Unlock the Secrets of Cause and Effect in 'The Book of Why': A Guide to the New Science of Cause and Effect

Italo de Pontes Oliveira: Os exemplos práticos são muitos bons, o inglês é de fácil entendimento. Recomendo.

Brazil on May 04, 2023

Snehlata: Satisfied but gave fuel to my urge to learn about causality. An amazing book overall

India on Apr 23, 2023

CG: Detallado en la exposición del tema.

Mexico on Jan 10, 2022

Olivier: A vivid question in epistemology is what comes first: data or models? An answer is that data fuel new models, and models fuel new experiments, which help discover new data. One says that observation is theory-laden. However, statistics and modern AI tend to rely exclusively on data. Not only correlation is not causation, but causation does not exist any more. The Book of Why is an attempt to replace causation in its due place. It is written by Judea Pearl, an expert in the domain of analyzing causation. The book is well written for the layman, though precise and informative. One may feel that sometimes the author lacks of modesty, but it is the price of asking an expert! The book makes a long tour of the question, through the history of statistics, and the pionneering works of Sewall Wright in genetics, to the modern frenzy of big-data and artificial intelligence. If for no other reason, this grand tour makes it worth reading the book.

France on Apr 16, 2021

Bai: Book by Judea Pearl, one of the leaders of causal inference who received a Turing award for inventing Bayesian networks. It assumes familiarity with probability and statistics, has a few equations (which can be skipped), with a level of technicality more than a popular science book and less than a textbook.

Causal inference is required because it's impossible to tell between causation and correlation from data alone, even with sophisticated deep learning techniques. For example, if birds chirp every day before sunrise, then a statistical method cannot tell you whether the birds are causing the sun to rise. Pearl gives three levels of causation, where each level can't be built up from tools of the lower levels.

- Level 1 -- Association: this is where most machine learning and statistics methods stand today. They can find correlations but can't differentiate them from causation.
- Level 2 -- Intervention: using causal diagrams and do-notation, you can tell whether X causes Y. The first step is to use this machinery to determine if a causal relation is possible from the data, then apply level 1 methods to compute the strength of the causality.
- Level 3 --...

Canada on Dec 26, 2019

Diana Senechal: The book's subtitle, The New Science of Cause and Effect, aroused both my skepticism and my curiosity: skepticism because I wondered how such a science could possibly be new, curiosity because I wanted to find out. The authors explain: Causal reasoning is ingrained in us and essential to our thinking, yet the human and social sciences often shy away from it, partly because they lack the proper models for its application. To stay on the safe side, people often speak in terms of "correlation" rather than causation. But this just evades the problem of causality, which can actually be described and tackled. The book shows how.

Reading it slowly, I reached the point where I could understand the explanations of the diagrams and formulas. I especially enjoyed Chapters 6 and 8 (on paradoxes and counterfactuals, respectively). Yet I was well aware, along the way, that to truly understand this subject--that is, to be able to create and apply causal models on my own--I would need to read the book several times, work through each of the examples, and then work independently on related problems. Even then, I could not guarantee that I would do this well, since causal reasoning requires...

United States on Jun 27, 2019

Aran Joseph Canes: The Book of Why is a popular introduction to Judea Pearl’s branch of causal inference. But it is also so much more.

Pearl has written many other textbooks introducing his graphical approach. But in this book, Pearl provides an engaging narrative of the history of causal inference, the important distinctions he sees in his branch and its importance for the future of Artificial Intelligence.

Briefly, Pearl views classical statistics as seriously flawed in not having developed a meaningful theory of causality. While able to demonstrate correlation, Pearl asserts that in classical statistics all relationships are two-way: that is 2x=3y+6 can also be written 3y=2x-6. We are left in doubt as to whether x causes y or y causes x.

Fundamentally, Pearl sees this problem as still plaguing all artificial intelligence and statistics. In its place, Pearl argues that the exact causal relationship between all variables should be explicitly symbolized in graphical form and only then can mathematical operations tease out the precise causal effect.

To be transparent, I am trained in the Rubin approach to causal inference and disagree with some of Pearl’s history and...

United States on May 17, 2018