I dunno, the link between low confidence and negative prior still seems to be missing here to a substanial extent. Like, where's the corresponding problem of low confidence, positive prior? It wouldn't seem to be mania from what you suggest. So what's going on here?
This was my immediate line of thinking too. The new idea here doesn’t seem to really be a synthesis of the two ideas at the start of the post any further beyond saying “depression is what you get when both these are true”. Mania being the corner of the box with low confidence, positive priors seems to fit really well with the theory.
Not to go evolutionary-psych here, but it seems like there are situations where it’s reasonable to be “normal” (a low-threat environment) and there are situations where it’s reasonable to be “low confidence, negative prior” (a high-threat environment), but I can’t think of any situations where it’s reasonable to be “low confidence, positive prior.” Generally being wrong about things in the negative direction will help you, while being wrong in the positive direction will hurt you. So it sounds plausible that the two variables would have a strong relationship.
Of course, being “low confidence, negative prior” all the time is really not good in our modern environment, but evolution wasn’t optimizing for that.
My personal experience of PTSD is that it's what happens after you "get got", but somehow survive. And you notice that no one else around you is adjusting their priors at all. And they don't even want to acknowledge you exist, because that might lead them to start asking questions about what the actual frequency of tigers-in-bushes is, and that might interfere with them "getting theirs". And every time you see a bush, your brain malfunctions and dumps a load of memories and emotions into your mind; not the tiger jumping from the bush, but instead, your friends running away and abandoning you.
I think 'low confidence, positive prior' is really useful if you're someone powerful. It allows you to wield your power to your own advantage recklessly, without being concerned about hurting people or facing consequences, because you don't notice that you're hurting them and your power insulates you from consequences anyway.
I feel like I've seen this personality style in a lot of successful C-level managers and entrepreneurs I've known over the years, and I feel like it matches how such people are often depicted. Of course, when those depictions are framed by neurotics or normal people, they often depict those powerful people as *not caring* about the people they hurt; but I think it's much more accurate, in most cases, that they simply don't notice.
Which is what would happen with 'low confidence positive prior'.
How is this behavior not "high confidence, positive prior"? Low confidence would be more likely to make you pay attention to how you're actually affecting people. It would reduce the importance of the positive prior.
I confess I'm a little confused about the math. In Bayes' Theorem (the basis of predictive coding) there's no coefficient you can apply to determine the importance of evidence. The posterior is purely a function of the prior and the plausibility of the evidence under different hypotheses. But I'm imagining that in this conception of depression or mania, you would, let's say, calculate the posterior P(H|E) in the normal way and then arbitrarily decide that P(H|E,depression) = P(H) + a*(P(H|E)-P(H)), where a < 1 if you're depressed and a >1 if you're feeling powerful/reckless.
I think this is basically equivalent to using "confidence" in its everyday sense. Am I wrong?
The prior *is* the term that determines the importance of evidence: the stronger your prior is, the less your belief is shifted by a given amount of evidence.
I think what your model is broadly correct there. 'Low confidence' being a < 1. So positive prior and low confidence would update priors less than an ideal reasoner, this matches somebody not noticing that you are affecting people negatively, they don't update their positive prior downward a sufficiently large amount on seeing evidence for it
"Usually cheerful but kind of unstable person"? "The kind of schizophrenic who hears nice voices"?
It may just be that our current world is sufficiently varied and chaotic that low prior weight is less damaging than high prior weight. It's the "move fast and break things" of mental models, while mania is the Theranos investor of mental models and depression is the "never start a business at all". Maybe the modern world just rewards agility in a way that screws people with strong priors except in the rare cases those priors lead them in exactly the right direction.
If you accept that there are many more way for things to go wrong than go right (Anna Karenina principle) I don't see any problem linking low confidence and negative prior.
Having low confidence has the effect of bringing probabilities closer to 50/50, so if each of the small things that can go wrong has a high chance of going right this can rapidly compounding effects. Say that 10 things need to go right with 95% prior probability, that means the chance of things going right overall is 0.95^10 = ~3/5. But if because of low confidence the perceived prior goes to 90%, that means the overall chance goes to ~1/3. The low confidence got transformed into a negative prior.
High impulsiveness perhaps? narcissism? can be the sort of person that doesn't get negative clues. thinks everything is great, everything will go just right, and everybody loves him, even when that's not the case - but also doesn't get disappointed when it's not the case because they don't really process that. they just don't notice negativity.
It sure sounds like a rare description, but i think the Ev-Psy explanation someone else gave in this thread explains fairly well why this one would rarer.
That doesn't fit in how children are fast learners ? If anything, children have much higher confidence than adults, it's just that their memory has a lot less weight compared to new evidence ? (Though maybe 'low memory content' is explanation enough, so they have similar confidence to adults ?)
I don't think "confidence in neural predictions" is part of the theory anymore (i.e. the theory espoused by this post).
Mathematically, I interpret "low confidence" as "the brain uses Bayes' Theorem correctly, but then pushes all probabilities toward uncertainty" (which, for a binary problem, means pushing all probabilities toward .5). This is *not* about having heavier weights on priors or on evidence.
The theory espoused in this post (I think) is that priors are weighted more heavily than they should be. So if the "correct" weighting is given by Bayes' theorem:
p(H|D) = p(H) * p(D|H)/p(D)
...then a depressed person is using something like:
p(H|D, depression) = p(H|D) + a(p(H|D) - p(H))
Note: this is the same as what dogiv wrote in their comment.
The difference is that "low confidence" as defined above would result in being less effective at everything, because all of your predictions would be biased towards uncertainty. Your parameter estimates are biased in expectation, meaning you don't converge on truth.
In contrast, a heavy weight for priors would result in requiring more data to learn (i.e. all priors update slower than they should). Your parameter estimates are unbiased in expectation (you do converge on truth).
A person with a heavy prior (meaning "highly-weighted prior") will make good predictions when their prior is close to objective truth. If I have a heavy prior that the world is bad, and the world *is* bad, then my predictions will be great. However, a "low confidence" person with a prior that the world is bad will still make somewhat-bad predictions, because their low confidence pushes them towards uncertainty about whether the world is good or bad.
(Lastly, a person with "light priors" would have high-variance, unbiased predictions. They converge on truth but their predictions are always swingy. Does this describe a known psychiatric abnormality?)
On a separate but related point, Scott writes:
"Looking over the date after you get home, you hyperfocus on the one really embarrassing thing you said, and any good parts of it mysteriously slip your mind."
Humans selectively attend to evidence (I imagine this is doubly true for manic/depressed humans). Scott doesn't explicitly say this, but it's important to call out, because he's assigning some significance to this fact here. This isn't about a neural implementation of Bayes' theorem that underweights evidence. This is about ignoring evidence altogether.
People look for evidence that fits their worldviews. This happens at a level of processing that we can be consciously aware of (or at least catch ourselves doing, if we reflect upon it). So if the "heavy priors" theory is true and explains the unconscious phenomena mentioned here, I think there are also higher-level things going on that contribute to someone being stuck at a certain set point for p(the world is bad). This would help explain why priors are not inevitably updated toward reality after enough data is gathered.
Thanks for asking that, because I'm not entirely sure! My thoughts:
The down-weighting of evidence that Scott talks about in this post is fundamental to the person's neural processing and has a variety of sensory, memory, and motor effects. AFAIK you can't "become aware of it" and thereby make it go away. You can't say "oh I'm just seeing the world in gray because of my silly brain down-weighting evidence. Now that I realize what's happening, I can see the world in color again."
The "ignoring evidence" that I was thinking of is at a higher level of cognition. It is something a person can consciously notice (e.g. "I'm interpreting this person's actions negatively because I am in a bad mood") and do something about ("I'm going to consciously think of positive interpretations in order to de-bias myself.") Down-weighting evidence is not fundamental to this person's neural processing. They probably see the world in color and have normal memory and ability to feel their body.
So the predictions differ:
A person with a depressed mood but who has otherwise normal neural processing should be able to regain a positive mood more easily than someone who downweights evidence, as long as they attend to evidence. This can be done using cognitive behavioral strategies.
