Very Late Review: Antifragile (2012) by Nassim Taleb

(c) 'Accidental Fish', 2013

"Nothing convinces us of our capacity to make choices — nothing sustains our illusion of freedom — more than our ability to regularise our behaviour. nothing is more capable of destroying our interest and our pleasure in what we do.

If it is the predictable that stupifies us and the unpredictable that terrorises us, what should we do? If we are always caught between risk and resignation, between confidence and catastrophe, how can we decide what to do next?

— Adam Phillips

My problem is what my mother kept telling me:
I'm too messianic in my views.


The most ambitious and messy book in his idiosyncratic four-volume work of evolutionary epistemology, the 'Incerto'. (It is Fooled By Randomness, Black Swan, Bed of Procrustes, and yonder.) The former three books are largely critical, hacking away at theory-blindness, model error, and the many kinds of people he sees as possessing unearned status (economists, journalists, consultants, business-book writers): this is the upswing, a chaotic attempt to give general positive advice in a world that dooms general positive advice.

Every other page has something worth hearing, for its iconoclasm, or a Latin gobbet, or catty anecdote, if not something globally and evidently true. I think he is right about 30% of the time, which is among the highest credences I have for anyone. I only think I am 35% right, for instance.* But a core point of his system is that his approach should work even given our huge and partially intractable ignorance.

The core point, repeated a hundred times for various domains:

In real life, many systems deteriorate without an irregular supply of stressors (non-fatal negative events), and actually benefit from them by constructively overreacting. By robbing such 'antifragile' systems of stressors, modern approaches to managing them do damage in the guise of helping out.**

This observation leads to his grand theory of everything: every system is either fragile (damaged by volatility), robust (resistant to damage from volatility), or antifragile. This isn't a trivial distinction, because each has formal properties that allow us to change arrangements to, firstly, prevent explosions, and then to gain from chance volatility.

Biology is definitely one of these antifragile systems***; his case that, absent gross financialisation, the global economy would be one is convincing too; and the idea's at least plausible when applied to the cultivation of virtue or existential strength in a single person. The danger with this - an indissoluble danger because there can be no general strategy to avoid it - is that in welcoming constructive stress we'll miss the point at which the welcomed dose turns destructive (where fasting starts to atrophy, where training becomes masochism, where critique becomes pogrom, where sink-or-swim encouragement turns abuse).

* This claim is remarkable for both its extreme vagueness and apparent arrogance. Here is a post to handle the former fact. And the latter:

It might strike you as beyond arrogant to assume that you just so happen to be the most reliable inference device in the world, but that doesn't (have to) follow from my claim above, which results from the trivial thought “I believe my own beliefs most”, instead.

(Consider the converse: if I came to view anyone as more reliable than me, the rational thing to do would be to incorporate their truer views (and, better, their methods) until I again thought of myself as at least their equal. So, either one believes the superficially arrogant position “I believe my beliefs most” – or else one must believe that one is incapable of adapting enough to superior methods when faced with them, or else one must believe that one cannot know which methods are best. So the above assumption is more about having a high opinion of rational adjustment than impossible egotism, I think.

Good news! We can now calibrate ourselves, at least for the most sensational and available predictions using this cool thing.^

Finally!: The fully-unpacked, properly defensible assumption might be something more like: “I am the agent that I know to be most transparently reliable or unreliable; I assume I’m adjusting properly to better methods; as such I have at least equal confidence in my own belief set, compared to the best known alternative agent's.”)

^ You might wonder if this argument suggests that I should have 100% confidence in my beliefs. No; even if I was the best inferrer, I would suffer uncertainty because of the opacity of my errors: that is, I know I'm often not right but don't know exactly whereabouts I'm not right. Also from the unsystematic internal PredictionBook every non-psychopath has ("wisdom is knowing you'll be an idiot in the future"). And another source of uncertainty is down to the unknowable (like what stocks will crash next week).

I do worry that, whatever my particular self-credence estimate is, the whole approach is subtly wrong somewhere – since "40%" gives the impression that I think of myself as a worse guide to the world than dumb chance^^ – but I think it works. Particularly if much of the missing 60% is made of safe scepticism rather than errors.

^^ For binary event spaces – but, really, how many of those are there in real life?

** He credits the formal basis of all this to Jensen's inequality, in a chapter which might be the clearest expression of the idea there is.

