05 September 2021

The purpose of the system is what it does

William Gibson quote-tweeted someone musing about supply chains in a context of "is this going to delay necessary change?" and it makes me want to scream into the void.

You can't control an economy in detail past a certain size.

What you can do is pick your tradeoffs; the long-supply-chain economy we've got is that way because it maximizes profits not through detailed control (which, remember, does not work) but by biasing the system to have small numbers of any particular community of practice, which tends to improve returns on investment in any particular community of practice.  If there's one factory on earth making titanium cookware, investing in the factory is both more reliable (there's demand, this is the whole supply) and more profitable (there's no competition).  This pattern extends to everything.  It relies on cheap shipping, so that shipping resin from the US Gulf Coast to China makes sense as a way to manage your industrial inputs.

Cheap shipping relies on nothing going wrong; if stuff goes wrong, you have to build in margin for it, and margin costs.  Competitive shipping prices -- one of the few areas where real competition remains -- rip out all the margin.  ("Just-in-time scheduling" is a palatable term for "zero-resilience manufacturing"; it's a cost optimization that deliberately choses systemic fragility to raise profits.  This was and is policy, just as much as not paying labour is policy.)

The other thing is that you're substituting a reliable schedule for having a system.  Systems have feedback.  Price signals are not sufficient feedback. (Prices can't reflect events people do not know about.  "Where does the hurricane land?", and so on, aren't predictable.  Resiliency has a cost, but it's a structural cost, you can't price it into a low-resiliency system.)

Necessary change is NOT "infrastructure resilience".  Necessary change is stuff like performing the experiment to answer "how small an economy with zero fossil carbon inputs can provide critical services?" (the stuff like refrigeration and anesthetized dentistry and a comms infrastructure).  It's a completely different world (at a somewhat higher elevation).

We can't get there from here; the only route involves coming down off this local maximum, and the sooner you do it, the less it costs.  The more deliberately you do it, the more survivable it is. (With machine agriculture, failed supply chains can result in a failed food supply just as easily as general agricultural failure due to climate shifts.)

Society is a machine.  The machine we've got is there to make sure the rich get richer.  It is not capable of doing anything else. (As we can see from the comprehensive failure of a pandemic response, including the "I have no obligations to something that doesn't exist" responses about social duty.)  It's absolutely not capable of resilience.

We need something resilient.  We can have it as soon as we decide we're going to have it, instead of this.  It's not an "and" question; it's an instead.


Tim McDermott said...

It surprises me (but shouldn't) that 60 some years after Lorenz discovered chaos, that there are folks who have never heard of self-organizing complexity. Even highly educated folks in highly responsible positions. How is it that the economists chasing a general equilibrium model of the economy have never learned that any system with non-linear feedback has chaotic regions in its state space. There is no equilibrium in real economic systems; I suspect there are only two attractors, hyper inflation and disinflation.

I spent 40 years in software development. In that time there were three paradigm shifts that rendered skills in in the previous approach obsolete: structured, object-oriented, and agile/craftmanship. I know how it feels to be de-skilled by progress. (I don't count O-O as progress; more like snake oil). So I have some sympathy for economists and others who are threatened by complex dynamic systems. But we should all have the grace to suck it up and learn new tricks or get out of the way.

Graydon said...

+Tim McDermott
So much of the study and discipline of economics is about expectations of control; do this, and employment increases. Do that, and interest rates go down.

(These are already extremely narrowly constrained constructions of "an economy"; constraining the system down to whatever narrow thing you can build a control mechanism for is a recognized way to make something controllable, but maybe not the best thing for an economy.)

I suspect that the expectation and requirement of control prevents anyone from getting up and saying "the economy is a chaotic system", just like "national economies are mythological" is not said in those words.

(O-O is progress if you're trying to build tools for unknown (and possibly unknowable) requirements. I don't much think it's general progress myself.)

Although there's a certain grim prospect of widespread starvation because someone didn't ship the whole order to the one factory on earth that makes the thing which makes the thing which makes the thing to inject the bearing grease (or whichever) forcing people to acknowledge the whole "it's only controllable if we can pretend enough, and it doesn't have to let us" part.

