3
Reasons for Skepticism, Part II: Explanation
while the objections in the previous chapter provide grounds for
considerable skepticism on specific points, there are also three broader
objections to consider regarding the explanations for a decline that have
been offered. They are somewhat more amorphous than my objections
to the data, but given how broadly they apply to the logic and evidence
of the decline-of-war, thesis, they are in many ways even more damning.
The first objection has to do with the fact that the international
system is a complex system. It consists of many countries, which
themselves consist of people, and the interactions of those people and
those countries in the context of the international system produces
behavior that can’t be understood or predicted just by examining
people or countries in isolation. For Professors Pinker and Mueller,
human beings are the heroes of the decline-of-war story: Our increasing
aversion to violence has, they claim, produced a more peaceful world. As
appealing as that story is, it ignores the very significant role of the social,
domestic, and international context with which those people act—
contexts that scholars have repeatedly shown to have a big influence on
behavior. Once we incorporate those contexts into the story, it becomes
much harder to believe that human pacifism, even if it is on the rise,
can really play a big role in influencing war at the international level.
The second objection is that the focus in Professor Pinker’s work in
particular on the spread of Renaissance humanism and empathy results
in a story that’s a bit too neat and clean to be true. The Enlightenment
gave rise to a remarkable diversity of ideas, not all of which lend
47
48 REASONS FOR SKEPTICISM, PART II: EXPLANATION
themselves to peace. Only focusing on those that do runs the risk
of circular reasoning—concluding that Enlightenment ideas caused
peace because we only examine the ideas that are prevalent in peaceful
communities. Moreover, the best-case examples for the argument that
Enlightenment humanism causes peace, developed Western countries,
have been shown to be very unrepresentative of the broader global
community.
The third objection is that virtually none of the data analyses
supporting the decline-of-war thesis takes into account the role of
chance. Even if the baseline rate of something (like conflict initiations
over time, or battle deaths in war) remains the same over time, there
are still random fluctuations from one year to the next or from one
war to the next. We need to know on average how big those random
fluctuations should be if we want to know whether the baseline rate
has really changed. Statisticians have provided us with an excellent and
diverse array of tools to use in order to answer exactly this question, but
for the most part those tools have not been brought to bear.
The Complexity of International Behavior
A complex system is a system that contains interacting agents whose
behavior differs from that which one would expect solely based on their
characteristics. To put it more succinctly (if less precisely), the behavior
of the whole is more than just the sum of the behavior of the parts.1
Humans exhibit complex behavior even in very small groups. A brief
story will illustrate this point. Years ago, a company started putting out
motivational posters—simple photographs with thick black borders,
one big word (“Teamwork,” perhaps), and a one-line motivational
saying underneath it, designed to be framed and put up in workplaces to
inspire the people who work there. A website calling itself Despair, Inc.
immediately started producing parodies of those posters, with funny
lines designed to undermine motivation. If my social media feed is
any indication, the parodies soon became vastly more popular than the
originals.
My favorite parody poster is one that features a photo of a circle of
business-attired people extending their hands inward to the center, one
on top of the other. The wording at the bottom reads, “MEETINGS:
THE COMPLEXITY OF INTERNATIONAL BEHAVIOR 49
None of Us Is as Dumb as All of Us.” If you’ve spent much of your
professional life in meetings or on committees, this probably resonates
with you. I certainly have, and I think of this poster whenever I find
myself wondering how groups that contain such smart individuals can
arrive at some of the decisions that come out of them.
That’s complexity. Would any one of us have concluded that this is
the best decision? Nope! Are we satisfied with it? Not entirely. Crucially,
if you had asked each of us what decision we thought made sense prior
to the meeting, could you have predicted the aggregate outcome? Most
likely not, unless we were deciding a very simple matter.
Countries, of course, are much more complex than committees, and
the international system is more complex than countries.2
So even if
it could be established that norms of peaceful behavior are spreading
throughout the international system, the decline-of-war thesis makes
a huge logical leap when it assumes that those attitudes will translate
directly into a decrease in violence in the international system. The
context of interaction matters a lot to the outcome, and there are a lot of
contextual “layers” between individual human beings and international
conflict that could overwhelm the impact of peaceful dispositions. I
explore three of those contextual layers below—one at the level of the
individual, one at the level of the country as a whole, and one at the
level of the international system.
Dispositions and Situations
Most people overestimate the impact of dispositions and downplay or
ignore the impact of situations in their explanations of human behavior.
If we hear what sounds like a woman in an adjacent room falling down
and crying out in pain, for example, our inclination is to believe that
we’d get up and find out whether she needs help. More than that, we
tend to believe that other people’s behavior—whether or not they get
up to help—is a reflection of their character. Kind people (like us!) will
get up to find out whether the woman in the next room needs any
assistance, but only cold or indifferent people will ignore her.
In fact, as Bibb Latané and Judith Rodin (1969) demonstrated in a
landmark study, whether or not people get up to help depends crucially
on situational factors. Latané and Rodin recruited male undergraduates
50 REASONS FOR SKEPTICISM, PART II: EXPLANATION
at Columbia University to come into an office and fill out a detailed
marketing questionnaire. Some filled out the survey alone; others did
so in the presence of another student who was ostensibly doing the same
thing but in reality was part of the experiment. While they were doing
so, they were able to hear the female “market research representative” in
the next room shuffling papers, opening and closing drawers, and so on.
Four minutes into the experiment, they heard the representative climb
up on a chair to reach for a stack of papers and then come crashing to
the ground as the chair collapsed. They then heard the representative
say, “Oh, my God, my foot... I... I... can’t move it. Oh... my ankle. I...
can’t get this... thing... off me.” If the students didn’t intervene, they
heard the representative’s cries of pain gradually subside until she got
up and limped out of the room. If there was another student present as
part of the experiment, he did not get up to help and responded only
minimally to anything that the experimental subject said or did.
