Nassim Taleb on the black swans of war

[Warspite]

It's been a long morning, and I maybe can't really understand what Taleb is saying.

So correct me if I am wrong: the fact that conflict and violence has measurably been decreasing doesn't make the world 'a more peaceful place' because the substantial possibility remains of a future, highly destructive conflict tied in some way to past and/or current events?

Maybe I am just too soaked in the assumptions of my field of conflict studies, but I am struggling to think of anyone working in foreign affairs or international security that doesn't know this to be true - even those of us who have read Pinker's work and taken it on board. Anyone aware of the existence of nuclear weapons, for instance, will know that conflict - a single conflict event - has the potential to wipe out in minutes what the Second World War took six years to do. Anyone watching Ukraine over the last two years, as an other example, will have an instinctive appreciation for the dangerously destabilising effect a seemingly self-contained civil political dispute can have on military security of an entire continent.

And like others have suggested in this thread already, other variables have an important impact on the incidence of war and violence, on which a lot of work has been done.

It seems to me, again I could be wrong, that Taleb and his co-author are using statistical theory to make a dubious point that we cannot consider the world peaceful because of the potential impact of a potential conflict. I think this mixes up 'peaceful' with 'stable'. The world in 1913 was peaceful; but it was not stable.

The TL;DR version - they're playing Languish-semantics because they don't like Stephen Pinker's book

Also:

Regarding the role of statistical tools, Taleb says, "Physicists get the point. But social scientists do not get it. Their statistical tools do not work for what we call fat tails, as defined as something prone to black swan events," he says. "They can write whatever narrative they want. But we call that journalistic. That's not statistical."

Is it really the case that someone so intelligence as Taleb could fail to realise that modelling human behaviour, particularly historical human behaviour with scant or unreliable data sources, is not at all like modelling physical systems?

[Warspite]

I find the general premise that wars are random events questionable. The long peace theory is not based on mathematical or chronological considerations, but on alleged or perceived changes to human interaction in the last decades. As human interaction is not random, but based on will a pure mathematical analysis could fall short.

I can't think of any quantitative political scientists that model wars as random events.

[Admiral Yi]

Here's my take on what he's saying:

History doesn't have nice, evenly distributed wars of average magnitude.  Things can go peacefully for a while the all hell breaks loose.  So the fact that we're in a peaceful stretch right now doesn't necessarily mean that the underlying distribution has changed.

Which, as an aside, is not at all the same thing as saying the frequency and magnitude of war are exactly the same now as they have always been.

[DGuller]

So correct me if I am wrong: the fact that conflict and violence has measurably been decreasing

Let's stop right here.  The bolded assertion is what is being challenged.  Taleb's argument is that wars are so statistically volatile that it is not statistically supportable to claim that war frequency has measurably been decreasing.  He's not saying that it's increasing or staying the same, he's just saying that you can't tell with the data we have, if you properly account for the historical volatility of war frequencies.

[DGuller]

Here's my take on what he's saying:

History doesn't have nice, evenly distributed wars of average magnitude.  Things can go peacefully for a while the all hell breaks loose.  So the fact that we're in a peaceful stretch right now doesn't necessarily mean that the underlying distribution has changed.

Which, as an aside, is not at all the same thing as saying the frequency and magnitude of war are exactly the same now as they have always been.

This is exactly right.

[Warspite]

So correct me if I am wrong: the fact that conflict and violence has measurably been decreasing

Let's stop right here.  The bolded assertion is what is being challenged.  Taleb's argument is that wars are so statistically volatile that it is not statistically supportable to claim that war frequency has measurably been decreasing.  He's not saying that it's increasing or staying the same, he's just saying that you can't tell with the data we have, if you properly account for the historical volatility of war frequencies.

Ok. Then we need to see how Taleb codes "war" because what has been considered under the label "war" has a huge impact on how violent the world seems.

As a practical example: consider that measures of violent death in El Salvador did not decrease after the signing of the Chapultepec peace accord; on the contrary, they even increased. Yet the "war" was considered to have stopped; the event was finished. Other highly violent societies slip in and out of "war" based on reasons of political labeling, not underlying reality.

What Pinker's work tried to show was that stripping away the labels the world has been becoming less violent - and thus more peaceful.

[grumbler]

I think this criticism misses the mark as well.  Statistics can be and is applied successfully to many events that are not truly random (if there even is such a thing as a truly random event).  Car accidents are not truly random either, they're a result of two drivers' decisions that eventually made them occupy the same space at the same time.  The key is whether the chain of events is so complex that the outcome is not truly predictable due to butterfly effects except in extremely short time horizons.

Okay, I can see I overstated the "random" element, but the argument that, even if wars are demonstrably less frequent over the last 70 years, such evidence is meaningless because of the "fat tail" of a distribution still implies an element of randomness.

[frunk]

Okay, I can see I overstated the "random" element, but the argument that, even if wars are demonstrably less frequent over the last 70 years, such evidence is meaningless because of the "fat tail" of a distribution still implies an element of randomness.

Taleb quotes this research, but it is a useful example.

The Physical Limits of Trick Shots in Billiards

The joker in the deck is gravity, a force that no one can entirely "screen out," no matter where they are in the universe. It makes a difference in the path of molecules and the path of billiard balls. For the first collision of a billiard ball, we can control the variable so well that we don't really have to think of gravity as anything other than the force holding the ball on the table. We control all the variables that matter: the placement and force of the hit. After a couple of collisions, we're less able to determine where the balls go. Even on an idealized surface, there are many options depending on the exact force with which the balls meet, and the forces acting upon them. After six or seven collisions, you don't just have to worry about the gravity of the Earth, but of the gravity of the people walking around the table. Exactly where these people are, and the gravitational pull their mass exerts on the balls, will determine whether the balls go one way or another. This means that, unless a pool player can carefully weight the people around the table, determine where they stand, there's no possible way for anyone to be certain of the trajectory of a ball after six or more collisions.

The issue isn't the influence of randomness, it's the difficulty of computing an accurate result due to the amount of data needed.  Determining the likelihood of war probably takes an amount of data and calculation that is well beyond any computer for the foreseeable future.

[Warspite]

Some interesting initial debates are already happening on the blogosphere about this paper. See for instance:
http://wmbriggs.com/post/16012/

[The Minsky Moment]

Regarding the role of statistical tools, Taleb says, "Physicists get the point. But social scientists do not get it. Their statistical tools do not work for what we call fat tails, as defined as something prone to black swan events," he says.

The assumption here is that statisticians can only work with normal distributions which is false.

Yes and no.  Obviously there are skewed distributions, but you may not have enough data to fit them accurately.  The more skewed the distribution, the more data you need.

There are statistical tools for dealing with non-normal distributions, including fat-tailed distributions.  So Talib is just wrong.  He didn't say normal distributions were more tractable, or easier to deal with, or cet. par. require less data to reach significant conclusions, all which would be justifiable.  He said statistical tools "do not work," no qualification. 

[Admiral Yi]

No Joan.  He said *their* statistical tools don't work.

[The Minsky Moment]

No Joan.  He said *their* statistical tools don't work.

Referring to all "social scientists" !
It's not like physicists have some magic set of statistical tools that no one else knows about or uses.

[The Brain]

I have a massive tool.

[grumbler]

I have a massive stool.

FYP, Ed.

[The Brain]

I wish.