Jerzy's Reviews > Super Crunchers: Why Thinking-by-Numbers Is the New Way to Be Smart

Super Crunchers by Ian Ayres
Rate this book
Clear rating

by
57901
's review
Jan 26, 08

bookshelves: cog-sci-and-psych, statistics, economics
Read in January, 2008

Lots to think about. Highly recommended.

The cheesy subtitle makes it sound like a bad self-help book, but it's really a good, broad overview of the different ways that people are starting to use statistics more heavily in their decision-making. Basically, he explains that people are able to collect and store loads more data than ever before, so the "Super Crunchers" are fitting very accurate equations to the data -- and many of their formulas are better at making accurate predictions than seasoned experts with good intuition.
In a way, it's similar in style to Malcolm Gladwell's "Blink" but the opposite in its moral: Gladwell says to trust your intuition more than you think, and Ayres says to trust (experts') intuition less than we might think.

One thing that struck me was the idea of whether to trust the data and the resulting equation even when you don't understand the underlying mechanisms that the equation describes. For example, in the 1840s, a doctor named Semmelweis found that there was a huge drop in mortality rates at hospitals if doctors and nurses washed their hands. The germ theory of disease was not known yet, so Semmelweis couldn't explain *why* washing your hands would help keep your patients alive - he just knew it would. But many doctors ridiculed Semmelweis and his theory, complaining that hand-washing was a waste of time.
It's clear to us in retrospect that those other doctors should have listened to the man with the data instead of continuing to unwittingly kill their patients by refusing to wash their hands. But still today, too often, doctors tend rely on their intuition and the theories they learned in med school years ago, even if new studies have come out that show a different diagnosis or treatment would be better.
The same argument goes on in other fields. Ayres describes a scripted, structured, rote-learning method for elementary school instruction called Direct Instruction (DI). This seems entirely contrary to the creative, open, exploratory, problem-solving, student-centered philosophy that many people (including myself) subscribe to. But even though DI seems wrong and illogical, apparently DI students tend to perform significantly better on all kinds of tests - even ones that "required higher-order thinking," says Ayres. So should we stick with our instincts about what's a good way to teach our kids? Or should we go with the method that's been proven to give better results, even if it's anathema to our philosophy?

Of course, you can't have this argument if you're not able to measure your desired endpoint numerically. So you can argue that student performance on standardized tests is not how we should be evaluating our schools, and then you can ignore DI's successes that way. But then are we sure we're not just being like those 1840s doctors, saying "I expect that shouldn't work, so I'll ignore it," and possibly hampering millions of children's educational potential by ignoring something that does work?
Ayres clearly believes that expert opinions and theories are well and good when you can't crunch the data; but if good data is available and can be crunched, by all means trust it over the experts. Why argue about whether or not method A *should* be better than method B, when you can find out which one *does* perform better? It's a valid point and an important problem to think about.

Also, apparently some people have found ways to accurately predict the profits of a movie based solely on aspects of the script (ignoring the performance of actors, director, cinematography, etc). It's weird to think that the movie studios might begin rewriting scripts to maximize the expected profit based on an equation. Won't that take all the art out of it? But of course, Ayres points out, the studios already make many "artistic" decisions based purely on commercial (rather than artistic) merit. So all the equation does is improve the quality of those commercial decisions. If they're going to mess with your wonderful screenplay anyway, they may as well mess with with in the right way, instead of relying only on hunches about what works.

He has plenty of examples of cases where Super Crunching has been used for great good, as with MIT's Poverty Action Lab and similar groups that run experiments to find out what policies actually work in reducing poverty, and thus are able to convince governments to make the proven choice. On the other hand, there are also many scary examples of casinos crunching your personal data to figure out exactly how much money you can lose and still have a good time, or credit card companies learning exactly what interest rate to offer you to keep your business. Obviously businesses try to do this all the time - cutting costs and increasing revenue is what businesses are about, after all - but the degree to which they're now using your personal data (collated from many different sources, no less) is pretty intimidating.

Naturally, it sounds scary to think that our doctors, teachers, and everyone else should give up complete control over what they do to a bunch of formulas. But it's not all controlled by computers. At the top, somebody has to be in charge of the data, deciding what information to collect and what variables to put in the regressions and so on. So I wish Ayres talked more about the titular "Super Crunchers" themselves, not just their equations. Who becomes a super cruncher? What's it like to have a job where you create regression models, design experiments, and collect data? If this is the wave of the future, how should we train our students to be good at dealing with these huge data sets in a professional and responsible way?
likeflag

Sign into Goodreads to see if any of your friends have read Super Crunchers.
sign in »

No comments have been added yet.