In J.

Pearl, Causality: Models, Reasoning, and Inference, New York:

Cambridge University Press, pp. 401-428, 2009.

EPILOGUE

The Art and Science of Cause and Effect

A public lecture delivered November 1996 as part of

the UCLA Faculty Research Lectureship Program

The topic of this lecture is causality – namely, our awareness of what causes what in the
world and why it matters.
Though it is basic to human thought, causality is a notion shrouded in mystery, con-
troversy, and caution, because scientists and philosophers have had difficulties defining
when one event truly causes another.
We all understand that the rooster’s crow does not cause the sun to rise, but even this
simple fact cannot easily be translated into a mathematical equation.
Today, I would like to share with you a set of ideas which I have found very useful
in studying phenomena of this kind. These ideas have led to practical tools that I hope
you will find useful on your next encounter with cause and effect.
It is hard to imagine anyone here who is not dealing with cause and effect.
Whether you are evaluating the impact of bilin-
gual education programs or running an experiment
on how mice distinguish food from danger or spec-
ulating about why Julius Caesar crossed the Rubi-
con or diagnosing a patient or predicting who will
win the presidential election, you are dealing with
a tangled web of cause–effect considerations.
The story that I am about to tell is aimed at
helping researchers deal with the complexities of
such considerations, and to clarify their meaning.
This lecture is divided into three parts.
I begin with a brief historical sketch of the
difficulties that various disciplines have had with
causation.
Next I outline the ideas that reduce or elimi-
nate several of these historical difficulties.

401
402 Epilogue

Finally, in honor of my engineering back-

ground, I will show how these ideas lead to sim-
ple practical tools, which will be demonstrated in
the areas of statistics and social science.
In the beginning, as far as we can tell, causal-
ity was not problematic.
The urge to ask why and the capacity to find
causal explanations came very early in human
development.
The Bible, for example, tells us that just a few
hours after tasting from the tree of knowledge,
Adam is already an expert in causal arguments.
When God asks: “Did you eat from that tree?”
This is what Adam replies: “The woman whom
you gave to be with me, She handed me the fruit
from the tree; and I ate.”
Eve is just as skillful: “The serpent deceived me, and I ate.”
The thing to notice about this story is that God did not ask for explanation, only for
the facts – it was Adam who felt the need to explain. The message is clear: causal ex-
planation is a man-made concept.
Another interesting point about the story: explanations are used exclusively for pass-
ing responsibilities.
Indeed, for thousands of years explanations had no other function. Therefore, only
Gods, people, and animals could cause things to happen, not objects, events, or physical
processes.
Natural events entered into causal explanations much later because, in the ancient
world, events were simply predetermined.
Storms and earthquakes were controlled by the
angry gods [slide 2] and could not in themselves
assume causal responsibility for the consequences.
Even an erratic and unpredictable event such
as the roll of a die [3] was not considered a chance
event but rather a divine message demanding
proper interpretation.
One such message gave the prophet Jonah the
scare of his life when he was identified as God’s
renegade and was thrown overboard [4].
Quoting from the book of Jonah: “And the
sailors said: ‘Come and let us cast lots to find out
who is to blame for this ordeal.’ So they cast lots
and the lot fell on Jonah.”
The Art and Science of Cause and Effect 403

Obviously, on this luxury

Phoenician cruiser, “casting lots”
was used not for recreation but for
communication – a one-way mo-
dem for processing messages of vi-
tal importance.
In summary, the agents of
causal forces in the ancient world
were either deities, who cause
things to happen for a purpose, or human beings and animals, who possess free will, for
which they are punished and rewarded.
This notion of causation was naive, but clear and unproblematic.
The problems began, as usual, with engineering; when machines had to be con-
structed to do useful jobs [5].
As engineers grew ambitious, they decided that
the earth, too, can be moved [6], but not with a sin-
gle lever.
Systems consisting of many pulleys and
wheels [7], one driving another, were needed for
projects of such magnitude.
And, once people started building multistage
systems, an interesting thing happened to causal-
ity – physical objects began acquiring causal
character.
When a system like that broke down, it was
futile to blame God or the operator – instead, a
broken rope or a rusty pulley were more useful
explanations, simply because these could be replaced easily and make the system work.
At that point in history, Gods and humans ceased to be the sole agents of causal
forces – lifeless objects and processes became partners in responsibility.
A wheel turned and stopped because the wheel preceding it turned and stopped – the
human operator became secondary.
Not surprisingly, these new agents of causation took on some of the characteristics
of their predecessors – Gods and humans.
Natural objects became not only
carriers of credit and blame but
also carriers of force, will, and
even purpose.
Aristotle regarded explanation
in terms of a purpose to be the only
complete and satisfactory expla-
nation for why a thing is what it is.
404 Epilogue

He even called it a final cause – namely, the final

aim of scientific inquiry.
From that point on, causality served a dual role:
causes were the targets of credit and blame on one
hand and the carriers of physical flow of control
on the other.
This duality survived in relative tranquility [8]
until about the time of the Renaissance, when it
encountered conceptual difficulties.
What happened can be seen on the title page
[9] of Recordes’s book “The Castle of Knowl-
edge,” the first science book in English, published
in 1575.
The wheel of fortune is turned, not by the wis-
dom of God, but by the ignorance of man.
And, as God’s role as the final cause was taken
over by human knowledge, the whole notion of
causal explanation came under attack.
The erosion started with the work of Galileo [10].
Most of us know Galileo as the man who was brought before the inquisition and
imprisoned [11] for defending the heliocentric theory of the world.
But while all that was going on, Galileo also managed to quietly engineer the most
profound revolution that science has ever known.
This revolution, expounded in his 1638 book
“Discorsi” [12], published in Leyden, far from
Rome, consists of two maxims:
One, description first, explanation second –
that is, the “how” precedes the “why”; and
Two, description is carried out in the language
of mathematics; namely, equations.
Ask not, said Galileo, whether an object falls
because it is pulled from below or pushed from
above.
Ask how well you can predict the time it takes
for the object to travel a certain distance, and how
that time will vary from object to object and as the
angle of the track changes.
Moreover, said Galileo, do not attempt to an-
swer such questions in the qualitative and slippery
nuances of human language; say it in the form of
mathematical equations [13].
The Art and Science of Cause and Effect 405