Most of the students who were left to fill out the survey alone—
70%—got up and attempted to help the woman in the next room.
When there was another student there and that student failed to react,
however, that behavior changed drastically: only 7% of the subjects got
up to help. A relatively minor change in the situation—the existence of a
bystander who did nothing—was sufficient to produce a night-and-day
difference in behavior.
Our human tendency to attribute behavior in such situations to
the character of the individual in question rather than to the specifics
of the situation is so profound that psychologists refer to it as the
“Fundamental Attribution Error.”3
That it is an error is no longer in
much doubt:
Consider the following scenario: While walking briskly to a meeting
some distance across a college campus, John comes across a man
slumped in a doorway, asking him for help. Will John help him,
or will he continue on his way? Before answering such a question,
most people would want to know more about John. ... In fact,
however, nothing one is likely to know or learn about John would be
of much use in helping predict John’s behavior in the situation we’ve
just described. ... A half century of research has taught us that in this
situation, and in most other novel situations, one cannot predict with
THE COMPLEXITY OF INTERNATIONAL BEHAVIOR 51
any accuracy how particular people will respond. At least one cannot
do so using information about an individual’s personal dispositions or
even about that individual’s past behavior. (Ross and Nisbett, 1991, 2).
American isolationism provides an excellent illustration of the way
in which the fundamental attribution error can creep into scholars’
explanations of a country’s foreign policy behavior. The United States
between World War I and World War II is often said to have pursued
an isolationist foreign policy, due in large part to the noninterventionist
“mood” of its citizens. Much of the historical scholarship on isolationism
in that period and subsequently has focused on figuring out which
kinds of people—the less educated? The poor? Recent immigrants?
Republicans? People from the Midwest?—are more likely to express
isolationist sentiments.4
More recent research has shown, however, that this apparent
disposition toward nonintervention depends crucially on the situational
context. In the case of the interwar period, Americans’ inclinations not
to become involved in the growing European war hinged crucially on a
key situational factor: Very few Americans, even those at high levels,
believed that Germany constituted a real threat to the Continental
powers. The shockingly successful German invasion of France in May
and June of 1940 proved conclusively that Germany was a genuine
threat, and as a result American support for intervention skyrocketed.
While only 20–30% of Americans were willing to risk war in order
to aid England and France prior to the invasion, support for doing
so had cleared 70% by March 1941.5
Dispositional explanations for
isolationism cannot explain this change of heart: it’s extremely unlikely,
of course, that nearly half of the American population moved to the
coasts, switched political parties, or went to college in the nine months
following the fall of France.
It seems plausible, at least, that nonviolence is every bit as dependent
on situation as isolationism is. Indeed, that is the premise of political
theorist Hannah Arendt’s brilliant and controversial book Eichmann in
Jerusalem: A Report on the Banality of Evil. Arendt’s horrifying conclusion
was that the subject of her book, Nazi SS-Obersturmbannführer
Adolf Eichmann, far from being incomprehensibly evil, was driven
to commit atrocities by the circumstances of the time. While Arendt
52 REASONS FOR SKEPTICISM, PART II: EXPLANATION
stopped short of endorsing Eichmann’s own explanation—that his
orders compelled him to comply—she concluded that he was an ideologue
who was easily swept up in the Nazi movement and able to justify
actions that he otherwise would have found immoral because they
furthered the movement. He was, in her words, a “joiner.” She writes
that “May 8, 1945, the official date of Germany’s defeat, was significant
for him mainly because it then dawned upon him that thenceforward
he would have to live without being a member of something or other.”6
Far from being unusual, Eichmann’s susceptibility to influence struck
Arendt as being terrifyingly common.
Stanley Milgram, a psychologist at Yale, was struck by Arendt’s
conclusions and set up a series of experiments to explore the limits of
human obedience.7
The subjects of Milgram’s experiments were told
that a person in another room—supposedly another subject, but in
reality a confederate—would answer a series of questions given to him
by the subject. The actual subjects of the experiment, the “teachers,”
were asked to push a button whenever the “learner” in the next room
gave an incorrect answer. Pushing the button would administer an
electric shock. The subjects were told to use a dial to make each electric
shock stronger than the previous one, up to a maximum of 450 volts.
The shocks, according to the experimenter, would cause no lasting
damage. (This, at least, was true: the shocks were fictional. The button
activated a tape recorder that had been loaded with a pre-recorded
script.) At first, pushing the button produced the sort of sounds from
the next room that one might expect from someone who had just been
zapped unexpectedly with a minor shock. Before long, however, the
subject could hear the person in the next room banging on the wall and
demanding that the experiment be brought to an end. If the shocks
continued, the reactions from the next room became more intense
and the subject could hear pleading, as well as the mention of a heart
condition. Beyond a certain point, pushing the button produced no
sound from the next room at all. If the subject expressed discomfort at
the noises coming from the other side of the wall (as each did, sooner
or later), the experimenter simply urged the subject to continue.
Milgram and his colleagues expected that very few subjects would
administer apparently murderous voltages of electricity to another
human being simply because they were told to do so. In fact, 65% of
THE COMPLEXITY OF INTERNATIONAL BEHAVIOR 53
the participants did. They argued with the experimenter, who remained
bizarrely impassive in the face of the cries from the next room. Some
got up as if to leave but dutifully returned to their seats when told to do
so. But in the end, 65% of the subjects continued to increase the voltage
and push the button until no more sounds came from the next room.8
There is substantial evidence that context drives behavior outside of
the laboratory as well. Author Karl Marlantes is a decorated veteran of
the Marine Corps who served in the Vietnam War. His account of his
time there, published under the title What It Is Like to Go to War (2011),
is an extended meditation on the effects that war had on him.
Marlantes offers an account of an airstrike that is particularly
relevant. In it, he helped protect a team of Marines that had taken a hill
and were trying to defend it against their North Vietnamese pursuers
until a helicopter could arrive to pick up one of their gravely wounded
comrades. Marlantes called in air support in the form of some Marine
A-4s, which came loaded with bombs and napalm. They used both, and
the North Vietnamese soldiers were left scattered across the hill, charred
from the napalm, either dead or dying.