These strategies would *also* help someone who down-weights evidence, but the improvement would be slower because even after consciously attending to evidence the person's brain is still down-weighting it at a level they don't have conscious access to.
Conscious effort is known to be able to eventually, after a training period, affect subconscious behavior (as well as, with more or less effort, being able to override subconsciousness in the moment).
I wouldn't be so sure that ignoring evidence is happening on a purely (or even mostly!) conscious level, and that you can NOT train yourself to see the world more in color (see : 'enlightenment').
This also might explain why mania and creativity go together. If depression means you up-weight priors, then mania could mean that you down-weight priors. Your mind is more open.
I think the premise is too broad. If you start lumping things like memories and introspection and other cognition together with more basic sensory input such as vision or hearing, and call of it "senses", then it seems like the word "senses" might as well be synonymous with "anything your brain does".
To the extent that a prior can be described as a semistable attractor state in the energy landscape of the brain, there really isn’t any difference for changing the shape of that state between new data coming in from the nerves connected to your sensory organs, and the nerves elsewhere. The priors/evidence weighting balance simply describes how easily new data is able to dislodge that attractor state into a new configuration.
I am confused however how things such as memory and learning play into this. It seems like this framework would also predict depressed people to be poorer learners. Is that true? I’m not sure what it would say about memory.
“ Researchers in one 2013 study discovered that people with depression couldn’t identify objects on a screen that were identical or similar to an object they had seen previously. According to researchers, this suggests that memory can be diminished as a result of depression. Researchers in a 2015 studyTrusted Source came to a similar conclusion. They concluded that depression might cause short-term memory loss.”
Not really? It might make it synonymous with "anything that you consciously experience", but your brain does a vast amount of things that you are never consciously aware of. And your senses just being anything that transmits information to the global workspace seems like a reasonable enough definition.
I recall Thomas Metzinger making the point that the distinction between introspection and non-introspection is somewhat artificial; in both cases you are experiencing some data that a subsystem has output into your consciousness, it just happens that some of the data is tagged as "internal" and some of it "external". But that's just a tag on the data, and while it does to some extent track a real difference, it can also easily get confused (e.g. projecting your own assumptions on other people is common).
In reality, what feels like raw sensory data is always mixed in with internally generated high-level models of what one expects to see, and what feels like purely internally generated thought always has a causal link to *some* past experience.
Well, to an extent it shouldn't bee surprising if everything your brain does share various important systems-level features, because it's all instantiated on the same computing substrate.
Like, 'everything a computer does' can be expressed in the language of ones and zeros, logic gates and cycles, etc.
Whether or not it's semantically useful to call it all 'senses' is another issue, but I don't think that being overly broad is a problem in and of itself - we should expect a broad theory, at some low-enough level of analysis.
Examined in detail, there effectively is no "more basic sensory input" – I think maybe smell and taste might be the closest. But _vision_ is incredibly sophisticated (and complicated) and is, in large part, actively _predicted_ by the brain and not some kind of 'raw data'.
This is fascinating. I have a child whose depression is exactly like this. But I’d like to suggest a different style of depression, one where the pt downplays or “forgets” positive priors, and focuses unduly on the (negative) now. Here, I think, the disconnect is more cognitive than sensory, and may respond to highly tactical approaches such as cbt.
On the "date" hypothetical: not that my experience of depression is universal, but my experience of that scenario would go as follows:
I am at a -9. I go on a neutral date, which is a 0. Typically (though not always), I have a brief elation period--I feel the date went better than it likely did, say a +2 or a +3. And then abruptly my mood shifts back. I begin to feel anxious and/or paranoid: maybe this dude is a serial killer, and he was charming and into me because he wants to kill me, or maybe he's got a wife and he's a cheater, or maybe he's repulsed by me but really hard up and so he's gonna go along with it until he gets some. Sometimes the anxiety is only the realization that one good date means there will be *another* date, and it just seems like too much work to repeat that performance.
Then, and more importantly, I forget. I forget that the experience was good and can only recall the experience with difficulty. I cannot overstate how much my experience of depression involves forgetting even basic facts about my life. If I feel bad about not being married, I will *literally forget* the people I've dated, and other people/things must prompt me before I remember. If I feel like a failure, I will *literally forget* how much money I make, or how my bosses have given me glowing performance reviews. I don't know that it's anything to do with weighting--at least for me, it seems like the issue is that, outside an extremely short window, the new experience (and many old experiences) is failing to be considered at all. The prior cannot update.
Have you tried writing down a detailed list of good things to be reminded of in such cases? for example, good dates, good feedback, etc (especially ones you have a special reason to believe are honest) - and then go to that list whenever you feel like you might be forgetting or discounting evidence?
It's something i thought of doing myself, but haven't actually done since i rarely feel that way or have trouble coming up with those examples from memory. but i think if i had the problem of often forgetting about them and feeling bad i would make that list.
Yes. Scott doesn't seem to be picking up that people negatively downgrade things further and ignore positive evidence.
We know that people do this with beliefs all the time. I'm surprised that Scott treats beliefs as a naive comparison of belief and evidence when we know the vast majority of people do not work like that.
I think that this is, in part, an attempt to come up with a theory about how that works. Have a linear function for the prior in the example (ie -9 + actual sensory data) is probably wrong but is simple enough to illustrait a point about veteran reasoning.
Reality may be more in line with a discrete system where things fall in categories of very bad, bad, meh, ok, good etc. If the prior is more heavily weighted enough, then new information will fail to update the prior out of the discrete category is in.
Or it could be any number of other update functions. But the core idea is that there is a failure to weight the evidence vs expectation in such a way that expectation dominates and you get stuck in a depressive state.
Agreed. The way depressed people (generally) seem to think is much more like an attractor/quantum state/gravity well where you need energy to move to a new stable state, and anything less than that energy doesn't move you even slightly.
Move someone up 0.9 and it rounds down to 0. Move up a whole point and you're great - but you can slide back into a lower orbit again because the sucking void beneath you continues to exert its pull.
Trying to test the theory: naively, we expect a prior in a negative direction to increase attention to evidence/sensory inputs in the OPPOSITE, namely positive direction, because that is the definition of surprise: you pay more attention to things you did not expect.
However, depressed people despite their supposedly negative prior seem to dwell much more on negative experiences than positive ones, and also have heightened pain and light sensitivity, so that would contradict our prediction. If you already expect bad things to happen to you AND your attention to sensoral inputs is lower, why would you be able to tolerate pain LESS? In some ways, depression seems to make people numb, like you explained, but in others more sensitive.
OTOH, I am not really sure whether my understanding of what a prior entails is correct: If you fear tigers, you will pay more attention to sounds that could be coming from tigers than other people, so it is probably more compicated.
I don't think the pain thing is necessarily a contradiction. The model could explain it as your sense data reporting a neutral pressure or mild discomfort, and your higher-level cognition favoring a prior towards any given sensation being unpleasant.
I'm not sure how this squares with the other sensory data, though. If I'm right, I'd predict that the reduced olfactory capability reported in depressed patients wouldn't extend to unpleasant smells, but I don't know if the current olfactory tests cover this angle.
For what it's worth, some types of depression I've had have made me tolerate pain less (mostly the "general malaise" sorts), and some types have made me tolerate pain more (mostly the "sitting motionless like a lump" sorts). It might also depend on the type of pain.
My depression definitely has a physical component. I don't know how I would score on any of those tests, but I know I feel incredibly lethargic and nothing is interesting. At all. Meditation and exercise don't help - one of my worst episodes came when I was going through a daily bootcamp.
Ketamine assisted therapy briefly worked for a few weeks but the effect faded. I never found any kind of trauma behind it - my therapist remarked that I was pretty unusual compared with his other patients.
Thanks also did some somatic therapy at a weekend workshop. Parts of it I liked but again, no lasting benefit.
I've had good luck with SSRIs but I have to change the drug every few years.