It is hard for us to appreciate today how strange

that idea sounded in 1638, barely 50 years after
the introduction of algebraic notation by Vieta. To
proclaim algebra the universal language of science
would sound today like proclaiming Esperanto the
language of economics.
Why would Nature agree to speak algebra? Of
all languages?
But you can’t argue with success.
The distance traveled by an object turned out
indeed to be proportional to the square of the time.
Even more successful than predicting out-
comes of experiments were the computational as-
pects of algebraic equations.
They enabled engineers, for the first time in
history, to ask “how to” questions in addition to
“what if” questions.
In addition to asking: “What if we narrow the
beam, will it carry the load?”, they began to ask
more difficult questions: “How to shape the beam so that it will carry the load?” [14]
This was made possible by the availability of methods for solving equations.
The algebraic machinery does not discriminate among variables; instead of predicting
behavior in terms of parameters, we can turn things around and solve for the parameters
in terms of the desired behavior.
Let us concentrate now on Galileo’s first
maxim – “description first, explanation second” –
because that idea was taken very seriously by the
scientists and changed the character of science
from speculative to empirical.
Physics became flooded with empirical laws
that were extremely useful.
Snell’s law [15], Hooke’s law, Ohm’s law, and
Joule’s law are examples of purely empirical gen-
eralizations that were discovered and used long
before they were explained by more fundamental
principles.
Philosophers, however, were reluctant to give
up the idea of causal explanation and continued to
search for the origin and justification of those suc-
cessful Galilean equations.
For example, Descartes ascribed cause to eter-
nal truth.
406 Epilogue

Liebniz made cause a self-evident logical law.

Finally, about one hundred years after Galileo,
a Scottish philosopher by the name of David Hume
[16] carried Galileo’s first maxim to an extreme
[17].
Hume argued convincingly that the why is not
merely second to the how, but that the why is to-
tally superfluous as it is subsumed by the how.
On page 156 of Hume’s “Treatise of Human
Nature” [18], we find the paragraph that shook up
causation so thoroughly that it has not recovered
to this day.
I always get a kick reading it: “Thus we re-
member to have seen that species of object
we call flame, and to have felt that species of
sensation we call heat. We likewise call to mind
their constant conjunction in all past instances.
Without any farther ceremony, we call the one
cause and the other effect, and infer the existence of the one from that of the other.”
Thus, causal connections according to Hume are the product of observations. Cau-
sation is a learnable habit of the mind, almost as fictional as optical illusions and as
transitory as Pavlov’s conditioning.
It is hard to believe that Hume was not aware
of the difficulties inherent in his proposed recipe.
He knew quite well that the rooster crow
stands in constant conjunction to the sunrise, yet it
does not cause the sun to rise.
He knew that the barometer reading stands in
constant conjunction to the rain but does not cause
the rain.
Today these difficulties fall under the rubric of
spurious correlations, namely “correlations that
do not imply causation.”
Now, taking Hume’s dictum that all knowl-
edge comes from experience encoded in the mind
as correlation, and our observation that correlation
does not imply causation, we are led into our first
riddle of causation: How do people ever acquire
knowledge of causation?
We saw in the rooster example that regular-
ity of succession is not sufficient; what would be
sufficient?
The Art and Science of Cause and Effect 407

What patterns of experience would justify call-

ing a connection “causal”?
Moreover: What patterns of experience con-
vince people that a connection is “causal”?
If the first riddle concerns the learning of
causal connection, the second concerns its usage:
What difference would it make if I told you that a
certain connection is or is not causal?
Continuing our example, what difference
would it make if I told you that the rooster does
cause the sun to rise?
This may sound trivial.
The obvious answer is that knowing “what
causes what” makes a big difference in how we act.
If the rooster’s crow causes the sun to rise,
we could make the night shorter by waking up
our rooster earlier and making him crow – say, by
telling him the latest rooster joke.
But this riddle is not as trivial as it seems.
If causal information has an empirical meaning beyond regularity of succession, then
that information should show up in the laws of physics.
But it does not!
The philosopher Bertrand Russell made this argument [19] in 1913:
“All philosophers,” says Russell, “imagine that causation is one of the fundamental
axioms of science, yet oddly enough, in advanced
sciences, the word ‘cause’ never occurs. . . . The
law of causality, I believe, is a relic of bygone age,
surviving, like the monarchy, only because it is er-
roneously supposed to do no harm.”
Another philosopher, Patrick Suppes, who ar-
gued for the importance of causality, noted that:
“There is scarcely an issue of ‘Physical Re-
view’ that does not contain at least one article us-
ing either ‘cause’ or ‘causality’ in its title.”
What we conclude from this exchange is that
physicists talk, write, and think one way and for-
mulate physics in another.
Such bilingual activity would be forgiven if
causality was used merely as a convenient commu-
nication device – a shorthand for expressing com-
plex patterns of physical relationships that would
otherwise take many equations to write.
408 Epilogue

After all! Science is full of

abbreviations: We use “multiply x
by 5” instead of “add x to itself 5
times”; we say “density” instead of
“the ratio of weight to volume.”
Why pick on causality?
“Because causality is differ-
ent,” Lord Russell would argue. “It
could not possibly be an abbrevi-
ation, because the laws of physics
are all symmetrical, going both
ways, while causal relations are
unidirectional, going from cause to
effect.”
Take, for instance, Newton’s law:

f  ma.

