The Book of Why

The field of causality explains how one can quantify direct and indirect effects of causes with the help of formulas and path diagrams. Previously science shied away from naming direct causal relationships, because it was too hard to unpick all the biasing effects that influence studies and data collections. All patterns were named correlations, nothing was causation.

Only random-controlled trials (double-blind, with a control group, careful selection of participants..) are a clean way of experimenting. But RCT studies are expensive and time-consuming, and sometimes would be unethical. And now we live in a world of so much observational (big) data. It would be terrible, if we'd continue to make mistakes based on wrong assumptions about that data. The book explains the methods of causal inference, that help to rid data of hidden distortions (by controlling for confounders).

The reading experience of this book was rather fluctuating. There is a certain arrogance that bleeds through when the author squabbles about this new field that he so clearly helped establish. Then in some sections in the middle the book just drifts too far into jargon, giving up all attempts at stringing the casual popsci reader along (in my opinion). But occasionally the book is very descriptive, explains examples clearly, then it is a delight, and super informative to read. Absolutely a fascinating field.

To be honest I didn't understand most of statistics formula explained, however insights into use case of causal models and differences were truly amusing.