Complexity provides a very interesting, very readable overview of several areas of science, looking at the underlying principles behind emergent behavior in computer networks, biology, ecology, economics, and neuroscience. The basic premise is that much of what human beings find important about the universe cannot be reduced down to simple laws, or built up from elementary explanations. Rather, chaotic dependence on initial conditions and the behavior of small elements 'seeing' only their immediate neighborhood can create wildly divergent and unpredictable macroscale effects, yet effects which exhibit a different kind of order.

This is pop science at it's best, with a solid text backed up by deep footnotes. Mitchell's area of expertise is genetic algorithms. Her graduate thesis was a genetic algorithm for discovering analogical rules under Douglas Hofstadter (which reminds me that I really need to reread GEB when I have the time), and they're the main subject area of the book, but lots of other topic, chaos, power laws, cellular automata, get explanations as well. Mitchell is optimistic and undogmatic about complexity. She acknowledges the deep divides between different definitions of terms, and a lack of really earthshattering applications, but hopes that complexity and information processing might prove a common language for 21st century science, in the way that calculus was a common language for 18th century science.