If you're looking for suggestions, you might try TMS (transcranial magnetic stimulation). There's no side effects that I've heard of, and it seemed to help. I got mine from a place that was also offering ketamine therapy, so you might try asking at your ketamine place and see if they do TMS, or know of anyone in the area doing it.
Briefly, for your third point under implications, that's been tried under the rubric "concreteness training" and appears at least in two studies to be fairly effective.
I am somewhat sceptical though. To me it falls under the Jordan Peterson "clean your room" category of advice that sounds good but doesn't really work in shifting the "attractor state" (to use your terms) - except maybe as a placebo. Both appear to be positive symptoms of a "deeper" shift that is not easily amenable to reverse engineering at the surface level, at least absent other important changes.
I had the same question. Having read a tiny bit of Karl Friston (confusing) and recently Michael Pollan's "How to Change Your MInd" I'm guilty of trying to integrate this subject and other examples of brain activity into the models for DMN and Markov blankets. The VDBEA piece seems to mesh well with a model where DMN dominance is overruling the sensory updates being discussed. Not so much with the Markov model - if the sensory input contradicts the prediction generated by priors why isn't the model updated accordingly?
How does this relate (if at all) to the account of autism that sees it as an unusually bottom-up style of information processing? Naively it would seem like a case of opposite weights on priors vs sensory experience, although I don’t know of any indication that autistic individuals are less likely to be depressed, and it doesn’t seem like something I would guess anecdotally either.
Also: some of the descriptions here about depression and trauma having to do with their having very negative priors; others have to do with weighting priors over new evidence. These seem logically distinct and in principle you could have something where someone has a high weight on very positive priors. But as you note happiness is associated with eg color - we don’t seem to have a lot of cases of abnormally happy people describing the world in greys. (Maybe people having mystical experiences that make them more detached from earthly experience? I guess there is a lot of language of happiness that implies floating, flying, various forms of ungroundedness.)
About the "bottom-up processing": I'd describe autism from PP perspective as a very rigid internal model structure that is optimized for high precision and is slow to update (contrast this with a very fluid worldview on the opposite side, related to say schizotypals, see the SSC post on the diametrical model of schizophrenia and autism for context). Byproducts of the rigid/highprecision structure are (again, my lax interpretation):
- you're probably very good in areas where high-precision reasoning is required (say mathematics) and very uncomfortable in areas where a high-precision model is not possible (say social interactions)
- as your model is slow to update, this gets better over time as you slowly incorporate more complex rules (afaik moderate autism, e.g. Aspergers, smoothes out over time as people get older and are maybe slowly building internal rules of coping strategies / picking places and situations they can handle / ...)
So from this model, "low certainty" in internal predictions should not be an issue if you are on the Autism spectrum. BUT: this comes at the price of being hit really hard in case of unexpected situations, e.g. where your predictions fail. Two elaborations:
- I know a small child that is somewhere close to the autism spectrum (no medical diagnosis and this might all smooth out over time) and if things don't go as expected, THIS IS LIKE VERY BAD PHYSICAL PAIN for her. This matches my experience with many other people (most of which are not autism-diagnosed and - if at all - only losely connected to the spectrum!) where i've experienced many times how unforseen events (e.g. a small mishap or unexpected error) quickly spirals out of control in their head and they end up tormenting themselves mentally. If we think of autism as an extreme, this is a potential cause for a very bad mental pain spiral (which is potentially very difficult to escape).
- somewhere in the comments (or Scotts post?), there was a reference that "high prior, low update rate" not only refers to exterior senses but also your internal feeling. I could list a few anecdotal cases (the child mentioned above, a friend diagnosed with Aspergers) where this slowness / neglect of internal sense data lead to followup injuries (the friend mentioned just sprained his arm while "doing normal movements" and I could easily imagine that there were some subtle warning signs coming from his body earlier he missed). Combine this with the point above and you have another recipe for potential mental blaming cascades.
If we think of hyperpriors as the general structure of our internal model of the world (how the hierarchy is connected / unconnected, how much update happens to the model structure in case of failed predictions, ..., resulting e.g. in the individual personality traits that are pretty robust over lifetime), we can think of trauma as "an event that shreds the whole model structure into pieces". This would align quite nicely with personality traits being changed by traumatic events and would explain why trauma therapy looks like "slowly rebuilding your internal model of the world". Also if event memories are tied to an operational internal model of the world, this would explain both why we all have no toddler memories and memory failures after trauma.
I'm more confused about depression. The link to low confidence sounds very reasonable, but as far as I can tell from the PP framework it's not really clear how high/low motivation works. I need to reread the "Guyenet on motivation" and "chamber of guf"-posts on SSC, but to me it looks as if this is mostly a personality trait / evolutionary preset parameter stored somewhere in the hyperprior (maybe not as a constant but as a function that reacts more or less to internal parameters... really, I'm confused).
I wonder if much of the positive effect, and positive affect, of exercise are mediated by just being more in your body. Exercise pushes you to feel your body, for sure. (Who is that in the back row saying "torture"?)
I have been reading about feedback control systems lately. This makes me think of a PID controller - would strongly weighting priors be akin to a more prominent integral function?
I think it's more complicated than a PID controller. Think of a system that generates a hierarchy of controllers on different levels and then chooses on the fly which one to use, discard or update. The hyperprior is the general hierarchical structure and set of update rules (eg if something goes wrong, how much adjustment should you do?)
Dante might have been commenting on this in Inferno Canto 10. The damned are chronologically far-sighted; they can remember things from the past and have a sense of what will happen in the future, but the present is a haze. This turns into a merciful feature once judgment day comes and the future ends, because they'll just sort of reach a steady state. That also sounds a lot like "curl into a ball" depression, where you just want to stay under the covers forever and shut out all stimulus.
One quibble though: I’m not sure the final haze is meant to be a merciful feature of Hell. Aren’t we meant to think that the denial of all knowledge will be a further punishment? The other sinners seem to be punished more after the world ends: consider how the heretics will have their tombs covered after the final judgment. Dante admired Aristotelian scholasticism, and one of Aristotle’s most famous doctrines is that all men by nature desire to know. On that score, it seems difficult to think that Dante thinks God is being merciful in this situation.
To the issue of why thinking about the trauma does not make it go away on its own. Stress seems to really ingrain memories in a way that normal memories are not. So equally as powerful is needed to dislodge the bad memory and EMDR is somehow simulating a high stress environment.
This makes me wonder about aphantasia and depression. I see some claims that aphantasia can become a thing after a depressive or psychotic episode or brain injury. But I wonder if it goes the other way too.
Also,
> your brain artificially down-weights their signal, marks them as suspicious and untrustworthy.
I don't know about suspicious, but you talked about "maggots" in your Muggeridge review, and that's how I generally feel in the mornings: whatever I happen to be thinking or feeling is getting stamped with big DISGUST labels.
A recent breakthrough for me was noticing that I feel like I have a highly tuned empathic sense that notices all the suffering and imperfection in the world, and noticing that it's odd that it doesn't also detect all the joy and triumph. It was quite overwhelming (in a good way) to even begin to contemplate that. So I've been paying more attention to the "Humans, Fuck Yeah!" genre, eg., https://space-australians.tumblr.com/
"That is, the first time you bungee jump you're probably pretty scared. But you bungee jump a few times, nothing bad happens to you, you learn on a preconscious level that it's not dangerous, and then you're fine with it." Adam Savage would disagree: in this Q&A video (https://youtu.be/-AvhYMPvPR4), he describes that it actually gets *worse* the second time, at least if you do multiple jumps in a row (ex. to shoot a TV show).
I can’t quite reconcile this argument with certain examples of over-weighting the sensory data going badly. I’m thinking of the stereotypical hypochondriac obsessing over any minor bodily sensation, or anxious people who live in California and are hypervigilant about any tremor that could be an earthquake. It seems that in those cases there’s both an incorrect negative prior and too much attention paid to sensory data. Perhaps because in these examples there’s no way to “disprove” the prior (there’s no sensation that proves one is completely healthy apart from maybe the passage of time without any other sensations) so the two directions are not in conflict with each other, and it’s another possible stable state?
