January 10, 2020

War is collective, fat-tailed, prolonged; lulls are not de-escalation

From my cursory reading of Twitter political people on both the left and right, there is a widespread misunderstanding about the nature of these phases after a major military or foreign policy event. They do not represent a de-escalation, calming down, standing down, blowing off steam, retreat, or anything of the sort. All the tension is still there, if anything it is building up so that the next major spectacle will be of greater magnitude than those spectacles we have already observed.

The basic confusion comes from analogizing a collective antagonism with an individual one. If two individuals get in each others' faces at a bar, exchange a punch or two, and then walk away after a brief stand-off, then the whole thing may be over. We can breathe a sigh of relief that no more fighting will be taking place between them.

But collective phenomena generally have a positive feedback loop, whereby growth feeds into more growth, since there are multiple individuals impacted by each individual's actions. For example, if instead of an individual vs. an individual, it's a clan vs. a clan in a fight, then we don't conclude anything about "it's all over" once one round of fighting is done with. The side that lost may want revenge.

And in any case, when the other side injures or kills one of your side, you want to deter them by inflicting greater harm on them -- thus, one injury begets multiple injuries, each of which begets multiple injuries, as each slight must be deterred against by an ever greater magnitude of retaliation.

That process is like exponential population growth, where one individual begets multiple offspring, each of whom begets multiple offspring. Or the spread of an epidemic disease, where each infected person transmits the disease to multiple others, each of whom transmits it to multiple others.

The growth of these processes does not require the agents to be constantly engaged in the activity which produces the growth. Exponential population growth does not require the individuals to be constantly getting it on, or constantly popping out babies. When an epidemic disease is spreading, there may be a day or two when no new people are infected, maybe even a week.

We don't measure the growth of a human population on the time-scale of days, weeks, or even months. It takes 9 months from conception to birth, so we'd only start with the scale of years, and perhaps going by decades. Nor do we measure the rise of an epidemic disease on the time-scale of hours or days -- maybe weeks, months, even years, depending on the rate of transmission. On smaller time-scales, these processes would show all kinds of whipsaw, jagged rises and falls -- while on the relevant time-scale, we would observe much smoother patterns.

War is no different (nor is any form of collective conflict, from gangs and clans up to nation-states and empires). Warfare does not require the actors to be constantly invading, bombing, or besieging each other.

There are a variety of ways to measure the intensity of the conflict -- number dead, number injured, amount of property gained or lost, amount of territory gained or lost, and so on and so forth. No matter which one we pick, it will show the same positive feedback loop, based on the logic of retaliation.

These variables -- number of infected people during an epidemic, number dead in a collective conflict, etc. -- are distributed in a fat-tailed way, where most cases don't amount to much, but a few cases go out to the extremes -- and crucially, they fuel their own growth out into the extremes (autocatalytic). Most spats between nations don't lead to any number of dead -- some dork from one country gives another leader the cold shoulder at a summit, and although it's clearly a sign of cold relations between the two nations, it is not part of a war, and no one dies.

But then, there are the occasional cases where the spat is part of a self-fueling growth process, and one event leads to even greater events, perhaps resulting in dozens dead, hundreds, thousands, millions, or tens of millions. It is impossible to predict beforehand how far out into the extreme the process will go -- it takes on a life of its own. The point is, once it starts, it can easily increase by orders of magnitude, because it is mostly driven by its own growth and is not subject to a heavier dampening mechanism.

Even more disturbingly, these distributions have the property that no matter how far out into the extreme you have already gone, you are expected to go out even further. The clearest example is a power-law distribution, whose conditional expectation is given here (under the answer with the green checkmark). It says, given that the variable (X) has already exceeded some particular value (x), its expectation (E(X)) is a positive multiple greater than 1 of that already-exceeded value (where b/(b-1) is the multiplier).

To make up a numerical example, say the variable X is the number of dead in a war, and the particular value, x, that it's already exceeded is 10. Let's say the multiplier (b/(b-1)) works out to 1.5, or a 50% increase. Then we expect the number dead to reach 1.5 * 10 = 15. It's worse than our initial milestone led us to think.

But here's the sick thing -- this multiplier is not dependent on how far out into the extreme you are. Its formula does not include any measure of how far it is from the average. You could be far out into the extreme -- 10,000 already dead -- and still, the multiplier tells you to expect 1.5 times that, or 15,000 dead. If you reach 1 million dead, expect 1.5 million. If you reach 1 billion dead, expect 1.5 billion.

For thin-tailed distributions, like the normal or bell-curve distribution, the further out toward the extreme you reach, the less further you can expect to go. For really extreme values, the expectation is that you go no further at all -- E(X) = x, conditional on having reached a very extreme value x. For example, if you tested into MENSA as a child with a 130 IQ, you can expect to keep that same value all the way through adulthood -- your initial IQ isn't going to fuel its own growth to 150, 180, 210, and beyond. One IQ point does not beget multiple IQ points, each of which begets multiple points, and so on.

