volatility –just consider that a certain structure of

COMMENTS ON COMMENTS uncertainty about the standard deviation will

Nassim Nicholas Taleb, April 22, 2007 produce some fat tails. If v is stochastic, with
variance v2, and v2 stochastic with variance v3, and
N- I sent copies of The Black Swan to 5 so on you can end up at the limit with a Levy
practitioners and 50 academics; 50 practitioners Process (some conditions: depends on if the
read it (I don’t know how), but only 5 academics nested vs are declining).
did so.
My rule of thumb: up to close to 5% probability, a
But I was lucky: among the academics was one of Gaussian with randomized std (stoch-vol) does
the most qualified humans on the planet, the the job. Beyond you need something else.
statistician Andrew Gelman. He is not just a
statistician but a thinker of statistics, so deep into I use Student T that effectively produce scalable
the subject that if he says something that tails of the form P>X – K x-α where α is the scale
contradicts my ideas, my “prior” is that he is right exponent and the degree of freedom. Or consider
and I am wrong. Luckily for me his notes are that you can decompose any distribution into a
more complementary –he lives in the “dual” space Gaussian with a number of Poisson jumps. The
of Bayesian statistics and this will prompt me to stochasticity is in the parameter α but we get
reexpress some of the ideas in that framework. good intuitions from that.