Is that necessarily overweighting the sensory data? It seems like you could be hyperaware of every minor bodily sensation or tremor without attributing it to sickness or an earthquake. If you were updating on information correctly, what should happen is that you'd notice that you're picking up on all such minor signs all the time without them actually being associated with anything bad, so you'd continue to be aware of them but they wouldn't cause you to freak out.
I think what you're describing is actually a case of having a high prior on there being an earthquake or illness and that prior then "looking for" anything that would be compatible with that story, and then not updating on the sensory evidence that repeatedly says "no illness" or "no earthquake".
Wonderful food for thought, thanks a lot. So far, my prejudice about psychology is that they mostly correlate things with other things. And most of the time, questionnaires with other questionnaires. So it is refreshing to see people developing proper theories that might lead to improved therapy
Wonderful food for thought, even if the theory turns out to be wrong or incomplete. So far my prejudice about academic psychology (not psychotherapists!) is that they correlate stupid questionnaires most of the time. It's so refreshing to see that people actually think and develop theories and models that may even improve therapeutic work.
>> manic people say the world seems oversaturated, even brighter and more colorful than usual
As a person with unipolar mania, this effect is *awesome* (and overwhelming). I get so stimulated walking or biking about and seeing little bits of color pop out at me, like flowers or plants that I've never noticed before on a common route. It's like a low-level psychedelic trip.
Albert Hofmann, the Swiss chemist who first synthesized LSD, wrote a book called "LSD: My Problem Child". He recounts an incident where, on a hike as a child, the world suddenly appeared much more vibrant and alive, a religious experience, for reasons that were mysterious to him. He compared his experiences on LSD to that moment. The Magic Theater scene in "Steppenwolf", by Herrmann Hesse (written many years before Hofmann discovered LSD) is the best description I've ever run across of an LSD trip. Something about those German Swiss, I suppose.
I think you meant NMDA antagonists rather than NDMA antagonists. Though I can see why you'd have MDMA on the mind when typing about trauma; by all accounts MDMA-assisted psychotherapy is incredibly effective for PTSD.
I'd rephrase that bungee jumping example. Let's say I went bungee jumping *without proper safety procedures*. I was scared, but everything went OK. I did it a few more times, I learned on a preconscious level that bungee jumping without proper safety procedures is not dangerous, and I was fine with it. So was everyone else, so I felt quite normal. Then something bad happened, something that impacted me directly, something rare enough that most people haven't seen it, let alone had it happen to them or near them. All of a sudden I'm no longer fine with bungee jumping without proper safety procedures. I have a different idea in my head of what proper safety procedures are, a different idea than pretty much everyone else out there has. It's hard for me to relate to them: they all think I'm paranoid and worrying too much, and I think they're all ignoring truths that are inconvenient to them at the moment. "Sweet summer child" is a good start, but combine that with one of Lovecraft's protagonists who just barely survived and encounter with Something, and that gets closer to how I feel. There's only one friend who I can really be open with; she was one of the people who cut ties with me as I started going downhill, but a bit later she had an eye-opening experience of her own, and now we can actually have serious discussions about stuff, even though we don't always agree on everything. That more than anything else has made me feel as though I'm not completely insane.
To put it another way, what was Julius Caesar's prior for "being stabbed by a friend" 1 day before his death, and 1 millisecond before his death? Which was more accurate? Could he have saved his life if he'd changed his prior 1 day earlier? But would it also have had a negative impact on his ability to form friendships afterwards? If he saw a therapist 1 year later, would the therapist conclude that he worried too much, and should trust more easily? Would he have lived a happier, healthier life if he'd somehow recognized what would happen, taken steps to neutralize it, and then lived the rest of his life without that event having any impact on his personality?
All that said, EMDR is amazing, although I feel like it's borderline witchcraft. My operating theory is that, given how we apparently rewrite memory every time we access it (last I heard, anyway), EMDR is a way to access and rewrite PTSD memories without rewriting the full emotional power, leading toward a more normal emotional level associated with the PTSD memories. But I try not to think too much about it, in case it makes the EMDR stop working. It feels a bit sacrilegious to consciously avoid thinking about something, but this is by far the best chance I've seen at my ever having a relatively normal life again, and I don't want to mess it up. :-/ My main complaint is that it's best done in-person, and I'd just started to make progress when covid-19 hit. So no more EMDR for me for a while, and even if my therapist gets trained for online EMDR, it doesn't seem to work as well for cases like mine.
Also, something you didn't mention, and something I hadn't heard of: part of my EMDR has involved my therapist asking me to pay attention to where I'm feeling stress in my body, after I revisit traumatic memories. And it does happen, I usally feel tense and sore somewhere, like in my shoulders, or neck, or back, or head, usually spots around 1" in diameter. So that may connect with the somatic therapy angle.
I also wonder a bit about the back-and-forth eye tracking of the fingers, and whether that has any relation to the pendulum used in classic hypnosis. Not that I feel particularly hypnotized, but would I know? :-)
And I realize here that I'm conflating some things. There's the "flashback"-like dump of emotions that I get in response to certain stimulus, which causes several seconds of highly distorted thinking, and then has a diminishing effect on my mind for at least half an hour. And then there's the set of beliefs about the world which I think follow fairly logically from events, but which are also reinforced by the "flashbacks", and which are closely-enough associated with the stimuli for the "flashbacks" that it's hard to examine them too closely without triggering more "flashbacks".
Perhaps a comparison would be between a vegetarian who doesn't eat meat, and feels a bit sad when meat-eating is going on, and will share their philosophy when asked, and a vegetarian who gets a rush of intense negative emotions every time they're around meat, to the point where they can only shop for food in vegetarian-only grocery stores, and hearing co-workers talk about a barbecue can make them unproductive for an entire day. Maybe if the PTSD is fixed (to whatever extent possible), the second one will stop being a vegetarian and start enjoying bacon again. Or maybe they'll just be like the first one.
In that account, Daniel Ingram gives the example of someone walking on a street where they're completely safe, when they suddenly recall the memory of something upsetting. The memory then keeps making them feel more and more upset, as it causes stress and that stress response keeps making the upsetting memory more salient, and so on.
Whereas for someone enlightened, they retain such a strong sense of awareness of their current environment that memories of past events just can't get any hold. One can still remember them and think about them, but one is also constantly aware of the fact that they are just memories and not a sign of something in the environment that's actually threatening, so they cannot go low-grade loopy and keep retriggering the same upset. In your terms, the sensory data keeps being high-precision enough to outweigh the priors in past memories.
I think the way to square this circle is by considering 'priors' to be some kind of global statistic that gets updated and "memories" as an active recall, i.e. indirectly experiencing them through something like 'conscious awareness'. The priors aren't memories – they're a 'cached' calculation, even if, in a very real sense, both they and 'memories' (actively recalled) are derived from the same original 'sensory data'.
I wonder if neurologically, priors aren't more about the brain's chemical balance ? (And maybe somewhat about large-scale neuronal architecture ?)
While memories are about the low-scale physical architecture of neuronal links, and memory recall is electronic, but obviously affected by chemical balance.
Maybe the 'weight' given to priors is encoded in "the brain's chemical balance" but my naïve understanding is that there's no way to encode priors in it exclusively.
I haven't read up on the details of this kind of stuff in a long time so I don't really know that much (or can't recall any details I once 'knew').
In a general sense, both priors and 'memory' are both memory in the sense of information. But just based on introspection and other's peoples reports about what we're calling here 'priors' and 'memory', accessing priors must be, effectively, a much cheaper computation to perform, i.e. 'cached' calculations. 'Memory' seems to, often, involve much more involved 'active recall', which seems (to me, and apparently many other people) something more like 'replaying sensory data'. I _believe_ that there's strong evidence that that's a fairly accurate summary of what happens neurologically, e.g. the same parts of the brain that process sensory data in 'real time' are also used during 'memory recall'. We very much do in a sense 'relive' our memories.
That's a good warning to repeat and I try to remain cognizant of it in these kinds of discussions!
