Interesting, good new thoughts and questions... (I just finished Pollan's "How to Change Your Mind" last weekend, so it's on my mind. So to speak.) I was wondering about the H1A/H2A receptors as well, vis-a-vis SSRI/SNRIs, not being a trained anything except musician, and my question boils down to this: as I understand it, the inhibitor drug's purpose is to inhibit re-uptake, to free up serotonin or norepinephrin so that ostensibly you have more to go around, so you ...what? Feel or think more easily? But they are functioning by inhibiting the receptors, which may be receivers for other things as well, right? So what might a person be missing by inhibiting the receptors?
I had a recent experience in meditation that made me think of prediction error: i felt like my body was both 'really big' and 'really small' at the same time. I eventually interpreted this as being something like 'the error bars on my body have blown up'. I figured my brain must be computing 'where my body is in space', that this computation must have some accompanied 'error' signal with it, and (maybe from the act of relaxing while staying alert?) my brain eventually said 'huh, i guess this could maybe be anywhere'
I've noticed, too, when my kids start fussing aggressively, if i'm really mindful, i can notice some feeling like "this loud noise should not be happening now" - which looks like prediction error, and thus suffering.
As for why money feels good (even though it's a big surprise), i think this is probably because it changes a bunch of other predictions about future needs being met. And, anecdotally, i've found that i sometimes have a negative emotional response when i see bitcoin going up dramatically in one day, even though i like bitcoin and hold a position in it. I think this is because it isn't really updating predictions about my future needs being met (i expect these to be met regardless), and instead is just a general 'some unexpected thing happened' type of prediction.
> one remaining problem here is why and how some prediction errors get interpreted as rewards
This happened to me a few years ago and it wasn't a pleasant experience. My career took a very sharp, but positive turn and it made me realize that my priors on myself were really off. I ended up in a position that I was shooting for but got there about a decade earlier than I thought I would. It made me realize how low my self-esteem had been, how bad at relationships I was (long story how this fits), and a few other matters.
I have one other experience of a prediction error turning positive in investing. It made me realize that I had just gotten lucky, that I didn't really know what I was doing, and that I should just stick to index funds (though perhaps index funds are no longer the best long-term vehicle if certain economists are to be believed).
I have more experience with mushrooms than most people who have used mushrooms, and only one experience with LSD, and found the two experiences to be so drastically different that I would have a hard time accepting generalizations about how psychedelics as a class should be used to attack problem-solving.
N=1, but mushrooms have been great for dealing with internal issues. I don't think I would be able to solve any technical issues on them (analytical thinking is challenging on mushrooms over 2ish grams), but they are fantastic for chasing threads of emotions and experiences, and for interacting with sources of meaning. LSD felt much more cerebral, euphoric, and neurotic.
The feeling of thinking on both of them is also quite different. With your eyes closed, thinking on mushrooms generally feels like it occurs in the space of "consciousness", like in meditation. Thoughts, sounds, patterns, physical sensations merge into a kind of chaotic space. Thinking on LSD (only one experience here, but decent dose) is pretty much just thinking - you're in your head like you are during your day-to-day, but with pretty patterns and a greater willingness to explore new ideas.
I imagine we do research into the impact of psychedelics a disservice when discussing "psychedelics" vs. the specific substances.
>First of all - predictive coding identifies suffering with prediction error. This conflicts with common sense.
This is something I really don't get. The predictive coding model seems super interesting and have a lot of explanatory power but it seems weird to suppose that it is a global model that explain everything brain related. Why identify suffering with prediction error, instead of the more common sens idea that prediction error contributes/is a type of suffering?
It is often said in the fire poi/"flow arts" community that practicing on acid - and other psychedelics, but mostly acid - is god's own cheat code. This matches up very nicely with George's theory about the proper use of psychedelics.
The only pedantic completion I'd have to this is that I don't think all "introspection" done on psychedelics results in weirdness.
For example, you could introspect vision and figure out: When I see a repeating pattern I don't actually see the "true" number of objects, instead, my mind kind of infers it as a repeating pattern and translates it into however many objects it wants, so I end up perceiving the number of stairs based on my rate of exhaustion, or more/fewer posts in a fence depending on how hard I focus on it
But if you introspect on concepts where there's no "reasonable" explanation to be found or no particular insight left to gain.
For example death, or meaning or consciousness itself, or "true love" or whatever. If you introspect specifically because those concepts are bothering you, then it gets weird. This is not to say doing so is always bad, I used to be afraid of death (like, existential dread lasting half an hour every few days or weeks for 10+ years afraid) and after 500mcg and 14 hours that basically stopped... maybe I got stuck with an unadaptive mental model as a tradeoff, but then I'm fine with the tradeoff from a negative utilitarian perspective. I know many other people that walked through these kinds of issues on psychedelics. But there's certainly a point where it gets weird.
In other words, I think there are facets of the mind that it can introspect with a small chance of it going bonkers (e.g. sight, sound), and facets where the risk is high (especially given repeated trips, especially given high dosages and or predispositions for certain types of psychosis).
As an aside, I find it interesting that quitting alcohol abuse has a high success rate by either joining a strongly religious community (e.g. AA) or taking psychedelics, which would fit the idea of "build a weird mental model to allow your frontal cortex to intervene in a very well-formed set of patterns". But then again, those are small n studies, and psychedelics seem to work for everything in small n studies in ways that really contradict the behaviours of users (e.g. 80% efficacy in terms of smoking cessation)
I think that psychiatry as a discipline sometimes overemphasises the role of specific receptor subtypes and underemphasises the role of where in the highly specialised structure of the brain those receptors are, in terms of the type of cognition they are involved in.
In terms of reward and reward prediction error, I also feel that there is a wealth of research on dopamine and the dopaminergic system being more involved in this.
If we recall at all of the brains dopamine is produced in the substantia infra and ventral tegmental area, and then projects into the limbic system, via the mesolimbic pathway and into the rest of the cortex, frontal cortex first, via the mesocortical pathway, then we can consider the effects of dopamine on reward.
There is a wealth of data on dopaminergic stimulation of the nucleus accumbens (limbic) as being involved in reward and reinforcement learning. There is also evidence of the orbitofrontal cortex connections to the limbic system as being critically involved in reward, but also reward anticipation, and reward prediction errors. Rolls writes extensively on this for example. De Youngs studies on brain volume and personality find that OFC volume and extroversion are positively correlated, which he interprets as fundamentally reward seeking.
