Everything is obvious

There are some really great summaries of this book on Goodreads: I like Deb's. Reviews like this have to be long because this book talks about so much stuff, but they are most useful after you've read the book. I'd like to highlight some things for people who haven't read it yet, and yet which might help those who have read it appreciate it more.

Duncan Watts is one of my heroes for two achievements: (1) expanding, with colleagues, the state of the art in path-dependent non-ergodic random processes theory, and (2) running, with colleagues, the famous MusicLab experiment, which demonstrated the power and real-world applicability of the first result. MusicLab, to me, is one of the great experiments of the last hundred years, and Watts himself has written a cogent summary of its meaning and relevance---highly, highly recommended, no matter where you think your interests lie, I guarantee you will find this article arresting. a more modern and appy description of the experiment is given by Kieran Healy on Flappy Bird. (I don't really have a great link for the first, other than Watts' previous books, [b:Small Worlds: The Dynamics of Networks Between Order and Randomness|373169|Small Worlds The Dynamics of Networks Between Order and Randomness|Duncan J. Watts|https://d.gr-assets.com/books/1348625225s/373169.jpg|363086] and [b:Six Degrees: The Science of a Connected Age|818170|Six Degrees The Science of a Connected Age|Duncan J. Watts|https://d.gr-assets.com/books/1348346215s/818170.jpg|804036].)

These two achievements establish him as more than a psychologist studying WEIRD (western, educated, industrialized, rich, democratic) college student volunteers. His research stints at Yahoo and Microsoft speak to a modern, digital-oriented approach to social science: his research is something hard-nosed engineering types like me cannot scoff at, no matter how silly we think sociology in general is. For these reasons, it was a no-brainer for me to pick this book up.

This book is composed to two parts, which take turns in the book. First is Watts's summarization and condensation of the combined literatures of cognitive psychology, behavioral economics, data science, but also philosophies of history and epistemology, forecasting and backcasting, and others from a very diverse set of academic pursuits. Summaries like Deb's (mentioned above) are valuable in helping readers keep track the sheer variety of experiments and analyses and hypotheses that Watts has professionally curated for a reader's consideration.

Interleaved with this extremely wide but carefully selected net of topics is Watts' commentaries on them, including his own experiments. And this is what I think some readers panning this book have missed. Because of the size, diversity, and technicality of the first part (discussing others' works), some people think that's all there is in this book: a summary of things they've read in Kahneman, Gigerenzer, Nozick and Rawls, and (my favorite) Arthur Danto. Watts has been humble in interleaving his own commentaries without much fanfare through the broader discussion, and the true value of the book is in what he, as someone who lives and breathes data-driven social experimentation, has to say.

For example, I have just reread the discussion of rational choice theory in chapter two. This is the idea of "homo economicus", bread and butter to classical economists, and today perhaps best known through the work of Freakonomics. Watts' details the theory and its many triumphs: how puzzling things like cheating teachers and lazy realtors and poor gang members and gift-reciprocating tribespeople playing Ultimatum can be explained, can be rationalized by finding the incentives people face. Once we unearth the potentially surprising and unexpected set of incentives at play in any given situation, puzzling phenomena can be reduced to rational actors maximizing their utility---as natural and simple as water finding its own level.

Watts is certainly not alone in criticizing rational choice theory, and he spends time giving a "litany" of cognitive biases that hamstring rationality, summarizing others' work: priming, framing, anchoring, availability, motivated reasoning, loss aversion, &c. But then he mounts an unusual (and possibly novel?) attack: one of the big problems with rational choice theory is that it can only explain phenomena in hindsight, and is terrible at predicting them, simply because the set of potentially relevant factors affecting a particular situation is so large (potentially infinite). Thus, while we should celebrate research that identifies the set of incentives at play in a specific behavior, we should also recognize that it won't help us with predicting behaviors in situations even slightly different.

It may be the case that this is just due to the youth of this field, and that at some point in the future, the number of incentives in their combinatorial complexity starts shrinking. But we're not there yet. And we may never get there.

This is to give an example of Watts' criticism with one specific topic. Other reviews give you a summary of just how many topics there are in this book, and you should know that for most of them you can find primary sources, but nearly all of them have this unique Wattsian twist: a criticism, an experimental extension, some connection to a totally different field that hasn't been considered before.

Therefore, this book has much novel content in it. It is much more than a summary of the state of the art in cognitive science or epistemology. Read it and be sure to pay attention to the humbly-offered fresh insights.