Top critical review
Reviewed in the United States on March 18, 2013
Fisher describes his book writing process as making notes of what interests him, and then writing to a deadline. It shows. "The Perfect Swarm" touches on lots of interesting work, and Fisher does try to relate everything to everyday life. Even when the everyday life example is not a direct application of an algorithm, it can give the reader a feel for it.
The deadline is what gets Fisher into trouble (to put it nicely). Fisher has a huge fear of writing above the reader's comprehension. In chapter one he uses a box for some more difficult material so that some readers may skip the box (an approach Sharon Blakeslee uses so effectively). Fisher does not use a box again until the penultimate chapter. Instead he refers to his end of book notes, which have some nuggets mixed in with simple references and odds and ends, none of these notes footnoted, even the references - you are supposed to just read through them (one note seems to discredit the scientist, Milgram, Fisher happily discusses in the main text). Fisher also avoids technical "details", so I was left with basic questions about the algorithms he discusses. Actually, Fisher seems to prefer lists to algorithms, and many of these are quite boring.
Fisher is often imprecise and perhaps it is just as well he avoids much technical material. In discussing Benford's law he refers to logarithms beginning more often with low digits, but as he states, the logarithms are random - this causes the numbers to which they refer, such as numbers in an accounting statement, to more often begin with low digits. In explaining Ramsey's Law, Fisher omits the "at least" repeatedly (p.164) as in the number of people necessary for there to be either (at least) "m" mutual acquaintances or "n" mutual strangers". In discussing degrees of separation he talks of 100 people, each of them knowing 100 people (p.110) without qualifying that there has to be no overlap in the people who are known for his example to work. When Fisher discusses "group think" he confounds cases where the actors have financial incentive to group think (bankers before financial crisis), or are culturally conditioned (ethnic bias) to group think which arises directly from shorter term peer influence. In discussing the problems of over fitting models with too many variables he muddies the basic point by suggesting that such variables really help explain the past rather than mistaking chance for causality. Based on a stock market heuristic that worked for only 6 months Fisher tries to make a point.