I don’t know the exact distribution of DA vs 5HT receptors in these regions (and I’m not sure who does), but common sense understanding of the emotional impact of serotonergics vs dopaminergics would be that dopaminergics produce a euphoria (reward) as well as highly motivating people (reward expectation) to move, socialise, or even do otherwise unrewarding tasks such as cleaning ones kitchen or revising for a exam when one has ADHD.
I’d also speculate that dopamine is actually the key neurotransmitter signalling pleasant surprise (a positive reward prediction error) and unpleasant shock (a negative reward prediction error). None of these emotions are produced by people taking serotinergics such as SSRIs.
In contrast, I think that both SSRIs primarily stimulating 5HT1A and psychedelics primarily stimulating 5HT2A are actually involved more in *changing* the weighting of neuronal connections which we’re making those prediction errors, through plasticity which psychedelics in particular have been shown to increase.
Studies on gambling back this idea up too. In rat brains, an unexpected reward is accompanied by a DA release whilst a loss was associated with a 5HT release. The interpretation here was that DA affected plasticity by making the Britons that fire together, wire together making the animal feel like “I really want to do that movement or behaviour again!” and thus explaining DAs role in habit formation and addiction, whereas 5HT is involved in increasing plasticity by reducing signal weights, when a prediction error is detected. Thus, the serotonergic system is involved in the recognition at some level that “this prediction did not work, I need to change my behaviour and do something different” which is massively enhanced if a person is taking SSRIs or psychedelics.
Therefore, I’d be wary of attribution of cause to 5HT for prediction error, but more if an effect of prediction error in producing changes in signal weighting via plasticity to change behaviour.
If you’ve ever used CBT to treat depression or anxiety you know how people make negative prediction errors, and how helping people to recognise this and change their behaviour is the key to treatment, and how an SSRI can really assist this process. SSRIs don’t produce reward (euphoria), reward prediction (motivation) or reward prediction error (surprise/shock) which I think are all dopaminergic. SSRIs help people change the reward prediction errors they are making that terrible things will happen, and thus help them to change their thinking and behaviour
-If you get $1 million because you're an ordinary middle-class person and a crypto billionaire semi-randomly decides to give you $1 million one day, you will be very happy.
Do we actually know being "very happy" would be the typical instant reaction? I once won $10,000, about 25% of my income at the time, in some promotional lottery I hadn't realized I had been entered in. I was going through a miserable time in life and my reaction was to be angry and annoyed, because it felt like a cruel joke. Only after a few weeks, when I used some of the money to buy a new guitar, did I feel happy to have it.
Maybe that was like stepping in a cold shower and suffering at first but then getting used to it? Or maybe because I was depressed, I didn't have a normal reaction?
I think it mattered that I didn't even know I was entered in the lottery. If someone buys a lottery ticket, perhaps a part of them predicts they will win. Otherwise, why did they buy the ticket?
Fits with Buddhist ideas; you don't suffer because you are in pain, you suffer because you don't think of pain as something that should be happening to you.
Surely what makes a cold shower decrease in pain over time is not an adjustment in your predictions, but a physical adjustment in the contrast in skin temperatures.
> one remaining problem here is why and how some prediction errors get interpreted as rewards
Objectively unlikely events could easily confirm our model of the world rather than violating it, if the model was suitably wrong to begin with.
Plausibly, our money-centric culture—and in particular our parasocial relationships with affluent pseudo-peers—causes our brains to maintain a prediction amounting to "I am wealthy, or soon will be, and any indication to the contrary is some terrible mistake." The conflict of this prediction with reality creates suffering, but we get stuck unable to update away from it, because our media habits keep giving us more fake evidence that it's true (which is maybe not really our fault since whole industries are dedicated to ensuring that very thing). A sudden windfall, then, won't feel surprising; it will feel like the natural order restored.
Now I'm no historian, but I'd guess that if you're a premodern farmer and you have an exceptional crop, your model already represented that as something that could and indeed *ought to* happen; although it was surprising in some statistical sense, you aren't surprised—the world feels more correct rather than less. But if the king rides up and hands you a fistful of glittering gold, aren't you more likely to be suspicious and scared than happy? Surely this is some mistake or trick or trap; surely the other villagers will turn on you, or bandits will kill you for it, or something—anything—will punish you, because suddenly you're outside the natural order.
> At its limit, this theory says that all action takes place through the creation and resolution of prediction errors - I stand up by "predicting" on a neurological level that I will stand up, and then my motor cortex tries to resolve the "error" by making me actually stand.
Thinking out loud: there are several programming languages whose core primitive operation is some version of this. Prolog being a conventional example.
In Prolog and other languages incorporating such features, 'programming' takes the form of, first, writing 'facts', and then second, writing underspecified facts. The program executes by "resolving the error" by figuring out which values need to be what to make your underspecified facts true. This is used a lot for logic programming (it's in the name, after all).
If you described this programming language as "resolving pattern-match error" to execute, that would be technically correct. Perhaps "prediction error" functions similarly in cognition. Perhaps, at some level what is technically happening is prediction error being resolved, but in practice that is an obtuse way of framing it
Off topic: you comment on 2 subjects I find interesting. 1. how do our brains choose goals - crude answer we feel tension when how we want the world to be is different from how it is; 2. what do drugs do to our brains. It would be awesome if you wrote a piece on how the 2 overlap.
So we have: A. how the world is; B. how I perceive the world; C. what I want the world to be; D. how much the disparity upsets me, E what I am doing about it. We can acheive our goals by moving the world closer to how we want it to be, looking at things in a different way, by changing what we want, by stopping caring about the difference. What is the attraction of different drugs? Which of these 4 ways of reducing that tension do they address? e.g. heroin - at a guess B and D; cocaine at a guess B and E. But I know nothing in these fields, what does the attraction of different drugs tell us about the model of motivation and consciousness?
> one remaining problem here is why and how some prediction errors get interpreted as rewards... This suggests there's still something we don't understand about prediction error and suffering
I think if you were describing a theory you were less attracted to, then instead of saying that this is one remaining thing we don't understand, then you'd say that this is a class of counterexamples that totally blows up the theory.