My comments are in red; his are in Black. The problem of course is that our degrees of
freedom in fitting models explode. Your error
about the estimate will be monstrous. If you can
April 9, 2007 do anything combining Gaussians, how you
combine them is infinite.
Nassim Taleb's "The Black Swan"
Effectively without a strong specification of a
OK, I finished reading it and transcribing my probability distribution our knowledge of the tails
thoughts. They're the equivalent of about 20 blog degrades very fast as the event is more remote.
entries (or one long unpublishable article) but it Me may know a bit about 1 in 10, not 1 in 10000,
seemed more convenient to just put them in one and certainly not about 1 in 106 unless we have a
place. very large sample and reasons to take it seriously.
As I noted earlier, reading the book with pen in The Black Swan to me comes from our inability to
hand jogged loose various thoughts. . . . The have a full grasp of the tails in decision making –
book is about unexpected events ("black swans") all we can do is avoid them.
and the problems with statistical models such as
the normal distribution that don't allow for these That said, I admit that my two books on statistical
rarities. From a statistical point of view, let me say methods are almost entirely devoted to modeling
that multilevel models (often built from Gaussian "white swans." My only defense here is that
components) can model various black swan Bayesian methods allow us to fully explore the
behavior. In particular, self-similar models can be implications of a model, the better to improve it
constructed by combining scaled pieces (such as when we find discrepancies with data. Just as a
wavelets or image components) and then chicken is an egg's way of making another egg,
assigning a probability distribution over the Bayesian inference is just a theory's way of
scalings, sort of like what is done in classical uncovering problems with can lead to a better
spectrum analysis of 1/f noise in time series. For theory. I firmly believe that what makes Bayesian
some interesting discussion in the context of inference really work is a willingness (if not
"texture models" for images, see the chapter by eagerness) to check fit with data and abandon
Yingnian Wu in my book with Xiao-Li on applied and improve models often.
Bayesian modeling and causal inference. More on black and white
(Actually, I recommend this book more generally;
it has lots of great chapters in it.) My own career is white-swan-like in that I've put
out lots of little papers, rather than pausing for a
N - Of course you can do a lot by combining few years like that Fermat's last theorem guy.
Gaussians –in fact everything. Take stochastic Years ago I remarked to my friend Seth that he's
followed the opposite pattern: by abandoning the On page xviii, Taleb discusses problems with
research-grant, paper-writing treadmill and social scientists' summaries of uncertainty. This
devoting himself to self-experimentation, he reminds me of something I sometimes tell political
basically was rolling the dice and going for the big scientists about why I don't trust 95% intervals: A
score--in Taleb's terminology, going for that black 95% interval is wrong 1 time out of 20. If you're
swan. studying U.S. presidential elections, it takes 80
years to have 20 elections. Enough changes in 80
On the other hand, you could say that in my
years that I wouldn't expect any particular model
career I'm following Taleb's investment advice--
to fit for such a long period anyway. (Mosteller
my faculty job gives me a "floor" so that I can
and Wallace made a similar point in their
work on whatever I want, which sometimes
Federalist Papers book about how they don't trust
seems like something little but maybe can have
p-values less than 0.01 since there can always be
unlimited potential. (On page 297, Taleb talks
unmodeled events. Saying p<0.01 is fine, but
about standing above the rat race and the
please please don't say p<0.00001 or whatever.)
pecking order; I've tried to do so in my own work
by avoiding a treadmill of needing associates to More generally, people (or, at least, political
do the research to get the funding, and needing commentators) often live so much in the present
funding to pay people.) that they forget that things can change. An
instructive example here is Richard Rovere's book
In any case, I've had a boring sort of white-swan
on Goldwater's 1964 campaign. Rovere, a
life, growing up in the suburbs, being in school
respected political writer, wrote that the U.S. had
continuously since I was 4 years old (and still in
a one-and-a-half-party system, with the
school now!). In contrast, Taleb seems to have
Democrats being the full party and the
been exposed to lots of black swans, both positive
Republicans the half party. Yes, Goldwater lost big
and negative, in his personal life.
and, yes, the Democrats did have twice the
N- I have certainly had many Black Swans in my number of Senators and twice the number of
life, s.a. wars, terminal diseases, a bestseller, and Representatives in Congress then--but, actually,
financial condition, but I wonder also if it is not from 1950 through 1990, the Republicans won or
because I introspected that I found them. Many tied every Presidential election (except 1964).
people of my background who experienced the Hardly the performance of a half-party.
same war, then went on to do well in finance do
N – My big problem is that we rarely make
not believe in my ideas –the fragility of
decisions in real life based on “probability” but
counterfactual outcomes. The big Black Swan is
rather probability times impact –the first moment
“being you” –quite an existential proposition. The
(or higher ones). So probability does not truly
odds of any of us being are so small...
matter on its own... I am writing a piece called
Chapter 2 of The Black Swan has a (fictional) “the fallacy of probability” on that.
description of a novelist who labors in obscurity
Knowing what you don't know, and
and then has an unexpected success. This
omniscience is not omnipotence
somehow reminds me of how lucky I feel that I
went to college when and where I did. I started The quotes on page xix remind me of one of my
college during an economic recession, and in favorites: "It ain't what you don't know that gets
general all of us at MIT just had the goal of you into trouble. It's what you know for sure that
getting a good job. Not striking it rich, just getting just ain't so" (Mark Twain?). I actually prefer the
a solid job. Nobody I knew had any thought that version that says, "It's what you don't know you
it might be possible to get rich. It was before don't know that gets you into trouble." Also Earl
stock options, and nobody knew that there was Weaver's "It's what you learn after you know it all
this thing called "Wall Street." Which was fine. I that counts."
worry that if I had gone to college ten years later,
On page xx, Taleb writes, "What you know cannot
I would've felt a certain pressure to go get rich.
really hurt you." This doesn't sound right to me.
Maybe that would've been fine, but I'm happy
Sometimes you know something bad is coming
that it wasn't really an option.
but you can't dodge it. For example, consider
95% confidence intervals can be certain diseases.
irrelevant, or, living in the present

© Copyright 2007 by N. N. Taleb.