The rules of algebra permit us to write this law in a wild variety of syntactic forms,
all meaning the same thing – that if we know any two of the three quantities, the third
is determined.
Yet, in ordinary discourse we say that force causes acceleration – not that accelera-
tion causes force, and we feel very strongly about this distinction.
Likewise, we say that the ratio f/a helps us
determine the mass, not that it causes the mass.
Such distinctions are not supported by the
equations of physics, and this leads us to ask
whether the whole causal vocabulary is purely
metaphysical, “surviving, like the monarchy . . .”.
Fortunately, very few physicists paid atten-
tion to Russell’s enigma. They continued to write
equations in the office and talk cause–effect in the
cafeteria; with astonishing success they smashed
the atom, invented the transistor and the laser.
The same is true for engineering.
But in another arena the tension could not go
unnoticed, because in that arena the demand for
distinguishing causal from other relationships was
very explicit.
This arena is statistics.
The story begins with the discovery of corre-
lation, about one hundred years ago.
The Art and Science of Cause and Effect 409

Francis Galton [20], inventor of fingerprinting

and cousin of Charles Darwin, quite understand-
ably set out to prove that talent and virtue run in
families.
Galton’s investigations drove him to consider
various ways of measuring how properties of one
class of individuals or objects are related to those
of another class.
In 1888, he measured the length of a person’s
forearm and the size of that person’s head and
asked to what degree one of these quantities can
predict the other [21].
He stumbled upon the following discovery: If
you plot one quantity against the other and scale
the two axes properly, then the slope of the best-fit
line has some nice mathematical properties. The
slope is 1 only when one quantity can predict the
other precisely; it is zero whenever the prediction
is no better than a random guess; and, most re-
markably, the slope is the same no matter if you
plot X against Y or Y against X.
“It is easy to see,” said Galton, “that co-relation must be the consequence of the
variations of the two organs being partly due to common causes.”
Here we have, for the first time, an objective
measure of how two variables are “related” to each
other, based strictly on the data, clear of human
judgment or opinion.
Galton’s discovery dazzled one of his disci-
ples, Karl Pearson [22], now considered to be one
of the founders of modern statistics.
Pearson was 30 years old at the time, an ac-
complished physicist and philosopher about to turn
lawyer, and this is how he describes, 45 years later
[23], his initial reaction to Galton’s discovery:
“I felt like a buccaneer of Drake’s days. . . .
“I interpreted . . . Galton to mean that there was
a category broader than causation, namely corre-
lation, of which causation was only the limit, and
that this new conception of correlation brought
psychology, anthropology, medicine, and sociol-
ogy in large parts into the field of mathematical
treatment.”
410 Epilogue

Now, Pearson has been de-

scribed as a man “with the kind of
drive and determination that took
Hannibal over the Alps and Marco
Polo to China.”
When Pearson felt like a buc-
caneer, you can be sure he gets his
bounty.
The year 1911 saw the publica-
tion of the third edition of his book
“The Grammar of Science.” It con-
tained a new chapter titled “Contin-
gency and Correlation – The Insuf-
ficiency of Causation,” and this is
what Pearson says in that chapter:
“Beyond such discarded fundamentals as ‘matter’ and ‘force’ lies still another fetish
amidst the inscrutable arcana of modern science, namely, the category of cause and
effect.”
And what does Pearson substitute for the archaic category of cause and effect? You
wouldn’t believe your ears: contingency tables [24].
“Such a table is termed a contingency table, and the ultimate scientific statement of
description of the relation between two things can always be thrown back upon such a
contingency table. . . .
“Once the reader realizes the nature of such a table, he will have grasped the essence
of the conception of association between cause and effect.”
Thus, Pearson categorically denies the need for an independent concept of causal
relation beyond correlation.
He held this view throughout his life and, ac-
cordingly, did not mention causation in any of his
technical papers.
His crusade against animistic concepts such as
“will” and “force” was so fierce and his rejection
of determinism so absolute that he exterminated
causation from statistics before it had a chance to
take root.
It took another 25 years and another strong-
willed person, Sir Ronald Fisher [25], for statis-
ticians to formulate the randomized experiment –
the only scientifically proven method of testing
causal relations from data, and to this day, the one
and only causal concept permitted in mainstream
statistics.
The Art and Science of Cause and Effect 411

And that is roughly where things stand today.

If we count the number of doctoral theses, re-
search papers, or textbook pages written on
causation, we get the impression that Pearson still
rules statistics.
The “Encyclopedia of Statistical Science” de-
votes twelve pages to correlation but only two
pages to causation – and spends one of those pages
demonstrating that “correlation does not imply
causation.”
Let us hear what modern statisticians say about
causality.
Philip Dawid, the current editor of “Biomet-
rika” (the journal founded by Pearson), admits:
“Causal inference is one of the most important,
most subtle, and most neglected of all the prob-
lems of statistics.”
Terry Speed, former president of the Biomet-
ric Society (whom you might remember as an ex-
pert witness at the O. J. Simpson murder trial),
declares: “Considerations of causality should be
treated as they have always been treated in statis-
tics: preferably not at all but, if necessary, then
with very great care.”
Sir David Cox and Nanny Wermuth, in a book published just a few months ago,
apologize as follows: “We did not in this book use the words causal or causality. . . . Our
reason for caution is that it is rare that firm con-
clusions about causality can be drawn from one
study.”
This position of caution and avoidance has par-
alyzed many fields that look to statistics for guid-
ance, especially economics and social science.
A leading social scientist stated in 1987: “It
would be very healthy if more researchers aban-
don thinking of and using terms such as cause and
effect.”
Can this state of affairs be the work of just one
person? Even a buccaneer like Pearson?
I doubt it.
But how else can we explain why statistics,
the field that has given the world such powerful
concepts as the testing of hypothesis and the
412 Epilogue

