Saturday, February 11, 2012

Another Extremely Fragile Definition On the Way

A new podcast gives us some insight, and quite possibly the entire plot, of Nassim Taleb's new book "Anti-fragility" due out later this year. When I say plot, I mean paragraph, quite possibly. And when I say paragraph I mean sentence. When I say sentence, I mean word. The word is "anti-fragility".

What struck me about this fascinating interview covering diverse topics (such as the effects on bone mass of interstellar travel - use it or lose it buddy) was the enormous effort Taleb had gone to in designing this new word. It is essential work, for as noted in The Black Swan and Fooled by Randomness, there is sometimes a dearth of words in the English language. For instance, there is no single word for anti-intellectual intellectual, someone who adopts populist tactics to attack anything remotely rigorous.

Yet sometimes there are many words for carefully delineated concepts related to missing data, bayesian estimation, subtle biases, sampling curiosities, statistical paradoxes and so forth. These are better replaced with a single word, like "silent evidence", that most people can at least remember, even if it isn't clear to anyone precisely what we are supposed to do with it.

Taleb's spectacular success in coining words and phrases rests on a massive pool of unconsidering humans who are suckers for pathetically thin observations like "use it or lose it". This massive reservoir was the topic of Maureen Tkacik's essay on Taleb's original promoter, Malcolm Gladwell. And it is not drying up. That might have something to do with a species known as the journalist, perhaps the only sector of society with no training in anything whatsoever who have nothing better to do than humor wafer thin piffle like the following:

"Then we get to the idea ... You take a car. You drive it against the wall at a tenth of a mile per hour, a hundred times or a thousand times. Have you done any harm, no. A tenth of a mile per hour won't harm you. The car will have some damage but it won't harm you. Now, drive the same car once at a hundred miles per hour...." 

This kind of thing gets a sympathetic giggle from interviewer Ross Roberts. Or perhaps he was just trying to keep the conversation going. You can almost hear the wheel's spinning in the guy's head, however, as he tries desperately to squeeze something profound out of the interview. But Taleb presses on just in case we didn't get the point, or appreciate his "universal notion of anti-fragility".

"This coffee cup I have on my desk has suffered a lot of shocks... but if I let it fall to the floor it will break ... you see?" 

Well yes Taleb, I think we do see, but I suggest those airline engineers continue to check for tiny fractures in turbines and other accumulated damage nonetheless.  You have to admire the boldness though, for not even Gladwell would dare repeat the same concept quite so many times. He would at least wrap it in a fascinating anecdote, or choose the three out of five hundred cases in a medical study that went against the trend (the exceptions that prove the rule). In Gladwell's defense he never claims a long gestation period whereas Taleb makes a point of his labor. Taleb carries metaphors on his back for decades for us - a weary selfless journey. It is kind of Taleb to finally dump these on us, though it might be the case that trading equity options for all those years has caused brain damage. Perhaps it causes one to fixate on unbelievably inane triviality like the fact that a payoff for a call option is non-linear. Use it or lose it.

But what of Taleb's previous attempts to create universal concepts? The Extremistan/Mediocristan distinction was laid before us in The Black Swan and was interpret as meta-guidance for applied mathematicians. Actually it was meta-nonsense for people who had never applied mathematics, never intended to apply mathematics, but wanted to think deep thoughts about other people who applied mathematics ... but let's set that aside and examine it on it's merits. Taleb presented a perfectly clear instruction manual. The world has linear and non-linear phenomena. But mathematicians have to be careful, he argued, not to use the wrong kind of mathematics. You wouldn't want to use linear mathematics to model a non-linear problem best treated with fractals, power laws and so forth now, would you? And I suppose you wouldn't want to use fractals to model a linear system either - though Taleb is highly skeptical that any exist so that is a decidedly less important topic.

Take web page popularity. It follows a Zipf distribution, landing it squarely in the groovy world of power laws. So it would be a massive mistake to try to apply linear mathematics to it, right? To me more precise I'd say it would be a massive mistake, a violation of a universal principal no less, to apply Linear Algebra to it, much less a singular value decomposition which is about as close to the epicenter of "linear mathematics" (to humor that ridiculous phrase) as one could surely get. Strange then, that Larry Page sought fit to do so when inventing Page Rank, the algorithm distinguishing Google from Yahoo that powered the greatest commercial success in recent history.

What are we to make of the non-linear/linear meta-advice, and for that matter the rest of Taleb's ranting?  Use it or lose it, one presumes, but I just quite get my head around the former possibility. Maybe that's because I've been banging my head against The Black Swan to gently, for too long. It's non-linear. One big whack and everything will be clear.

No comments:

Post a Comment