You've got three examples of the "prediction errors -> suffering", of which one fits (losing money), one doesn't fit (gaining money) and one only fits if you torture the definitions of words until they submit (feeling pain). Most of the other examples of prediction errors that I can think of don't seem to cause suffering; e.g. when watching a movie it's not at all unpleasant when the camera cuts to another angle.
I think there's a lot of good in predictive processing, but this particular "prediction errors -> suffering" aspect seems a bit of a dead end to me.
I find the idea that we shouldn't use psychedelics for introspection a bit troubling: I've been studying up on psychocybin based psychotherapy lately, and it seems to get very promising results in treating chronic depression and anxiety. Yet the entire model of therapy is based on internal introspection: you give people an extremely high dosage of psychocybin, and then you put a blindfold on them and play gentle music so they aren't distracted by the outside world and can focus on internal introspection. Could this be a dangerous therapy model to pursue? The studies done Dr. Roland Griffiths seem to show good outcomes, but I'm not sure how well you can measure "weirdness" after treatment.
> the 10% of Americans who use psychedelics mostly don't end out as weird as they did
Huh. All the people I know who are really into psychedelics also have very eclectic beliefs about religion or humanity or whatever. Maybe you're right about dosing (note the "really into", not "10% of the people I know"). Or maybe we've all learned to be ok with some types of cognitive weirdness in a way that it's acceptable to be a yogi barista inventing your own religion and nobody really makes a big deal out of it anymore.
2c on Active Inference Framework and reward/suffering: (context: I have developed an AIF-based model that's being published in Entropy soon, so I can speak to the math/theory side of it - less so to the psych/neurology angle)
Friston is generally cavalier with his writing and this seems to me like an example of that. If you're abstractly just interpreting animals as simple self-regulating systems (trying to stand up, or stay fed, etc) whose only ambition is to maintain homeostasis, the prediction error / suffering metaphor works great, in the following sense: the reward function is endogenous, so any behavior is a self-fulfilling prophecy. But if you're considering a sophisticated agent that's composed of multiple subsystems, capable of foresight/delaying reward, attaching value to symbols, etc, then when isolating an individual loop/subsystem, you must accept that in general the reward function will be exogenous (provided by whatever the relevant desire-generating subsystem is), and the discrepancy between desired and observed/predicted future can be arbitrarily large. This reduces the philosophical issue to an empirical or modeling issue: given a system composed of multiple interacting AIF subsystems, what is the specification of desire that produces the "correct" behavior in each subsystem?
Note also that, while this decomposition is totally compatible with the highest-level system being (approximately) only concerned with its homeostasis, it does not require it, and indeed, it is a fair question whether any real-life animals fit that bill. Indeed, Gaia theorists may well argue that only the biosphere as a whole can count as an endogenously motivated system, with each individual animal or biome being at least partially exogenously motivated...
Did MK-ULTRA yield any useful information? It might be interesting to know how people react to LSD without knowing they are taking it. I realize this program was insanely evil but it still may have had interesting results.
"This is good insofar as you're suffering less, but bad insofar as you've adjusted to stop caring about a bad thing or thinking of it as something that needs solving [...]."
I'd expect lowering prediction-error==suffering to zero is a complete cure for depression. Think ugh-fields. The less horrible they feel, the easier they are to address.
If Terrible Issue X doesn't compel you to action with the Terrible Feelings gone, perhaps X was never worth addressing in the first place.
I know you’ve been writing about predictive coding for years but some ideas stick out to me now.
Driving is pleasurable (and being a passenger less so) because I get to make and fulfill a lot of correct predictions.
The musicianship in sound engineering is a process of perceiving and resolving prediction errors. You hear the gap between what the mix is and what it “wants” to be, then you move your fingers on the board to bring it in line.
Software design shares this “is vs. wants to be” structure although the feedback loop is much slower. “Wants to be” is the same kind of intuitive, black box oracle. This confuses and frustrates those who do not have the same kind of oracle inside themselves and yearn for some kind of logical, rules-based system to substitute for it.
Sure, math is fun or if you’re a Brit I guess maths are fun, but I don’t think the problems you are concerned with here are going to yield to some nth order predicate calculus.
There is a reason they call it meta (beyond) physics. William James knew this. He used nitrous oxide to get a better ‘understanding’ of Hegel.
I think if you insist on sticking to the strictly rational you aren’t going to get where you want to go.
I think the explanation for the last one is very simple: Our brain has a set of limiting priors because evolutionary, they helped in advancing human society. It was just not good for society to have lots of individuals stepping outside of the beaten path.
Psychedelics widen your thinking by reducing these priors. This is very interesting (and in many cases helpful) to the individual, but historically not to society because peasants should do their field work and cleaners should clean rooms instead of thinking about the meaning of life.
So, the weird psychedelicists are weird from a position of usefulness to society, but maybe not from their own position. And this weirdness doesn't happen as much anymore because society has - at least in some progressive areas - adapted to psychedelics. We now know how to steer psychedelicists so they stay useful to society. Just look at all the post experience integration trainings and sitter guidance that are out there
Quoting George: "I found it curious that people who take psychedelics for introspection usually end up as religious cranks or burnouts. While people that take psychedelics because "they are fun" don't seem to experience many negative side effects."
- - -
For certain definitions of "introspection," this sounds to me like "People who expect a substance to remedy their existential malaise end up disappointed, but generally satisfied people who use a substance to have a bit more fun usually have fun."
I have a chronic pain condition so I get a lot of opportunity to analyze the experience of pain. Most kinds of pain can in fact be zeroed out completely with sustained application of attention. The trick is, unintuitively, to focus on the pain as precisely as possible. The pain sensation will dissolve and disaggregate into a bunch of distinct sensory signals, none of which carries a "suffering" valence. I think what's happening is you're devoting your whole consciousness to "predicting" what the next moment of pain-sensation will be like, and having done that, the pain does indeed become incorporated into your overall prediction, and thus stops being pain. In keeping with this model, the trick only works if the sort of pain you're attending to is a static (or regularly pulsing) kind.
As soon as your attention wavers, the pain comes back, because you stop actively predicting it. I think pain is a relatively unique sort of signal that resists being ignored, for obvious reasons.