design of experiment, would give up so early on

causation?
One obvious explanation is, of course, that cau-
sation is much harder to measure than correlation.
Correlations can be estimated directly in a sin-
gle uncontrolled study, while causal conclusions
require controlled experiments.
But this is too simplistic; statisticians are not
easily deterred by difficulties, and children man-
age to learn cause–effect relations without running
controlled experiments.
The answer, I believe, lies deeper, and it has to
do with the official language of statistics – namely,
the language of probability.
This may come as a surprise to some of you, but
the word cause is not in the vocabulary of prob-
ability theory; we cannot express in the language
of probabilities the sentence, mud does not cause
rain – all we can say is that the two are mutually correlated or dependent – meaning that
if we find one, we can expect the other.
Naturally, if we lack a language to express a certain concept explicitly, we can’t ex-
pect to develop scientific activity around that concept.
Scientific development requires that knowledge be transferred reliably from one study
to another and, as Galileo showed 350 years ago, such transference requires the preci-
sion and computational benefits of a formal language.
I will soon come back to discuss the importance of language and notation, but first I
wish to conclude this historical sur-
vey with a tale from another field in
which causation has had its share of
difficulty.
This time it is computer sci-
ence – the science of symbols – a
field that is relatively new yet one
that has placed a tremendous em-
phasis on language and notation and
therefore may offer a useful per-
spective on the problem.
When researchers began to en-
code causal relationships using
computers, the two riddles of causa-
tion were awakened with renewed
vigor.
The Art and Science of Cause and Effect 413

Put yourself in the shoes of this robot [26] who

is trying to make sense of what is going on in a
kitchen or a laboratory.
Conceptually, the robot’s problems are the
same as those faced by an economist seeking to
model the national debt or an epidemiologist at-
tempting to understand the spread of a disease.
Our robot, economist, and epidemiologist all
need to track down cause–effect relations from
the environment, using limited actions and noisy
observations.
This puts them right at Hume’s first riddle of
causation: how?
The second riddle of causation also plays a role
in the robot’s world.
Assume we wish to take a shortcut and teach our robot all we know about cause and
effect in this room [27].
How should the robot organize and make use of this information?
Thus, the two philosophical riddles of causation are now translated into concrete and
practical questions:
How should a robot acquire causal information through interaction with its envi-
ronment? How should a robot process causal information received from its creator–
programmer?
Again, the second riddle is not as trivial as it might seem. Lord Russell’s warning
that causal relations and physical equations are incompatible now surfaces as an appar-
ent flaw in logic.
For example, when given the information, “If the grass is wet, then it rained” and
“If we break this bottle, the grass will get wet,” the computer will conclude “If we break
this bottle, then it rained” [28].
The swiftness and specificity
with which such programming bugs
surface have made Artificial Intel-
ligence programs an ideal labora-
tory for studying the fine print of
causation.
This brings us to the second part
of the lecture: how the second riddle
of causation can be solved by com-
bining equations with graphs, and
how this solution makes the first
riddle less formidable.
414 Epilogue

The overriding ideas in this solution are:

First – treating causation as a summary of be-
havior under interventions; and
Second – using equations and graphs as a math-
ematical language within which causal thoughts
can be represented and manipulated.
And to put the two together, we need a third
concept: treating interventions as a surgery over
equations.
Let us start with an area that uses causation
extensively and never had any trouble with it: en-
gineering.
Here is an engineering drawing [29] of a circuit
diagram that shows cause–effect relations among
the signals in the circuit. The circuit consists of
and gates and or gates, each performing some log-
ical function between input and output. Let us ex-
amine this diagram closely, since its simplicity and
familiarity are very deceiving. This diagram is, in
fact, one of the greatest marvels of science. It is
capable of conveying more information than mil-
lions of algebraic equations or probability functions or logical expressions. What makes
this diagram so much more powerful is the ability to predict not merely how the circuit
behaves under normal conditions but also how the circuit will behave under millions of
abnormal conditions. For example, given this circuit diagram, we can easily tell what
the output will be if some input changes from 0 to 1. This is normal and can easily be
expressed by a simple input–output equation. Now comes the abnormal part. We can
also tell what the output will be when we set Y to 0 (zero), or tie it to X, or change this
and gate to an or gate, or when we perform any of the millions of combinations of these
operations. The designer of this cir-
cuit did not anticipate or even con-
sider such weird interventions, yet,
miraculously, we can predict their
consequences. How? Where does
this representational power come
from?
It comes from what early econ-
omists called autonomy. Namely,
the gates in this diagram represent
independent mechanisms – it is
easy to change one without chang-
ing the other. The diagram takes
advantage of this independence and
The Art and Science of Cause and Effect 415

describes the normal functioning of the circuit us-

ing precisely those building blocks that will remain
unaltered under intervention.
My colleagues from Boelter Hall are surely
wondering why I stand here before you blather-
ing about an engineering triviality as if it were the
eighth wonder of the world. I have three reasons
for doing this. First, I will try to show that there is
a lot of unexploited wisdom in practices that en-
gineers take for granted.
Second, I am trying to remind economists and
social scientists of the benefits of this diagram-
matic method. They have been using a similar
method on and off for over 75 years, called struc-
tural equation modeling and path diagrams, but
in recent years they have allowed algebraic con-
venience to suppress the diagrammatic represen-
tation, together with its benefits. Finally, these di-
agrams capture, in my opinion, the very essence of causation – the ability to predict the
consequences of abnormal eventualities and new manipulations. In S.Wright’s diagram
[30], for example, it is possible to predict what coat pattern the guinea-pig litter is likely
to have if we change environmental factors, shown here as input (E), or even ge-
netic factors, shown as intermediate nodes between parents and offsprings (H). Such
predictions cannot be made on the basis of algebraic or correlational analysis.
Viewing causality this way explains why scientists pursue causal explanations with
such zeal and why attaining a causal model is accompanied by a sense of gaining “deep
understanding” and “being in control.”
Deep understanding [31] means knowing not merely how things behaved yester-
day but also how things will behave under new hypothetical circumstances, control
being one such circumstance. Inter-
estingly, when we have such under-
standing we feel “in control” even if
we have no practical way of control-
ling things. For example, we have
no practical way to control celes-
tial motion, and still the theory of
gravitation gives us a feeling of un-
derstanding and control, because it
provides a blueprint for hypotheti-
cal control. We can predict the ef-
fect on tidal waves of unexpected
new events – say, the moon being
hit by a meteor or the gravitational
constant suddenly diminishing by a
416 Epilogue

