Book review: The Drunkard's Walk
Photo credit:
Cover of The Drunkard's Walk at Amazon.com The Drunkard's Walk: How Randomness Rules Our Lives by Leonard Mlodinow.
The human brain has a powerful capability to spot patterns. It's so good at spotting patterns, it can spot patterns that aren't even there. It's this fact that makes randomness--the topic of Mlodinow's new book--so interesting and so confusing.
Let's say you're a subject in this experiment: A scientist has rigged up a light to flash red or green, and arranged things so that it flashes red twice as often as green. You guess whether it will be red or green--and you get a reward every time you're right. What do you do?
Well, if you're a rat, you observe that the red is the most common and then always guess red--that maximizes the reward. If you're a human, though, you study the sequences and try to figure out a pattern. If there is a pattern, and it's not too complex for you to figure it out, this can be a winning strategy--you can get the reward every time. If there is no pattern, though, you'll be routinely beaten by the rat.
Mlodinow's book is a fascinating excursion into the realm of random events, and a surprisingly interesting history of the mathematical tools that people have come up with to analyze events that are or might be random. (It's also the only book that has ever made me think, "Gee, I should get a good statistics text--the math here is really interesting.")
These issues matter in the real world. Take, for example, drug testing. Let's say you have a test for (illegal or performance-enhancing) drug users that's a pretty good test--it detects 90% of the drug users while only fingering an innocent person 0.1% of the time.
Suppose you give this test to 100 student athletes at a school who are suspected of drug use. If 20 of the students actually are drug users, odds are that you'll catch 18 of them, and do so without accusing an innocent student.
Suppose, though, that you give the same test to a much larger population that includes very few drug users--perhaps a large technical facility with 20,000 scientists and engineers, only a dozen of whom are drug users. Look at what your results are now: You successful detect 10 or 11 of the drug users, but you also "detect" 20 people who aren't drug users at all--you've got more false positives than true positives! (People who don't understand randomness make this sort of mistake all the time.)
In addition to providing an interesting walk through the history of people's understanding of randomness, and an introduction to the mathematics of dealing with these problems, Mlodinow provides some useful tips on being successful in a world where random chance has a lot to do with success. (The gist of which is summed up in a quote from IBM pioneer Thomas Watson: "If you want to succeed, double your failure rate.")
Some things in life are random and other things aren't. The Drunkard's Walk is filled with interesting stories of just how far astray people can go when they don't have the tools and the insight to figure out which situation they're in.
Editor's note: Mr. Mlodinow has generously offered 7 of our lucky readers a free copy of his new book. For a chance to win, just pop over to this forum thread before 6/30/08 and leave us a comment.
Disclaimer: The links and mentions on this site may be affiliate links. But they do not affect the actual opinions and recommendations of the authors.
Wise Bread is a participant in the Amazon Services LLC Associates Program, an affiliate advertising program designed to provide a means for sites to earn advertising fees by advertising and linking to amazon.com.
Hows this for random: I was on a book design blog this morning that i have never visited before, and it showcased this book cover. Weird.
I love books about how our minds work! I signed up for the contest/giveaway. Thanks for the informative review!
Thanks for the review, and the change to win a copy!
Judging from the comments on the "win this book" thread, there are many people like me who have been regulars here for a long time but just needed something like this to convince us to register. You've started a great registration campaign!
This book sounds a lot like CHANCE. Interesting topic.
This book sounds like Innumeracy. I'm a big maths geek, so I love this sort of stuff. Although, I'm always concerned that I'll be the defendant in a trial and be wrongfully convicted because neither the jury, nor the expert witness understood statistics. Has actually happened in the UK to at least 2 people - the Royal Statistical Society complained about the trials, but that wasn't pointed out until they were up for appeal.
I haven't read it, but I've also heard that Fooled by Randomness is good; maybe I'll read these two back to back and see what comes up. Against the Gods also has admirers, but IIRC it was a trifle disappointing. There are also books about more general cognitive biases, e.g. Cialdini's Influence, that may be useful to frugal folk.
(The links are to LibraryThing, since that is the service I subscribe to. They are available on Amazon, &c.)
This sounds absolutely facinating. I think I need to read the whole thing.
Thanks, too, for the pointers to other books on related topics--I'll check them out.
Even in this example, if you run a second test on these then you will still have 9-10 positive results that are correct and then .02 false positives. Each subsequent run seriously increased the effectiveness without taking nearly as much out of the correct positives.
Sounds like a very interesting book, maybe I should start reading more on statistics and such, I loved that class in school.
That's a good example of how randomness matters.
If the false positives are random, then repeating the test would (probably) finger a completely different cohort of 20 victims--one would have to be very unlucky to be falsely fingered twice. On the other hand, if the false positive was due to some systemic factor, you might get a false positive every time.
So, it's important to understand which factor is at play.
Also, it may well be that they're already running the test twice, in order to get the 0.1% false positive rate, in which case you might have to run it four times to get the false positive rate down to less than one person in the population.
And, of course, once you accept that some fraction of false positives is okay, you're also accepting that, no matter how small the fraction is, you're going to have some people who lose out, even if the odds are millions-to-one against. (After all, the odds are millions-to-one against winning the lottery, but someone wins one somewhere, almost every week.)
Sounds like a fascinating book, and I'd love to know which episodes of STNG he wrote as well.
A quick check of the Leonard Mlodinow page at IMDB.com gives his Star Trek credits as:
I was driving home from yet another evening work event a few weeks ago feeling bad for myself and wondering why I ended up working where I work when NPR did an interview with the writer and his book. I heard it just when I needed to and it lifted my spirits. I'm doing everything I can to land a new great job and the interview inspired me to "double my failure rate" which to me meant doubling the amount of work I'm doing to secure the kind of job I want. Thanks for the inspiration.