The Book of Why: The New Science of Cause and Effect

by Judea Pearl

My emphasis on language also comes from a deep conviction that language shapes our thoughts. You cannot answer a question that you cannot ask, and you cannot ask a question that you have no words for. As a student of philosophy and computer science, my attraction to causal inference has largely been triggered by the excitement of seeing an orphaned scientific language making it from birth to maturity.

I especially want to highlight the role of data in the above process. First, notice that we collect data only after we posit the causal model, after we state the scientific query we wish to answer, and after we derive the estimand. This contrasts with the traditional statistical approach, mentioned above, which does not even have a causal model.

Another advantage causal models have that data mining and deep learning lack is adaptability. Note that in Figure I.1, the estimand is computed on the basis of the causal model alone, prior to an examination of the specifics of the data. This makes the causal inference engine supremely adaptable, because the estimand computed is good for any data that are compatible with the qualitative model, regardless of the numerical relationships among the variables.

First, very early in our evolution, we humans realized that the world is not made up only of dry facts (what we might call data today); rather, these facts are glued together by an intricate web of cause-effect relationships. Second, causal explanations, not dry facts, make up the bulk of our knowledge, and should be the cornerstone of machine intelligence. Finally, our transition from processors of data to makers of explanations was not gradual; it was a leap that required an external push from an uncommon fruit. This matched perfectly with what I had observed theoretically in the Ladder of Causation: No machine can derive explanations from raw data. It needs a push.

The successes of deep learning have been truly remarkable and have caught many of us by surprise. Nevertheless, deep learning has succeeded primarily by showing that certain questions or tasks we thought were difficult are in fact not. It has not addressed the truly difficult questions that continue to prevent us from achieving humanlike AI. As a result the public believes that “strong AI,” machines that think like humans, is just around the corner or maybe even here already. In reality, nothing could be farther from the truth.

The goal of strong AI is to produce machines with humanlike intelligence, able to converse with and guide humans. Deep learning has instead given us machines with truly impressive abilities but no intelligence. The difference is profound and lies in the absence of a model of reality.

The rewards of having a causal model that can answer counterfactual questions are immense. Finding out why a blunder occurred allows us to take the right corrective measures in the future. Finding out why a treatment worked on some people and not on others can lead to a new cure for a disease. Answering the question “What if things had been different?” allows us to learn from history and the experience of others, something that no other species appears to do. It is not surprising that the ancient Greek philosopher Democritus (460–370 BC) said, “I would rather discover one cause than be the King of Persia.”

I totally agree with Yuval Harari that the depiction of imaginary creatures was a manifestation of a new ability, which he calls the Cognitive Revolution. His prototypical example is the Lion Man sculpture, found in Stadel Cave in southwestern Germany and now held at the Ulm Museum (see Figure 1.3). The Lion Man, roughly 40,000 years old, is a mammoth tusk sculpted into the form of a chimera, half man and half lion.

Nonetheless, I think that Turing was on to something. We probably will not succeed in creating humanlike intelligence until we can create childlike intelligence, and a key component of this intelligence is the mastery of causation.

This ability to emulate interventions by smart observations could not have been acquired had the statistical properties of Bayesian networks not been unveiled between 1980 and 1988. We can now decide which set of variables we must measure in order to predict the effects of interventions from observational studies. We can also answer “Why?” questions. For example, someone may ask why wiggling A makes C vary. Is it really the direct effect of A, or is it the effect of a mediating variable B? If both, can we assess what portion of the effect is mediated by B?

To navigate these new waters, we will have to understand the ways in which orthodox statisticians have learned to address causation and the limitations of those methods. The questions we raised above, concerning the effect of interventions, including direct and indirect effects, are not part of mainstream statistics, primarily because the field’s founding fathers purged it of the language of cause and effect. But statisticians nevertheless consider it permissible to talk about causes and effects in one situation: a randomized controlled trial (RCT) in which a treatment A is randomly assigned to some individuals and not to others and the observed changes in B are then compared. Here, both orthodox statistics and causal inference agree on the meaning of the sentence “A causes B.”