[owaty]

To counter some of the comments here, I absolutely loved the book and went on to recommend it to all my scientist friends. While it may get a bit technical for the lay audience, it should be within reach for a typical scientist or IT person. I wish our society had a better understanding of causality—that would raise the level of many important discussions.

Being a long-time fan of Gelman (and having studied his Bayesian Data Analysis textbook), I am baffled and disappointed that he doesn't seem to understand Pearl's ideas. In his linked 2009 post[1], he wrote: "I’ve never been able to understand Pearl’s notation: notions such as a “collider of an M-structure” remain completely opaque to me." I wonder if, after reading this book accessible even to non-statisticans, he still doesn't understand it.

[1]: https://statmodeling.stat.columbia.edu/2009/07/05/disputes_a...

[mindcrime]

To counter some of the comments here, I absolutely loved the book and went on to recommend it to all my scientist friends.

Likewise (well, other than not really having any "scientist friends"). I loved this book, think Pearl has some amazingly valuable ideas, and found the book relatively accessible even though I'm not a statistician. I won't claim to have understood every detail on the first reading, but I got enough out of it to feel like I'll understand it all after a couple of follow on readings, plus consulting Pearl's other books.

I wish our society had a better understanding of causality—that would raise the level of many important discussions.

Absolutely.

[sjg007]

To be fair Pearl does not speak in the same terms as the statisticians. Collider is a weird word and an M structure is even weirder. Graphical models suffer similarly.