Steven Pinker and the need for modest claims in speculative science writing

By Nathan Geffen, 9 January 2015

"Never mind the headlines. We've never lived in such peaceful times." This is the subtitle of an article by Steven Pinker and Andrew Mack published in Slate in December. They write, "The world is not falling apart. The kinds of violence to which most people are vulnerable-homicide, rape, battering, child abuse--have been in steady decline in most of the world. Autocracy is giving way to democracy. Wars between states--by far the most destructive of all conflicts--are all but obsolete. The increase in the number and deadliness of civil wars since 2010 is circumscribed, puny in comparison with the decline that preceded it, and unlikely to escalate."

Pinker assembled a mass of data for his 2012 tome The Better Angels of Our Nature to make the case that the rate at which humans kill each other, whether in war or peacetime, has come down through the ages. His wide knowledge of the statistics of murder going back thousands of years, that he presents in Better Angels is impressive. At nearly 700 pages excluding references, it is impossible for a short article like this to summarise adequately Pinker's proposed explanations for the decline of homicide. Key causes, he argues, are the increasing monopoly on violence by states, and that humanity is undergoing a civilising process as evidenced by a multitude of factors from seemingly trivial matters like improved table manners to the development of heftier values, such as the concept of human, and even animal, rights.

But there are problems with the way Pinker presents his thesis that provide lessons for how science should and shouldn't be written. Better Angels suffers from a severe lack of modesty -- and the Slate article even more so. I do not mean modesty in its usual sense; whether Pinker and Mack are humble or braggarts is irrelevant to whether or not their claims are true. I mean rather that Better Angels at times ignores evidence that contradicts its central claims, and fails to acknowledge the limitations of evidence supporting them. This doesn't mean the book's central claims are false. On the contrary I am mostly persuaded that the rate of homicide has declined and that these are much better times for a greater proportion of humanity than many centuries past. Nor is this a problem from the bookshop popular science shelves that is confined to Pinker's book, but it does offer an excellent case study in immodest science writing.

Before justifying my criticisms I confess that I am in no position to affirm or contest most of the book's data. However, I am well acquainted with two of the conflicts discussed in Better Angels and the Slate article: Iraq and Israel. Pinker's analysis of the former is misleading, and of the latter, shallow.

In the Slate article, Pinker and Mack state that there have been 150,000 civilian deaths in Iraq between 2003--the year the US-led war began--and 2014. As in Better Angels Pinker appears to back the estimate of Iraq Body Count, which primarily uses newspaper reports to tally deaths. But this new and controversial methodology of tallying civilian war deaths constitutes no more than a lower bound, and has a multitude of shortcomings. Pinker and Mack have backed a war mortality estimate that is convenient for their thesis, but quite likely wrong.

Now estimating the number of violent deaths in Iraq, civilian or otherwise, since the US invaded is notoriously difficult and highly contested, as explained in this Mother Jones article. A multitude of studies covering different time periods have been done. Pinker and Mack's number is on the very low-end of the scale, and implausible. At the high-end one study estimated over a million violent deaths by 2007.

The most virulently debated estimate comes from a survey published in The Lancet. It estimated about 600,000 violent deaths in Iraq, both civilian and combatant, from 2003 to 2006. In Better Angels Pinker dismisses this study in a couple of superficially argued paragraphs. He presents neither the readily available arguments defending the study, nor other surveys, such as a World Health Organisation one, which estimated lower than the Lancet but nevertheless much higher than Iraq Body Count. Nor does Pinker list or explain the criticisms of the Iraq Body Count estimates. The Wikipedia article on casualties of the Iraq war does a far better job of describing how complex matters are than either Better Angels or the Slate article.

What we learn from the widely divergent estimates of violent deaths in Iraq is that even in a time of sophisticated data collection and instant access to information it is incredibly hard to estimate war casualties. The task must be even harder for wars preceding our generation. But a plethora of statistics presented in Better Angels rests on these kinds of doubtful estimates. Faced with this kind of uncertainty, Better Angels would have done well to present its claims with greater circumspection.

About the Palestinian-Israeli conflict, Pinker and Mack write in their Slate article that despite last year's attack on Gaza, the "Israel-Palestine conflict was once a far more dangerous Israel-Arab conflict. Over the course of 25 years, Israel fought the armies of Egypt, Syria, and Jordan five times, with more than 100,000 battle deaths, and in 1973 both Israel and the United States put their nuclear forces on high alert in response to the threat. For the past 41 years there have been no such wars, and neither Egypt nor any other Arab regime has shown an interest in starting one."