*** (In particular species-level evolution, but also organism-level health.)


Some pigeonholes you might think to put all this in:

  • Conservative? Yes; but a good-hearted Burkean (“Antifragility implies that the old is superior to the new… What survives must be good at serving some purpose that time can see but our eyes and logical faculties can’t capture…”). Most people are conservative over some things (e.g. the natural world; we just happen to call that conservationism instead). Also approves of any high technology that removes anything he views as a disease of civilisation, like these things were supposed to be. So, in general, conservative only in the sense that existential risk people are.*

  • Economic conservatism. Only sort of; he's a trader, and would have speculation free to flow provided that deposit banks were nationalised first, and prioritises deficit reduction in a way we associate with conservatives but which e.g. Sweden pulled off without any lasting social justice sting. More formally he is against centralisation on both moral and technical grounds; that is likely a principle with some conservative effects, justified, in theory, by its keeping us alive. (Life-critical politics.)

  • Laissez faire? No: he recommends radical change to e.g. science funding, but no decrease. Big fan of Switzerland’s government, read into that what you will. He sees “optionality”, an originally financial concept, as the solution to fragility risks and the key to success in every domain there is. This isn’t at all as economistic as it sounds; the sacred and the humane somehow fit perfectly into his core rationalist agenda, persistence through change.

  • Social conservatism? No sign; no discussion of discrimination. Some people think such abstention is oppressive, but they are probably wrong.

  • Social Darwinist? Nah.

  • Bioconservative? Absolutely; he describes himself as the ‘diametric’ opposite of Ray Kurzweil, and he’s in full uproar over the global risk posed by synthetic biology (and recently fleshed out this horror in highly rigorous terms).

  • Anti-intellectual? Not at all! Only anti-academia, and they still do not represent the whole of quality intellectual life. Hates irresponsible ‘canned methods of inference’ too (statistical significance, etc).

  • Lacrimist? (That is, does he glorify suffering?) Not quite. He certainly views comfort as vitiating. His opposition to transhumanism is too quick and doesn't take the moral challenge of a world of pained beings seriously enough, for me.

  • Macho? Hm. Well, nature has made certain challenging actions optimal. Amusing proto-paleo attitude, too:
    I, for my part, resist eating fruits not found in the ancient Eastern Mediterranean (I use “I” here in order to show that I am not narrowly generalizing to the rest of humanity). I avoid any fruit that does not have an ancient Greek or Hebrew name, such as mangoes, papayas, even oranges. Oranges seem to be the postmedieval equivalent of candy; they did not exist in the ancient Mediterranean.

* His work fits the x-risk paradigm very well, but he developed his edifice in complete isolation from them, and has an uncompromising scepticism about expected value that might not make cross-overs all that fruitful.


How original is the core point, really?

Well, who cares? His claim is that he had to invent the word 'antifragile', not the idea. He says, idiosyncratically, that Seneca and Nietzsche had the nub of the idea, and Jensen the formal essence; Darwin certainly did too. "Resilience engineering" and in computing, 'defensive programming' (b. 1998) and 'self-healing systems' (b. 2001) are at least on the same track, though not getting beyond a lively sort of robustness. But I doubt that most systems can become antifragile - e.g. it's hard to imagine an antifragile jet engine (one that harvests bird strikes for fuel, or soot cleaning)? So maybe it's only the grand generalisation to all design that's new.


Gripes: His footnotes are collected by theme rather than linked to his claims directly, which makes it so difficult to follow up his sources that his credibility suffers. He namedrops, which is not the same as showing his working. I would really like to see his backing for his cool claims (about e.g. an irregular sleep pattern as a good thing, or things like ‘I suspect that thermal comfort ages people’), but it’s hidden away and often one-study. (Again: apparently one-study, since his working is not easily on show.)

He has a surprisingly high opinion of Steve Jobs – who I view as a grand example of an empty suit: there are 9 references to Jobs’ hokey shark-wisdom, (where Gigerenzer and Mandelbrot get 8, Jensen gets 7, Marx 7). Does Jobs really count as a ‘practitioner’ with ‘skin in the game’? Eh.

His homebrew jargon starts to drag – some sentences are wholly composed of his neologisms plus a barrel of articles and prepositions. (I used the glossary early and often.) Repetitive: tells what he’ll tell you, tells you he’s told you. Some passages really suffer from his wholesale hostility to copy-editing; there are some flatly bad sentences here. And he namedrops a lot, more than fair attribution of ideas – there are several passages that are just lists of people he likes (e.g. p.257-8).

I don't see that it's worthwhile to criticise his arrogant style; it's what animates his points, and he never uses it on weak targets.

Lastly, he sometimes makes of a system’s persistence the highest good. (Where its persistence is to be contrasted with mere stability.) This is in tension with his wonderful emphasis on artistic and quasi-sacred values elsewhere in the book.

But it talks about everything, is historically wide-eyed, relentlessly rational, and often funny. And the method-worldview-style it suggests might stop life crushing us utterly.


No comments:

Post a Comment