But I generally use _computation_ metaphors, which, admittedly, have the weakness that they're not always easily distinguishable from 'computer metaphors', given that they share many of the same technical terms. You're right that brains are _very_ different than the architecture of what most people understand 'computers' to mean.
But from both a mathematical and physics perspective, the brain is just another 'computer', i.e. it must have some mechanism for storing information, retrieving it, and performing computations on that information. From that perspective (especially the mathematical side), the 'implementation details' are 'trivial'. Obviously, the implementation details are actually interesting and important, but they also don't fundamentally change the relevant 'computer science' of how the brain _has_ to work, albeit in an extremely abstract (and impractical) sense.
Was funny, but it made me notice that I don't think I'd ever thought of the 'awareness' usually associated with meditation and some drugs as literal sensory awareness before. Apparently I'd biased that word towards awareness of not-actually-physically-there sensations (mystical, hallucinatory, or metaphysical).
I wonder if that insight might help one to approach it in a more effective way.
(And then I remember I'm prone to things like health anxiety, making the prospect of trying to feel things *more* seem analogous to reaching out to touch a stove that might be very hot...)
It can be both. Extreme awareness of what you are looking at can also help you see small distortions in it. Think about looking through a screened window casually versus looking through a screened window very precisely at a pattern on the other side. At a casual glance, the screen just becomes a grayish blur over the image, but a precise glance lets you see the shape and size of the screen cells.
- (Somewhat improperly called ?) 'Mindfulness meditation' where you're trying to be as aware of your 'corporeal' senses as possible, suppressing everything else.
- More 'traditional meditation', where, on the contrary, you're trying to suppress your 'corporeal' senses, but *also* your memories (which Scott puts into the 'senses' bin), and even your thoughts, so I guess what is left to focus on are the processes by which the brain itself works ?
I first learned of the Sniffin' Sticks Test when I saw it in a paper investigating anosmia in COVID-19 patients. I thought it was just something they made up for the study but no, apparently it's a common benchmark of olfaction.
Anyone have the actual best references for the color perception - depression link? This post just links to an older post, which links to a paper by Chekroud, a much more general review. I ask because this seems like the sort of embodiment study not to survive the replication crisis, and I'd be interested to see what sort of control groups and measures they used.
Relatedly, I looked into the attention bias literature (anxiety, OCD) a while back, and it's not very solid (much p hacking, little evidence that biases play a causal role).
Quite interesting though i wouldn't lean on EMDR too much: not just because all psychotherapies are kind of equally effective (wampold, 2015) but it's been shown that EMDR doesn't require eye movements at all, and it's not particularly extra-effective in controlled trials. In my opinion, is enough with any kind of attention splitting process so you get old fashioned exposure plus interference with your emotional arousal. In any case, in my opinion any psychotherapy involves some degree of exposure (+ acceptance?) of felt/sensed disturbance plus s certain reprocessing of it (and priors as well) towards new narratives. Just an idea.
Congrats from Spain on extremely amazing blog and smart commentators.
How do you reconcile this with the fact that the most effective and best supported trauma therapies (Prolonged Exposure and Cognitive Processing Therapy) both involve revisiting traumatic memories over and over in excruciating detail?
The literature to date suggests EMDR works independent of the eye movements, i.e. it is the exposure piece that is doing the work. It is probably a mistake to spend too much time trying to reconcile theories with van der Kolk's stuff; the years have not been kind in finding supporting evidence for it and the trauma researchers do not take it very seriously. At the very least, his work is not some well-established, gold standard, obviously true framework.
These therapies, by taking place in the safest possible environment, would be either slowly pushing the prior to be more favorable and/or give more weight to the immediate senses (rather than memory senses) ?
Interesting. I wonder if this is part of why people with migraines are at such high risk for major depression. The risk goes up about 3.2x the general population once you get a migraine diagnosis:
I have chronic migraines that cause a lot of sensory/nervous system hyperactivity, so I literally can't trust most of my senses most of the time. Obviously dealing with the pain and disability that comes with migraines can't help here, but it seems like over time have actually unreliable sensory data would take it's toll. Interestingly, the new CGRP blockers are the only medications I've taken that get rid of the weird sensory issues I get when I get a migraine. All the other medications just address the pain part. Massage helped, though I could never quite figure out why. I think it helped put me back in touch with my body more effectively.
This discussion may be adjacent to some of the discussions surrounding predictive coding and fatigue in exercise.
This discussion paper frames fatigue in exercise as a response to a scenario where sensory information cannot be trusted: https://psyarxiv.com/nqxz2/
The argument is essentially that, when exercising at high intensity, the allostatic mechanisms that regulate various systems up and down the stack starting making bad predictions and fail to keep various parameters (heart rate, blood CO2 levels, contractile potential, etc.) within acceptable bounds.
So, the organism experiences "fatigue" as a way of returning to a situation in which sensory information can be trusted — aka "not exercising really hard."
In the case of depression, it seems that the low trust of sensory information could create a similar state where the organism looks to "shelter and rest" in order to return to a situation in which sensory information regains trustworthiness.
As mentioned, it seems like the depression attractor state could potentially be found through some combination of all of these phenomena.
t seems to me that depression starts with a lack of control in the quality of thoughts and then establishes itself as a phenomenon in brain chemistry. Thus, after a certain stage, the neurotransmitters are compromised and chemical treatment becomes necessary. However, if we have the maturity to realize that everything starts with the quality of thoughts and we are able to maintain good mental hygiene, depressive states can be preventively prevented
To the question of perceived -6 event still being better than an expected -9 event and yet not resulting in proper updating: It’s possible that being depressed imposes an additional -3 penalty on every experience simply because being depressed is itself actively unpleasant. Sort of like how my brain doesn’t automatically subtract away the pain of a migraine when judging whether I enjoyed a movie while suffering through one; the migraine was part of the experience, and so the movie seems worse. Depression is painlike, in addition to all the other stuff. In this case, the depression is part of the experience, and I think it sucks even worse than the sensory dulling effect accounts for. When you add up the sensory dulling, the update downweighting, the intrinsic suckiness of your experiential frame, and the low prior, it’s possible that you could still manage to undershoot your unrealistically low expectations.
I am not sure we can separate 'priors' and 'evidence' so neatly. Perception is actively constructed by the brain so our here-and-now perception ('evidence') is shaped by priors as well. Ambiguous cues will be interpreted according to our priors. Occasionally, the data will be both less ambiguous and incongruous with our priors but misperceptions are fairly common experiences, so it is not surprising that when our experience is incongruous with our expectations, we are at least sceptical that things are what they appear to be. This scepticism is probably fairly adaptive a lot of the time: "if the deal seems too good to be true, then it probably is"-kind-of-thing.
Haven’t read the paper yet to know how this might work with positive emotion (i.e. mania), but it seems like there might be a handy explanation here for how meditation leads to awakening/enlightenment/etc. I mean, if meditation in those with depression or PTSD can help correct the underweighting of the sensory channel, couldn’t it cause overweighting in normal people? And, as I assume the paper suggests with mania, wouldn’t overweighting the sensory channel make you feel pretty good assuming your life isn’t absolute garbage? I.e that decent coffee you have every day tastes really good (which is something “enlightened” people have actually described to me). And might those experiences be self reinforcing in the same way as negative ones are with depression? And, if you’re overweighting the sensory channel, couldn’t you convince yourself of some pretty odd beliefs that might correspond with the bare facts of the sensory channel, i.e. that’s there’s really no concrete sense of self to be found and thus the Buddhist conception of noself?
Thank you. You article made me recall & reassess my personal experience of (mild) depression. Hope it will add on to your observations. What you are describing, seems familiar at least.
I experienced sensor issues, as in - less colour, less taste in food, less enjoynment of life.
I experienced negative perception in work, communications with people, or simply reading or watching films.
Overall, it feels like all cells in your body tell you "i want to die" instead "I want to live".
