human-meaning

Is Violence Really Declining?

In 1945, two atomic bombs killed approximately 200,000 people in Hiroshima and Nagasaki. In the 80 years since, nuclear weapons have killed no one in war. Zero.

Is this evidence that humanity is getting better? Or are we just living in the gap between catastrophes?

This question—whether violence is genuinely declining—turns out to be one of the most contentious in social science. The data seems clear. The interpretation is anything but.

The Claim

Steven Pinker, in his 2011 book The Better Angels of Our Nature, made the most comprehensive case for declining violence. His argument spans multiple timescales:

The Pacification Process. The transition from hunter-gatherer societies to states reduced violence dramatically. Pre-state societies had homicide rates orders of magnitude higher than modern states—15-25% of deaths from violence, compared to less than 1% today. Archaeological evidence supports this: skeletal remains from prehistory show rates of violent trauma far exceeding anything in recorded history. The state, whatever its other crimes, seems to suppress private violence.

The Civilizing Process. Within states, homicide rates have fallen steadily for centuries. Medieval Europe had murder rates 30-50 times higher than today. The decline correlates with the consolidation of state power, the rise of commerce, and changes in manners and self-control. This is sociologist Norbert Elias's thesis: as courts became centers of power, elites developed self-restraint, and those habits gradually diffused through society. What was once normal—dueling, blood feuds, casual brutality—became shameful.

The Humanitarian Revolution. The 18th and 19th centuries saw the abolition of judicial torture, public execution, slavery, and cruel punishments. Practices that were once normal became abhorrent.

The Long Peace. Since 1945, major powers haven't fought each other directly. This is historically unprecedented—great powers used to fight constantly. Before World War II, a European great power war occurred about once per generation. In the eight decades since, zero. The United States and Soviet Union armed themselves to annihilation but never fired a shot at each other directly.

The New Peace. Even civil wars and genocides have declined since their Cold War peaks. The 2000s and 2010s saw fewer war deaths than any decade in recorded history. The conflicts that do occur—while devastating to those involved—are smaller in scale, shorter in duration, and less deadly per capita than the wars of previous centuries.

The Rights Revolutions. Violence against minorities, women, children, and animals has declined dramatically within living memory. What was once acceptable is now criminal.

Pinker marshals enormous amounts of data across centuries and societies. The pattern seems undeniable: by virtually every measure, humans are less violent than they used to be.

Why This Matters

If Pinker is right, the implications are profound:

Progress is real. We're not just cycling through the same patterns of violence. Something has actually changed about human society—or human nature. The Enlightenment project of improving humanity through reason wasn't naive idealism—it was working.

Institutions matter. States, commerce, literacy, international organizations—these weren't just nice to have. They fundamentally altered human behavior. Understanding which institutions reduced violence could help us build more of them, strengthen the ones we have, and avoid dismantling what works.

The future can be better. If violence declined for reasons we can understand, we might be able to extend those trends. Peace isn't just luck—it's engineering.

Pessimism is wrong. The common belief that things are getting worse—that the past was more peaceful and people used to be better—is empirically false.

But if Pinker is wrong—or if the picture is more complicated—then:

We're complacent. Believing violence has permanently declined might blind us to risks we should take seriously.

We're measuring the wrong things. Deaths from direct violence might be falling while suffering from other sources increases.

We're in a lull, not a trend. What looks like a trajectory might just be fluctuation. The next catastrophe could be around the corner.

The stakes are high. Our policy choices, our institutions, our sense of what's possible—all depend on whether the optimist story is true.

The Skeptics

The decline narrative has attracted fierce criticism:

The Statistical Critique

Nassim Taleb and Pasquale Cirillo argue that Pinker's statistics are fundamentally flawed. War deaths follow a fat-tailed distribution—most years have few deaths, but occasional catastrophes dominate the total. In such distributions, trends are meaningless.

Imagine flipping a coin that's mostly fair but occasionally explodes. You can't conclude the coin is getting safer just because it hasn't exploded recently. The lack of recent catastrophes doesn't predict future catastrophes—and extrapolating from recent data is the same error that led people to conclude housing prices could never fall nationwide.

This is known as the problem of "silent evidence." We observe the years without catastrophe because we're alive to observe them. We don't observe the alternative timelines where nuclear war happened in 1962 or 1983. Survivor bias makes peace look more robust than it is.

Taleb's argument is technical but devastating if correct: you literally cannot conclude that violence is declining from the data Pinker presents. The sample is too small, the variance too high, the tail risks too extreme. We would need thousands of years of data to have statistical confidence, and we have less than a century of the "Long Peace."