N – You are right and I think that I am right: it is length by a jump every few days; they don't grow
a reduced form/structural form type of statement. the same amount every day).
Creativity is not (yet) algorithmic N- Now that I am done with the book, I realize
that more things come from jumps and fractures
On page xxi, Taleb says how almost no great
that I ever imagined ...
discovery came from design and planning. This
reminds me about a biography of Mark Twain that
I read several years ago. Apparently, Twain was
Aha
always trying to create a procedure--essentially,
an algorithm--to produce literature. He tried I was also reminded of the fractal nature of
various strategies, collaborators, etc., but nothing scientific revolutions--basically, at all scales
really worked. He just had to wait for inspiration (minutes, hours, days, months, years, decades,
and write what came to mind. centuries, . . .), science seems to proceed by
being derailed by unexpected "aha" moments.
Also on page xxi, Taleb writes "we don't learn
(Or, to pick up on Taleb's themes, I can anticipate
rules, just facts, and only facts." This statement
that "aha" moments will occur, I just can't predict
would surprise linguists. It's been well
exactly when they will happen or what they will
demonstrated that kids learn language through
be.)
rules (as can be seen, for example, from
overgeneralizations such as "feets" and Liberals and conservatives
"teached"). More generally, folk science is
On page 16, Taleb asks "why those who favor
strongly based on categories and natural kinds--I
allowing the elimination of a fetus in the mother's
think Taleb is aware of this since he cites my
womnb also oppose capital punishment" and "why
sister's work in his references. (A recent example
those who accept abortion are supposed to be
of naive categorization in folk science is in the
favorable to high taxation but against a strong
papers of Satoshi Kanazawa.)
military," etc. First off, let me chide Taleb for
N- You are right: I meant that we did not learn deterministic thinking. From the General Social
rules for large deviations/Extremistan. I use you Survey cumulative file, here's the crosstab of the
sister’s work for the psychology of induction: why responses to "Abortion if woman wants for any
is it that we are prepared to make adaptive reason" and "Favor or oppose death penalty for
inductive inferences in some domains and not murder":
others?
40% supported abortion for any reason. Of these,
Recognition, prevention, and saltatory 76% supported the death penalty.
growth
60% did not support abortion under all conditions.
On page xxiii, Taleb writes that "recognition can Of these, 74% supported the death penalty.
abe quite a pump." Yes, but recall all those
This was the cumulative file, and I'm sure things
scientists whose lives were shortened by two
have changed in recent years, and maybe I even
years (on average) from frustration at not
made some mistake in the tabulation, but, in any
receiving the Nobel Prize!
case, the relation between views on these two
N- Yes indeed I call for the elimination of prizes issues is far from deterministic!
because of such side effects/ creation of a
But getting back to the main question: I don't
gradiant...
think it's such a mystery that various leftist views
On page xxiv, "few reward acts of prevention": (allowing abortion, opposing capital punishment,
I'm reminded of our health plan in grad school, supporting a graduated income tax, and reducing
which paid for catastrophic coverage but not the military) are supposed to go together--nor is it
routine dental work. A friend of mine actually had a surprise that the opposite positions go together
to get root canal, and eventually got the plan to in a rightist worldview. Abortion is related to
pay for it, but not without a struggle. women's rights, which has been a leftist position
for a long time. Similarly, conservatives have
On page 10, Taleb writes, "history does not crawl,
favored harsher punishments and liberals (to use
it jumps." This reminds me of the evidence on
the U.S. term) have favored milder punishments
saltatory growth in infants (basically, babies grow

© Copyright 2007 by N. N. Taleb.

