Cause and Effect

Judea Pearl and Dana Mackenzie delve into causality and its role in science and AI

Published 5 years ago on Jul 14, 2018 2 minutes Read

In ordinary language, the question “Why?” has at least two versions. The first is straightforward: you see an effect, and you want to know the cause. Your grandfather is lying in the hospital, and you ask, “Why? How could he have had a heart attack when he seemed so healthy?” But there is a second version of the “Why?” question, which we ask when we want to better understand the connection between a known cause and a known effect. For instance, we observe that Drug B prevents heart attacks. Or, like James Lind, we observe that citrus fruits prevent scurvy. The human mind is restless and always wants to know more. Before long we start asking the second version of the question: “Why? What is the mechanism by which citrus fruits prevent scurvy?” This chapter focuses on this second version of “why.”

The search for mechanisms is critical to science, as well as to everyday life, because different mechanisms call for different actions when circumstances change. Suppose we run out of oranges. Knowing the mechanism by which oranges work, we can still prevent scurvy. We simply need another source of vitamin C. If we didn’t know the mechanism, we might be tempted to try bananas.

The word that scientists use for the second type of “Why?” question is “mediation.” You might read in a journal a statement like this: “The effect of Drug B on heart attacks is mediated by its effect on blood pressure.” We want to ask certain typical questions about a mediator: Does it account for the entire effect? Does Drug B work exclusively through blood pressure or perhaps through other mechanisms as well? The placebo effect is a common type of mediator in medicine: if a drug acts only through the patient’s belief in its benefit, most doctors will consider it ineffective. Mediation is also an important concept in the law. If we ask whether a company discriminated against women when it paid them lower salaries, we are asking a mediation question. The answer depends on whether the observed salary disparity is produced directly in response to the applicant’s sex or indirectly, through a mediator such as qualification, over which the employer has no control.

All the above questions require a sensitive ability to tease apart total effects, direct effects (which do not pass through a mediator), and indirect effects (which do). Even defining these terms has been a major challenge fo rscientists over the past century. Inhibited by the taboos against uttering the word “causation,” some tried to define mediation using a causality-free vocabulary. Others dismissed mediation analysis altogether and declared the concepts of direct and indirect effects as “more deceptive than helpful to clear statistical thinking.”

This is an extract Judea Pearl and Dana Mackenzie's The Book of Why published by Allen Lane