"To do mathematics is to be at once touched by fire and bound by reason. Logic forms a narrow channel through which intuition flows with vastly augmented force."
How Not to be Wrong is easily the best book I've read in several months. (Maybe that doesn't sound too impressive, but it's actually pretty high praise.) It might be appropriately called How to Avoid Lying with Statistics and Occasionally Discover Truths, but that wouldn't be as catchy. Ellenberg gets straight to the point with real-life cases of problems that need the hell solved out of them, like whether to believe in God, judging whether a public health problem is becoming a crisis, and why elections are doomed to always be ambiguous depending on how you count votes. His examples enliven crucial concepts like Pascal's Wager, Euclidean versus non-Euclidean geometries, Bayesian inference, and the limitations of statistical "significance." My favorite chapter ties together how some MIT undergrads earned a stable income playing a (now-defunct) state lottery, the development of techniques to draw in perspective on flat paper, and the simultaneous invention and discovery error-correcting code. Ellenberg's presentation is eminently approachable: his engaging, humorous, and occasionally inspiring prose is accompanied by just a few equations (but enough of the right ones) and some simple yet explosively illuminating graphs. Over and over, this book reveals how same maths describe a lot of different problems, and the connections we can draw between them can powerfully multiply our skills at being not-wrong--and sometimes actually being right.
If you like Nate Silver, Martin Gardner, Radiolab, or simply not being wrong, you should give this book a try. You will discover something awesome and become smarter AND wiser, and probably even sexier.