What I understood over time, and what it took to get back to healthy:
1. Self-perception. I am a very happy and vivid person overall, so it took me months to realize smth is not okay, as my self-perception and reality do not match anymore. Tiredness, eager to be upset over anything, conflicts, irritability etc.
It`s hard to accept that it`s a problem. You keep thinking, its tiredness, and it will pass soon. I started having more walks, eating fruits, taking D3 etc after a while, which helped too. But those phisical habits are hard to get back to. It`s not that you do not understand it`s important. You just don`t have any energy, and you have to retrain yourself back to it for months. This is I think one of the reasons why depressed people get so angry over "take a walk & eat chocolate" advice.
2. Psychodinamics, hormons, neural activity changes drastically. You constantly feel everything is not working, you gonna loose, you gonna fail. It takes a huge effort to set up a self-feedback loop, to separate what you can reasonably assess is true, and your impressions about it.
Eventually I ended up spending 1-2h a day every morning for 1 year, just talking to myself, remembering my usual life perception & logic, reactions to similar situations before, reasoning over scenarios & things to do, and distancing within my mind from the constant negative "we all gonna die" feeling. It`s what I think you describe on ranks. I reranked myself basically.
3. Broken external feedback loop. When depressed, you react differently, and you provoke slightly different reactions from people or even attract similar more often. You keep getting negative signals not just within, but from outside, and at certain point it entangles so much you do not know who you are and if you ever going to be okay and if you can trust people.
Same as before - I tried to distant myself from my state, so to speak, to see where my reactions are more or less reasonable, and where it`s a negative spark. And then go out to talk to people, do some work etc - while contiously holding this framework, that I might overreact or take things more negative tnan they are. And read feedback from that perspective. Plus openly ask questions - to the point of being blunt, "did you mean to offend me or i misunderstood you entirely?". Again, retraining myself.
4. Dependance. You grow dependant on emotions, cos they are rare. Whatever makes you feel is important (90% not valid or true). It`s extremely unhealthy and shakes the psycho. It takes huge effort & discipline to stop reacting, to play down emotions, concentrate on how you feel, and think, and reason. This part of the equation is a major destabilizer, and less control you have over yourself, longer you can`t handle the whole thing. That was one of the reasons I stayed away from any farma; transfering my dependancy to chemicals didnt seem like a good idea.
5. Being stupid and in the i do not know are the most wonderous states of mind, but they are the hardest to get back to. Experience and knowledge becomes a problem & limitation. Nomally you learn and re-learn easily, and it`s a part of - i think - just being curious and open. In depression you grab ahold to what you know, and try to apply it to a changed situation. Basically, everything that you know is limiting you from getting better, while kinda protecting from being hurt further. Reconstructing, shifting the assemblage point seems like an evident way through, but it`s pretty much impossible to pull off. I attempted that after about a year of just reframing myself towards less negative view of the world.
6. Self-destructive & destructive relationships. Really hard to fix while in depression. Feedback loop makes all worse many times over and toxic relations become unrecognizable. I eventually chose solitude, and got back to being around pretty much anyone except close family after a few months or so. Cleaned out about a half of my networks afterwards too.
Strangely, afterwards I realized that I had a sense of happiness within depression, tho somewhat hidden. At the very same time I remember a feeling, that I will always be that sad and gray, and my life won`t go back to mornal - which I somehow accepted.
I started to see people who are not doing okay - I could now recognize face muscles, changed mimics, changed reactions, slight differences in behaviour - things I did not see or recognize before. It really helped to connect to friends and family better, care of them better.
Whole experience was powerfull, cos it teaches to take in all bad and all good, to be perceptive over a wide range of feelings and emotions, but not be ruled by those as much. Take life in, but not be destroyed by it. Be able to sit in the dark and not be scared of it. Be present for people around. Well, those kinds of things.
Thanks for your articles, btw. I genuinely enjoy the way you think.
First, if "quieter" sensory input leads a pessimist into ever greater depression via the proposed feedback loop, shouldn't it also lead an optimist to euphoria? How would we know if there were lots of people made super-happy people by the same sensory deficit?
Second, if a preference for abstract thought over sense experience caused depression--without the balancing euphoria I just proposed--wouldn't we find that, historically, rationalists (say, philosophers and theologians) had a strong tendency to be depressed, while empiricists (say, modern scientists) had very little depression?
I'm still pretty confused by the numerical dating example. I kept trying to think of solutions, but they all came back to "the brain uses coarse-grained binning for good/neutral/bad events when reinforcing hyperpriors". Anybody has something better than that?
If stress from subordinated SEC status degrades the frontal cortex, then your ability to reassess your priors is damaged. Feeling able to appreciate the beauty or what-not, again, comes with not being shat-upon on a daily basis. Seems obvious to me what the limitations of this “epiphany” are.
Is part of the problem a failure of actually update priors? People in general aren't actually very good Bayesians - and there are perhaps good evolutionary reasons. No matter how many times the thing you saw out of the corner of your eye turns out to be a tree stump and not a snake, you still want to hold a decent prior that it might be a snake. So if your date was a -6 not a -9 as per your prior, you might not actually update your prior, because, you never know, snakes!
I'm not a scientist but anecdotally speaking ketamine therapy saved my life. I can't say I was ever clinically depressed but I was incredibly anxious, prone to panic attacks, and had horribly low self esteem due to some negative childhood experiences. Even though I was professionally successful and had a vibrant social life I felt worthless, lonely, and unlovable. After a few months of ketamine infusions it was like my brain could suddenly let go of those negative childhood memories and accept the evidence that, at this present moment, I am a decent person with many people who care about me. I honestly wouldn't have believed it until it happened to me but it does feel like the ketamine rewired my brain to stop dwelling on prior negative experiences to the exclusion of sensory evidence. I also had a strong fear of driving due to a friend being in a terrible car accident when I was a teenager - I didn't drive a car for over 10 years. The idea of driving was inconceivable. Even riding in a car as a passenger would give me a panic attack 50% of the time. A few months after completing the ketamine theory something shifted in me and suddenly driving didn't seem so scary. Now I own a car again and drive on a daily basis and it's like my entire world has opened up. I have no idea how this happened except that I suddenly stopped caring about that one horrible memory that made driving seem impossibly dangerous and started letting myself be swayed by the fact that most people I know drive on a daily basis and are perfectly fine.
Anyway, fascinating stuff. I'm a true believer in ketamine therapy now and I look forward to seeing where it goes in the future. It feels like my brain was a train stuck hurtling down a very old, narrow, and broken track and ketamine turned it into a car capable of changing lanes, making sharp turns, and backing up when necessary.
The people I know who tend toward depression also have a narrowed bandwidth of memory of prior experience, too. But I really would characterize it as narrowed, not merely negative. They don't remember very positive experiences but they also don't seem to remember very negative, either. It's like everything gets compressed into this one narrow stratum of negative experience and anticipating and feeling at a certain level X on the emotional dial, and everything above or below gets folded into that channel. Like the brain is doing an FFT back and forth with only 64 points.
This accords well with my experiences with meditation. I find that, occasionally, I'm able to experience my thoughts not as an inner monologue, but as a phenomenon that I'm experiencing, independent of my consciousness. Those rare intervals coincide with the largest improvements in my mood and ability to deal with depression and anxiety.
>You update your prior some bit of the way between -9 and -6. Do this enough times, and you should be able to get your prior all the way back to 0, ie eliminate your depression. Right?
I can't comment on the medicinal aspects of this, but here are some mathematical thoughts.
The odd-ratio version of Bayes theorem is, for B meaning "the world is bad", ¬B meaning "not B" (ie the world is good), and G being the evidence of a good date (or a good anything):
P(B|G)/P(¬B|G) = P(G|B)/P(G|¬B) . P(B)/P(¬B).
What does it mean to have sticky priors? The prior odds are P(B)/P(¬B); the updated odds are P(B|G)/P(¬B|G). The term that updates is thus P(G|B)/P(G|¬B). Sticky priors means that that ratio is closer to 1 than it should be.
So what is P(G|B)/P(G|¬B) ? It's the ratio of P(G|B) (how likely is a good date given a bad world?) to P(G|¬B) (how likely is a good date given a not-bad world?)