At the time, Marlantes writes, he was elated at having saved his team.
Looking back, he writes, “I now think of what was ‘the enemy’ as human
beings, so I find it hard to crow about burning them to death.” He
makes a point of noting that he now feels an empathy for his enemy
that he did not feel at the time—the same empathy that, according to
Pinker, has grown along with the spread of Renaissance humanism.9
The key question is, what differences does that empathy make?
Looking back on his experience at the time, with the benefit of years and
the empathy that they provide, would Marlantes do anything differently
today? By his own account, he would not:
I’d still do the same thing, only I would be aware of a horrible
dilemma. I would be much more reluctant to use napalm now,
knowing I could get the job done a lot more humanely with bombs.
But scrambled aircraft arrive on station loaded with what they’re
loaded with. Once I had decided to be in that situation, I couldn’t
then decide that the team should sacrifice itself for my misgivings
about using napalm. ... Ideally, I would hope that, in spite of the
adrenaline, I’d at least stay conscious of a terrible sadness while I
54 REASONS FOR SKEPTICISM, PART II: EXPLANATION
burned these people. But burn them I would. (Marlantes, 2011,
41–42).
To his credit, Marlantes is ruthlessly honest. Looking back, he does
feel empathy for the other human beings on the hill. He no longer sees
them as the impersonal, inhuman enemy. But these other human beings
are doing their best to kill his human beings, the soldiers on his side of
the war. And in that situation, while empathy might produce a lingering
sadness, it would not prevent him from using napalm to kill the North
Vietnamese soldiers. The requirements of the situation, in other words,
outweigh the dispositional compulsions of empathy.
It might seem that I’ve chosen an easy case to use to demonstrate
the power of situations over dispositions because war is an unusually
compelling environment. That’s probably true. But it’s also the environment
that’s most relevant to the decline-of-war argument. Ignoring
the power of the situation is a bad idea in general when we’re trying
to explain human behavior, but it’s really a bad idea when we’re trying
to explain how people behave when the lives of their comrades or their
fellow citizens are on the line.
So, the first point to make (and I hope I have not made it in tedious
detail, but I do think that it bears strong emphasis) is that dispositions
alone rarely predict behavior. Even a soldier looking back at his actions
in Vietnam with sadness and empathy recognizes the fact that he would
have to do precisely the same thing today if he were to find himself
in the same situation. A majority of ordinary Americans administered
what they thought were lethal doses of electricity to their fellow human
beings when someone in a position of authority asked them to do so. We
don’t like to think that we, ourselves, are so malleable as to be capable
of these sorts of behavior, but decades of research point unambiguously
to the conclusion that we are.
Strategic Interaction
Even if we entirely leave aside the power of the situation in determining
human actions and assume, contra lots and lots of evidence, that
individual dispositions translate unproblematically into actions, we
immediately run into another problem: People and countries react
strategically to one another’s actions. A change in the behavior of
THE COMPLEXITY OF INTERNATIONAL BEHAVIOR 55
one actor may result in a change in the behavior of others, and the
outcome of those changes may be the opposite of what we’d expect. For
that reason, there’s no guarantee that more peaceful preferences in the
context of strategic interaction will actually lead to more peace. They
may actually do the opposite.
One of the most well-known examples of this outcome is the onset
of the First World War. When Austrian Archduke Franz Ferdinand
was assassinated and the Austro-Hungarian Empire contemplated war
against Serbia, Germany had to balance the merits of backing an ally in
what they hoped would be a short Austro-Serbian war in the Balkans
against the risk of a broader Continental war. The German decision
to back Austria was based in large part on the mistaken belief that the
other powers would not intervene: As the Saxon envoy in Berlin put
it, “England is absolutely pacific and France as well as Russia likewise
do not feel inclined towards war” (Clark, 2013, 515). Perversely, if the
British or the Russians had been more aggressively interventionist early
on in the crisis, Germany would almost certainly have tempered its
support for Austria-Hungary, and the Continent might have remained
at peace.10
Of course, one could also cite plenty of examples of cases in which
a bellicose foreign policy also got a country into war: I don’t mean
to imply that an increase in pacifism inevitably increases the risk of
conflict! Rather, when we take the context of strategic interaction into
account, we can’t simply draw a direct connection between an increase
in pacifism, on the one hand, and less conflict on the other. We have to
take the calculations of other actors into account, and those calculations
could easily lead to more aggressive actions that would increase the
probability of war.
Professors Andrew T. Little and Thomas Zeitzoff lay out a detailed
theoretical model that nicely captures the logic of how countries’
preferences and calculations can coevolve when war becomes more or
less costly and countries become more or less pacifistic as a result. The
logic is a bit more complex than what I’ve laid out above, but the
upshot is similar: as countries become less willing to fight, they do
more actual fighting. The authors argue specifically that the logic of their
argument undermines the decline-of-war thesis: “[E]ven if we accept
that conflict is declining over time as a result of an increasing relative
56 REASONS FOR SKEPTICISM, PART II: EXPLANATION
value of cooperation, we can not infer from this fact that people have
become more peace loving” (Little and Zeitzoff, 2017, 17).
Structure and Anarchy
As important as the immediate situation and the logic of interaction can
be in determining an individual’s behavior, there are other contexts that
matter as well. To international relations theorists, perhaps the most
important of these is international anarchy. The word “anarchy,” in this
context, does not connote chaos or disorder; rather, it refers to the lack
of any overarching political authority in the international system. While
an experimenter’s office or a war zone may be the context within which
people interact, anarchy is the context within which countries interact.