It's a long argument to make here, but the homicide rate is not the key issue in Palestine. It's much safer to live on the West Bank than in South Africa. You are less likely to be murdered, whether by Palestinians or Israelis, in Ramallah or Bethlehem than Cape Town (where I live) or Johannesburg. The problem is state oppression of Palestinians, primarily in Gaza and the West Bank, but also in Green Line Israel. This has got progressively worse over the last 40 odd years. By only looking at the homicide rate in the Israeli-Palestinian conflict, Pinker and Mack fail to get to the essence of what is wrong there.

The Israeli-Palestinian conflict is also instructive in showing that while homicide as a marker for the state of the world is useful, it only takes you so far. Brazil, Mexico and South Africa have much higher recorded homicide rates than North Korea, Saudi Arabia or Iran; that doesn't make the latter better societies; I suspect most people would argue quite the opposite.

Others have pointed out further examples of cherry-picking and misleading claims in Better Angels. For example:

• Craig Lerner explains that in data Pinker presents to make his argument that the two world wars of the 20th century were not the worst in history, he compares them with atrocities that took place over much longer periods of time, such as the mideast slave trade which spanned 1,200 years. "Historians a millennium from now, if they exist, will likely collect several wars of the last century (World War I, the Russian Civil War, World War II, etc.), as a single event. So aggregated, the 'Germanic Wars of the Twentieth Century' emerge ... as one of the deadliest conflicts in history," writes Lerner.
• David Bentley Hart criticises Pinker for calling attention "to the many brutal features of much of medieval life" but not being evenhanded and acknowledging the "moral goods that medieval society achieved despite its material privations," several of which Hart lists.
• A fundamental argument in Pinker's book is that it is the rate of homicide, not the gross number of homicides that is the relevant factor to consider in humanity's march to peace. On the face of it this seems obvious. As the human population grows, so too will the number of homicides. But matters aren't that simple. As Hart writes, "a remote Inuit village of one hundred souls where someone gets killed in a fistfight is not twice as violent as a nation of 200 million that exterminates one million of its citizens. And even where the orders of magnitude are not quite so divergent, comparison on a global scale is useless, especially since over the past century modern medicine has reduced infant mortality and radically extended life spans nearly everywhere (meaning, for one thing, there are now far more persons too young or too old to fight)."
• Edward Hermann and David Peterson point out an interesting sleight of hand in Better Angels. In a graph depicting what Pinker claims are the random distribution of wars from 1800 to 1950, Pinker invites readers to "Cover up the two outliers with your thumb". The two outliers are of course the two world wars, which if you leave them uncovered substantially lessens the apparent randomness Pinker wants his readers to see.
• To show how increasingly humane we're becoming, Pinker makes particularly superficial arguments on our changing ethic towards animals. Vegetarianism and animal rights are on the rise he argues in Better Nature. But he glosses over the worldwide rise in meat consumption, and with it factory farming practices that accentuate animal suffering. It's all very well that more people care about animal rights nowadays, but turning Pinker's own thesis back on him, the rate (and absolute number) of human-inflicted animal suffering is increasing dramatically. (I'm not a vegetarian, but I am simply pointing out the weakness of this aspect of Pinker's book).

The critique of Pinker by Herman and Peterson is particularly caustic, and occasionally compelling. Better Angels is a "propaganda windfall for the imperial bloc" that "could only be purchased with a denial of reality," they write. This is not entirely fair. Arguing that human rights struggles and global institution building has made the world safer and nicer does not make one an apologist for imperialism. And Pinker's views on Islam are a bit more nuanced than Herman or Peterson give him credit for. Nevertheless, these writings of Pinker and Mack--as well as Herman and Peterson--suffer from far too much mixing of personal political views with consideration of empirical questions. When readers take the time to read Better Angels, they might reasonably expect a fair description of the evidence both for and against the claim that violence has declined, as well as an honest acknowledgement of the limits of the available data. What a reader, or at least this one, does not want is an indecipherable mixture of the writer's political opinions and biases with the data, that results in a distorted, substantially incomplete presentation.

Herman and Peterson describe Better Angels as a terrible book. I disagree. It is an important albeit seriously flawed book that is worth reading. But the Slate article with Mack really is terrible. It emphasises very short-term trends, which Better Angels largely avoids, and cannot be predicted by Pinker's explanations for the decline of violence. It is a mush of superficial political musings. For example, the Russia-Ukraine conflict is rosily analysed in a short paragraph.

Pinker has cherry-picked, at least in some important parts of his book, as well as the Slate article. Inconvenient facts are discarded too easily, so that claims are made without the requisite scientific modesty they require. None of this disproves that the long-term rate of violence is declining, nor the explanations for this that Pinker (and Mack) advance, but the evidence in at least some important cases has not been presented with the scientific modesty required by the limitations and inconsistency of the data.