So even if the depressed person feels "that kinda-good date is unlikely given a bad world", this will not update their priors if they also feel "that kinda-good date is unlikely given a not-bad world".
So there are two phenomena that can be in play: P(G|B) can be high (kinda-good dates are likely, even in a bad world - here we get all the stuff that says the date is not that good, or that good dates will lead to heartbreak anyway, and so on).
Or P(G|¬B) can be low - kinda-good dates are not likely, even in a good world. This could be unreasonable expectations; "if the world was good, this date would be much better".
The same result holds in the standard Bayes theorem, of course (the odds-ratio version and the standard version are equivalent). But I think it's easier to see in the odd-ratio version.
On a person note, as a less-depressed person than once I was, I can anecdotally confirm the poor sense of smell and the reduced autobiographical memory.
So I think you're saying that people could have biased estimations of the likelihood ratio, such that all likelihood ratios are closer to 1 than they should be?
My reasoning: if depressed people have a high posterior for p(the world is bad), which they would if they had a high prior for it and were slow to update, then the depressed person will have a negative view about the future. What are my chances of having a good date, given how bad the world is? p(good date | bad world) = low.
Right? I feel like this is highly related to the premise of the post, although the post focuses more on unconscious predictions that underlie all brain computation.
So, the premise is that the brain *accurately* predicts likelihood (the technical term, not the layman's term). In the above example, the brain predicts that p(G|B) is low.
But you're suggesting that the brain can't accurately predict likelihood (or at least, likelihood ratios). If the brain doesn't accurately predict likelihood, the depressed person wouldn't have a problem. They would think "the world is bad but it's *totally possible* to have great dates anyway". And that's... not what happens.
Or, if you think the brain can predict likelihood but specifically gets likelihood *ratios* wrong... I dunno, do we think the brain models straight to the ratio when doing bayesian updates? Or does it do the equivalent of modeling the two likelihoods and dividing them?
...I agree that the former would have the same result as the math I linked to earlier - unbiased parameter estimates that converge slowly.
>So I think you're saying that people could have biased estimations of the likelihood ratio, such that all likelihood ratios are closer to 1 than they should be?
Maybe, maybe not. I was playing with mathematics to find a reason why the update would not happen. The math here suggests one possibility; that can translate into something in psychology that might be worth investigating. Since it comes directly from the math, it would be fascinating... if true.
There are many other reasons the update maybe wouldn't happen; this is just one possibility among many.
I fit this model, and it sounds quite useful. There is a difficulty in assessing evidence and being able to control the perceived danger. The danger is the thrashing of internal demons. Since being wrong on the assessment creates incredible emotional pain, the tendency is to not take anything of value from the experience.
When Scott is describing how the brain supposedly assesses an event based on priors and evidence he is assuming that the event is assessed once with priors and evidence acting as weighted multipliers.
There's another way this could happen. The brain could be performing an additive operation every time either a prior or sensory evidence enters consciousness. The level of consciousness is what assigns lower weights for sensory evidence.
I think his objection is somewhat silly on the part of the priors. Why would a prior be assessed only once? And only after the date? If the priors are only assessed in expectation of something they should be separate from the sensory evidence of the actual date.
If we see priors and sensory evidence acting on emotionality on this additive, separate way, we can explain the detrimental effects of rumination. Scott's proposal of a multiplier would always have rumination ending up at level 0 emotipnality.
So how does the memory loss fit into all this? If depression causes one to overweigh priors vs senses, why does this result in the experience that formed those priors being erased?
What makes you think that it has been erased, compared to has never been strongly fixed in the first place or being just hard to access at this point in time ?
Hi Scott, have you come across the Reconsolidation of Traumatic Memories (RTM) technique from Richard Gray for curing PTSD? If the website is to be believed, what research that has been done - three trials with @30 patients - they average 85-95 diagnosis elimination, far exceeding the efficacy of emdr etc.
> ...focus on trying to get the brain to pay unusually precise attention, under unusually safe conditions, to something it usually represses...
This is psychoanalysis! The free associative technique produces repressed contents, and the presence of the analyst makes the analysand feel safe, plus their desire to "please" the analyst (transference), beyond a mere desire to "get well", produces more motivation to actually uncover repressed contents.
And Wilhelm Reich in the 1920s tried to tie this practice into the body, as per the second idea, with, uh, questionable results.
Happy to have stumbled upon this article! correlates with my revisiting Bayes theorem yesterday (evidence does not determine beliefs; evidence updates beliefs, i.e. the evidence and hypothesis are weighed against memory or current beliefs) as well as SEP's entry on Karl Popper. Neither of these exactly relate to trauma, but their critical approaches: precise results (Bayes), and real, falsifiable theories (Popper) may contribute something to the problem of dealing with trauma.
I quote the relevant passage *"[Popper] argues that the Baconian/Newtonian insistence on the primacy of ‘pure’ observation, as the initial step in the formation of theories, is completely misguided: all observation is selective and theory-laden—there are no pure or theory-free observations." If this -- that there must be a 'selective and theory-laden' approach to observation -- is applied to trauma, would the problem not expand/explode one hundred fold? I ask the question on the assumption that theories of human nature are wanting in precision, repeatability...
Also, I disagree with Sniffboy. The link between low confidence and negative prior seems to me about as strong as the one between high confidence and positive prior.
Thank you for opening discussion.
*I only cite Popper because I don't yet fully understand Bayes theorem.
On the part you were a little bit resistant about, could you posit that the brain doesn't actually ever compare the experience to the prior in the step where it updates its new priors? Then you could have: Prior of -9, 'objectively' neutral experience: 0, but weighted between 0 and -9, let's say -3. In the step of updating priors, instead of comparing the -9 expectation with the -3 reality, it might simply record '-3' : bad experience, revise priors downwards. This would awkwardly approximate Bayesian reasoning, which might be the best our brains can mange. It would also be consistent with human brains being generally quite innumerate.
I really like Scotts take on borderline personality disorder. How does ”to loose priors” in that context align with this model of ”to rigod priors” in this context?
If most cases of depression can resolve themselves by averaging out the difference between overly pessimistic priors and data (your -9 expectation + 0 experience = -6 perception), then it follows the cure for most depressions is to have as many experiences as possible to dilute the bad priors. Go out and ovewhelm the system with new inputs that force an update.
Incidentally, that's precisely what depression prevents you from doing.
One simplistic summary: depression, anxiety, and ptsd represent an over-reliance on priors in their constant sense-making negotiation with evidence. The antidote is to bolster our sensitivity to evidence, and decrease the power of the priors.
Framed in that way, this sounds mighty close to the overall project of rationalism. Because the techniques here are relational and embodied, they celebrate forms of evidence that rationalism doesn't usually focus on, but the project is the same. Are there clues here to what a more embodied rationalism might look like?
In contrast, a dogmatic community will over-rely on their collective priors and assign unusually low precision to empirical evidence. Can we interpret that as a kind of collective trauma/depression/anxiety, even if the individuals themselves don't have those conditions?
Late to the party but wonder how (if?) this relates to sensory deprivation tanks... some kind of reset or a variable removal which can help...? Or maybe the whole thing is bullshit :)
Very interesting! This seems to be missing an important piece, though. I would expect universal low sensory precision to translate to impulsivity or hyperactivity, whereas in fact in depression people seem to be more lethargic and it's harder rather than easier to accomplish movement.
As per https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/ IV.5, movement is based on prediction of action. When your prediction mismatches the world, either action happens to bring the world into accordance with prediction, or prediction updates to match the world. Low precision sensory data would mean that it gets thrown out more often in favor of the prediction, i.e., it biases the system towards acting on any impulse/prediction/whim that arises. This also explains hyperactivity in fruit flies with ADHD (https://www.ted.com/talks/david_anderson_your_brain_is_more_than_a_bag_of_chemicals?language=en), albeit with a different mechanism (decreased signaling of surprisal rather than lower sensory precision).