The importance of anarchy lies in the incentives that it produces for
the countries that make up the international system. In the absence of a
global police force capable of righting wrongs as serious as invasion and
conquest, no country can be assured of its own survival. Each must
prepare, to some degree, for the possibility of conflict with others,
no matter how peaceful its intentions. Because those intentions can
never be known with certainty and countries concerned about their
own survival are often risk averse, that preparation in and of itself can
be seen as threatening. This situation, in which actions taken to increase
the security of one country prompt reactions that increase tensions and
create conflict, is known in the international relations literature as the
“security dilemma.”11
Without a doubt, the classic modern work on the impact of
anarchy on international politics remains Kenneth Waltz’s Theory of
International Politics. While Waltz did not deny that human nature
and domestic politics can play a role in determining a country’s foreign
policy,12
he argued that they were typically swamped by the logic of
survival under anarchy.
The state among states, it is often said, conducts its affairs in the
brooding shadow of violence. Because some states may at any time
use force, all states must be prepared to do so—or live at the mercy
of their militarily more vigorous neighbors. Among states, the state
of nature is a state of war. This is meant not in the sense that war
constantly occurs but in the sense that, with each state deciding for
THE COMPLEXITY OF INTERNATIONAL BEHAVIOR 57
itself whether or not to use force, war may at any time break out. ...
Among men as among states, anarchy, or the absence of government,
is associated with the occurrence of violence. (Waltz, 1979, 102).
The idea that anarchy stood in the way of a reduction in international
warfare was also endorsed by none other than Professor Norbert Elias,
the German sociologist whose work on the “civilizing process” is one of
the biggest influences on Professor Pinker’s book:
At the international level there is no overarching power to prevent
a stronger state from invading a weaker state to demand taxes and
obedience from its citizens and so de facto to annex the weaker state.
Nobody can prevent a mighty state from doing this except another
mighty state. And if such states exist they live in constant fear of
each other, in the fear that their rivals could become stronger than
themselves.13
Waltz’s book prompted countless counterarguments, modifications,
clarifications, extensions, and heated debates. Robert Axelrod (1984)
argued that cooperation among self-interested actors could easily evolve
despite anarchy, thanks to the promise of future interactions—the socalled
“shadow of the future.” Robert Keohane (1984) argued that
an evolving web of international institutions was increasingly able to
facilitate international cooperation. Later, Alexander Wendt (1992) took
a considerably more radical position, arguing that, as the title of the
article puts it, “anarchy is what states make of it”: anarchy is not
an inherent feature of the international system but is rather a social
construction that is amenable to reinterpretation.
It is no exaggeration to say that these and related works laid the
foundation for international relations theory for decades. A summary
of the research that came in their wake would take multiple chapters.
The main point, though, is that much of the debate centered on the
impact of anarchy on the behavior of the countries that make up the
international system and how that impact might be mitigated.
At a minimum, the logic of international anarchy calls into question
the claim that a nonviolent public will automatically produce a
nonviolent foreign policy. As the old Latin adage goes, Si vis pacem,
para bellum (If you want peace, prepare for war). Even strong advocates
58 REASONS FOR SKEPTICISM, PART II: EXPLANATION
of nonviolence can find themselves compelled to advocate military
means to defend the national interest, and the defense of the national
interest is a constant concern in the absence of world government.
Self-defense can easily be mistaken for aggressive intent and produce
precisely the conflict that it is meant to deter. For the most part,
Pinker and Mueller, the main advocates of the norms-of-nonviolence
explanation, simply don’t address international anarchy or explain how
the peaceful dispositions of the citizens of a country could prevail in an
international context that compels countries to risk conflict.14
The Spread of Enlightenment Humanism
Another reason to be skeptical of the argument that underpins the more
ambitious variant of the decline-of-war thesis is that the causal story that
drives it—the spread of Renaissance humanism and empathy—is a bit
too neat and clean to be true. The idea that human progress marches
steadily in a single direction, albeit with occasional slight stumbles
and reversals, is an example of what the eminent British historian
Sir Herbert Butterfield (1965, v) called “the whig interpretation of
history”:
the tendency in many historians to write on the side of Protestants
and Whigs, to praise revolutions provided they have been successful,
to emphasise certain principles of progress in the past and to produce
a story which is the ratification if not the glorification of the present.
Regardless of the debate, Butterfield (1965, 5) writes, “The historian
tends in the first place to adopt the whig or Protestant view of the
subject, and very quickly busies himself with dividing the world into
the friends and enemies of progress.”
Whiggish accounts of history suffer from multiple shortcomings. In
the first place, they tend to view the past through the lens of the present
and overemphasize those elements of history that are comprehensible
and sympathetic to their authors. For that reason, the understanding of
the past that they convey is biased in favor of the ideas and institutions
of the present. That bias tends to produce a narrative that emphasizes
historical progress toward those ideas and institutions rather than the
usually more complex, usually more alien understanding of history as
it actually happened.
THE SPREAD OF ENLIGHTENMENT HUMANISM 59
Such accounts are also teleological, meaning that they tend to portray
history as progressing toward a particular goal rather than as being
driven toward an uncertain future by the ideas, resources, and passions
of the peoples who comprise it. Hindsight is always 20/20, as the saying
goes, but Whiggish hindsight is even more acute and considerably more
narrow. Historians generally struggle against this artificial clarity, and
with good reason.
Professor Pinker is clearly aware both of the charge of Whiggery and
of its implications. He (2011a, 692) writes:
The metaphor of an escalator, with its implication of directionality
superimposed on the random walk of ideological fashion, may seem
Whiggish and presentist and historically naïve. Yet it is a kind of
Whig history that is supported by the facts.
Much of this book is devoted to an evaluation of that last claim, so I’ll
leave it to the reader to decide whether or not it’s warranted. I will note
in passing that it seems to me to be an odd defense: Few if any Whig
historians don’t believe their accounts to be supported by the facts.
