Extreme Darwinism at a nation state level
Empires and nation states rise and fall.
Consider the consolidation of Chinese political units between 2000 BC and 221 BC:
- 2,000 BC: 3,000 polities (Xia dynasty – Three Dynasties period)
- 1,500 BC: 1,800 polities (Shang dynasty)
- 1,200BC: 170 polities (Western Zhou dynasty)
- 770 BC: 23 polities (Eastern Zhou dynasty – Spring and Autumn (770-476 BC) period)
- 475 BC: 7 polities, (Eastern Zhou dynasty – Warring States (475-221 BC) period)
- 221 BC: 1 polity (Qin dynasty)
Basically, the Qin started a meritocracy experiment. They assigned high-level military positions based on competence rather than kinship ties. They won more battles. Soon, other polities began to copy their meritocratic recruitment process. Those that didn’t were conquered. Later on, the Qin started assigning government positions based on merit too. Any peasant that can read and write could sit an exam and work their way up to very senior positions. Meanwhile, most other states around the world didn’t do this until much later. This is supposedly one reason why Asians value studying so much. Education is a path out of poverty.
With morals aside and only looking at the mechanisms, I’d argue this is natural selection at a nation state level. Some were more politically organised than others. Some adapted. Others died. Over time, the politically organised became the structural norm.
Now consider the Spanish conquest of the Aztecs. Long story short: deception, fear, steel weapons, guns, and most catastrophically, germs. Is this also natural selection? Well, many would argue that the circumstances are materially different in this example. There’s a lot more factors at play. It’s not that simple.
Another consideration. Many pacific island nations are at the mercy of rising sea levels. If they die off, is that just disorder feeding into antifragility? Is it an insult to antifragility to step in and help them? I’m sure most would of course not. Because rising sea levels was not their fault. It was the industrial nations with high energy usage per capita, and high capita. But the weak pays the biggest price. Okay, so if it’s not someone’s fault they don’t deserve to be sacrificed for the greater whole? What about the hundreds of thousands peasants and soldiers in during the Warring States period in China. Was it their fault?
You can see where this is going. But I think this post is long enough.
These are not easy questions. The more you look into it the more tangled you realise it is. The more you learn, the more you realise how complex the world is.
5. Closing thoughts
- Antifragility is a powerful mental model
- Antifragile is beyond the resilient., It gains from disorder, volatility, uncertainty, stress, time etc.
- It’s applicable in so many domains
- An antifragile system needs some fragile components
- Modernity unfortunately artificially denies the antifragile from its much needed disorder
- This is particularly notable in the current economic paradigm – the relentless pursuit of efficiency
- Non-linear means accelerating harm or loss. Many mistake non-linear to be linear.
- Convex is the favourable type of linear. It wins more than it loses from volatility. The antifragile is convex.
- Black Swans are always out there.
- When you have optionality, you want dispersion. Because you have favourable asymmetry and you don’t care about average outcomes, only the most favourable ones.
- But, as with all mental models, we need to be wary of its limitations and misinterpretations
- The notion of antifragility can easily be misused and abused by justifying extreme utilitarian solutions
I’ll now wrap up by going back to why this book made such an impact on me so you can see where I’m coming from.
5.2. Personal development
Why was this the right book at the right time for me?
Exactly five years ago, I was the national strategy and operations director for a not-for-profit, a youth leadership development organisation in Vietnam. Unimaginably rewarding but excruciatingly demanding. Shit happened everyday. There were always multiple fires burning. It never ended.
During such times, I lived by the energy bank concept. Start the day at a neutral zero. Some things/ moments/ thoughts etc are energy gains/deposits: powerful conversations, small wins, good food etc. And some things are energy drains/withdrawals: arguing with idiots, when someone breaks a promise, shit weather etc. Hopefully the day ends with a net positive. Usually it’ll reset. But sometimes the balance can rollover to the next day and accumulate.
