Taka's Reviews > The Drunkard's Walk: How Randomness Rules Our Lives

The Drunkard's Walk by Leonard Mlodinow
Rate this book
Clear rating

's review
Dec 24, 2011

it was amazing
bookshelves: 2-nonfiction-you-must-read, iu-fall-2011, popular-science, 2015

Even better the second time--

This little book is just so good—not only does it give you just enough math to make you feel curious and satisfied, it tells a ripping good story about probability theory and statistics, providing along the way compelling portraits of the eccentric scientists and mathematicians who contributed to the fields. This time, I wanted to refresh my memory of all the thorny problems probability and statistics give us (we are really, really bad at intuiting probability, as psychologists have again and again shown us).

One good refresher among many was the fallacy we make in dealing with conditional probability, mostly prominently manifested in conspiracy theories and paranoid thoughts: these events happened, therefore there is a huge conspiracy. Or close to home (for me at least): an agent hasn't gotten back to me yet, therefore she must not like my work. Probability-wise these are based on the wrong probabilities, and logically, these are equivalent to the fallacy of affirming the consequent (if P then Q, Q, therefore P). So from the valid, highly probable inference, "If there is a huge conspiracy, these events happen" or "If an agent doesn't like my work, she will not respond for a long time," we see the consequent—these events happened, or the agent hasn't responded in a long time—and draw the mistaken conclusion that there is a huge conspiracy. Or the agent doesn't like my work. What's wrong with this is that there are so many possible reasons why a series of events occurred other than due to a huge conspiracy, or why the agent hasn't gotten back to me in a long time (she's just busy!). In probability terms:

the probability that she he doesn't respond to me in a long time GIVEN she doesn't like my work

is high and valid, whereas:

the probability that an agent doesn't like my work given she has not responded to me in a long time

is low (because there could be all sorts of reasons why she hasn't responded to me).

Just to hammer it home, this can be illustrated with a simple example:

If you are human, you eventually die.

is a perfectly valid conditional, but

You (pointing at a squirrel) eventually die, so you (the squirrel) must be human.

is definitely not.

More importantly, what I for some reason failed to write about in my 2012 review and totally forgot about until I reread the book is Yale sociologist Charles Perrow's normal accident theory Mlodinow mentions in the last chapter, how disasters in complex systems occur when many little human mistakes just happen to coincide at just the wrong (or right—depending on your perspective) time. I was so interested in this theory that I actually bought the seminal book by Perrow himself (and duly put on the shelf for "read immediately").

Excellent, excellent book.

[Read 1/5/2012] Awesome--

This book made me admire what modern statistics—a topic I couldn't care less—is capable of doing and convinced me, like Taleb's The Black Swan and Burton Malkiel's Random Walk Down Wall Street how randomness really rules our lives and it's important to recognize chance events and not mistakenly assign them some causality that's not there. The history of probability theory and statistics Mlodinow tells in this book is nothing short of fascinating, and I was floored by the answers to some of the problems he so deftly presents.

For example:

1) there are three doors. Behind one of them is a treasure, and behind two are geese. You pick a door. The host of the show opens one of the doors you've picked and show geese behind it. Is it better to switch your choice?

The answer: yes. You will increase your probability of winning from 1/3 to 2/3. Why? Read the book to find out why.

2) The Attorney's Fallacy. Take the O.J. Simpson trial. The prosecutor argued O.J. Simpson was an abusive husband. The defense attorney Alan Dershowitz then argued that the probability of an abusive husband killing his wife is so low, the prosecutor's argument for O.J.'s propensity for violence is misguided. In more detail:

4 million women are battered annually by their husbands and boyfriends in the U.S.
Yet in 1992, a total of 1,432 women (or 1 in 2,500) were killed by their husbands or boyfriends.
Therefore, few men who beat their wives or girlfriends go on to murder them.

Convincing, but that's not the relevant probability. The relevant probability is rather: the probability that a battered wife who was murdered is murdered by her abuser. And of all the battered women murdered in 1993 in the U.S.some 90% were killed by their abuser.

Then there's the reassuring implication that success comes to you largely by random—publication, prizes, business success, fame, etc.—and that means the longer we persevere, the better our odds are of succeeding. As an aspiring writer, this non-deterministic paradigm of looking at the world has helped me boost my confidence and determination.

A must-read.

10 likes · flag

Sign into Goodreads to see if any of your friends have read The Drunkard's Walk.
Sign In »

Reading Progress

December 24, 2011 – Shelved
Started Reading
January 5, 2012 – Shelved as: iu-fall-2011
January 5, 2012 – Shelved as: popular-science
January 5, 2012 – Shelved as: 2-nonfiction-you-must-read
January 5, 2012 – Finished Reading
June 7, 2015 – Shelved as: 2015

Comments (showing 1-3)

dateUp arrow    newest »

message 3: by Soo (new) - added it

Soo Ha! I can always count on you to point me towards books I wouldn't have looked at on my own. ;)

Taka I live to serve, Soo :)

message 1: by sanny (new) - added it

sanny You have some very interesting non-fiction reads and I'd love to see what you're reading or recommending next. I was quite fascinated by randomness after Taleb's books and this one would be on my to-buy-list. Thanks for the convincing review!

back to top