factor of 2 – and, just as important,

the gravitational theory gives us the
assurance that ordinary manipula-
tion of earthly things will not con-
trol tidal waves. It is not surpris-
ing that causal models are viewed
as the litmus test for distinguishing
deliberate reasoning from reactive
or instinctive response. Birds and
monkeys may possibly be trained to
perform complex tasks such as fix-
ing a broken wire, but that requires
trial-and-error training. Deliberate
reasoners, on the other hand, can
anticipate the consequences of new
manipulations without ever trying
those manipulations.
Let us magnify [32] a portion of the circuit diagram so that we can understand why
the diagram can predict outcomes that equations cannot. Let us also switch from logi-
cal gates to linear equations (to make everyone here more comfortable), and assume we
are dealing with a system containing just two components: a multiplier and an adder.
The multiplier takes the input and multiplies it by a factor of 2; the adder takes its input
and adds a 1 to it. The equations describing these two components are given here on the
left.
But are these equations equivalent to the diagram on the right? Obviously not!
If they were, then let us switch the variables around, and the resulting two equations
should be equivalent to the circuit shown below. But these two circuits are different.
The top one tells us that if we physically manipulate Y it will affect Z, while the bottom
one shows that manipulating Y will affect X and will have no effect on Z. Moreover,
performing some additional algebraic operations on our equations, we can obtain two
new equations, shown at the bot-
tom, which point to no structure at
all; they simply represent two con-
straints on three variables without
telling us how they influence each
other.
Let us examine more closely the
mental process by which we deter-
mine the effect of physically ma-
nipulating Y – say, setting Y to 0
[33].
Clearly, when we set Y to 0,
the relation between X and Y is no
longer given by the multiplier – a
The Art and Science of Cause and Effect 417

new mechanism now controls Y, in

which X has no say. In the equa-
tional representation, this amounts
to replacing the equation Y  2X
by a new equation Y  0 and solv-
ing a new set of equations, which
gives Z  1. If we perform this
surgery on the lower pair of equa-
tions, representing the lower
model, we get of course a different
solution. The second equation will
need to be replaced, which will
yield X  0 and leave Z uncon-
strained.
We now see how this model of intervention leads to a formal definition of causation:
“Y is a cause of Z if we can change Z by manipulating Y, namely, if after surgically
removing the equation for Y, the solution for Z will depend on the new value we sub-
stitute for Y.” We also see how vital the diagram is in this process. The diagram tells us
which equation is to be deleted when we manipulate Y. That information is totally
washed out when we transform the equations into algebraically equivalent form, as
shown at the bottom of the screen. From this pair of equations alone, it is impossible to
predict the result of setting Y to 0, because we do not know what surgery to perform –
there is no such thing as “the equation for Y.”
In summary, intervention amounts to a surgery on equations (guided by a diagram)
and causation means predicting the consequences of such a surgery.
This is a universal theme that goes beyond physical systems. In fact, the idea of
modeling interventions by “wiping out” equations was first proposed in 1960 by an
economist, Herman Wold, but his teachings have all but disappeared from the economics
literature. History books attribute this mysterious disappearance to Wold’s personality,
but I tend to believe that the reason goes deeper: Early econometricians were very careful
mathematicians; they fought hard
to keep their algebra clean and for-
mal, and they could not agree to
have it contaminated by gimmicks
such as diagrams. And as we see
on the screen, the surgery operation
makes no mathematical sense with-
out the diagram, as it is sensitive to
the way we write the equations.
Before expounding on the prop-
erties of this new mathematical op-
eration, let me demonstrate how
useful it is for clarifying concepts
in statistics and economics.
418 Epilogue

Why do we prefer controlled

experiment over uncontrolled stud-
ies? Assume we wish to study the
effect of some drug treatment on
recovery of patients suffering from
a given disorder. The mechanism
governing the behavior of each
patient is similar in structure to the
circuit diagram we saw earlier.
Recovery is a function of both the
treatment and other factors, such as
socioeconomic conditions, life
style, diet, age, et cetera. Only one
such factor is shown here [34].
Under uncontrolled conditions, the choice of treatment is up to the patients and may
depend on the patients’ socioeconomic backgrounds. This creates a problem, because we
can’t tell if changes in recovery rates are due to treatment or to those background fac-
tors. What we wish to do is compare patients of like backgrounds, and that is precisely
what Fisher’s randomized experiment accomplishes. How? It actually consists of two
parts, randomization and intervention.
Intervention means that we change the natural behavior of the individual: we separate
subjects into two groups, called treatment and control, and we convince the subjects to
obey the experimental policy. We assign treatment to some patients who, under normal
circumstances, will not seek treatment, and we give a placebo to patients who otherwise
would receive treatment. That, in our new vocabulary, means surgery – we are severing
one functional link and replacing it with another. Fisher’s great insight was that con-
necting the new link to a random coin flip guarantees that the link we wish to break
is actually broken. The reason is that a random
coin is assumed to be unaffected by anything we
can measure on a macroscopic level – including,
of course, a patient’s socioeconomic background.
This picture provides a meaningful and formal
rationale for the universally accepted procedure of
randomized trials. In contrast, our next example
uses the surgery idea to point out inadequacies in
widely accepted procedures.
The example [35] involves a government offi-
cial trying to evaluate the economic consequences
of some policy – say, taxation. A deliberate de-
cision to raise or lower taxes is a surgery on the
model of the economy because it modifies the con-
ditions prevailing when the model was built. Eco-
nomic models are built on the basis of data taken
over some period of time, and during this period
The Art and Science of Cause and Effect 419