for a long time also. The graduated income tax worth. Being in debt for $10 million and thus
favors the have-nots rather than the have-mores, being "too big to fail" is (almost) equivalent to
and the military is generally a conservative having $10 million in the bank.
institution. Other combinations of views are out
N- Donald Trump was long an option when he
there, but I don't agree with Taleb's claim that the
started his big thing –Manufacturers Hanover
left-right distinction is arbitrary.
(from whom he borrowed heavily) was taking his
N – You are right about the determinism. I think huge risk for pennies. He had the upside; they
that I used statistics from a French Catholic “poll” kept the downside.
about guillotine from around the debate on its
The discussion on page 112 of how Ralph Nader
abolition around 1974. So I will rephrase the
saved lives (mostly via seat belts in cars) reminds
metaphor of clustering (there are so many of
me of his car-bumper campaign in the 1970s. My
these).
dad subscribed to Consumer Reports then (he still
My first shadow of non-agreement is on the “it is does, actually, and I think reads it for pleasure--it
not a mystery”. Are we sure we are not fitting an must be one of those Depression-mentality
explanation to the result? I want to think some things), and at one point they were pushing
more on that. Maybe you are right, and some heavily for the 5-mph bumpers. Apparently there
traits cluster naturally. I myself am both a hyper- was some federal regulation about how strong car
humanist and have strong libertarian tendencies bumpers had to be, to withstand a crash of 2.5
and tend to find this rare, mostly because selfish miles per hour, or 5 miles per hour, or whatever--
people who become libertarians may do so the standard had been 2.5 (I think), then got
because of their selfishness. raised to 5, then lowered back to 2.5, and
Consumer's Union calculated (reasonably
Picking pennies in front of a
correctly, no doubt) that the 5 mph standard
steamroller
would, in the net, save drivers money. I naively
On page 19, Taleb refers to the usual investment assumed that CU was right on this. But, looking at
strategy (which I suppose I actually use myself) it now, I would strongly oppose the 5 mph
as "picking pennies in front of a steamroller." standard. In fact, I'd support a law forbidding
That's a cute phrase; did he come up with it? I'm such sturdy bumpers. Why? Because, as a
also reminded of the famous Martingale betting pedestrian and cyclist, I don't want drivers to
system. Several years ago in a university library I have that sense of security. I'd rather they be
came across a charming book by Maxim (of gun scared of fender-benders and, as a consequence,
fame) where he went through chapter after stay away from me! Anyway, the point here is not
chapter demolishing the Martingale system. (For to debate auto safety; it's just an interesting
those who don't know, the Martingale system is to example of how my own views have changed.
bet $1, then if you lose, bet $2, then if you lose, Another example of incentives.
bet $4, etc. You're then guaranteed to win exactly
N- I agree. There are always second order effects.
$1--or lose your entire fortune. A sort of lottery in
reverse, but an eternally popular "system.") Three levels of conversation, or, why
lunch at the faculty club might
Throughout, Taleb talks about forecasters who
(sometimes) be more interesting than
aren't so good at forecasting, picking pennies in
hanging out with chair-throwing
front of steamrollers, etc. I imagine much of this
traders
can be explained by incentives. For example,
those Long-Term Capital guys made tons of On page 21, Taleb compares the excitement of
money, then when their system failed, I assume chair-throwing stock traders to "lunches in a drab
they didn't actually go broke. They have an university cafeteria with gentle-minded professors
incentive to ignore those black swans, since discussing the latest departmental intrigue." This
others will pick up the tab when they fail (sort of reminds me of a distinction I came up with once
like FEMA pays for those beachfront houses in when talking with Dave Krantz, the idea of three
Florida). It reminds me of the saying that I heard levels of conversation. Level 1 is personal:
once (referring to Donald Trump, I believe) that spouse, kids, favorite foods, friends, gossip, etc.
what matters is not your net worth (assets minus Level 2 is "departmental intrigue," who's doing
liabilities), but the absolute value of your net what job, getting person X to do thing Y, how to

© Copyright 2007 by N. N. Taleb.

