Taleb refutes Pinker on war

A long time back I criticized Pinker’s “better angels of our nature”, arguing that he mistakes leftism for progress and goodness, arguing that criminal violence by citizens against each other is at extraordinarily high levels, that state violence against citizens in the form of imprisonment and policing is at levels unprecedented except when a state is crushing a hostile enemy population in a condition of war or near war, and that levels of warfare between states are pretty much what they always have been, arguably considerably worse.

For example Pinker complacently observes that the Victorians were shocked and horrified by a crime wave, but neglects to observe that this crime wave consisted of one mugging in London every few months – which crime wave never went away, but instead people got used to it, and then it got vastly worse, and people got used to it again, and then it got vastly worse still, and people attempted to abandon much of their cities to savages, and then the crime wave followed them, and there is now no safe area in London.  The idea of the inner city as some kind of jungle is new, starting in the late nineteen forties, early fifties.  Early in the twentieth century, the idea that the affluent and respectable might have to abandon vast expanses of wealth and property, of huge, beautiful and high status buildings where once the wealthy and fashionable lived, to the vandalism and depravity of savages would have been as unimaginable as wolves and bears prowling the streets of London to devour passers by.

My criticism of his argument on war is that war is a bursty phenomenon, sometimes there are a lot of mighty big wars, and sometimes, when one hegemon has the upper hand, or several hegemons  remember the last big war too well, not many wars.  This peace lasts until the dominant hegemon weakens, or people forget how bad the last big war was, forget how easy it is to start wars, and how hard it is to stop them, whereupon they go at it again.  And the generation that remembers the last big war is now mostly dead.

Taleb, arguably the worlds leading expert on the statistics of bursty phenomena, makes the same argument in a more scientific fashion backed by statistics.   War follows a power law with an exponent substantially less than one and substantially greater than zero, rather than a normal distribution, meaning that risk is dominated by large rare events – the risk of losing life and property in a big war is far greater than the risk  of  losing life and property  in a small war, even though small wars are common and big wars are rare.

Pinker tells us.

wars between great powers and developed nations have fallen to historically unprecedented levels. This empirical fact has been repeatedly noted with astonishment by many military historians and international relations scholars…

Taleb tells us that because war follows a power law rather than a normal distribution, if one analyzes the level of warfare using statistics appropriate to a normal distribution, at any given time, chances are it has either fallen to historically unprecedented levels, or a great war has broken out and one is too busy trying to stay alive to do statistics.

With a power law phenomenon, recent experience almost always massively under estimates the risk of large rare events, recent experience is almost always nicer than experience over a longer period.  Until it is not.

31 Responses to “Taleb refutes Pinker on war”

    • jim says:

      Yes, seen it, don’t think it is responsive, a mere stream of ink signifying nothing. An actual reply would necessarily have to contain actual statistics.

      Pinker says:

      Yes, that’s exactly what I point out: great-power wars became steadily more destructive from 1500 through 1945. The turning point tat marks the onset the Long Peace was in 1945, not 1914.

      Seventy years of peace is not very long at all, not a “long peace”. It is arguably the median gap between episodes of horrifically destructive great power warring.

      So if great power wars became steadily more destructive from 1500 to 1945, seventy years of near peace is not nearly long enough to break the trend towards greater and greater state violence.

      As Taleb points out, for any seventy years in the last five centuries, chances are that that seventy years was far more peaceful than the average, because the average includes infrequent events of enormous violence.

      For most of the past five centuries, people have had just as good evidence for a “long peace” as Pinker has, until suddenly they did not any more.

      • Dr. Faust says:

        Around 80 years is the standard between American wars. That’s just long enough for most of the veterans to die off and for everyone to forget about war.

        If you read psychohistory they would say that a lead up to war is prefaced by a depiction of mothers as killers signifying a country’s eagerness to fight.

  1. Max says:

    hegamon s/b hegemon

  2. Irrelevant says:

    Even calling war “bursty” rather understates the problems with prediction here. Unless you’ve got some magic method of controlling for population density and prevalent style of warfare, the total war deaths function = the Chinese war deaths function + noise, which makes it even more uninformative for “civilizational risk” or whatever we’re actually trying to measure than it otherwise would be.

  3. Rasputin says:

    I enjoyed most of this, but your statement “…and there is now no safe area in London” is WAY out. I live there, and while there were areas that you wouldn’t want to be in late at night a decade or so ago, most of those areas have been massively gentrified over the last 5 years. From Hackney through Brixton, the trend which you are talking about has been thoughly reversed, and you wouldn’t get that much change out of £500,000 for a nice one bedroom flat.

  4. Lars Grobian says:

    Most of this makes sense, but you are wrong about crime. Have you seen hbd*chick’s posts about the declining murder rate over the centuries in England and Western Europe, starting in the Middle Ages? Victorian London is hardly the default state of the human race.

    Of course that makes Pinker wrong too, if he wants to pretend violent crime is on a continuing decline since AD 1300 or whenever. The murder rate stopped declining decades ago, and there’s nothing in sight offering any hope that the reversal is temporary.

  5. OldStudent says:

    Is Lyme Disease an offensive or a defensive bio weapon?
    I understand it hijacks the amygdala.