While the charge of Whiggery is a fairly abstract (one might even say
“academic”) objection, it has real consequences. I briefly explore three
below: an unwillingness to recognize those aspects of Enlightenment
thinking that aren’t especially pacifistic; the tension between perpetual
peace and just war; and the fact that, even now, Enlightenment
values are limited to a fairly small and unrepresentative part of the
globe.
The Dark Side of the Enlightenment
The Enlightenment was an eighteenth-century movement that gave
rise to some of the most powerful and enduring ideas of the modern
era, many of which—representative governance, freedom, progress,
an emphasis on reason, tolerance—are so fundamental to liberal
democracy that many citizens simply take them for granted. It would be
absurd to argue that the Enlightenment did not represent a substantial
leap forward in human progress or that much of the betterment of the
human condition that has taken place over the past two centuries does
not owe a very considerable debt to the ideas that came out of the
Enlightenment.
60 REASONS FOR SKEPTICISM, PART II: EXPLANATION
That said, the Enlightenment gave rise to a tremendous diversity
of ideas, not all of which turned out to be either liberal or especially
benign. The philosopher John Gray (2015), in his critique of Better
Angels, makes this point more expertly than I will here, but the upshot
of my complaint is this: The Enlightenment produced a big, complex,
often contradictory body of ideas that don’t all lead to outcomes that
are liberal, peaceful, or progressive.
Exhibit A for this argument has to be Jean-Jacques Rousseau.
Rousseau, one of the best-known of all Enlightenment thinkers, was
also one of the most articulate critics of progress. He is most prominently
associated with the argument that modernism, science, and
progress are corrosive to morality. Johann Gottfried Herder, undeniably
an Enlightenment philosopher, is perhaps the most central philosopher
of nationalism, an idea that cuts directly against the universalism that
is typically associated with Enlightenment thought and that is rarely
associated with the spread of peace. Georg Wilhelm Friedrich Hegel’s
emphasis on freedom derived from the writings of such philosophers
as Rousseau and Immanuel Kant, but in Elements of the Philosophy of
Right he argued that true freedom can only be achieved via a strong
state—an idea that would be anathema to modern libertarians but was
welcomed by twentieth-century fascists as well as by Karl Marx, who
adapted Hegelian ideas about historical progress to produce an entirely
different brand of totalitarianism. If nothing else, the Cold War should
have dispelled any notion that Enlightenment values will save us from
conflict: Marx was nothing if not a child of the Enlightenment, yet his
disciples clashed, dangerously and often and at staggering cost, with
those of Locke and Kant.
The argument that Enlightenment humanism leads directly to peace
also ignores the distinctly illiberal ideas that arise as a reaction to the
shortcomings of Enlightenment thought and against the modernity
that it ushered in. This point is the central theme of Pankaj Mishra’s
2017 book Age of Anger, which argues that much of the violence of
the past two centuries has its roots in the anger of the people left
behind by modernity and the philosophies and demagogues to which
they were drawn. Similarly, Mark Lilla’s The Shipwrecked Mind is
an extended meditation on the power of political nostalgia and the
THE SPREAD OF ENLIGHTENMENT HUMANISM 61
reactionary ideas to which it gives rise, while Jan-Werner Müller’s What
Is Populism? highlights the antipluralist undercurrents that both arise
from and erode liberal democracy. The bottom line in each case is
that, even in democratic, free-trading societies that reflect the ideals
of Enlightenment humanism, some people end up a lot happier than
others, and disaffection can ripen into conflict.
Why do these examples matter? I’m not arguing that Kant didn’t
write Perpetual Peace or that the values of liberal democracy owe
nothing to the Enlightenment. But when Professor Pinker argues
that “liberalism, modernity, cosmopolitanism, the open society, and
Enlightenment values always have to push against our innate tribalism,
authoritarianism, and thirst for vengence” (Edsell, 2017), he misses
the point that much of that tribalism and authoritarianism owe their
present-day expression either directly or indirectly to the Enlightenment
as well. That, in turn, raises the very real possibility that the relationship
that Pinker and others see between Enlightenment values and peace
is either circular or wrong. If the only Enlightenment ideas that are
acknowledged as such are those that haven’t given rise to war, the
proposition that Enlightenment values cause peace becomes true by
definition. If we look at the full range of ideas that came out of
the Enlightenment as well as those that have arisen as a reaction
against the more liberal ones, it becomes hard to sustain the claim that
Enlightenment values inevitably lead to peace.
Indeed, scholars who have taken a broader and more nuanced view
of the Enlightenment have often found it to be a mixed blessing at best
when it comes to peace. Some even conclude that it has done more
harm than good. Roger Osborne, after a sweeping review of Western
civilization in which he chronicles ideologies, extremism, genocide, war,
and a shocking range of examples of what Robert Burns called “man’s
inhumanity to man,” offers a forceful and not atypical summary:
There remains a belief, particularly among liberal westerners, that
[the present moment] is simply a short-term crisis brought on by
the hypocritical piety of certain leaders. There is even an idea that
the current situation has been brought about by irrational, religiousbased
ideas, and that a healthy dose of rationalism will put us back
on course. The history of the last 2,500 years, and the last 150 years in
62 REASONS FOR SKEPTICISM, PART II: EXPLANATION
particular, shows that this is an illusion. The fundamental western
belief that there are rational ways of organizing the world which
will bring benefit to all has been at the root of every human-made
catastrophe that has overtaken us. (Osborne, 2006, 491–492).
