I did a quick ctrl-f through the first few chapters of the book, and Norbert Wiener is mentioned just a couple times, and only in the context of his original work. I think that's a serious omission, given that Wiener wasn't just the coiner of the terms "cybernetics" and "control systems", but spent a huge amount of effort relating that to psychology and psychological issues.
From ChatGPT's summary: "Wiener's Cybernetics: Or Control and Communication in the Animal and the Machine (1948) extends feedback analysis to the nervous system and to psychiatric disorders (e.g., schizophrenia as a breakdown of feedback)."
Ignoring this part of Wiener's work means that SMTM also ignored all the work that followed after it, like Karl Friston's work (zero mentions in SMTM as far as I can tell), which spends a lot of time discussing happiness in the context of control systems. Again, from ChatGPT: " In “What is Mood? A Computational Perspective” (Clark, Watson & Friston 2018) , mood = a hyper-prior over expected precision; momentary happiness corresponds to downward shifts in expected surprise."
I'd really want an explanation from SMTM why they omitted this, because, at the very least, this is a serious failure to engage with the literature.
I have no training in psychology, but here are some things I would expect psychology to be able to discover if it is a useful paradigm:
1. What governors exist, like you said
2. What classifications of governors are there, and how can you tell them apart?
3. What influences set points?
4. Can a new governor be constructed? What would be required for me to build a 'buy healthy food' control system, and what would it's characteristics be?
5. Can a governor be destroyed?
6. How isolated are governors? Do governors leak error signals between each other? Can I need to pee so badly that I get lonely, or eat a sandwich so good I forget to breathe?
this was fantastic — just finished a degree in systems design engineering & the connection between control systems theory and “psychology troubleshooting“ feels intuitive but seeing it laid out so well connected a few new pieces in my brain. Thanks for pulling it all together!!
Hmm, a very interesting idea. Thinking of it like a thermostat is what much of the medical and psych industry already does. In other words cause -> effect. This misses the extreme level of complexity that is part of the human system. For example to borrow the thermostat analogy. It is too hot. Let's point a fan at the thermostat when in reality it was installed in by a huge window and sits in the sun all day. What is the correct "fix"? Oh, you're depressed, here take a SSRIs or some other option. What is really happening is that effect is being addressed (or the error in the thermostat model) but not the cause.
Really interesting how this builds off the one idea I know from Slime Mold Time Mold - a cohesive theory as to why people have a hard time losing weight. They argue that there's a control system in your body that is basically a steady-state weight (again, like a thermostat) and that environmental factors may raise that 'temperature' from what would otherwise be our default settings per our individual genetics.
So basically you are suggesting we look at the nervous system as being composed of interacting control systems, similarly to how other body systems maintain homeostasis. This makes sense to me, but it also doesn’t seem new: a cybernetic model of psychology, as you call it, is otherwise known as a functionalist theory of the mind. If this is a new paradigm, how do you distinguish it from the psycho-functionalist project associated with people like Jerry Fodor? One thing that might be different is the idea that there’s just be one overarching teleological target controlling all cognitive functions, namely health in the broadest sense of the term. I think that’d a promising way to think about things, and likely to deliver much explanatory power, but I’m not sure if that’s exactly what you’re proposing.
This is an amazing article, and one that naturally fits in with my own understanding of my brain, so i believe in it's accuracy! For those with utilitarian inclinations who say the line "happiness is not an emotion" and worried that utility might be completely thrown out the window, I've actually written an article about exactly how to handle this: https://www.lesswrong.com/posts/4PovnstgKAELqAtbn/an-ethical-framework-to-supersede-utilitarianism
The TLDR is that utility shouldn't be about maximizing happiness, but instead minimizing the errors in these control systems.
I feel personally assaulted by this post because when you asked what a personality psychologist would say a "trait" was, I gave the *exact damn answer you did.* Screw you, Adam, you can't be an expert personologist as well, you're already good at your own things!
But I do want to observe a few weak points in this piece. First, the notion that behaviorism is “dead” and lies “at the bottom of the ocean” doesn’t hold up. It's more like Skinner wanted us to believe Behaviorism was the whole body, and we've learned it's just an arm. That arm has sprouted itself a hand (e.g., social learning theory) and now there's other arms (e.g., cognitive psychology) and while we may continue searching for the rest of the body (I don't know how this became a weird forensic reference) it's not like we're still not using the arm. Behaviorism still gets used in training animals (and people, if you count ABA and CBT!).
