Nate Silver runs FiveThirtyEight.com, the polling site most accurate in the last election. His new book, The Signal and the Noise: Why Most Predictions Fail But Some Don't is a clear, well-reasoned, quite accessible book about the probabilistic nature of life. He discusses why weather forecasting predictions have gotten better over the decades, but earthquake forecasting hasn't, and why the majority of studies in medicine and in many other fields are probably fatally flawed in their statistical interpretation of data. Fundamentally the issue is one of separating signal from noise, of detecting the real signal (if there is one) and not being deluded by the meaningless noise that surrounds the signal.
This is a great book to read just after reading Nassim Taleb's recent book Antifragile, since both deal with much the same topic, how we humans misunderstand randomness and frequently misinterpret noisy data.