I’m not a scholar of the Enlightenment by any means. But I’m pretty
sure that a discussion of the Enlightenment that only covers the parts
about science, reason, progress, and peace does not do the subject
justice. Pinker’s disagreement with a broad swath of Enlightenment
scholars on the nature and implications of the Enlightenment is
evident in Enlightenment Now, which offers the bizarre spectacle of
the Johnstone Family Professor in the Department of Psychology at
Harvard University taking a populist position against the intellectuals
who disagree with him. Perhaps predictably, the book was savaged by
precisely the Enlightenment scholars who would normally be expected
to cheer him on.15
Just War and Perpetual Peace
It’s difficult to imagine any ideology of progress leading to the
abolition of war, simply because wars are often fought to further
somebody’s notion of human progress. Even people who subscribe
to Enlightenment ideals believe in the concept of a just war, or a war
that is considered to be morally justifiable.16
For that reason, the moral
underpinnings of Western liberalism can make people more willing
to fight under some circumstances than they otherwise would be. As
Rory McCann’s character, Sandor Clegane, put it in Game of Thrones,
“Lots of horrible shit in this world gets done for something larger than
ourselves.”17
One category of liberal just war is humanitarian intervention, or
intervention to protect the human rights of people under threat. As
Professor Gary Bass (2008) demonstrates in his engaging study of the
subject, the spread of liberalism in Europe corresponded with the rise of
humanitarian intervention in practice, starting with the Greek civil war
in the 1820s. Recent examples include NATO’s involvement in Kosovo
in 1999 and Libya in 2011, as well as, arguably, the present conflict with
Islamic State forces.
THE SPREAD OF ENLIGHTENMENT HUMANISM 63
While some commentators do see a trend away from American
humanitarian intervention starting with the George W. Bush administration
(Kim 2003), global trends are moving in precisely the opposite
direction. In 2005, the member states of the United Nations ratified
a new doctrine known as the Responsibility to Protect (R2P). While
in many ways R2P establishes a formal framework for humanitarian
intervention that relies as much as possible on nonviolent means, it
also establishes as a fundamental principle of international law not just
the right of the international community to violate the sovereignty of
independent states in order to protect their populations from human
rights abuses but their actual responsibility to do so. Given many states’
present human rights practices, if taken seriously this principle could
lead to widespread intervention that would not have been countenanced
or even contemplated in the absence of liberal Enlightenment norms
and values. Professor Page Fortna (2013, 569) makes this point succinctly
in her review of Professor Goldstein’s book:
I support R2P on ethical grounds, and think it will lead to a more
just world, but there is a tension here that goes unremarked. Another
word for military intervention, even if its motive is humanitarian, is
war. R2P may well increase violence rather than reduce it.
R2P goes beyond intervention, of course. If intervention fails to
address the root causes of human rights violations or, worse, destabilizes
a state and makes future violations more likely, it will have accomplished
little. NATO’s intervention in Libya, for example, while hailed as a
“model intervention” at the time (Fortna, 2013, 569), left a greatly
destabilized country where what little security there was was provided
by rival militias.
Accordingly, as the International Coalition for the Responsibility to
Protect puts it, “advocates around the world have embraced [R2P] as
a full spectrum of responsibility: from the responsibility to prevent, to
react, and to rebuild.”18
When, as is often the case, the government
of the country in question is responsible for human rights violations,
measures up to and including regime change and nation building,
à la Iraq and Afghanistan, plausibly fall under the umbrella of
peacebuilding.19
64 REASONS FOR SKEPTICISM, PART II: EXPLANATION
Again, these are not necessarily bad goals. Muammar Gaddafi was
massacring his people with breathtaking brutality in order to avoid the
fate experienced by other leaders during the Arab Spring, and stopping
him was the right thing to do. We cannot pretend, however, that
stopping him did not amount to military intervention, or that the price
of stopping him itself isn’t still being paid in blood. A good cause may
justify the use of violence, but it doesn’t erase the fact of it.
The Study of WEIRD People
By far the best examples of a spread of peace in the decline-of-war
literature come from the Western industrialized countries in the post–
World War II era. One of Pinker’s (2011a, 249–251) most memorable
passages has to do with the number zero: No country has used nuclear
weapons on another since 1945, no Great Powers have fought one
another since 1953, no interstate wars have been fought by European
countries, no countries have conquered parts of other countries by
force (those last two have to be updated in light of Russia’s 2014 war
with Ukraine), and so on. Most of these zeros involve the advanced
industrialized countries that Pinker argues are at the vanguard of the
spread of Renaissance humanism.
This narrow focus omits a lot of people, and there’s no reason to
believe that the people it includes are representative of the rest. In
fact, there are very good reasons to believe that they’re not. In 2010,
psychologists Joseph Henrich, Steven J. Heine, and Ara Norenzayan
(2010, 61) wrote what has since become a very widely cited paper on
exactly this issue. Titled “The Weirdest People in the World?,” the article
pointed out that behavioral scientists regularly draw conclusions from
studies of people who are WEIRD—Western, Educated, Industrialized,
Rich, and Democratic. WEIRD people, the authors argue, are “among
the least representative populations one could find for generalizing
about humans” when it comes to traits like cooperation, fairness,
and moral reasoning—precisely the sorts of traits that Pinker argues
underpin the decline of war.
As you’ll see in detail later in the book, I don’t think that the claims
of a recent decline of war among WEIRD people are entirely wrong.
But WEIRD people aren’t the only ones who have managed to put a
damper on international conflict: other “islands of peace” have existed
THE ROLE OF CHANCE 65
at different points in time, and many of them can’t be attributed to
WEIRDness.
The Role of Chance
My final reason to be skeptical of the major arguments and findings that
support the decline-of-war thesis is more abstract but no less important
than those that I’ve listed so far: They generally don’t take into account
the role of chance.
We don’t tend to think about the role of chance much when we’re
contemplating international conflict. We tend to assume that because
things happened a certain way, they had to happen that way. We see
history as a concrete series of events, and if a particular kind of event
like the use of force between two countries happened more during one
part of history than it did during another, that’s all we need to know.
That’s the only way it could have happened.
A more sophisticated variant of this argument is grounded in the
distinction between populations and samples. The wars that we’re
interested in are bounded by concrete dates. That means that we have,
not a sample of wars, but rather the complete population of wars that
occurred during those historical periods. If we have the complete population,
one might argue, why are we doing statistical inference at all?