of time taxes were lowered and raised in response

to some economic conditions or political pressure.
However, when we evaluate a policy, we wish to
compare alternative policies under the same eco-
nomic conditions – namely, we wish to sever this
link that, in the past, has tied policies to those con-
ditions. In this setup, it is of course impossible to
connect our policy to a coin toss and run a con-
trolled experiment; we do not have the time for
that, and we might ruin the economy before the
experiment is over. Nevertheless the analysis that
we should conduct is to infer the behavior of this
mutilated model from data governed by a nonmu-
tilated model.

I said should conduct because you will not

find such analysis in any economics textbook. As
I mentioned earlier, the surgery idea of Herman
Wold was stamped out of the economics litera-
ture in the 1970s, and all discussions on policy
analysis that I could find assume that the mutilated
model prevails throughout. That taxation is under
government control at the time of evaluation is assumed to be sufficient for treating tax-
ation as an exogenous variable throughout, when in fact taxation is an endogenous vari-
able during the model-building phase and turns exogenous only when evaluated. Of
course, I am not claiming that reinstating the sur-
gery model would enable the government to bal-
ance its budget overnight, but it is certainly
something worth trying.

Let us now examine how the surgery interpre-

tation resolves Russell’s enigma concerning the
clash between the directionality of causal rela-
tions and the symmetry of physical equations. The
equations of physics are indeed symmetrical, but
when we compare the phrases “A causes B” versus
“B causes A,” we are not talking about a single
set of equations. Rather, we are comparing two
world models, represented by two different sets of
equations: one in which the equation for A is sur-
gically removed; the other where the equation for
B is removed. Russell would probably stop us at
this point and ask: “How can you talk about two
world models when in fact there is only one world
model, given by all the equations of physics put
together?” The answer is: yes. If you wish to
420 Epilogue

include the entire universe in the

model, causality disappears be-
cause interventions disappear – the
manipulator and the manipulated
lose their distinction. However,
scientists rarely consider the en-
tirety of the universe as an object of
investigation. In most cases the sci-
entist carves a piece from the uni-
verse and proclaims that piece in –
namely, the focus of investigation.
The rest of the universe is then con-
sidered out or background and is
summarized by what we call bound-
ary conditions. This choice of ins
and outs creates asymmetry in the way we look at things, and it is this asymmetry that per-
mits us to talk about “outside intervention” and hence about causality and cause–effect
directionality.
This can be illustrated quite nicely using Descartes’ classical drawing [36]. As a
whole, this hand–eye system knows nothing about causation. It is merely a messy
plasma of particles and photons trying their very best to obey Schroedinger’s equation,
which is symmetric.
However, carve a chunk from it – say, the object part [37] – and we can talk about
the motion of the hand causing this light ray to change angle.
Carve it another way, focusing on the brain
part [38], and lo and behold it is now the light ray
that causes the hand to move – precisely the oppo-
site direction. The lesson is that it is the way we
carve up the universe that determines the direc-
tionality we associate with cause and effect. Such
carving is tacitly assumed in every scientific in-
vestigation. In artificial intelligence it was called
“circumscription” by J. McCarthy. In economics,
circumscription amounts to deciding which vari-
ables are deemed endogenous and which exoge-
nous, in the model or external to the model.
Let us summarize the essential differences be-
tween equational and causal models [39]. Both use
a set of symmetric equations to describe normal
conditions. The causal model, however, contains
three additional ingredients: (i) a distinction be-
tween the in and the out; (ii) an assumption that
each equation corresponds to an independent
mechanism and hence must be preserved as a
The Art and Science of Cause and Effect 421

separate mathematical sentence;

and (iii) interventions that are inter-
preted as surgeries over those mech-
anisms. This brings us closer to
realizing the dream of making
causality a friendly part of physics.
But one ingredient is missing: the
algebra. We discussed earlier how
important the computational facili-
ty of algebra was to scientists and
engineers in the Galilean era. Can
we expect such algebraic facility to
serve causality as well? Let me
rephrase it differently: Scientific
activity, as we know it, consists of
two basic components:
Observations [40] and interventions [41].
The combination of the two is what we call a laboratory [42], a place where we con-
trol some of the conditions and observe others. It so happened that standard algebras
have served the observational component very well but thus far have not benefitted the
interventional component. This is true for the algebra of equations, Boolean algebra,
and probability calculus – all are geared to serve observational sentences but not inter-
ventional sentences.
Take, for example, probability theory. If we wish to find the chance that it rained, given
that we see the grass wet, we can express our question in a formal sentence written like
that: P (Rain | Wet), to be read: the probability of Rain, given Wet [43]. The vertical bar
stands for the phrase: “given that we see.” Not only can we express this question in a
formal sentence, we can also use the machinery of probability theory and transform the
sentence into other expressions. In our example, the sentence on the left can be trans-
formed to the one on the right, if we find it more convenient or informative.
But suppose we ask a different question: “What is the chance it rained if we make
the grass wet?” We cannot even
express our query in the syntax of
probability, because the vertical bar
is already taken to mean “given that
we see.” We can invent a new sym-
bol do, and each time we see a do
after the bar we read it given that we
do – but this does not help us com-
pute the answer to our question,
because the rules of probability do
not apply to this new reading. We
know intuitively what the answer
should be: P (Rain), because making
422 Epilogue