The Definitional Critique

Johan Galtung introduced the concept of "structural violence"—harm embedded in social arrangements rather than inflicted by identifiable actors. When people die from poverty, lack of healthcare, or environmental degradation, they're victims of violence even if no one pulled a trigger.

By Galtung's measure, violence might be increasing even as direct killing decreases. A world where fewer people are murdered but more people die preventable deaths hasn't become less violent—it's just changed the form of violence.

Critics argue that Pinker's framework is too narrow, counting only direct physical harm while ignoring the violence of systems. Colonialism, for instance, killed millions through famine, disease, and economic extraction—but much of this wouldn't show up in murder statistics. The Bengal famine of 1943 killed three million people under British rule—was that violence? If a pharmaceutical company prices a drug beyond what dying patients can afford, and they die, is that violence?

The definitional question isn't just academic. If violence is declining only because we've changed what counts, that's a very different story than genuine moral progress. We might be outsourcing violence rather than reducing it—pushing it onto distant populations, future generations, or other species.

The Temporal Critique

How long is the trend? If you measured violence from 1939-1945, you'd conclude humanity was doomed. If you measure from 1990-2020, you'd conclude peace is inevitable. The time window determines the answer.

The Long Peace is now 80 years old. That's a long time in human memory but a short time in historical terms. The Pax Romana lasted two centuries. The Concert of Europe maintained relative peace for a century. Both ended in catastrophe.

Are we on a genuine new trajectory? Or are we just living in another interval—one that could end as suddenly as the others?

The question is especially pressing given that we now possess weapons capable of ending civilization. Previous peaceful intervals ended in wars that killed millions. If this one ends similarly, it might kill billions—or everyone. The stakes of being wrong about the trend have never been higher.

The Data

Let's look at what we actually know:

Homicide rates have declined dramatically. This is the best-documented trend. Medieval Europe had rates of 20-40 per 100,000; modern Western Europe has rates of 1-2 per 100,000. The decline is real and large.

War deaths per capita have declined. More contested, but the data suggests fewer people are dying in wars relative to population. The 20th century's world wars were catastrophic, but they were also anomalous—most centuries saw steady low-level warfare rather than occasional total war.

Torture and cruel punishment have been abolished in most places. This is observable and recent. Within living memory, public executions and judicial torture were eliminated from most societies.

Violence against previously vulnerable groups has declined. Women, children, minorities, LGBT people—all face less legal and social violence than they did a generation ago, at least in developed countries. Domestic violence, child abuse, and hate crimes, while still present, are now illegal and socially stigmatized in ways that would have been unimaginable to previous generations.

But:

The 20th century was extraordinarily violent. 100-200 million people died from violence in the 20th century. This is more than any previous century in absolute terms. Pinker argues per capita rates are what matter; critics argue absolute numbers matter too.

We may be in a statistical lull. The distribution of war deaths has extreme variance. Decades of peace followed by catastrophic war is the historical norm, not evidence of a trend.

New forms of violence may be emerging. Cyber warfare, terrorism, environmental destruction—these don't fit neatly into historical categories. An attack on infrastructure could kill thousands without anyone firing a shot. Climate change may displace hundreds of millions. Are we measuring the right things?

What Would Change Our Minds?

The debate is partly empirical and partly conceptual:

Empirically: Another major war would obviously shift the statistics. So would a sustained period of low violence. At some point, the Long Peace becomes long enough to constitute evidence of a real shift.

Conceptually: The disagreement is about what violence means, what counts as evidence, and how to handle uncertainty. Pinker and Taleb aren't just reading the data differently—they have different theories of probability itself.

Causally: Even if violence has declined, we need to know why. If it's due to nuclear deterrence, we should worry about proliferation. If it's due to commerce and interdependence, we should worry about deglobalization. If it's due to changes in values, we might be more optimistic.

The Takeaway

Is violence really declining? The honest answer: it's complicated.

The data on homicide and some forms of violence is strong—rates have fallen dramatically over centuries. The data on war is more ambiguous—we may be in a statistical lull rather than a genuine trend.

The conceptual questions are even harder. What counts as violence? How do we handle tail risks? What time horizon matters?

This series will explore these questions in depth. We'll examine Pinker's optimist case, Taleb's statistical critique, Galtung's structural violence, and the various theories of why violence might be declining.

The goal isn't to declare a winner. It's to understand what we can know, what we can't, and what the implications are for how we live.

Is Violence Really Declining?

The Claim

Why This Matters