~~Moved from GR~~
The Tipping Point
by Malcom Gladwell
This is a brilliant and fantastically well-written popular science account of the importance of networks and complexity in our everyday lives. Even though I do research in this area, the narrative grabbed and held my attention. Every concept is illustrated and told via beautifully embedded and incredibly entertaining stories. At the same time, the book's style is so clear and simple that I would love to hand it around to all the non-computer scientists I know who ask me about my research. Although certainly far from scientific, this book certainly is more eloquent, entertaining, and informative than I think a scientist could ever be.
In his narrative, Gladwell simplifies and streamlines all of the topics it discusses--in fact, to put it in the book's own terms, The Tipping Point itself is a "connector" and a "salesman," taking the abstruse knowledge of the "maven" academics and translating it to mainstream. It did indeed create an epidemic of interest in the influence of the networks around us. However, one difficulty I have with the narrative is Gladwell's absolute certainty in the theories he supports. Possibly this is because I am all too aware of the well-documented failures in the hypotheses he presents as fact. Basically every story he tells concludes with "x succeeded because they did y." Although it makes for a coherent narrative, it made me wince. One of the most important rules of statistics (and one I think it is very important for non-scientists to grasp!!!) is that correlation does not imply causation. But Gladwell ignores this. He argues his points are indisputably right and uses correlations (and even worse, cherry-picked examples that behave the way he theorizes they should) to imply causation.
For example, he talks about the Broken Window Theory--that all the little details like graffiti create a tipping point that ends up creating the big crimes, and that fixing the windows will solve crime. He uses NYC, when Guiliani cleaned the streets and the crime radically decreased, as an example of where it was "proved." But he fails to note that NYC also introduced zero-tolerance for the "big" crimes at the same time, and that when LA tried to use graffiti-targeting tactics, they just didn't work. Thinking that correlation implies causation is bad, but cherry-picking only the data that correlates is even more problematic.
The book reminded me of The Drunkard's Walk: How Randomness Rules Our Lives for a rather strange reason: the books are polar opposites in their approaches to data. Mlodinov sets out to show that everything in life has randomness, and that although it is human nature to ascribe meaning to everything, often times, there's no reason for the way things are except randomness. Gladwell, on the other hand, believes that everything has an underlying story. In my opinion, both of these approaches are seductive, but dangerous. I would love to read a book co-written by these two: together, they would be the perfect mixture of belief in an underlying model and scepticism of the data's value.
Overall, Gladwell is a fantastic storyteller. His narrative is cohesive, entertaining, and convincing. This is a wonderful read both for individuals familiar with social network theory and people who are allergic to mathematics or computers.