the grass wet does not change the

chance of rain. But can this intuitive
answer, and others like it, be derived
mechanically, so as to comfort our
thoughts when intuition fails?
The answer is yes, and it takes
a new algebra. First, we assign a
symbol to the new operator “given
that we do.” Second, we find the
rules for manipulating sentences
containing this new symbol. We do
that by a process analogous to the
way mathematicians found the
rules of standard algebra.
Imagine that you are a mathematician in the sixteenth century, you are now an expert
in the algebra of addition, and you feel an urgent need to introduce a new operator, mul-
tiplication, because you are tired of adding a number to itself all day long [44]. The first
thing you do is assign the new operator a symbol: multiply. Then you go down to the
meaning of the operator, from which you can deduce its rules of transformations. For
example: the commutative law of multiplication can be deduced that way, the associative
law, and so on. We now learn all this in high school.
In exactly the same fashion, we can deduce the rules that govern our new symbol:
do (.). We have an algebra for seeing – namely, probability theory. We have a new op-
erator, with a brand new outfit and a very clear meaning, given to us by the surgery
procedure. The door is open for deduction, and the result is given in the next slide [45].
Please do not get alarmed, I do not expect you to read these equations right now,
but I think you can still get the flavor of this new calculus. It consists of three rules that
permit us to transform expressions involving actions and observations into other expres-
sions of this type. The first allows us to ignore an irrelevant observation, the third to
ignore an irrelevant action; the sec-
ond allows us to exchange an ac-
tion with an observation of the same
fact. What are those symbols on
the right? They are the “green
lights” that the diagram gives us
whenever the transformation is
legal. We will see them in action
on our next example.
This brings us to part three of
the lecture, where I will demon-
strate how the ideas presented thus
far can be used to solve new prob-
lems of practical importance.
The Art and Science of Cause and Effect 423

Consider the century-old debate

concerning the effect of smoking on
lung cancer [46]. In 1964, the Sur-
geon General issued a report link-
ing cigarette smoking to death,
cancer, and most particularly lung
cancer. The report was based on
nonexperimental studies in which
a strong correlation was found be-
tween smoking and lung cancer,
and the claim was that the corre-
lation found is causal: If we ban
smoking, then the rate of cancer
cases will be roughly the same as
the one we find today among non-
smokers in the population.

These studies came under severe attacks from the tobacco industry, backed by some
very prominent statisticians, among them Sir Ronald Fisher. The claim was that the
observed correlations can also be explained by a model in which there is no causal con-
nection between smoking and lung cancer. Instead, an unobserved genotype might exist
that simultaneously causes cancer and produces an inborn craving for nicotine. Formally,
this claim would be written in our notation as: P(Cancer | do (Smoke))  P(Cancer),
meaning that making the population smoke or stop smoking would have no effect on the
rate of cancer cases. Controlled experiments could decide between the two models, but
these are impossible (and now also illegal) to conduct.

This is all history. Now we enter a hypotheti-

cal era where representatives of both sides decide
to meet and iron out their differences. The tobacco
industry concedes that there might be some weak
causal link between smoking and cancer and rep-
resentatives of the health group concede that there
might be some weak links to genetic factors. Ac-
cordingly, they draw this combined model, and the
question boils down to assessing, from the data,
the strengths of the various links. They submit
the query to a statistician and the answer comes
back immediately: impossible. Meaning: there is
no way to estimate the strength from the data, be-
cause any data whatsoever can perfectly fit either
one of these two extreme models. So they give
up and decide to continue the political battle as
usual. Before parting, a suggestion comes up: per-
haps we can resolve our differences if we measure
some auxiliary factors. For example, since the
424 Epilogue

causal-link model is based on the

understanding that smoking affects
lung cancer through the accumula-
tion of tar deposits in the lungs, per-
haps we can measure the amount of
tar deposits in the lungs of sampled
individuals, and this might provide
the necessary information for quan-
tifying the links. Both sides agree
that this is a reasonable suggestion,
so they submit a new query to the
statistician: Can we find the effect
of smoking on cancer assuming that
an intermediate measurement of tar
deposits is available? The statistician comes back with good news: it is computable and,
moreover, the solution is given in closed mathematical form. How?

The statistician receives the

problem and treats it as a problem
in high school algebra: We need
to compute P (Cancer), under hy-
pothetical action, from nonexperi-
mental data – namely, from expres-
sions involving no actions. Or: We
need to eliminate the “do” symbol
from the initial expression. The
elimination proceeds like ordinary
solution of algebraic equations – in
each stage [47], a new rule is ap-
plied, licensed by some subgraph
of the diagram, eventually leading
to a formula involving no “do” symbols, which denotes an expression that is computable
from nonexperimental data.

You are probably wondering

whether this derivation solves the
smoking–cancer debate. The an-
swer is no. Even if we could get
the data on tar deposits, our model
is quite simplistic, as it is based on
certain assumptions that both par-
ties might not agree to – for in-
stance, that there is no direct link
between smoking and lung cancer
unmediated by tar deposits. The
model would need to be refined
The Art and Science of Cause and Effect 425

then, and we might end up with a

graph containing twenty variables
or more. There is no need to panic
when someone tells us: “you did
not take this or that factor into ac-
count.” On the contrary, the graph
welcomes such new ideas, because
it is so easy to add factors and mea-
surements into the model. Simple
tests are now available that permit
an investigator to merely glance at
the graph and decide if we can com-
pute the effect of one variable on
another.
Our next example illustrates how a long-standing problem is solved by purely graph-
ical means – proven by the new al-
gebra. The problem is called the
adjustment problem or “the covari-
ate selection problem” and repre-
sents the practical side of Simpson’s
paradox [48].
Simpson’s paradox, first no-
ticed by Karl Pearson in 1899, con-
cerns the disturbing observation
that every statistical relationship
between two variables may be re-
versed by including additional fac-
tors in the analysis. For example,
you might run a study and find that
students who smoke get higher grades; however, if you adjust for age, the opposite is true
in every age group, that is, smoking
predicts lower grades. If you fur-
ther adjust for parent income, you
find that smoking predicts higher
grades again, in every age–income
group, and so on.
Equally disturbing is the fact
that no one has been able to tell us
which factors should be included
in the analysis. Such factors can
now be identified by simple graphi-
cal means. The classical case dem-
onstrating Simpson’s paradox took
place in 1975, when UC-Berkeley
426 Epilogue