Hmmmm, seems like high H2TA (- Optimism - Plasticity - Sensitivity to stimuli - Ease of learning) may also explain why some people get so STUCK in high gain-high cost/loss situations such as living with an (originally amazing and still occasionally great) abusive partner. They keep trying to figure it out, make it work, and believe at some deep molecular level that this can be accomplished. They may have successfully applied these strategies to many many other situations in their lives. They may also assume that their partner also sees the world this way, and is also working hard to improve things (and the abuser may have figured out how to encourage this assumption).
I've seen Michael Pollan's book referenced here, he mentions research indicating down-regulation of the default mode network during psychedelics - does this fit with the induction of a 'bias towards thinking of problems as solveable'? (Maybe.)
The classic paper on the subject (Berg et al. (1998) in Molecular Pharmacology) is one of my personal all-time favourites within the field, with very nice graphs illustrating the difference between various agonists.
I've taken LSD and shrooms a couple dozen times over the past three or four years, "just for fun" — and my beliefs are essentially the same as they were before I started this (still a boring old stick-in-the-mud materialist who insists on skepticism re: Illuminati and Hidden Masters and psychic powers and so forth).
On the other hand, out of two friends who took psychedelics with the intention of doing some "deeper" introspection than I engaged in, one is just the same — still grounded and reasonable — and the other went totally whacky. Not sure this supports the hypothesis in the post... but n = 3, of course.
My two principle LSD experiences as a teenager were playing the first mass effect game and watching the sci-fi channel Dune and children of dune TV miniseries in a weekend acid binge around 2007.
Anecdotal effects include a decade long lasting sci fi space optimism infusion, an intrinsic faith in the padishah emperor, and the vague knowledge that Butlerian jihad against the AIs leads to space Jesus thousands of years later.
Not sure, if I understand everything correctly - but I think @scott says that antidepressants (think SSRI) push button one which let's your brain get happier with the world as it is. While psychedelics would push button two which let's you get motivated to solve the world's problems right now. I can only tell about my own experience in taking SSRI since 9 months now - they seam to push both buttons inside my brain. I'm way less concerned with anything, getting way easier to sleep after 45+ years of having hard times every evening with endless circles of thoughts. And I'm so muchess reluctant to start anything, I have a lot higher drive and motivation to do the first step of any small to large scale change I'll find is necessary. My impression is, that I'm much less stuck in toxic thinking like "every tiny piece is in some kind connected with ever other piece - so no way to find out where to begin with". In generall I've become a much more pragmatic person with a higher motivation to get things done now and getting less distracted by the endless number of tiny things related.
I have always had this narrative that my cold tolerance shot up inexplicably after starting SNRIs, so that's my selective anecdata related to these theories.
"First of all - predictive coding identifies suffering with prediction error. This conflicts with common sense."
Predictive coding does not actually make this claim, it's solely an operational interpretation of the bayesian brain hypothesis. Though some researchers, like Thomas Metzinger have suggested that emotional affect is related to the rate at which uncertainty is resolved within the context of the bayesian brain hypothesis.
Nonetheless, if it is true that prediction error is related to affect, it's important to understand that prediction error in this context exists at multiple levels in the brain heirarchy. If we have a discussion about you stabbing my arm and you subsequently do so, I will have successfully squashed prediction error at the low levels of the visual perception hierarchy. But the higher levels, extending out to those those that might exist in the association cortex, have a probabilistic prior that includes something like "My self model dislikes bodily harm", and in the behavioural (active inference) context, "My self model is not being harmed". This is a prior that has probably developed on an evolutionary timescale. Thus by allowing you to stab my arm I would actually be behaving in a way that would have dramatic higher expected prediction error (expected free energy).
I don’t know anyone who takes LSD at parties regularly. It is too intense and lasts too long for folks who just want to unwind for an evening. Festivals maybe. But most people I know who regularly take it are more like growth mindset people who just want to experience new things and see what else the brain can do. Not people trying to fix themselves.
I've taken a number of psychedelics specifically for introspection and I can't say they modified my beliefs much, partly probably because I never felt belief that isn't backed up by concrete action can really do anything. But they definitely helped me out to figure out what I needed to do to change my life for the better. I'd say the "crank or burnout" transition can only happen if you have some sorts of prior mystic view of the world that elevates religious or supernatural "secret" types of knowledge over the observed material reality, which might've been more common in the last century, pre-Internet, around when these substances first surfaced.
And as for how acid and transhumanism come together...I didn't have to take those drugs to obtain transhumanist beliefs - rather, I had transhumanist beliefs in the first place that then led me to explore chemically altered states of consciousness just in case they could give rise to any interesting insights, which I'd say they did in a mostly practical way - made me change some of my hobbies, study certain topics and switch occupations for the better. My overall outlook and preferences stayed exactly the same, yet I started to feel like I could be doing more interesting/productive things with my time than I normally used to.
I guess my advice to would-be users is simply to avoid accepting ideas you get on trips in the long-term unless they also happen to entirely make sense when fully sober. You're just vividly exploring alternate possibilities that may or may not make sense in the end. :)
My initial inchoate thought (which I haven’t considered further, just finished work, tired); I wonder how this fits in psychotic depression, BPD or rapid cycling BPAD?
I find it frustrating when authors don't say whether they are talking about presynaptic or postsynaptic 5-HT1A receptors, as the presynaptic receptors are known to have multiple functional activities aside from their "primary" role of negative feedback.
When publishing work related to 5-HT2A, speculations should also consider
1. the role of 5-HT2C receptors, and perhaps the 5-HT(2A)-(2C)
2. how the ligands used in animal studies such as DOI have a much greater affinity for heterodimeric receptors like 5-HT(2A)-mGluR(2), -D(2), -CB(1), etc and often should not be conflated with native serotonergic activity.
Interesting, good new thoughts and questions... (I just finished Pollan's "How to Change Your Mind" last weekend, so it's on my mind. So to speak.) I was wondering about the H1A/H2A receptors as well, vis-a-vis SSRI/SNRIs, not being a trained anything except musician, and my question boils down to this: as I understand it, the inhibitor drug's purpose is to inhibit re-uptake, to free up serotonin or norepinephrin so that ostensibly you have more to go around, so you ...what? Feel or think more easily? But they are functioning by inhibiting the receptors, which may be receivers for other things as well, right? So what might a person be missing by inhibiting the receptors?