get money for Z--basically, level 2 is all about agree with this and do it all the time in my
money. Level 3 is impersonal things: politics, research. But what about other aspects of life?
sports, research, deep thoughts, etc. When For example, I was reading The Black Swan,
talking with Dave, I resolved to minimize level 2 which I knew ahead of time would contain lots of
conversation and focus on the far more important information that I already agreed with. Should I
(and interesting) levels 1 and 3. Level 2 topics instead read a book on astrology? In practice, I'm
have an immediacy which puts them on the top of sure this would just confirm my (true) suspicion
the conversational stack, which is why I made the that astrology is false, so I'm kinda stuck.
special effort to put them aside. Anyway, it struck
N- The Black Swan is a skeptical statement that
me in reading page 21 of Taleb's book that chair-
destroys claimed knowledge, but does not come
throwing stock traders have much more
to a standard empirical solution except negatively
interesting level 2 conversations (compared with
(we know what is wrong). We know that astrology
professors or even grad students), and quite
is either wrong or unusuable or “charlatanism”
possibly they have better level 1 conversations
(Popper, according to some criterion). However
also--but I'd hope that the level 3 conversations
the solution is decision making by choosing your
at the university are more interesting. Being on
situations...
campus, I'm used to having all sorts of good level
3 conversations, but I find these harder to come Note one thing I edited out of the book because it
by in other settings. Probably it's nothing to do is not intuitive: you need to look for
with the depth of these other people, just that I counterexamples for the usual, but you can use
find it easier to get into a good conversational verification for the unusual. A White Swan not
groove with people at the university. In any case, taking place is a Black Swan. So the confirmation
I try (not always successfully) to keep discussion is a little more complicated.
conversations away from "the latest departmental
Rare events and selection bias
intrigue."
The footnote on 61 reminded me of a talk I saw a
N- Andrew, you are in the tails of academic
couple years ago where it was said that NYC is
success, with interesting people around you. I
expected to have a devastating earthquake some
had no guarantee of making it there –it was a
time in the next 2000 years.
remote bet in my mind at the time. And I get my
ideas backwards: first in the field, then in my On page 77, Taleb says that lottery players treat
library. odds of one in a thousand and one in a million
almost the same way. But . . . when they try
But my real agenda was to be left alone mature
making lottery odds lower (for example, changing
my ideas and work on my craft...
from "pick 6 out of 42" to "pick 6 out of 48,"
Riding the escalator to the stairmaster people do respond by playing less (unless the
payoffs are appropriately increased). I attribute
The story on page 54 about the people who ride
this not to savvy probability reasoning but to a
the escalator to the Stairmasters reminds me that,
human desire not to be ripped off.
where I used to work, there was a guy who
carried his bike up the stairs to the 4th floor. This On page 102 and following, Taleb discusses
always irritated me because it set an unfollowable selection bias. I also recommend the article by
example. For instance, one day I was on the Howard Wainer et al. (A Selection of Selection
elevator (taking my bike to the 3rd floor) and Anomalies); Deb Nolan and I also have a few in
some guy asked me, "You ride your bike for the our Teaching Statistics book.
exercise. Why don't you take the stairs?" (I
Then, on page 126, Taleb describes a conference
replied that I don't ride my bike for the exercise.)
he attended where his "first surprise was to
discover that the military people there thought,
behaved, and acted like philosophers [in the good
Confirmation bias, or, shouldn't I be
sense of the word] . . . They thought out of the
reading an astrology book?
bix, like traders, except much better and without
Around pages 58-59, Taleb talks about fear of introspection." He goes on to discuss why
confirmation bias and recommends that we look military officers are such good skeptical thinkers.
for counterexamples to our theories. I certainly But this seems like a clear case of selection bias!

© Copyright 2007 by N. N. Taleb.