Now, just because we "expect" a multiple of the value already reached, for fat-tailed distributions, doesn't mean we'll actually reach that -- it could land short, it could stop altogether, or it could exceed the expectation. For a self-driven growth process, you can't conclude it's over until it's over.

At the least, you need to observe falling values over the relevant time-scale -- if, one week after the next, the number of people infected in an epidemic keeps dropping, that's enough to conclude that it's no longer rising but is burning out. Population growth could hit a saturation level and neither grow nor shrink, or their way of life could be pulled out from under them and their numbers plummet year after year.

(That is more likely to see due to migration into and out of an area, a la the "boomtown to ghost-town" phenomenon, not so much due to changes in birth and death rates. But still, the relevant time-scale is years or decades.)

Other examples are the social adoption of fads, where each adopter can serve as a model to multiple imitators, each of whom can be copied by multiple imitators; or, as Art De Vany showed, the success of movies (and by assumption, or by related research I haven't read, other forms of entertainment), where a hit movie grows by each viewer passing on positive word-of-mouth to other people, each of whom may see the movie and pass on positive word-of-mouth to others.

You cannot predict beforehand how far the growth will go -- will some fad be adopted by only hundreds of thousands in the US, or by millions, tens of millions, etc. Or, will some hit movie hit $1 million in profit, $10 million, $1 billion, $10 billion, etc. The process has to work itself out, and only when you see fewer and fewer people practicing a fad on the time-scale of months or years, or the number of tickets sold falling week after week after week, do you conclude that its growing phase is over.

You certainly do not conclude that a smash-hit movie is "done for" just because its ticket numbers are low measured on a Wednesday, compared to last Saturday when it was doing gangbusters. That's part of the natural rhythm of when viewers go to the movies, making the time-scale of days irrelevant -- it's weeks.

Likewise, only a fool concludes that the Iran War is "over," or "cooling off," or "winding down," or even "hitting a pause" / "standing still". We don't measure such a process on the time-scale of days, so we know nothing about where the next value will be -- it hasn't even been a week since their missile attacks on the US base in Iraq. For an empire like the US facing off against a regional power like Iran, it can be no sooner than the scale of months, perhaps years, depending on how intense the war is.

As welcome as his direct appeals to Trump have been, Tucker Carlson -- the only anti-war voice in the mass media, let alone who have Trump's ear -- has not "talked Trump down off the ledge". He may be limiting the damage done per episode by convincing Trump to minimize his retaliation, but that is not the same thing as halting the growth process. It's going to keep growing, getting worse than it already has gotten.

Even if Tucker had no impact on Trump, the admin would still go through lulls in its overall escalation of conflict against Iran. Tucker is trying to heal only one of those infected by the anti-Iran disease within the US power structure, albeit an influential one -- all those other ones are still infected, and infecting others, including influential ones. That is not going to reverse the growth of an epidemic, which I'm sure he's under no illusions of doing. It requires a full-scale realignment of foreign policy to do that.

What does the trajectory look like in this case? Well, when Trump came into office, tensions were simmering down, owing to the Obama admin's nuclear deal -- although from a different perspective, that was just a temporary lull during a 40-year campaign to weaken them and absorb them into our sphere of influence.

In any case, Trump staffed a bunch of Iran hawks into his admin, which was the first milestone. Then he abrogated the nuclear deal, another major milestone. Then he came this-close to striking Iran last summer when they shot down one of our drones. Now we have reached two major milestones almost simultaneously -- Trump ordering the assassination of their top military leader and national hero, Soleimani, and Iran retaliating by hitting US bases in Iraq with their missiles. Both of those were first-of-their-kind events between the two nations.

On the relevant time-scale, we'll have to wait over the next months, or up to a year, to see if the conflict has begun to simmer down. So far, all we have observed is growth upon growth upon growth -- notwithstanding the lulls in between each of those milestones, which again are like the ticket sales during Monday, Tuesday, and Wednesday, between weekends.

So far, the near-term looks only to escalate, since the Trump admin has imposed further sanctions on Iran, and Iran has said the strikes this week were only the beginning of the campaign to drive the US empire out of the Middle East. Even if nothing happens for the next 6-12 months, that does not mean things have cooled down -- lack of escalation does not mean improvement, it may mean things have reached a saturation level, or it may be a longer-than-usual lull before it explodes all the more violently when the next milestone event is reached.

In either of those cases, all the tension from before is still there, threatening to blow up in the future -- the tension will not have dissipated, until we see major concessions from one or both sides, such as the US pulling more troops out of Iraq, shutting down bases in the Middle East, slashing the funds for such activities, and the like.

In the meantime, don't mistake lulls for the dissipation of tension. This is not an acute beef between two individuals in the heat of the moment -- it is a chronic conflict between large-scale collective actors. And remember the property of the power-law distribution -- for each milestone you have already reached, expect it to go some degree further, and that multiplier will not diminish no matter how extreme the milestone may be.

No comments:

Post a Comment

You MUST enter a nickname with the "Name/URL" option if you're not signed in. We can't follow who is saying what if everyone is "Anonymous."