  6. Michael says:

    I like this jim
    There are more real crimes committed, more people in jail, and vastly more actions that the state deems crimes – and if what the state deems a crime is not truly a crime, then that is state violence.
    theres also a website devoted to cataloging states killing people i want to say he calls it democide its been a while.

    • Lesser Bull says:

      Violent crime stats are skewed in the US because the most violent are locked up and not usually charged for assaults and rapes in jail. They are often just dealt with administratively.

  7. OldStudent says:

    I notice that only prisons are mentioned.
    You neglect to mention the violence of electroshocking
    And the forced medication and lobotomizing of the
    “Mentally Ill”, abetted by Police Enforcers.
    Cui Bono from these practices?

    Je Suis Frances

  8. […] rule. Re-Colonization and Future Orientation. Where public policy misses biology. Where Pinker got lost. Religion and morality (topologically related). The death of trust. America’s secessionist […]

  9. Wars do not “follow” power laws. Wars are caused by men. At best we can say that our knowledge of the uncertainty of the frequency of wars can be quantified by some probability model.

    Taleb suggests his model is better than Pinker’s, and he’s almost certainly right. But Taleb, like Pinker, reify their models and ascribe the cause of wars to be the probabilities given by the models.

    I have much more here: http://wmbriggs.com/post/16012/

    See also the comments by “Ye Olde Statistician” about the quality of the data.

    • I think you need to spend some time running some monte carlo simulations. Pinker is trying to use a few monte carlo histories and trying to make a statement of the “generator”, when we need the entire history to do that.

      The rest of your argument is not really a rebuttal but trying to be against the entire philosophy of probability/statistical inference in general, and is still not spelled out enough to make sense.

      “The only reason to build a model of violence—and it’s a darn good reason—is to predict how many dead bodies we expect to create in the future (so we know where not to be). But given all the unquantifiable uncertainties mentioned, I would have very little confidence in that model.”

      Or every other reason to build a ‘model’ other than “prediction”. The rest of your blog posts are cool, because anyone who knows probability and philosophy of probability past entry-level discussions is cool. However, I struggle to see how it warranted a blog post because it hardly made any point at all. Either you missed the point of his posts or you were just trying to write a post with backlinks to all your articles and got a bit carried away, your continual use of “prediction” worries me.

      Btw to cut to the chase if we have a back to back via blog posts, yes I know of cox’s axioms, franklin’s book is on my shelf, etc etc we can just skip to the meat of the discussion.

      “Wars do not “follow” power laws. Wars are caused by men. At best we can say that our knowledge of the uncertainty of the frequency of wars can be quantified by some probability model.”

      This statement is an empty verbalism. It’s a highly acute misphrasing, at best its like saying velocity is not the integral of acceleration because there are no “integrals”.

    • and furthermore the idea that everyone thinks probability is “just a number” or quantifiable (uncertainty in your phrasing) is a myth, see Marsay’s entire website dedicated to debunking this trope


      • jim says:

        Obviously there are always unknown unknowns, and we can be pretty sure that the likelihood of war is dominated by unknown unknowns. If war intensity had many small causes of small effect, they would sum to a good approximation to a normal distribution, and war would have approximately constant intensity. In which case the low level of warfare since World War Two would mean that humans have become nicer or wiser.

        If however, wars are like avalanches, where the fact that one of the assassins of the archduke lucked out can set off something arbitrarily large, where small causes can have gigantic effects, then the low level of warfare since World War II is just the luck of the draw, and one day some trivial event will result in every nuclear power nuking everything in sight.

        Taleb shows that wars follow the kind of distribution that avalanches have. Which tells us a lot about our unknown unknowns.

        • Dave Marsay says:

          Hi, I’m the Marsay that SanguineE… refers to. The state of the art in modelling conflict used to be to fit stochastic models and extrapolate. This is limited for the reasons that Taleb gives, and his points are well taken by the appropriate communities. (The benefit of an unusual surname is that you can google me on this, if you want.)

          An alternative way of modelling, that came out of the Great War, is monitor how well one is doing at fitting the stochastic models and what assumptions one is having to make. My experience is that if you run through the data and pick out significant changes in ‘epoch’, these pretty much match up to those picked out by people more closely engaged. One has as similar situation in economics. There are those who hold that economies are stochastic and even ergodic and that 2008 had some improbable events, but this hardly seems meaningful given that no-one seems to have the slightest idea of the relevant conditional probabilities. The main thing is that whether stochastic and hence ‘objectively knowable’ but practically unknowable, or (as I propose) not actually stochastic, the implications for action are pretty much the same. In conflicts of all kinds, it is no use just trying to prevent a repetition of the last type of crisis.

          • jim says:

            The interesting question is, in the light of past behavior, how surprising is it that seventy years have gone by without a direct conflict between great powers?

            I don’t think, and Taleb does not calculate, that this is particularly surprising, and if it was surprising, I would attribute it to the Pax Atomica, the peace of terror, not to an unusual improvement in wisdom and virtue.

  10. Old Student says:

    Frances says there is Sweet Music Here.
    She will Always Love David
    This is Her Song

  11. […] Don’t worry, despair. More. Related: There will be no big event. Related: Taleb refutes Pinker on war. […]

  12. […] Don’t worry, despair. More. Related: There will be no big event. Related: Taleb refutes Pinker on war. […]

  13. […] Also from Jim, coverage of the burgeoning feud between Pinker and Taleb. […]

Leave a Reply