The military officers who come to an academic Also, as discussed in our Teaching Statistics book,
symposium are probably an unusual bunch. when teaching probability I prefer to use actual
random events (e.g., sex of births) rather than
N- That was a military symposium with no
artificial examples such as craps, roulette, etc.,
selection bias beyond the choice of employment in
which are full of technical details (e.g., what's the
that unit. In other words, a weaker bias than that,
probability of spinning a "00") that are dead-ends
but perhaps a bias nevertheless.
with no connection to any other areas of inquiry.
Losers lie In contrast, thinking about sex of births leads to
lots of interesting probabilistic, biological,
On page 118-119, there's a discussion of how
combinatorical, and evolutionary directions.
someone with a winning streak in life can think
it's skill, even if it's just luck and selection (that Overconfidence as the side effect of
the losers don't get observed). I'd like to add communication goals
another explanation, which is that people lie.
On page 14, Taleb discusses overconfidence (as
Someone who tells you he won ten straight times
in the pathbreaking Alpert and Raiffa study). As
probably actually won ten times out of fifteen.
we teach in decision theory, there's actually an
Someone who tells you he broke even probably is
easy way to make sure that your 95% intervals
a big loser. Etc.
are calibrated. Just apply the following rule: Every
On page 125, Taleb explains why the Fat Tonys time someone asks you to make a decision, spin a
get more Nobel Prizes in medicine than the Dr. spinner that has a 95% chance of returning the
Johns. I don't know if this is really true, but if it is, interval (-infinity, infinity), and a 5% chance of
I might attribute it to the Tonys' better social skills returning the empty set. You will be perfectly
(i.e., helping others be happy and getting people calibrated (on average). The intervals are useless,
to do what they want) more than their better however, which points toward the fact that when
ability to assess uncertainty. people ask you for an interval, you're inclined (for
Gricean reasons if no other) to provide some
N- Our common friend Dan Goldstein and I are
information. According to Dave Krantz, much of
starting a whole program to tests the intuitions of
overconfidence of probability statements can be
“Fat Tonys” –if “those who think outside the box”
explained by this tension between the goals of
present the regularities that we ascribe to them.
informativeness and calibration.
N – This becomes the classical Gigerenzer
Of fights and coin flips problem with laboratory effects (resembles the
ludic fallacy). I am moving away from
On page 127-128, Taleb discusses the distinction
overconfidence from laboratory experiments and
between uncertainty and randomness (in my
found tests that are more ecological. However,
terminology, the boxer, the wrestler, and the coin
field studies such as those behind the planning
flip). I'd only point out that coins and dice, while
fallacy show the same distortion. Remember that
maybe not realistic representations of many
we are in Extremistan where more unknowns can
sources of real-world uncertainty, do provide
hit us. Also Goldstein and I did the test on subject
useful calibration. Similarly, actual objects rarely
using impact (not probabilities) and the pilot study
resemble "the meter" (that famous metal bar that
shows the effect being severe in fat-tailed
sits, or used to sit, in Paris), but it's helpful to
variables. We also took care to avoid using
have an agreed-upon length scale. We have some
“probability” but described frequencies. Finally, I
examples in Chapter 1 of Bayesian Data Analysis
tried to avoid using “overconfidence” and used
of assigning probabilities empirically (for football
“epistemic arrogance” for that reason.
scores and record linkage).
My central doubt is linked to my skepticism about
N – The way I see it is that a framework that
results that deal with probabilities without working
accepts no metaprobabilities is defective, period.
with payoff, and those that do not distinguish
This is where a Bayesian like you will never accept
between type 1 and type 2 randomness.
the difference “Knightian nonKnightian”. The only
one I accept is qualitative: ludic/nonludic. On page 145, Taleb discusses the fallacy of
assuming that "more is better." A lot depends
here on the statistical model you're using (or

© Copyright 2007 by N. N. Taleb.