An analogy helps to clarify this position: Survey researchers typically
interview only a tiny fraction of a given population, so their conclusions
about the population as a whole (“Smith’s job approval rating is at 32%,
± 3%”) include some degree of uncertainty. But what if they interviewed
the entire population? They’d be able to say that Smith’s job approval
rating is exactly 33.12% (or whatever). There wouldn’t be any uncertainty
at all about the population’s views. Analogously, if you think of these
sets of wars as the complete population of wars that happened during
those two time periods, then it makes no sense to talk about uncertainty.
This reasoning doesn’t command a majority position among the statsand-war
crowd. In fact, I can’t think of any practitioner who has actually
espoused it. I think that’s because the overwhelming majority of us
believe, at least implicitly, that the history that we have observed was not
fated to happen the way that it did—that chance events could very easily
have changed the outcome. To return to the survey research example,
people’s answers to survey questions about things like approval ratings
66 REASONS FOR SKEPTICISM, PART II: EXPLANATION
are rarely so concrete that you’d get exactly the same answer under
different circumstances. If you ask me about President Smith on the
first sunny day of spring, when I’m out of my office and enjoying the
weather, I might tell you that Smith is doing a bang-up job. If you
ask me the same question in the dead of winter, before I’ve had my
morning coffee, I might give you a much less charitable assessment. If
that’s the case more generally, even a survey of the entire population
would produce results that contain some uncertainty because, if you
did it over again, you’d get a different number.
So it is with war. In any war, there are critical junctures at which
things might have gone a different way. Hitler’s Germany might have
been stopped cold by the French in 1940 and slowly rolled back to
Berlin. Or, as fans of The Man in the High Castle would point out, things
could have gone another way: As that alternative history goes, Giuseppe
Zangara could have succeeded in his 1933 attempt to assassinate Franklin
Delano Roosevelt, thereby eliminating both the New Deal and the
Lend-Lease Act and undermining the Manhattan Project to such an
extent that the Nazis develop atomic weapons first, bomb Washington,
and end up dividing North America with the Japanese Empire. One of
Paraguayan President Francisco Solano López’s military commanders
could have talked him out of his doomed war against Argentina, Brazil,
and Uruguay in the 1860s, or talked him into surrender once it was
clear that the war was going very poorly. Or, as I pointed out back
on page 26, the Cuban Missile Crisis could very easily have escalated
to nuclear war if one of the three Soviet officers on a Foxtrot-class
submarine had changed his mind. When you add up all of those little
chance events, it’s hard not to conclude that history could very easily
have played out differently—that some wars could have been far more
deadly and others far less so.
As these examples suggest, chance plays a prominent role in warfare
in at least two ways. First, while we can spot conflict-prone situations or
regions, we cannot with any certainty predict when conflict will actually
break out until it’s about to happen, and sometimes not even then. The
rate of conflict onset should be higher in conflict-prone situations, but
that’s a statistical generalization, not a point prediction.
The second way in which chance plays a role in warfare has to do with
escalation. Despite the fact that war has been dissected and analyzed in
THE ROLE OF CHANCE 67
hundreds if not thousands of scientific books and articles over the past
few decades, we simply don’t have any meaningful idea of how deadly a
war is going to be before it starts. The opposing parties can’t really have
a clear idea about how much blood has to be shed before a negotiated
settlement becomes possible. Worse, although many wars remain fairly
small, a very small number of wars become shockingly large—and we
don’t know with any degree of certainty which wars will fizzle out and
which ones will escalate to cataclysmic proportions.
This fact raises a real problem for analysts: How do we distinguish
the signal from the noise? Looking only at a series of events and nonevents,
how can we tell whether the fundamental relationships that
govern the thing that we’re interested in (the amount of underlying
tension that produces conflict, or the escalatory dynamics that turn a
few small wars into really big ones) have changed, or whether we’re
just looking at the kind of chance variation that we’d expect to see over
time?
This problem is exacerbated by one of those odd quirks in human
psychology: We try pretty hard to spot patterns in random data.
There is even a name for this predilection: apophenia, the human
tendency to perceive meaning in random or meaningless information.
The constellations are an obvious example of people’s ability to pick out
meaningful patterns from a random array of stars. The persistence of
horoscopes, too, is a testament to our capacity for seeing patterns in
our daily lives where (sorry) none exist. The number of people seeing
images of Jesus in everything from toast to marmite to sliced potatoes
is so well known that it has inspired a “grilled cheezus” sandwich
press.
In short, we humans simply can’t be trusted to eyeball something
and decide whether or not it’s the result of chance. We need statistical
inference to keep us honest. Statistical inference is designed with exactly
this problem in mind. Its value lies in its ability to estimate, very
precisely, how uncertain we are about our estimate of a number, whether
that number is the rate of conflict initiation or the propensity of
conflicts to escalate. Once we know how uncertain our estimates are,
we can answer the central inferential question in the decline-of-war
literature: How can we reliably differentiate changes in violent behavior
from random noise?
68 REASONS FOR SKEPTICISM, PART II: EXPLANATION
Randomness in Thick-Tailed Distributions
Perceiving patterns in randomness is an especially acute problem when
we’re dealing with thick-tailed distributions, like the distribution of
battle deaths in war. Thick-tailed distributions have the characteristic
that the overwhelming majority of observations are relatively small in
magnitude and the small number of very large ones are extraordinarily
large. Because smaller events are so incredibly common, they’re very
likely to trigger our apophenia and make us conclude that there’s
a pattern there when there really isn’t. Because larger events are so
spectacularly large, moreover, they can make us think that we’re looking
at a new pattern when we really aren’t.
Consider, for example, household income in the United States,
which is measured by the Census Bureau’s ongoing American Community
Survey (ACS). Figure 3.1 shows a sample of household incomes
from the ACS’s 2009–2013 Public Use Microdata Sample (PUMS),
an anonymized subset of the actual survey that is made available for
researchers. This particular figure shows the incomes of 1.7 million
of America’s 123 million households. This is a notoriously thick-tailed
distribution: while the majority of the households report modest or
even negative income, the top 1% earn upward of $430,000 per year.