But after reading Antifragile, I rigged my energy bank. Pretty much like what central banks are doing right now… I forced what used to energy drains to flip into energy gains. Instead of letting shit drain my energy, I figured why not re-train my perception of reality to gain from it. Let shit that flies at you strengthen you rather than hurt you. Don’t just survive the storm, thrive in it. Instead of weathering challenges, welcome them. Better yet, proactively seek them. Disorder and volatility is bitter medicine. Unpleasant now, but I know this is good for me.
“Transmit positivity throughout uncertainty.”
(This was one of the ‘leadership development characteristics’ explicitly defined by that not-for-profit I worked for that I valued the most.)
“The problem is not the problem, the problem is how you see the problem.”
(Don’t know where I first got this from. But I went around saying it all the time, during my director tenure. And successfully annoyed lots of people with it.)
Going further, when I looked back at all my shitty moments in life, I realised I always come out in a better position. A better person. Sometimes, the most effective way forward was by taking a much needed a slap in the face. Wake up. Grow up.
“If you’re going to eat a shit sandwich, don’t nibble.”
(Ben Horowitz, Hard Thing About Hard Things)
Of course, this is not easy. If it were easy, everyone would do it and there’s no point in writing this. So I’m still learning. But the consciousness of the idea and intent was the start that made all the difference.
5.3. Antifragility as a complexity science mental model
So I valued Antifragile for its merits for practical application to personal development. But so many other books, articles, podcasts etc influenced me too. Mark Manson’s article Subtle Art of Not Giving a Fuck , for example, also influenced me a lot around the same time (in 2014). This was years before it turned into an immensely popular book.
So what made Antifragile so different? Why did it stand out so much?
Only recently I started to understand why.
It was my first proper glimpse into complexity science. An introduction to how its concepts serve as mental models (and meta mental models) that help with understanding the world. It had subconsciously made me see things with complexity lens all this time without explicitly knowing it was just complexity science. Complexity includes topics like:
- Modular silos vs systems view
- Domain dependence vs transdisciplinarity
- Non-linear vs linear
- First order effects vs second order effects
- Chaos theory
- Scaling laws and fractals
- Emergence, self-organising, cellular automata
- Information and computation
- …and more
But I could only appreciate this complexity science value the second time I read the book (in 2018). When I actually put the mental effort into digesting the technical/mathematical supporting concepts. Or, maybe I tried back in 2014 and I was just too stupid to get it. (This is why I wrote an 8k word blog post that could have been much shorter had I took out and further simplified the more technical sections. But I didn’t want to deny you from potentially getting as much out of it as I did.)
The interesting thing about studying complexity is, it puts you on this path to want to learn more horizontally. History, economics, business, philosophy, sociology, mathematics, science, finance, psychology, etc. That’s what it is. The book was kind of like the catalyst that turned my interest in how the world worked into an obsession.
5.4. If you wish to continue exploring
This post is by no means a summary of the book. It’s more like study notes for some of the key ideas in the book. Maybe I briefly covered on half or so.
It may not be the right book at the right time for you as it was for me. But hey, if you found the concepts interesting, I’d encourage you to consider exploring it further. Read the actual book.
“…This makes this book my central work. I’ve had only one master idea, each time taken to its next step, the last step—this book—being more like a big jump.” – Taleb
Thank you for reading. I hope this post serves you as well as it has served me in writing it.
If you liked Antifragile, you may like his other bestsellers too:
- Antifragile, Nassim Taleb (2012)
- Fooled by Randomness, Nassim Taleb (2001)
- Black Swan, Nassim Taleb (2007)
- Skin in the Game, Nassim Taleb (2018)
Found this helpful? You can return the love by either:
- buying me a coffee
- purchasing books via the links in my book list (or by clicking on the book cover images above)
- signing up for a free Audible trial (great for walks and drives)
- signing up for a free Kindle unlimited trial (recommend investing in a Kindle if physical books aren’t for you)
- signing up for a free Amazon Prime trial to help reduce your book costs
- sharing this post with your friends
Drop your email below for exclusive subscriber-only TLDR summaries of new posts and more.