was investigated for sex bias in

graduate admission. In this study,
overall data showed a higher rate of
admission among male applicants;
but, broken down by departments,
data showed a slight bias in favor
of admitting female applicants. The
explanation is simple: female appli-
cants tended to apply to more com-
petitive departments than males,
and in these departments, the rate of
admission was low for both males
and females.
To illustrate this point, imag-
ine a fishing boat with two different nets, a large mesh and a small net [49]. A school
of fish swim toward the boat and
seek to pass it. The female fish try
for the small-mesh challenge, while
the male fish try for the easy route.
The males go through and only fe-
males are caught. Judging by the
final catch, preference toward fe-
males is clearly evident. However,
if analyzed separately, each indi-
vidual net would surely trap males
more easily than females.
Another example involves a
controversy called “reverse regres-
sion,” which occupied the social
science literature in the 1970s. Should we, in salary discrimination cases, compare
salaries of equally qualified men
and women or instead compare
qualifications of equally paid men
and women?
Remarkably, the two choices
led to opposite conclusions. It
turned out that men earned a higher
salary than equally qualified wom-
en and, simultaneously, men were
more qualified than equally paid
women. The moral is that all con-
clusions are extremely sensitive to
which variables we choose to hold
constant when we are comparing,
The Art and Science of Cause and Effect 427

and that is why the adjustment prob-

lem is so critical in the analysis of
observational studies.
Consider an observational
study where we wish to find the
effect of X on Y, for example, treat-
ment on response [50]. We can
think of many factors that are rele-
vant to the problem; some are
affected by the treatment, some are
affecting the treatment, and some
are affecting both treatment and
response. Some of these factors
may be unmeasurable, such as
genetic trait or life style; others are measurable, such as gender, age, and salary level.
Our problem is to select a subset of these factors for measurement and adjustment so
that, if we compare subjects under the same value of those measurements and average,
we get the right result.
Let us follow together the steps that would be required to test if two candidate mea-
surements, Z1 and Z2, would be sufficient [51]. The steps are rather simple, and can be
performed manually even on large graphs. However, to give you the feel of their mech-
anizability, I will go through them rather quickly. Here we go [52–56].
At the end of these manipulations, we end up with the answer to our question: “If X
is disconnected from Y, then Z1 and Z2 are appropriate measurements.”
I now wish to summarize briefly the central message of this lecture. It is true that
testing for cause and effect is difficult. Discovering causes of effects is even more dif-
ficult. But causality is not mystical or metaphysical. It can be understood in terms of sim-
ple processes, and it can be expressed in a friend-
ly mathematical language, ready for computer
analysis.
What I have presented to you today is a sort
of pocket calculator, an abacus [57], to help us
investigate certain problems of cause and effect
with mathematical precision. This does not solve
all the problems of causality, but the power of
symbols and mathematics should not be underes-
timated [58].
Many scientific discoveries have been delayed
over the centuries for the lack of a mathematical
language that can amplify ideas and let scientists
communicate results. I am convinced that many
discoveries have been delayed in our century for
lack of a mathematical language that can handle
428 Epilogue

causation. For example, I am sure that Karl Pear-

son could have thought up the idea of randomized
experiment in 1901 if he had allowed causal dia-
grams into his mathematics.
But the really challenging problems are still
ahead: We still do not have a causal understanding
of poverty and cancer and intolerance, and only
the accumulation of data and the insight of great
minds will eventually lead to such understanding.
The data is all over the place, the insight is
yours, and now an abacus is at your disposal, too.
I hope the combination amplifies each of these
components.

Thank you.

Acknowledgments

Slide 1 (Dürer, Adam and Eve, 1504 engraving) courtesy of the Fogg Art Museum, Har-
vard University Art Museums, Gift of William Gray from the collection of Francis Calley
Gray. Photo by Rick Stafford; image copyright © President and Fellows of Harvard Col-
lege, Harvard University. Slide 2 (Doré, The Flight of Lot) copyright William H. Wise
& Co. Slide 3 (Egyptian wall painting of Neferronpe playing a board game) courtesy of
the Oriental Institute of the University of Chicago.
The following images were reproduced from antiquarian book catalogs, courtesy of
Bernard Quaritch, Ltd. (London): slides 4, 5, 6, 7, 8, 9, 15, 27, 31, 36, 37, 38, 40, 42,
and 58.
Slides 10 and 11 copyright The Courier Press. Slides 13 and 14 reprinted with the per-
mission of Macmillan Library Reference USA, from The Album of Science, by I. Bernard
Cohen. Copyright © 1980 Charles Scribner’s Sons.
Slide 16 courtesy of the Library of California State University, Long Beach. Slides
20 and 22 reprinted with the permission of Cambridge University Press. Slide 25: copy-
right photograph by A. C. Barrington Brown, reproduced with permission.
Slide 30: from S. Wright (1920) in Proceedings of the National Academy of Sciences,
vol. 6; reproduced with the permission of the American Philosophical Society and the
University of Chicago Press. Slide 57 reprinted with the permission of Vandenhoeck &
Ruprecht and The MIT Press.
NOTE: Color versions of slides 19, 26, 28–29, 32–35, and 43–56 may be downloaded
from http://www.cs.ucla.edu/~judea/.