The Book of Why
| Company type | Nonfiction book |
|---|---|
| Industry | Publishing |
| Founded | 2018 |
| Founder | Judea Pearl, Dana Mackenzie |
| Products | Book on causality and causal inference |
The Book of Why: The New Science of Cause and Effect is a nonfiction publication from 2018, authored by computer scientist Judea Pearl alongside writer Dana Mackenzie. It explores the concept of causality and causal inference through both statistical and philosophical lenses, targeting a broad readership.
Summary
The volume consists of an introduction followed by ten chapters.
Introduction: Mind over Data
The opening section critiques early statistical approaches from the 20th century for their failure to establish cause-and-effect relationships. It then introduces the 'Causal Revolution,' a transformative period in the mid-1900s that brought forth innovative conceptual frameworks and mathematical techniques for understanding causality.
Chapter 1: The Ladder of Causation
In this chapter, the 'ladder of causation' is introduced, outlining three distinct tiers of causal reasoning. The initial tier, called 'Association,' deals with identifying links between variables without implying cause—such as noting that a rooster’s crow coincides with sunrise. Conventional statistical methods, including correlation and regression, function at this stage.
The second tier, 'Intervention,' examines how intentional changes influence outcomes, addressing questions like "Does smoking raise the risk of lung cancer?" This stage explicitly involves causal mechanisms beyond mere associations.
At the top level, 'Counterfactuals' investigates hypothetical alternatives—pondering scenarios like "Would Kennedy still be alive if Oswald had not assassinated him?" This level represents the most sophisticated form of causal analysis.
Chapter 2: From Buccaneers to Guinea Pigs: The Genesis of Causal Inference
This chapter reviews the early statistical contributions of Francis Galton and Karl Pearson, highlighting Pearson's influence in restricting discussions about causality within statistics. It also introduces causal analysis through path diagrams, emphasizing the pioneering work of Sewall Wright.
Chapter 3: From Evidence to Causes: Reverend Bayes meets Mr Holmes
Here, Bayes' Theorem and Bayesian Networks are presented and linked to causal diagrams, laying the groundwork for understanding probabilistic approaches to causal inference.
Chapter 4: Confounding and Deconfounding, or, Slaying the Lurking Variable
This section clarifies the concept of confounding and demonstrates how causal diagrams help detect confounders and their effects. It stresses that although randomized controlled trials (RCTs) can effectively remove confounding biases, carefully designed causal models may sometimes serve as alternatives to RCTs.
Chapter 5: The Smoke-filled Debate: Clearing the Air
Focusing on historical controversies regarding smoking and lung cancer, this chapter reviews positions held by Abraham Lilienfeld, Jacob Yerushalmy, Ronald Fisher, and Jerome Cornfield. Despite observed correlations, skeptics attributed these to confounding factors to challenge claims of causation. The authors illustrate how contemporary causal reasoning validates cigarettes as a cause of cancer.
Chapter 6: A Wealth of Paradoxes
This section delves into various paradoxes such as the Monty Hall Problem, Simpson's paradox, Berkson's paradox, and Lord's paradox, demonstrating how applying causal inference can clarify these puzzling statistical phenomena.
Chapter 7: Surpassing Adjustment: Mastering the Peak of Intervention
Addressing the 'second tier' on the ladder of causation, this chapter explains how causal graphs are utilized to evaluate the impact of interventions—for instance, the effect of smoking on lung cancer. It introduces analytical techniques like the 'front-door criterion' and 'do-calculus'. The chapter includes two illustrative examples: John Snow’s discovery linking cholera to contaminated water sources, and investigations into the connection between cholesterol levels and heart attack risk, incorporating instrumental variables for determining causal effects.
Chapter 8: Counterfactuals: Exploring Alternate Realities
Focusing on the highest level of the causation ladder—counterfactuals—this chapter presents structural causal models that facilitate reasoning about hypothetical scenarios beyond conventional statistical methods. Practical applications in fields such as climate science and legal contexts underscore their significance.
Chapter 9: Mediation: Uncovering Underlying Mechanisms
This chapter concentrates on mediation, examining the pathways through which causes bring about effects. It highlights research including Barbara Stoddard Burks’ studies on childhood intelligence, Chicago public schools’ implementation of an 'algebra for all' initiative, and the use of tourniquets in treating battlefield injuries.
Chapter 10: Big Data, AI, and Fundamental Questions
The concluding chapter explores the importance of causal reasoning within big data analytics and artificial intelligence (AI). It addresses the philosophical challenge of enabling AI systems to reflect on their own decisions, a process that inherently requires counterfactual—and thereby causal—reasoning.
Reviews
Judea Pearl’s official website provides scientific context, sample excerpts, corrections, and 37 reviews related to The Book of Why.
Jonathan Knee offered a favorable critique in The New York Times, describing the book as "illuminating," while noting that some parts may prove "challenging" for readers without a background in mathematics.
Tim Maudlin presented a mixed assessment in The Boston Review, lauding it as a "splendid overview" but expressing reservations about separating counterfactuals within the ladder of causation and its approach to free will. He also pointed out earlier contributions by researchers like Clark Glymour, suggesting Pearl might have streamlined his efforts by acknowledging them. Pearl responded that he was aware of this prior work during his writing.
Zoe Hackett’s review in Chemistry World was positive, emphasizing the book’s intellectual demands and recommending it particularly for philosophy of science students and introductory statistics courses.
Lisa R. Goldberg delivered an in-depth technical critique in Notices of the American Mathematical Society.