I had a recent experience in meditation that made me think of prediction error: i felt like my body was both 'really big' and 'really small' at the same time. I eventually interpreted this as being something like 'the error bars on my body have blown up'. I figured my brain must be computing 'where my body is in space', that this computation must have some accompanied 'error' signal with it, and (maybe from the act of relaxing while staying alert?) my brain eventually said 'huh, i guess this could maybe be anywhere'
I've noticed, too, when my kids start fussing aggressively, if i'm really mindful, i can notice some feeling like "this loud noise should not be happening now" - which looks like prediction error, and thus suffering.
As for why money feels good (even though it's a big surprise), i think this is probably because it changes a bunch of other predictions about future needs being met. And, anecdotally, i've found that i sometimes have a negative emotional response when i see bitcoin going up dramatically in one day, even though i like bitcoin and hold a position in it. I think this is because it isn't really updating predictions about my future needs being met (i expect these to be met regardless), and instead is just a general 'some unexpected thing happened' type of prediction.
> one remaining problem here is why and how some prediction errors get interpreted as rewards
This happened to me a few years ago and it wasn't a pleasant experience. My career took a very sharp, but positive turn and it made me realize that my priors on myself were really off. I ended up in a position that I was shooting for but got there about a decade earlier than I thought I would. It made me realize how low my self-esteem had been, how bad at relationships I was (long story how this fits), and a few other matters.
I have one other experience of a prediction error turning positive in investing. It made me realize that I had just gotten lucky, that I didn't really know what I was doing, and that I should just stick to index funds (though perhaps index funds are no longer the best long-term vehicle if certain economists are to be believed).
I have more experience with mushrooms than most people who have used mushrooms, and only one experience with LSD, and found the two experiences to be so drastically different that I would have a hard time accepting generalizations about how psychedelics as a class should be used to attack problem-solving.
N=1, but mushrooms have been great for dealing with internal issues. I don't think I would be able to solve any technical issues on them (analytical thinking is challenging on mushrooms over 2ish grams), but they are fantastic for chasing threads of emotions and experiences, and for interacting with sources of meaning. LSD felt much more cerebral, euphoric, and neurotic.
The feeling of thinking on both of them is also quite different. With your eyes closed, thinking on mushrooms generally feels like it occurs in the space of "consciousness", like in meditation. Thoughts, sounds, patterns, physical sensations merge into a kind of chaotic space. Thinking on LSD (only one experience here, but decent dose) is pretty much just thinking - you're in your head like you are during your day-to-day, but with pretty patterns and a greater willingness to explore new ideas.
I imagine we do research into the impact of psychedelics a disservice when discussing "psychedelics" vs. the specific substances.
The description of H2A stimulation sounds unsettlingly familiar.
>First of all - predictive coding identifies suffering with prediction error. This conflicts with common sense.
This is something I really don't get. The predictive coding model seems super interesting and have a lot of explanatory power but it seems weird to suppose that it is a global model that explain everything brain related. Why identify suffering with prediction error, instead of the more common sens idea that prediction error contributes/is a type of suffering?
It is often said in the fire poi/"flow arts" community that practicing on acid - and other psychedelics, but mostly acid - is god's own cheat code. This matches up very nicely with George's theory about the proper use of psychedelics.
Sorry, what is the link to George’s commentary? I think that’s what’s being excerpted in the quotes, but I can’t see the link to the original.
The only pedantic completion I'd have to this is that I don't think all "introspection" done on psychedelics results in weirdness.
For example, you could introspect vision and figure out: When I see a repeating pattern I don't actually see the "true" number of objects, instead, my mind kind of infers it as a repeating pattern and translates it into however many objects it wants, so I end up perceiving the number of stairs based on my rate of exhaustion, or more/fewer posts in a fence depending on how hard I focus on it
But if you introspect on concepts where there's no "reasonable" explanation to be found or no particular insight left to gain.
For example death, or meaning or consciousness itself, or "true love" or whatever. If you introspect specifically because those concepts are bothering you, then it gets weird. This is not to say doing so is always bad, I used to be afraid of death (like, existential dread lasting half an hour every few days or weeks for 10+ years afraid) and after 500mcg and 14 hours that basically stopped... maybe I got stuck with an unadaptive mental model as a tradeoff, but then I'm fine with the tradeoff from a negative utilitarian perspective. I know many other people that walked through these kinds of issues on psychedelics. But there's certainly a point where it gets weird.
In other words, I think there are facets of the mind that it can introspect with a small chance of it going bonkers (e.g. sight, sound), and facets where the risk is high (especially given repeated trips, especially given high dosages and or predispositions for certain types of psychosis).
As an aside, I find it interesting that quitting alcohol abuse has a high success rate by either joining a strongly religious community (e.g. AA) or taking psychedelics, which would fit the idea of "build a weird mental model to allow your frontal cortex to intervene in a very well-formed set of patterns". But then again, those are small n studies, and psychedelics seem to work for everything in small n studies in ways that really contradict the behaviours of users (e.g. 80% efficacy in terms of smoking cessation)
I think that psychiatry as a discipline sometimes overemphasises the role of specific receptor subtypes and underemphasises the role of where in the highly specialised structure of the brain those receptors are, in terms of the type of cognition they are involved in.
In terms of reward and reward prediction error, I also feel that there is a wealth of research on dopamine and the dopaminergic system being more involved in this.
If we recall at all of the brains dopamine is produced in the substantia infra and ventral tegmental area, and then projects into the limbic system, via the mesolimbic pathway and into the rest of the cortex, frontal cortex first, via the mesocortical pathway, then we can consider the effects of dopamine on reward.
There is a wealth of data on dopaminergic stimulation of the nucleus accumbens (limbic) as being involved in reward and reinforcement learning. There is also evidence of the orbitofrontal cortex connections to the limbic system as being critically involved in reward, but also reward anticipation, and reward prediction errors. Rolls writes extensively on this for example. De Youngs studies on brain volume and personality find that OFC volume and extroversion are positively correlated, which he interprets as fundamentally reward seeking.
