Okay, this took a bit for me to understand but it was quite cool. I will give an analogy using the perspective of a country as an economic agent as that is how I conceptualise this. (hopefully I can get it confirmed that this is indeed how it works)
It’s basically about a loss of agency by belief in a story of externalised control. You have two things in active inference, world model minimization by making your model closer to reality and action based minimization, essentially taking actions that make the environment more predictable. This second is where you describe things like agency.
What this essay is kind of saying is that these countries are collapsing what should be a two-channel optimization — perception and action — down to perception alone, losing optionality over time.
This is locally coherent as it makes things “grow” over time and the numbers the economists care about go up so it’s all good. But there’s a sneaky backdoor!
The country’s ability to take action atrophies as the diverse local economy that would generate new capabilities gets crowded out by commodity dependence. Yet the people who are supposed to be responsible for the country’s executive function also starts to believe that “Oh, I don’t see any problems, I should continue to sit in this dark room where there are no shocks” and so there is no prediction error and no model updates. (where the dark room is the view that GDP is going up and we all good.)
This is also a problem in the modelling of it itself since there’s money to be gained in believing in this story as an economist.
Badabing badaboom, Taleb’s black swan rolls around and you’re dead since you have little agency.
I don’t exactly think it’s collapsing to perception-only. The state is still acting, and still perceiving, but the dimensionality has been compressed. Often to two or three dimensions, GDP and inflation and unemployment (or in countries with a currency peg, replace inflation with bond rates).
Try designing a self driving car with a three pixel camera! A country is even more complex than a car—it contains many cars, for instance.
Meanwhile the actual people in the country are still doing high-dimensional things like raising children (a tremendous amount of perception, cognition, and action is needed even to raise kids somewhat badly), raising crops, finding mates, running logistics and delivery for groceries… but none of that registers in the system’s measurements, so people end up having to either try to evade the system’s unhelpful notice or treat it as a sucker to be fooled & extracted from, just to live good lifecycles or accomplish other goals like scientific research or tech development.
I think the portrayal of Augustus in I, Claudius might be a more illustrative example than a depressive in a dark room. He’s the apex of a system of narcissistic supply: everyone’s committed to treating him like he’s in charge when he’s in the room, but they don’t think he can be effectively informed or reasoned with, so they fight each other by trying to control his access to information so he makes the decisions they need him to make & thinks they’re all his ideas.
The measurement problem here is better covered in the linked piece on Solow convergence, and in Statisticism, which criticizes the tendency to fixate on low-dimensional indicators that aggregate lots of details as evidence to find out how things are going, like reading the newspaper and counting negative and positive stories.
Okay, this took a bit for me to understand but it was quite cool. I will give an analogy using the perspective of a country as an economic agent as that is how I conceptualise this. (hopefully I can get it confirmed that this is indeed how it works)
It’s basically about a loss of agency by belief in a story of externalised control. You have two things in active inference, world model minimization by making your model closer to reality and action based minimization, essentially taking actions that make the environment more predictable. This second is where you describe things like agency.
What this essay is kind of saying is that these countries are collapsing what should be a two-channel optimization — perception and action — down to perception alone, losing optionality over time.
This is locally coherent as it makes things “grow” over time and the numbers the economists care about go up so it’s all good. But there’s a sneaky backdoor!
The country’s ability to take action atrophies as the diverse local economy that would generate new capabilities gets crowded out by commodity dependence. Yet the people who are supposed to be responsible for the country’s executive function also starts to believe that “Oh, I don’t see any problems, I should continue to sit in this dark room where there are no shocks” and so there is no prediction error and no model updates. (where the dark room is the view that GDP is going up and we all good.)
This is also a problem in the modelling of it itself since there’s money to be gained in believing in this story as an economist.
Badabing badaboom, Taleb’s black swan rolls around and you’re dead since you have little agency.
I very much appreciate the active listening.
I don’t exactly think it’s collapsing to perception-only. The state is still acting, and still perceiving, but the dimensionality has been compressed. Often to two or three dimensions, GDP and inflation and unemployment (or in countries with a currency peg, replace inflation with bond rates).
Try designing a self driving car with a three pixel camera! A country is even more complex than a car—it contains many cars, for instance.
Meanwhile the actual people in the country are still doing high-dimensional things like raising children (a tremendous amount of perception, cognition, and action is needed even to raise kids somewhat badly), raising crops, finding mates, running logistics and delivery for groceries… but none of that registers in the system’s measurements, so people end up having to either try to evade the system’s unhelpful notice or treat it as a sucker to be fooled & extracted from, just to live good lifecycles or accomplish other goals like scientific research or tech development.
I think the portrayal of Augustus in I, Claudius might be a more illustrative example than a depressive in a dark room. He’s the apex of a system of narcissistic supply: everyone’s committed to treating him like he’s in charge when he’s in the room, but they don’t think he can be effectively informed or reasoned with, so they fight each other by trying to control his access to information so he makes the decisions they need him to make & thinks they’re all his ideas.
The measurement problem here is better covered in the linked piece on Solow convergence, and in Statisticism, which criticizes the tendency to fixate on low-dimensional indicators that aggregate lots of details as evidence to find out how things are going, like reading the newspaper and counting negative and positive stories.