Specifically for traits, the cybernetic framing just isn't so simple. Traits have to be recognized as emergent properties of genetic predispositions, reinforcement histories, cognitive schemas, and cultural narratives... they're multidetermined. If we're going to use the Big Five as an example, I can understand your explanation of extraversion as a social stimulation error correction; I can make one up for agreeableness (conflict/affiliation error sensitivity) and neuroticism (threat/danger error amplification) but I begin to hit trouble with things like conscientiousness (low thresholds for disorder?) and openness to experience (novelty-seeking governor?) and I think we both know that the Big Five are a necessarily limited list of traits as it is. It actually reminds me a bit of Walter Mischel's (1995; love him or hate him) IF-THEN contingencies, in which he believed each person's personality was simply their unique pattern of IF-THEN reactions. But I think cybernetics is one step better, if only because there will be fewer governors than there will be IF-THEN contingencies (in fact, each governor might be responsible for a number of particular IF-THENs, if we wanted to go that route, which I actually don't).
I almost think I'd want to take the paradigm a step further and say that (at least for personality, which is what I'm best qualified to address) the Free Energy Principle (FEP) is a better modeling paradigm than cybernetics. If cybernetics is like a thermostat, FEP is more like a weather forecasting service, and people are often not passive in that cybernetic-homeostatic way. Rather, we're constantly trying to predict future states based on current data and past patterns, minimizing prediction error (surprises!) by trying constantly to bring the world in line with our predictions.
So while cybernetics might explain why we feel uncomfortable when needs aren't met (hunger, social isolation, pain), FEP might explain why some people seek novelty, tolerate ambiguity, or rigidly pursue goals even when environments shift (e.g., conscientiousness might = tightly maintained long-range forecasts; openness might = flexible, exploratory model updates; neuroticism might = hypersensitive to threat predictions).
Regardless, here's the compliment underneath all my blah-blah: this article definitely got me thinking about psychology (and personality specifically) in an interesting way -- and I love the humility of knowing that "it might be very wrong" but as Carol Dweck always says, it only makes us "smack our lips [why, Carol?] and cry out 'I love a challenge!'"
Interesting article. I found this sentence particularly ironic "That’s why our best theory of personality performs about as well as the Enneagram, a theory that somebody just made up."
It is clearly a dig at the Enneagram, "a theory that somebody just made up". The first question is would that not be true of any and every theory? So why the put down? The irony is that the enneagram theory has these set points you describe. There are nine types whose dynamics can be understood as the result of the interplay between triads. Triads like which emotion is dominant, anger, fear, pride/shame, or which instinct is dominant, self-preservation, sexual and social, etc., etc. Sure, the enneagram has no academic imprimatur, but for instance Nancy McWilliams book on psychoanalytic types comes up with 9 types that are very good matches for the enneatypes. And we don't really know who came up with the theory, the earliest on record probably was inspired by a Sufi mystic tradition. The difficulty with the enneagram is precisely that the tests are not very good, which is unsurprising since a lot of a personality stems from unconscious inner conflicts. Might not be a bad idea to dig into the enneagram to consider the different set points identified in that theory.
Yeah, no, the enneagram is "something that someone just made up" in the sense that a) it has no theory behind it, not even that of "impressionistic research" (e.g., the Big Five are based on the lexical theory: that any word to describe a trait of personality must be encoded already in our language) and b) it offers no additional predictive utlity over other, more researched measures of personality and behavior (you can use other measures and get better results). And while we're at it, c) it's a typology, and typologies are terrible -- even clinical psychology is finally moving away from "you are either narcissistic or you're not" to "narcissism is a continuum anyone can be scaled on."
Things based on "mystic traditions" generally aren't good scientific concepts, though I'm sure there are cases where this isn't true.
I hope to live to see the day of "Inside Out 3" where the Make Sure You Spend Time with Other People System is broken and needs fixing.
Lmao and one of the new added emotions is actually just “ChatGPT Therapist”
I did a quick ctrl-f through the first few chapters of the book, and Norbert Wiener is mentioned just a couple times, and only in the context of his original work. I think that's a serious omission, given that Wiener wasn't just the coiner of the terms "cybernetics" and "control systems", but spent a huge amount of effort relating that to psychology and psychological issues.