I don’t know the exact distribution of DA vs 5HT receptors in these regions (and I’m not sure who does), but common sense understanding of the emotional impact of serotonergics vs dopaminergics would be that dopaminergics produce a euphoria (reward) as well as highly motivating people (reward expectation) to move, socialise, or even do otherwise unrewarding tasks such as cleaning ones kitchen or revising for a exam when one has ADHD.
I’d also speculate that dopamine is actually the key neurotransmitter signalling pleasant surprise (a positive reward prediction error) and unpleasant shock (a negative reward prediction error). None of these emotions are produced by people taking serotinergics such as SSRIs.
In contrast, I think that both SSRIs primarily stimulating 5HT1A and psychedelics primarily stimulating 5HT2A are actually involved more in *changing* the weighting of neuronal connections which we’re making those prediction errors, through plasticity which psychedelics in particular have been shown to increase.
Studies on gambling back this idea up too. In rat brains, an unexpected reward is accompanied by a DA release whilst a loss was associated with a 5HT release. The interpretation here was that DA affected plasticity by making the Britons that fire together, wire together making the animal feel like “I really want to do that movement or behaviour again!” and thus explaining DAs role in habit formation and addiction, whereas 5HT is involved in increasing plasticity by reducing signal weights, when a prediction error is detected. Thus, the serotonergic system is involved in the recognition at some level that “this prediction did not work, I need to change my behaviour and do something different” which is massively enhanced if a person is taking SSRIs or psychedelics.
Therefore, I’d be wary of attribution of cause to 5HT for prediction error, but more if an effect of prediction error in producing changes in signal weighting via plasticity to change behaviour.
If you’ve ever used CBT to treat depression or anxiety you know how people make negative prediction errors, and how helping people to recognise this and change their behaviour is the key to treatment, and how an SSRI can really assist this process. SSRIs don’t produce reward (euphoria), reward prediction (motivation) or reward prediction error (surprise/shock) which I think are all dopaminergic. SSRIs help people change the reward prediction errors they are making that terrible things will happen, and thus help them to change their thinking and behaviour
-If you get $1 million because you're an ordinary middle-class person and a crypto billionaire semi-randomly decides to give you $1 million one day, you will be very happy.
Do we actually know being "very happy" would be the typical instant reaction? I once won $10,000, about 25% of my income at the time, in some promotional lottery I hadn't realized I had been entered in. I was going through a miserable time in life and my reaction was to be angry and annoyed, because it felt like a cruel joke. Only after a few weeks, when I used some of the money to buy a new guitar, did I feel happy to have it.
Maybe that was like stepping in a cold shower and suffering at first but then getting used to it? Or maybe because I was depressed, I didn't have a normal reaction?
I think it mattered that I didn't even know I was entered in the lottery. If someone buys a lottery ticket, perhaps a part of them predicts they will win. Otherwise, why did they buy the ticket?
Fits with Buddhist ideas; you don't suffer because you are in pain, you suffer because you don't think of pain as something that should be happening to you.
Surely what makes a cold shower decrease in pain over time is not an adjustment in your predictions, but a physical adjustment in the contrast in skin temperatures.
> one remaining problem here is why and how some prediction errors get interpreted as rewards
Objectively unlikely events could easily confirm our model of the world rather than violating it, if the model was suitably wrong to begin with.
Plausibly, our money-centric culture—and in particular our parasocial relationships with affluent pseudo-peers—causes our brains to maintain a prediction amounting to "I am wealthy, or soon will be, and any indication to the contrary is some terrible mistake." The conflict of this prediction with reality creates suffering, but we get stuck unable to update away from it, because our media habits keep giving us more fake evidence that it's true (which is maybe not really our fault since whole industries are dedicated to ensuring that very thing). A sudden windfall, then, won't feel surprising; it will feel like the natural order restored.
Now I'm no historian, but I'd guess that if you're a premodern farmer and you have an exceptional crop, your model already represented that as something that could and indeed *ought to* happen; although it was surprising in some statistical sense, you aren't surprised—the world feels more correct rather than less. But if the king rides up and hands you a fistful of glittering gold, aren't you more likely to be suspicious and scared than happy? Surely this is some mistake or trick or trap; surely the other villagers will turn on you, or bandits will kill you for it, or something—anything—will punish you, because suddenly you're outside the natural order.
> At its limit, this theory says that all action takes place through the creation and resolution of prediction errors - I stand up by "predicting" on a neurological level that I will stand up, and then my motor cortex tries to resolve the "error" by making me actually stand.
Thinking out loud: there are several programming languages whose core primitive operation is some version of this. Prolog being a conventional example.
In Prolog and other languages incorporating such features, 'programming' takes the form of, first, writing 'facts', and then second, writing underspecified facts. The program executes by "resolving the error" by figuring out which values need to be what to make your underspecified facts true. This is used a lot for logic programming (it's in the name, after all).
If you described this programming language as "resolving pattern-match error" to execute, that would be technically correct. Perhaps "prediction error" functions similarly in cognition. Perhaps, at some level what is technically happening is prediction error being resolved, but in practice that is an obtuse way of framing it
Off topic: you comment on 2 subjects I find interesting. 1. how do our brains choose goals - crude answer we feel tension when how we want the world to be is different from how it is; 2. what do drugs do to our brains. It would be awesome if you wrote a piece on how the 2 overlap.
So we have: A. how the world is; B. how I perceive the world; C. what I want the world to be; D. how much the disparity upsets me, E what I am doing about it. We can acheive our goals by moving the world closer to how we want it to be, looking at things in a different way, by changing what we want, by stopping caring about the difference. What is the attraction of different drugs? Which of these 4 ways of reducing that tension do they address? e.g. heroin - at a guess B and D; cocaine at a guess B and E. But I know nothing in these fields, what does the attraction of different drugs tell us about the model of motivation and consciousness?
> one remaining problem here is why and how some prediction errors get interpreted as rewards... This suggests there's still something we don't understand about prediction error and suffering
I think if you were describing a theory you were less attracted to, then instead of saying that this is one remaining thing we don't understand, then you'd say that this is a class of counterexamples that totally blows up the theory.
You've got three examples of the "prediction errors -> suffering", of which one fits (losing money), one doesn't fit (gaining money) and one only fits if you torture the definitions of words until they submit (feeling pain). Most of the other examples of prediction errors that I can think of don't seem to cause suffering; e.g. when watching a movie it's not at all unpleasant when the camera cuts to another angle.