1500000
1000000
Householdrank
500000
$0 $500,000 $1,000,000
Household income
$1,500,000 $2,000,000
0
FIGURE 3.1 The distribution of a sample of American household incomes.
THE ROLE OF CHANCE 69
Even this sample understates the length of the tail in the distribution
of income: if the long tail on the right captured the annual household
income of software magnate Bill Gates, whose net worth increases by
about $5 billion every year,20
the rescaling of the graph would make it
impossible to distinguish anyone else’s income from zero.21
Now, let’s imagine that you’ve been hired by the Census Bureau to
contact people and ask them questions about their lives, their families,
their households, and so on as part of the ACS. You manage to reach
fifty respondents over the course of your first day, and you dutifully
record all of their answers. At the end of the day, it occurs to you that
the calls you’ve been making didn’t actually seem very random: your
respondents got less wealthy as the day went on. A quick plot of income
vs. respondent number for your fifty respondents (Figure 3.2) confirms
your impression.22
The first family you reached had a household income of about
$115,000—higher than average, to be sure, but not shocking. The next
family, though, had a household income of over $850,000, which
ended up being the highest income of the day. The third and fourth
households were unremarkable, but household five had an income of
over $190,000 and household six brings in over $525,000! Tallying it up,
you find that the first six households you reached had almost 40% of
Respondent number
Householdincome
0 10 20 30 40 50
$0
$200,000
$400,000
$600,000
$800,000
FIGURE 3.2 Illusory decline in average household income over the course
of fifty interviews.
70 REASONS FOR SKEPTICISM, PART II: EXPLANATION
the total income of all fifty combined. That seems incredibly unlikely
if this really is a random sample—and if it’s not, there might be reason
to worry about the entire survey. Should you contact your supervisor
and explain your concerns?
The answer, as you’ve almost certainly surmised, is “no.” It’s not
actually too unlikely that you’d get a couple of very wealthy respondents
early on. They’re rare, to be sure, but they’re not that rare. What’s more
important is the fact that the distribution of income is really skewed:
the largest incomes are a lot larger than the smaller ones. So when
you do encounter a few households with really high incomes, your
expectations for the rest of your calls are really out of whack. As the
day progresses, you don’t see any more really high incomes, and you
start to wonder whether something has changed. It’s normal to think
that it has—the downward trend in the incomes of your respondents
is pretty noticeable. But it really is an illusion: the fifty observations in
Figure 3.2 were drawn totally at random from the PUMS data. It just
looks a lot like a trend because the big outliers are so much bigger than
we intuitively think they should be by chance—and because we humans
are cursed with apophenia.
Now, with all that in mind, take another look at Figure 3.2. Once
you’ve done so, flip back to Figure 2.9, the graph of annual battle
deaths divided by world population. Keeping in mind the fact that
battle deaths, like household income, follow a thick-tailed distribution,
are you really confident that what you’re seeing is a downward trend in
the data and not just random noise? I’m not.
Another way to see this point is to take a look at Figure 3.3,
which shows what happens when you draw random numbers from
a normal distribution and a particular thick-tailed distribution called
a power law distribution (which, not coincidentally, provides a pretty
good fit to the data on the deadliness of war, as we’ll see in chapter
5). The value on the y-axis is the running mean of these series of
numbers: At observation 200, the running mean is the average of
the first two hundred observations. The dashed line represents the
average of the distribution from which these observations are drawn.
If you want to use a sample value as an estimate of that average, the
running mean should converge to the underlying average value pretty
quickly.
THE ROLE OF CHANCE 71
0 200 400 600 800 1000
5.0
5.5
6.0
6.5
7.0
Sample size
Normalsamplemean
0 200 400 600 800 1000
15
20
25
30
35
Powerlawsamplemean
FIGURE 3.3 Running means of random samples drawn from normal (top)
and power law (bottom) distributions as the number of observations
increases from 1 to 1,000. Dashed gray lines represent the true average,
or population mean.
We can see that in the case of the normal distribution, it does just
that. The initial average of your sample observations may be off by a fair
bit initially, but it very quickly converges to something close to the true
underlying average value and more or less stays there. This is why we
can get reasonable estimates of the characteristics of a population (on,
say, a survey) just by looking at a sample.
The sample average of the data drawn from a power law distribution,
by contrast, wanders all over the place and depends crucially on the
number of large observations that occur in the sample. If there are
more than we’d expect, the sample average will be too high. If there are
fewer, it’ll be too low. And because those large observations are really
rare events, it takes an incredibly long time for them to average out and
produce a sample mean that’s reliably anywhere near the value that we’re
trying to estimate. That’s why we can’t just use standard statistical tests,
like a difference-of-means test, to answer the question of whether war
72 REASONS FOR SKEPTICISM, PART II: EXPLANATION
is becoming less deadly. Fortunately, as we’ll see in chapter 5, there are
some nonstandard statistical tests that can be of use.
Specifics aside, the point is this: The only way we can really know
whether we’re looking at a trend or at random noise is to rely on
statistical tests to help us determine whether the number that we’re
trying to track, whatever it is, has changed more than we’d expect due to
chance. Despite the fact that this problem is at the heart of the declineof-war
debate, such tests are routinely disregarded. In two recent articles
in which there have been statistical tests of the battle-death data, the
results indicate that the apparent trend is very plausibly an illusion—
a normal pattern following the misleadingly large outliers of the late
1940s and early 1950s.23
Conclusion
This chapter and the previous one have laid out a handful of reasons
that have prompted me to be skeptical of the decline-of-war thesis.
Undermining the evidence in favor of that thesis is not, of course, the
same thing as showing that there is no decline: The absence of evidence
is not evidence of absence. It is much harder to answer the question of
whether or not war is in decline, and if so, when, where, and why. I
take up this challenge in the next three chapters.