From ChatGPT's summary: "Wiener's Cybernetics: Or Control and Communication in the Animal and the Machine (1948) extends feedback analysis to the nervous system and to psychiatric disorders (e.g., schizophrenia as a breakdown of feedback)."
Ignoring this part of Wiener's work means that SMTM also ignored all the work that followed after it, like Karl Friston's work (zero mentions in SMTM as far as I can tell), which spends a lot of time discussing happiness in the context of control systems. Again, from ChatGPT: " In “What is Mood? A Computational Perspective” (Clark, Watson & Friston 2018) , mood = a hyper-prior over expected precision; momentary happiness corresponds to downward shifts in expected surprise."
I'd really want an explanation from SMTM why they omitted this, because, at the very least, this is a serious failure to engage with the literature.
Does it talk about or build on the already existing Cybernetic Theory of Psychopathology by Colin G. DeYoung & Robert F. Krueger?
https://www.tandfonline.com/doi/abs/10.1080/1047840X.2018.1513680
https://psycnet.apa.org/record/2023-67353-002
Also Cybernetic Big Five Theory
https://www.sciencedirect.com/science/article/abs/pii/S0092656614000713
Lisa Feldman Barrets work on emotions and allostasis also relevant.
I for one, welcome our cybernetic overlords
I have no training in psychology, but here are some things I would expect psychology to be able to discover if it is a useful paradigm:
1. What governors exist, like you said
2. What classifications of governors are there, and how can you tell them apart?
3. What influences set points?
4. Can a new governor be constructed? What would be required for me to build a 'buy healthy food' control system, and what would it's characteristics be?
5. Can a governor be destroyed?
6. How isolated are governors? Do governors leak error signals between each other? Can I need to pee so badly that I get lonely, or eat a sandwich so good I forget to breathe?
this was fantastic — just finished a degree in systems design engineering & the connection between control systems theory and “psychology troubleshooting“ feels intuitive but seeing it laid out so well connected a few new pieces in my brain. Thanks for pulling it all together!!
Hmm, a very interesting idea. Thinking of it like a thermostat is what much of the medical and psych industry already does. In other words cause -> effect. This misses the extreme level of complexity that is part of the human system. For example to borrow the thermostat analogy. It is too hot. Let's point a fan at the thermostat when in reality it was installed in by a huge window and sits in the sun all day. What is the correct "fix"? Oh, you're depressed, here take a SSRIs or some other option. What is really happening is that effect is being addressed (or the error in the thermostat model) but not the cause.
Really interesting how this builds off the one idea I know from Slime Mold Time Mold - a cohesive theory as to why people have a hard time losing weight. They argue that there's a control system in your body that is basically a steady-state weight (again, like a thermostat) and that environmental factors may raise that 'temperature' from what would otherwise be our default settings per our individual genetics.
So basically you are suggesting we look at the nervous system as being composed of interacting control systems, similarly to how other body systems maintain homeostasis. This makes sense to me, but it also doesn’t seem new: a cybernetic model of psychology, as you call it, is otherwise known as a functionalist theory of the mind. If this is a new paradigm, how do you distinguish it from the psycho-functionalist project associated with people like Jerry Fodor? One thing that might be different is the idea that there’s just be one overarching teleological target controlling all cognitive functions, namely health in the broadest sense of the term. I think that’d a promising way to think about things, and likely to deliver much explanatory power, but I’m not sure if that’s exactly what you’re proposing.
This is an amazing article, and one that naturally fits in with my own understanding of my brain, so i believe in it's accuracy! For those with utilitarian inclinations who say the line "happiness is not an emotion" and worried that utility might be completely thrown out the window, I've actually written an article about exactly how to handle this: https://www.lesswrong.com/posts/4PovnstgKAELqAtbn/an-ethical-framework-to-supersede-utilitarianism
The TLDR is that utility shouldn't be about maximizing happiness, but instead minimizing the errors in these control systems.
I feel personally assaulted by this post because when you asked what a personality psychologist would say a "trait" was, I gave the *exact damn answer you did.* Screw you, Adam, you can't be an expert personologist as well, you're already good at your own things!
But I do want to observe a few weak points in this piece. First, the notion that behaviorism is “dead” and lies “at the bottom of the ocean” doesn’t hold up. It's more like Skinner wanted us to believe Behaviorism was the whole body, and we've learned it's just an arm. That arm has sprouted itself a hand (e.g., social learning theory) and now there's other arms (e.g., cognitive psychology) and while we may continue searching for the rest of the body (I don't know how this became a weird forensic reference) it's not like we're still not using the arm. Behaviorism still gets used in training animals (and people, if you count ABA and CBT!).