I think there's a lot of good in predictive processing, but this particular "prediction errors -> suffering" aspect seems a bit of a dead end to me.
I find the idea that we shouldn't use psychedelics for introspection a bit troubling: I've been studying up on psychocybin based psychotherapy lately, and it seems to get very promising results in treating chronic depression and anxiety. Yet the entire model of therapy is based on internal introspection: you give people an extremely high dosage of psychocybin, and then you put a blindfold on them and play gentle music so they aren't distracted by the outside world and can focus on internal introspection. Could this be a dangerous therapy model to pursue? The studies done Dr. Roland Griffiths seem to show good outcomes, but I'm not sure how well you can measure "weirdness" after treatment.
> the 10% of Americans who use psychedelics mostly don't end out as weird as they did
Huh. All the people I know who are really into psychedelics also have very eclectic beliefs about religion or humanity or whatever. Maybe you're right about dosing (note the "really into", not "10% of the people I know"). Or maybe we've all learned to be ok with some types of cognitive weirdness in a way that it's acceptable to be a yogi barista inventing your own religion and nobody really makes a big deal out of it anymore.
New to the neighborhood question: Has Michael Pollan’s book been hashed out here?
2c on Active Inference Framework and reward/suffering: (context: I have developed an AIF-based model that's being published in Entropy soon, so I can speak to the math/theory side of it - less so to the psych/neurology angle)
Friston is generally cavalier with his writing and this seems to me like an example of that. If you're abstractly just interpreting animals as simple self-regulating systems (trying to stand up, or stay fed, etc) whose only ambition is to maintain homeostasis, the prediction error / suffering metaphor works great, in the following sense: the reward function is endogenous, so any behavior is a self-fulfilling prophecy. But if you're considering a sophisticated agent that's composed of multiple subsystems, capable of foresight/delaying reward, attaching value to symbols, etc, then when isolating an individual loop/subsystem, you must accept that in general the reward function will be exogenous (provided by whatever the relevant desire-generating subsystem is), and the discrepancy between desired and observed/predicted future can be arbitrarily large. This reduces the philosophical issue to an empirical or modeling issue: given a system composed of multiple interacting AIF subsystems, what is the specification of desire that produces the "correct" behavior in each subsystem?
Note also that, while this decomposition is totally compatible with the highest-level system being (approximately) only concerned with its homeostasis, it does not require it, and indeed, it is a fair question whether any real-life animals fit that bill. Indeed, Gaia theorists may well argue that only the biosphere as a whole can count as an endogenously motivated system, with each individual animal or biome being at least partially exogenously motivated...
For those who would like to read George's review: https://cerebralab.com/Stress_and_Serotonin
Did MK-ULTRA yield any useful information? It might be interesting to know how people react to LSD without knowing they are taking it. I realize this program was insanely evil but it still may have had interesting results.
"This is good insofar as you're suffering less, but bad insofar as you've adjusted to stop caring about a bad thing or thinking of it as something that needs solving [...]."
I'd expect lowering prediction-error==suffering to zero is a complete cure for depression. Think ugh-fields. The less horrible they feel, the easier they are to address.
If Terrible Issue X doesn't compel you to action with the Terrible Feelings gone, perhaps X was never worth addressing in the first place.
I know you’ve been writing about predictive coding for years but some ideas stick out to me now.
Driving is pleasurable (and being a passenger less so) because I get to make and fulfill a lot of correct predictions.
The musicianship in sound engineering is a process of perceiving and resolving prediction errors. You hear the gap between what the mix is and what it “wants” to be, then you move your fingers on the board to bring it in line.
Software design shares this “is vs. wants to be” structure although the feedback loop is much slower. “Wants to be” is the same kind of intuitive, black box oracle. This confuses and frustrates those who do not have the same kind of oracle inside themselves and yearn for some kind of logical, rules-based system to substitute for it.
Sure, math is fun or if you’re a Brit I guess maths are fun, but I don’t think the problems you are concerned with here are going to yield to some nth order predicate calculus.
There is a reason they call it meta (beyond) physics. William James knew this. He used nitrous oxide to get a better ‘understanding’ of Hegel.
I think if you insist on sticking to the strictly rational you aren’t going to get where you want to go.
I think the explanation for the last one is very simple: Our brain has a set of limiting priors because evolutionary, they helped in advancing human society. It was just not good for society to have lots of individuals stepping outside of the beaten path.
Psychedelics widen your thinking by reducing these priors. This is very interesting (and in many cases helpful) to the individual, but historically not to society because peasants should do their field work and cleaners should clean rooms instead of thinking about the meaning of life.
So, the weird psychedelicists are weird from a position of usefulness to society, but maybe not from their own position. And this weirdness doesn't happen as much anymore because society has - at least in some progressive areas - adapted to psychedelics. We now know how to steer psychedelicists so they stay useful to society. Just look at all the post experience integration trainings and sitter guidance that are out there
Quoting George: "I found it curious that people who take psychedelics for introspection usually end up as religious cranks or burnouts. While people that take psychedelics because "they are fun" don't seem to experience many negative side effects."
- - -
For certain definitions of "introspection," this sounds to me like "People who expect a substance to remedy their existential malaise end up disappointed, but generally satisfied people who use a substance to have a bit more fun usually have fun."
This does not seem like it should be a surrpise.
I have a chronic pain condition so I get a lot of opportunity to analyze the experience of pain. Most kinds of pain can in fact be zeroed out completely with sustained application of attention. The trick is, unintuitively, to focus on the pain as precisely as possible. The pain sensation will dissolve and disaggregate into a bunch of distinct sensory signals, none of which carries a "suffering" valence. I think what's happening is you're devoting your whole consciousness to "predicting" what the next moment of pain-sensation will be like, and having done that, the pain does indeed become incorporated into your overall prediction, and thus stops being pain. In keeping with this model, the trick only works if the sort of pain you're attending to is a static (or regularly pulsing) kind.
As soon as your attention wavers, the pain comes back, because you stop actively predicting it. I think pain is a relatively unique sort of signal that resists being ignored, for obvious reasons.