implicitly using). With least squares, overfitting is hand, they were extremely theoretical, using
a real concern. Less so in Bayesian inference, but advanced mathematics to prove very subtle things
still it comes up with noninformative prior in probability theory, often things (such as the
distributions. An important--the important--topic strong law of large numbers) that had little if any
in Bayesian statistics is the construction of practical import. But when they did applied work,
structured prior distributions that let the data they threw all this out the window--they were so
speak but at the same time don't get afraid of using probability models that they would
overwhelmed by a flood of data. often resort to very crude statistical methods.
N – I meant that more information in life (not N- This is TBS! Episteme v/s techne. We are alas
statistics) is not necessarily better given our brain better at doing than knowing...
architecture and the cognitive costs of new
I'm only a statistician from 9 to 5
information and the confirmation bias. But I agree
that more is better if you are in Mediocristan or if I try (and mostly succeed, I think) to have some
you can process information without costs and unity in my professional life, developing theory
can adjust for the data mining effect that can that is relevant to my applied work. I have to
creep up when you have more variables. But you admit, however, that after hours I'm like every
need to have a strong prior. other citizen. I trust my doctor and dentist
completely, and I'll invest my money wherever the
Of taxonomies and lynx
conventional wisdom tells me to (just like the
In the discussion of Mandelbrot's work on page people whom Taleb disparages on page 290 of his
269, I'd also mention his models for taxonomies, book).
which have a simple self-similar structure without
Miscellaneous sociological thoughts
the complexities of the more familiar spatial
examples. Also, the story about the problems of Taleb's comment on page 155 about economics
Gaussian models reminds me Cavan Reilly's being the most insular of fields reminds me of this
chapter in this book, where he fits a simple story of the economist who said that economists
predator-prey model with about 3 parameters to are different than "anthropologists, sociologists,
the famous Canadian lynx data and gets much and public health officials" because economists
better predictions than the standard 11-parameter believe that "everyone is fundamentally alike"
Gaussian time series models that are usually fit to [except, of course, for anthropologists, etc.].
those data. Economists often do seem pretty credulous of
arguments presented by other economists!
Buzzwords
The reference on page 158 to dentists reminded
On page 278, Taleb rants against statistical
me of the dentists named Dennis.
buzzwords such as standar deviation and
correlation, and financial buzzwords such as risk. On page 166, Taleb disparages plans. But plans
This reminds me of my rant against the can be helpful, no? Even if they don't work out. It
misunderstood concept of "risk aversion." I have usually seems to me that even a poor plan (if
to write this up fully sometime, but some of my recognized as tentative) is better than no plan at
rant is here. all.
N – I would join you in the task. This brings us to N- I make plans. But I KNOW that they have an
the old Machina and Rothchild: risk aversion is error rate; and I know that I am vulnerable; and I
just what risk averse people do... know that a 25 year plan has trillions of times the
error rate of a 2 month plan.
It's all over but the compartmentalizin'
The key is identifying the domains where plans
On page 288, Taleb discusses people who
tend to fail.
compartmentalize their intellectual lives, for
example the philosopher who was a trader but The discussion on page 171 of predicting
didn't use his trading experiences to inform his predictions reminds me of the paradox, of sorts,
philosophy. I noticed a similar thing about some that opinion polls shift predictably during
of my collegues where I used to teach in the presidential nominating conventions (for evidence,
statistics department at Berkeley. On the one see here, for example), even though conventions

© Copyright 2007 by N. N. Taleb.

are very conventional events, and so one's shift in On page 209, Taleb writes, "work hard, not in
views should be (on average) anticipated. grunt work . . .". I have mixed feelings here. On
one hand, yes, grunt work can distract from the
On page 174-175, Taleb commends Poincare for
big projects. For example, I'm blogging and
not wasting time finding typos. For me, though,
writing lots of little papers each year instead of
typo-finding is pleasant. Although I am reminded
attacking the big questions. On the other hand,
of the expression, "there's no end to the amount
these little projects are the way I get insight into
of work you can put into a project after it's done."
the big questions. Getting in down and dirty,
N- You are first to hear this. Some typos in playing with the data and writing code, is a way
Fooled by Randomness have been there since that I learn.
2001: six years and half a million readers. The
N- The hardest work is in destroying a mental
best was Gladwell’s blurb “ is to conventional Wall
routine or push into a problem beyond the details.
Street wisdom what Luther’s ninety-nine theses
[not ninety five] were to the Catholic church” The mention on page 210 of Pascal's wager
stayed on the cover for a long time (two years) reminds me of the fallacy of the one-sided bet.
before a church historian wrote to me. I told her I'm hoping that now that this fallacy has been
that 4% error rate is very tolerable for us. named, people will notice it and avoid it on
occasion.
People do not read with an editor’s eye. They see
things differently. N- True. Ecologically we do not make simple bets.
Simple bets are rather laboratory contraptions.
The graphs on pages 186-187 have that ugly
Excel look, with unnecessary horizontal lines and The discussion on page 222 of capitalism,
weirdly labeled y-axes. In any case, they remind socialism, and attribution errors reminds me of
me of the game of "scatterplot charades" that I the saying that everybody wants socialism for
sometimes enjoy playing with a statistics class. themselves and capitalism for everybody else
The game goes as follows: someone displays a (and there's nothing more fun than spending
scatterplot--just the points, nothing more--and other people's money).
everyone tries to guess what's being plotted.
The discussion on the following page of the long
Then more and more of the graph is revealed--
tail reminds me of the conjecture about the "fat
first the axis numbers, then the axis labels--until
head" of mega-consumers.
people figure it out.
The footnote on page 224 about book reviews
I'm a little puzzled by Taleb's claim, at the end of
reminds me of a general phenomenon which is
page 193, that "to these people amused by the
that different reviews of the same book tend to
apes, the idea of a being who would look down on
have almost the exact same information. This
them the way they look down on the apes cannot
becomes really clear if you look up a bunch of
immediately come to their minds." I'm amused by
reviews on Nexis, for example. It can be
apes but can imagine such a superior being who
frustrating, because for a book I like, I'd be
would be amused by me. Why not?
interested in seeing lots of different perspectives.
N- We are different! In contrast, on the web the implicit rules haven't
been defined yet, so there's more diversity (as in
On page 196, Taleb writes, "a single butterfly
this non-review right here, or in these comments
flapping its wings in New Delhi may be the certain
on Indecision).
cause of a hurricane in North Carolina . . ." No--
there is no "the cause" (let alone, "the certain The comments on page 231 on the Gaussian
cause"). Presumably another butterfly somewhere distribution remind me of this story where even
else could've moved the hurricane away. Galton got confused about the tails of the
distribution as applied to human height.
Page 198: the chance of a girl birth is 48.5%, not
50%. On page 240, Taleb writes that Gauss, in using
the normal distribution, "was a mathematician
N-Bingo! But given that I did not know it, it is a
dealing with a theoretical point, not making claims
50-50 random number generator –Bayesian
about the structure of reality like statistical-
thinking.
minded scientists." I don't have my Stigler right

