Can drinking coffee help people live longer? What makes a stock’s price go up? Why did you get the flu? Causal questions like these arise on a regular basis, but most people likely have not thought deeply about how to answer them.
This book helps you think about causality in a structured way: What is a cause, what are causes good for, and what is compelling evidence of causality? Author Samantha Kleinberg shows you how to develop a set of tools for thinking more critically about causes. You’ll learn how to question claims, identify causes, make decisions based on causal information, and verify causes through further tests.
Whether it’s figuring out what data you need, or understanding that the way you collect and prepare data affects the conclusions you can draw from it, Why will help you sharpen your causal inference skills.
Chapter 1Beginnings: Where do our concepts of causality and methods for finding it come from?
What is a cause?
How can we find causes?
Why do we need causes?
Chapter 2Psychology: How do people learn about causes?
Finding and using causes
Chapter 3Correlation: Why are so many causal statements wrong?
What is a correlation?
What can we do with correlations?
Why isn’t correlation causation?
Multiple testing and p-values
Causation without correlation
Chapter 4Time: How does time affect our ability to perceive and reason with causality?
The direction of time
When things change over time
Using causes: It’s about time
Time can be misleading
Chapter 5Observation: How can we learn about causes just by watching how things work?
The limits of observation
Chapter 6Computation: How can the process of finding causes be automated?
Chapter 7Experimentation: How can we find causes by intervening on people and systems?
Getting causes from interventions
Randomized controlled trials
Are experiments enough to find causes?
Chapter 8Explanation: What does it mean to say that this caused that?
Finding causes of a single event
Explanation with uncertainty
Separating type and token
Causality in the law
Chapter 9Action: How do we get from causes to decisions?
This book is for readers with an interest in science who want a deeper understanding of the methods of causal thinking, its cognitive obstacles, and its potential for effecting change in the world. Little prior knowledge in mathematics or statistics is required, but that doesn't mean it is easy reading. Part of the reason is that, as the author warns in the first chapter, no definition of causality covers all cases, and each definition has counter-examples that another does not. Thus, thinking about causality requires the ability to keep in mind a wide variety of cases and counter-cases, some more subtle than others.
The book has three aspects that I found most compelling. The first one is the multitude of examples used to illustrate concepts or motivate their introduction. These examples range from the trivial every-day situation to famous court cases, from psychological experiments to clinical trials, from stock market data to paradoxes in physics. A second compelling aspect is the book's extensive discussion of three major traditions of causal thinking: philosophical, psychological, and mathematical/computational. All three traditions are needed to inject clarity into cognitive biases and puzzling statistical paradoxes. Thirdly, the book provides some valuable recipes for attacking causal problems. There is for example John Stuart Mill's list of methods for learning about causes (pages 79-85), and Bradford Hill's considerations for evaluating causal claims (pages 180-188).
A possible weakness is the book's occasionally meandering style of exposition. For instance, a new concept may be introduced with lots of examples instead of a clear definition. This is particularly frustrating in the longest chapter, on computation. I would have been hard pressed to give a simple definition of Granger causality after reading about it in this chapter, so I turned to Wikipedia for help. Most of the time the author succeeds in making her material accessible to a non-mathematical audience, but obviously this approach is not without limitation. The discussion of Simpson's paradox for example, is quite informative, but ends with the somewhat limited observation that level of granularity matters when looking at data, as does background knowledge about the underlying problem. While this is true, Simpson's paradox typically serves to draw a very fundamental distinction between statistical and causal thinking. Fortunately the book's bibliography is quite extensive and useful, and I was able to sharpen my understanding of the paradox by studying a cited paper by Judea Pearl. Overall the book succeeds in presenting a remarkable number of key ideas, and together with the bibliography this is an excellent starting point for further study.
Bottom Line Yes, I would recommend this to a friend
Great primer for newcomers, timely reminder for experts
from Leeds, England
About Me Designer, Educator
Easy to understand
Comments about oreilly Why:
As a humanities graduate in an increasingly data-driven world, I found this book a great introduction to the key concepts and methods of finding causes. The author has a refreshingly straightforward approach, bringing in wide-ranging examples from the domains of medicine, law, public policy and personal decision-making. She is also very clear about the practical and moral limits of current methods, in a way that should give big data's cheerleaders pause for thought. Reading this book has got me thinking about how I can better integrate qualitative and quantitative evidence in my practice as a service designer.
Bottom Line Yes, I would recommend this to a friend