Hmmmm, seems like high H2TA (- Optimism - Plasticity - Sensitivity to stimuli - Ease of learning) may also explain why some people get so STUCK in high gain-high cost/loss situations such as living with an (originally amazing and still occasionally great) abusive partner. They keep trying to figure it out, make it work, and believe at some deep molecular level that this can be accomplished. They may have successfully applied these strategies to many many other situations in their lives. They may also assume that their partner also sees the world this way, and is also working hard to improve things (and the abuser may have figured out how to encourage this assumption).
I've seen Michael Pollan's book referenced here, he mentions research indicating down-regulation of the default mode network during psychedelics - does this fit with the induction of a 'bias towards thinking of problems as solveable'? (Maybe.)
Any discussion of the effects of 5-HT[2A] agonist psychedelics has to include the (not-so-recently-discovered) fact that different GPCR agonists can produce different effects. It's not a simple one-dimensional "more or less"-scale - it's multi-dimensional. See for example https://juniorprof.wordpress.com/2010/09/21/agonist-directed-trafficking-of-receptor-stimulus-pharm-551a-berg-et-al-1998/
The classic paper on the subject (Berg et al. (1998) in Molecular Pharmacology) is one of my personal all-time favourites within the field, with very nice graphs illustrating the difference between various agonists.
I've taken LSD and shrooms a couple dozen times over the past three or four years, "just for fun" — and my beliefs are essentially the same as they were before I started this (still a boring old stick-in-the-mud materialist who insists on skepticism re: Illuminati and Hidden Masters and psychic powers and so forth).
On the other hand, out of two friends who took psychedelics with the intention of doing some "deeper" introspection than I engaged in, one is just the same — still grounded and reasonable — and the other went totally whacky. Not sure this supports the hypothesis in the post... but n = 3, of course.
My two principle LSD experiences as a teenager were playing the first mass effect game and watching the sci-fi channel Dune and children of dune TV miniseries in a weekend acid binge around 2007.
Anecdotal effects include a decade long lasting sci fi space optimism infusion, an intrinsic faith in the padishah emperor, and the vague knowledge that Butlerian jihad against the AIs leads to space Jesus thousands of years later.
Not sure, if I understand everything correctly - but I think @scott says that antidepressants (think SSRI) push button one which let's your brain get happier with the world as it is. While psychedelics would push button two which let's you get motivated to solve the world's problems right now. I can only tell about my own experience in taking SSRI since 9 months now - they seam to push both buttons inside my brain. I'm way less concerned with anything, getting way easier to sleep after 45+ years of having hard times every evening with endless circles of thoughts. And I'm so muchess reluctant to start anything, I have a lot higher drive and motivation to do the first step of any small to large scale change I'll find is necessary. My impression is, that I'm much less stuck in toxic thinking like "every tiny piece is in some kind connected with ever other piece - so no way to find out where to begin with". In generall I've become a much more pragmatic person with a higher motivation to get things done now and getting less distracted by the endless number of tiny things related.
Hope this makes sense to you.
I have always had this narrative that my cold tolerance shot up inexplicably after starting SNRIs, so that's my selective anecdata related to these theories.
I think you sound pretty naive about the subjective experiences and motivations of people utilizing psychedelics.
"First of all - predictive coding identifies suffering with prediction error. This conflicts with common sense."
Predictive coding does not actually make this claim, it's solely an operational interpretation of the bayesian brain hypothesis. Though some researchers, like Thomas Metzinger have suggested that emotional affect is related to the rate at which uncertainty is resolved within the context of the bayesian brain hypothesis.
Nonetheless, if it is true that prediction error is related to affect, it's important to understand that prediction error in this context exists at multiple levels in the brain heirarchy. If we have a discussion about you stabbing my arm and you subsequently do so, I will have successfully squashed prediction error at the low levels of the visual perception hierarchy. But the higher levels, extending out to those those that might exist in the association cortex, have a probabilistic prior that includes something like "My self model dislikes bodily harm", and in the behavioural (active inference) context, "My self model is not being harmed". This is a prior that has probably developed on an evolutionary timescale. Thus by allowing you to stab my arm I would actually be behaving in a way that would have dramatic higher expected prediction error (expected free energy).
I don’t know anyone who takes LSD at parties regularly. It is too intense and lasts too long for folks who just want to unwind for an evening. Festivals maybe. But most people I know who regularly take it are more like growth mindset people who just want to experience new things and see what else the brain can do. Not people trying to fix themselves.
I've taken a number of psychedelics specifically for introspection and I can't say they modified my beliefs much, partly probably because I never felt belief that isn't backed up by concrete action can really do anything. But they definitely helped me out to figure out what I needed to do to change my life for the better. I'd say the "crank or burnout" transition can only happen if you have some sorts of prior mystic view of the world that elevates religious or supernatural "secret" types of knowledge over the observed material reality, which might've been more common in the last century, pre-Internet, around when these substances first surfaced.
And as for how acid and transhumanism come together...I didn't have to take those drugs to obtain transhumanist beliefs - rather, I had transhumanist beliefs in the first place that then led me to explore chemically altered states of consciousness just in case they could give rise to any interesting insights, which I'd say they did in a mostly practical way - made me change some of my hobbies, study certain topics and switch occupations for the better. My overall outlook and preferences stayed exactly the same, yet I started to feel like I could be doing more interesting/productive things with my time than I normally used to.
I guess my advice to would-be users is simply to avoid accepting ideas you get on trips in the long-term unless they also happen to entirely make sense when fully sober. You're just vividly exploring alternate possibilities that may or may not make sense in the end. :)
My initial inchoate thought (which I haven’t considered further, just finished work, tired); I wonder how this fits in psychotic depression, BPD or rapid cycling BPAD?
I find it frustrating when authors don't say whether they are talking about presynaptic or postsynaptic 5-HT1A receptors, as the presynaptic receptors are known to have multiple functional activities aside from their "primary" role of negative feedback.
When publishing work related to 5-HT2A, speculations should also consider
1. the role of 5-HT2C receptors, and perhaps the 5-HT(2A)-(2C)
2. how the ligands used in animal studies such as DOI have a much greater affinity for heterodimeric receptors like 5-HT(2A)-mGluR(2), -D(2), -CB(1), etc and often should not be conflated with native serotonergic activity.