© Copyright 2007 by N. N. Taleb.

here, but I'd always understood that Gauss To conclude: it's fun (but work) to read a book
developed least squares and the normal manuscript with pen in hand. Also liberating that
distribution in the context of fitting curves to the book is already coming out, so instead of
astronomical observations. Sure he did lots of scanning for typos or whatever, I can just write
pure math, but he (and Laplace) were doing down whatever ideas pop up.
empirical science too.
N- To conclude it is fun (but not workl) to read a
N- I provide in the next paragraph the quote from review with pen in hand. It is also an honor.
Hardy who said “even Gauss” in his discussion of
P.S. Here are my thoughts on Taleb's previous
the aims of mathematics is to be useless. I talk
book.
about astronomy elsewhere.
I like Galileo's quote on page 257, "The great
book of Nature lies ever open before our eyes and Posted by Andrew at April 9, 2007 12:00 AM
the true philosophy is written in it. . . . But we
cannot read it unless we have first learned the
language and the characters in which it is written.
. . . It is written in mathematical language and the
characters are triangles, circles and other
geometric figures." As Taleb writes, "Was Galileo
legally blind?" Actual nature is not full of triangles
etc., it's full of clouds, mountains, trees, and other
fractal shapes. But these shapes not having
names or formulas, Galileo couldn't think of them.
He chose the natural kind that was closest to
hand. En el pais de los ciegos, etc.
N- This is Platonicity: pick the form that matches
the closest to you... and you have the cost of
mistaking the map for the territory.
On page 261, Taleb writes that in the past 44
years, "nothing has happened in economics and
social science statistics except for some cosmetic
fiddling." I'd disagree with that. True, I'm sure
you could find antecedents of any current method
in papers that were written before 1963, but I
think that developing methods that work on
complex problems is a contribution in itself.
There's certainly a lot we can do now that
couldn't be done very easily 44 years ago.
N- To me doing nothing is often far better than
accepting methods that do not “approximate”.
This is my idea about decision making under
ignorance. Going to the doctor before 1900,
particularly at the hospital La Pitié in Paris
quintupled your risk of death –but people were
blinded by the idea of “science”. The breakeven
happened well into the 20th century.
The FDA applies strict rules: no drug is strictly
better than a bad drug. But it took us 20 centuries
to reach such level of functional skepticism.
Reading with pen in hand

© Copyright 2007 by N. N. Taleb.