James said...

I've recently been refactoring procedural code to OO code with a marked improvement in modularity and maintainability. (No hit to performance. I keep my hands off code that's obviously been optimized down at C level.) So I'll stick up for O-O, though to a true object fanatic I'm probably far too willing to accept a procedural subtext.

Supply chain mechanisms are like serialization mechanisms in critical queueing contexts. If something goes slightly wrong there's a real probability of stuff going badly wrong in unexpected ways. The current disruptions in supply chains, which are mainly a real work version of the Beer Game as supply and demand shift plus having the wrong resources in the wrong place are nothing to what happens if ports have to close and waterways become unreliable for navigation for significant time periods as a result of storms, or for longer periods as a result of erosion or silting. Or just lack of workers (UK right now, but a simmering issue elsewhere).

Graydon said...

I think the "badly wrong in unexpected ways" is amplified by two things; there's one place to get whatever it is, on the one hand, and there's no resilience anywhere because it's been systematically removed, on the other. So even when the alternative source stays up, it hasn't got and probably can't build the capacity to replace lost output. And nobody has stock to keep things running for any appreciable time.

If you can't get the thing in time, something else doesn't ship in time. And it eventually produces an unbridgeable gap. I think that's going to happen much sooner than the rising sea or exceeds-design storms are going to take ports out of service. (Though I think ports are going to start going out of service for the reasons you give.)

Greg said...

Graydon: -

The thing, one of the things, about complex adaptive systems, is that they adapt. Adapt to compensate. And there is more than one layer of adaptation. OK, that's two things.

(*Rues, again, lending Sterman's Business Dynamics to his brother, lo these many years ago, so he can't re-read some things.*)

Adapt right up until they flip into a new regime, of course. But like Jørgen Randers of Limits To Growth: The 30-Year Update fame, I don't expect our global system to get to that point until the second half of this century, if then. Even with the current ripples from having a big transient applied.

So with logistics, there are lots of responses operating on different timescales. Putting up with it. Coercing workers to work more. Increasing shipping prices until arbitrageurs move containers from where they are to where they are needed. Increasing worker supply. Increasing worker productivity, perhaps by teleoperation. Building new drayage - ships, trains, and planes. Re-configuring structures - canals, roads, ports, railways, airports, distribution hubs. Re-orienting households' tastes away from services dependent on logistics. Changing the social status of logistics workers.

And with the "one critical component made in one anonymous factory": we've been here before, with computer hard drives a decade ago, and Ajinomoto build-up film in 2020 as recent examples. There really is a lot of waste on the demand side of the system, and this is one of its saving--er, features. And again, the "right to repair" movement among farmers is one complex-adaptive response to a similar felt constraint.

The "how do we get there from here?" question is interesting; looking at total fertility trends, we'll know the answer by 2200, if not earlier. We'll know we're on the way when there is increasing volatility (which includes having the situation quickly flip between appearing much better and much worse). Sufficiently complex adaptive systems seem to have endogenous simulated annealing.

Graydon said...


There's nothing about "adaptive" -- notionally the property of all organisms; to a first approximation, all organisms are extinct -- that says you are going to get through this.

The Carbon Binge looks like a demonstration that you can't trade habitability for posterity.

We haven't got adaptive systems. We've got a (collective) delusion that adaptation happens because humans have this enormous problem with narrative, and it's so much easier to impute agency than to recognize that we're getting change in a population due to differential death rates. Because we can recognize prospective risks, we get reactive systems, but that's not -- cannot be! -- something that's a systemic scale reaction at the next scale up. So no matter how carefully some large institution tries to preserve itself, it's not acting at species scale. (Or all the oil companies would have done something very different in the 1980s.)

So far, the Arctic Amplification Hypothesis' predictions have held. If those predictions continue to hold, we're looking at a lot of warming by the end of this century. Given that even the IPCC figures agriculture cannot make it past 2 C of warming, I don't think the long time frames where "most people have died" is not part of the prediction make any sense.