Specifically for traits, the cybernetic framing just isn't so simple. Traits have to be recognized as emergent properties of genetic predispositions, reinforcement histories, cognitive schemas, and cultural narratives... they're multidetermined. If we're going to use the Big Five as an example, I can understand your explanation of extraversion as a social stimulation error correction; I can make one up for agreeableness (conflict/affiliation error sensitivity) and neuroticism (threat/danger error amplification) but I begin to hit trouble with things like conscientiousness (low thresholds for disorder?) and openness to experience (novelty-seeking governor?) and I think we both know that the Big Five are a necessarily limited list of traits as it is. It actually reminds me a bit of Walter Mischel's (1995; love him or hate him) IF-THEN contingencies, in which he believed each person's personality was simply their unique pattern of IF-THEN reactions. But I think cybernetics is one step better, if only because there will be fewer governors than there will be IF-THEN contingencies (in fact, each governor might be responsible for a number of particular IF-THENs, if we wanted to go that route, which I actually don't).
I almost think I'd want to take the paradigm a step further and say that (at least for personality, which is what I'm best qualified to address) the Free Energy Principle (FEP) is a better modeling paradigm than cybernetics. If cybernetics is like a thermostat, FEP is more like a weather forecasting service, and people are often not passive in that cybernetic-homeostatic way. Rather, we're constantly trying to predict future states based on current data and past patterns, minimizing prediction error (surprises!) by trying constantly to bring the world in line with our predictions.
So while cybernetics might explain why we feel uncomfortable when needs aren't met (hunger, social isolation, pain), FEP might explain why some people seek novelty, tolerate ambiguity, or rigidly pursue goals even when environments shift (e.g., conscientiousness might = tightly maintained long-range forecasts; openness might = flexible, exploratory model updates; neuroticism might = hypersensitive to threat predictions).
Regardless, here's the compliment underneath all my blah-blah: this article definitely got me thinking about psychology (and personality specifically) in an interesting way -- and I love the humility of knowing that "it might be very wrong" but as Carol Dweck always says, it only makes us "smack our lips [why, Carol?] and cry out 'I love a challenge!'"
The def. of happiness there is satiation, satisfaction. Okay! For higher forms, may consult the Thomists and Aristotelians.
Are you trolling us with the "UCLA Monocle Lab" joke?
great article - feels like the closest thing i’ve seen to a first-principles approach in psych.
dear adam,
fascinating piece as always!
i like this a lot: "The point of a paradigm is to be wrong in the right direction."
thanks for sharing!
love
myq
Interesting article. I found this sentence particularly ironic "That’s why our best theory of personality performs about as well as the Enneagram, a theory that somebody just made up."
It is clearly a dig at the Enneagram, "a theory that somebody just made up". The first question is would that not be true of any and every theory? So why the put down? The irony is that the enneagram theory has these set points you describe. There are nine types whose dynamics can be understood as the result of the interplay between triads. Triads like which emotion is dominant, anger, fear, pride/shame, or which instinct is dominant, self-preservation, sexual and social, etc., etc. Sure, the enneagram has no academic imprimatur, but for instance Nancy McWilliams book on psychoanalytic types comes up with 9 types that are very good matches for the enneatypes. And we don't really know who came up with the theory, the earliest on record probably was inspired by a Sufi mystic tradition. The difficulty with the enneagram is precisely that the tests are not very good, which is unsurprising since a lot of a personality stems from unconscious inner conflicts. Might not be a bad idea to dig into the enneagram to consider the different set points identified in that theory.
Yeah, no, the enneagram is "something that someone just made up" in the sense that a) it has no theory behind it, not even that of "impressionistic research" (e.g., the Big Five are based on the lexical theory: that any word to describe a trait of personality must be encoded already in our language) and b) it offers no additional predictive utlity over other, more researched measures of personality and behavior (you can use other measures and get better results). And while we're at it, c) it's a typology, and typologies are terrible -- even clinical psychology is finally moving away from "you are either narcissistic or you're not" to "narcissism is a continuum anyone can be scaled on."
Things based on "mystic traditions" generally aren't good scientific concepts, though I'm sure there are cases where this isn't true.