(Also, the theory put forth by that link might explain low-functioning autism, but at best is exactly backwards for high-functioning autism. As described in https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/, the model of autism as overly narrow confidence intervals on predictions fits much, much better, IMO, than the model of generic failure of prediction.)
> so the system that should be habituating to the trauma is actually "learning" to become more and more afraid of it.
I'm suspicious of this. IIRC, Body Keeps the Score says that traumatic memories of war vets with PTSD were virtually unchanged ~40 years later, whereas the same memories of war vets without PTSD underwent the same sort of change and drift that normal memories do. I think in fact the "this is scary" response to re-experiencing the memory counts as sensory experience and is thus underweighted in the same way that normal experience is---the memory remains unchanged, rather than becoming either more or less scary.
Chatting with my friend Rivka, another component came up that seems quite related, though I'm not yet sure how it fits in. She told me that people with trauma tend to feel less in control of the world and their bodies. Since, IIRC, learning is best done interactively, when you can both observe the world and see the effects of the attempts you make to control it, it seems like half of this process is missing in trauma, where you feel like you can't control the world. It seems like this should somehow connect with the trouble updating on sensory experience, though I don't yet see the gears of how it fits in.
"This is the most exciting paper I've read so far this year and an important addition to my understanding of predictive processing and psychiatry in general."
Writing this as the *last sentence* is some pretty serious lede burying, Scott. Maybe put this sort of thing in the first paragraph.
"In reality they average 0, ie completely neutral. You're depressed, so you have a prior that they'll average -9, ie quite bad indeed. And you assess things based 66% on your prior, 33% on evidence. You go on an average date, ie 0. Given your prior of -9 and your weighting, you perceive it as having a valence of -6. You expected things to be -9 bad, but they were really only -6 bad; it seems you were overly pessimistic. You update your prior some bit of the way between -9 and -6. Do this enough times, and you should be able to get your prior all the way back to 0, ie eliminate your depression. Right?"
Maybe it is, but how many people who predict -9 dates go on dates? When my depression is at its worst, I definitely predict terrible outcomes to things that normally are at least a little positive. Or if I intellectually know they're usually positive, I'll hyperfocus on some small part that could be hard or negative, and be convinced that that part will overshadow any good things. Whether or not people tend to push through (and thus get re-calibrated) could be a major factor for which attractor states they tend towards.
What makes this theory particularly unintuitive (whether coming in with knowledge of Bayes or not) and hard is that 'memory' gets placed in the 'new experience' bin, rather than the 'prior experience' bin !
But I guess that these 'priors' are still a very specific kind of memory, in the 'architecture of the brain' sense ?
We're getting here in the hard (but fuzzy, it's *philosophically* hard) frontier between 'mind' and 'body'...
"In theory you should be able to do the same with traumatic memories: you think about the traumatic event, but since the trauma's over and thoughts alone can't harm you, you're fine and you habituate to the trauma and don't mind thinking about it anymore."
Anecdotally, when an IED hit my truck in Afghanistan this was exactly what I did whenever there was down time, starting a day or two after medevac. I would relive the explosion, the adrenaline hits, after the memory-experience was over I would check my surroundings and verify to myself that I was in a hospital bed and everything was fine, and then back in for another round. I repeated the process until I didn't get the adrenaline spikes; might have taken a hundred iterations over the span of a week.
I dunno, the link between low confidence and negative prior still seems to be missing here to a substanial extent. Like, where's the corresponding problem of low confidence, positive prior? It wouldn't seem to be mania from what you suggest. So what's going on here?
This was my immediate line of thinking too. The new idea here doesn’t seem to really be a synthesis of the two ideas at the start of the post any further beyond saying “depression is what you get when both these are true”. Mania being the corner of the box with low confidence, positive priors seems to fit really well with the theory.
Not to go evolutionary-psych here, but it seems like there are situations where it’s reasonable to be “normal” (a low-threat environment) and there are situations where it’s reasonable to be “low confidence, negative prior” (a high-threat environment), but I can’t think of any situations where it’s reasonable to be “low confidence, positive prior.” Generally being wrong about things in the negative direction will help you, while being wrong in the positive direction will hurt you. So it sounds plausible that the two variables would have a strong relationship.
Of course, being “low confidence, negative prior” all the time is really not good in our modern environment, but evolution wasn’t optimizing for that.
My personal experience of PTSD is that it's what happens after you "get got", but somehow survive. And you notice that no one else around you is adjusting their priors at all. And they don't even want to acknowledge you exist, because that might lead them to start asking questions about what the actual frequency of tigers-in-bushes is, and that might interfere with them "getting theirs". And every time you see a bush, your brain malfunctions and dumps a load of memories and emotions into your mind; not the tiger jumping from the bush, but instead, your friends running away and abandoning you.
I think 'low confidence, positive prior' is really useful if you're someone powerful. It allows you to wield your power to your own advantage recklessly, without being concerned about hurting people or facing consequences, because you don't notice that you're hurting them and your power insulates you from consequences anyway.
I feel like I've seen this personality style in a lot of successful C-level managers and entrepreneurs I've known over the years, and I feel like it matches how such people are often depicted. Of course, when those depictions are framed by neurotics or normal people, they often depict those powerful people as *not caring* about the people they hurt; but I think it's much more accurate, in most cases, that they simply don't notice.
Which is what would happen with 'low confidence positive prior'.
How is this behavior not "high confidence, positive prior"? Low confidence would be more likely to make you pay attention to how you're actually affecting people. It would reduce the importance of the positive prior.
I confess I'm a little confused about the math. In Bayes' Theorem (the basis of predictive coding) there's no coefficient you can apply to determine the importance of evidence. The posterior is purely a function of the prior and the plausibility of the evidence under different hypotheses. But I'm imagining that in this conception of depression or mania, you would, let's say, calculate the posterior P(H|E) in the normal way and then arbitrarily decide that P(H|E,depression) = P(H) + a*(P(H|E)-P(H)), where a < 1 if you're depressed and a >1 if you're feeling powerful/reckless.
I think this is basically equivalent to using "confidence" in its everyday sense. Am I wrong?
(I think I reversed the < and >. Oops.)
The prior *is* the term that determines the importance of evidence: the stronger your prior is, the less your belief is shifted by a given amount of evidence.
I think what your model is broadly correct there. 'Low confidence' being a < 1. So positive prior and low confidence would update priors less than an ideal reasoner, this matches somebody not noticing that you are affecting people negatively, they don't update their positive prior downward a sufficiently large amount on seeing evidence for it
"Usually cheerful but kind of unstable person"? "The kind of schizophrenic who hears nice voices"?
It may just be that our current world is sufficiently varied and chaotic that low prior weight is less damaging than high prior weight. It's the "move fast and break things" of mental models, while mania is the Theranos investor of mental models and depression is the "never start a business at all". Maybe the modern world just rewards agility in a way that screws people with strong priors except in the rare cases those priors lead them in exactly the right direction.
If you accept that there are many more way for things to go wrong than go right (Anna Karenina principle) I don't see any problem linking low confidence and negative prior.
Having low confidence has the effect of bringing probabilities closer to 50/50, so if each of the small things that can go wrong has a high chance of going right this can rapidly compounding effects. Say that 10 things need to go right with 95% prior probability, that means the chance of things going right overall is 0.95^10 = ~3/5. But if because of low confidence the perceived prior goes to 90%, that means the overall chance goes to ~1/3. The low confidence got transformed into a negative prior.
High impulsiveness perhaps? narcissism? can be the sort of person that doesn't get negative clues. thinks everything is great, everything will go just right, and everybody loves him, even when that's not the case - but also doesn't get disappointed when it's not the case because they don't really process that. they just don't notice negativity.
It sure sounds like a rare description, but i think the Ev-Psy explanation someone else gave in this thread explains fairly well why this one would rarer.
Low confidence, positive prior is childhood.
That doesn't fit in how children are fast learners ? If anything, children have much higher confidence than adults, it's just that their memory has a lot less weight compared to new evidence ? (Though maybe 'low memory content' is explanation enough, so they have similar confidence to adults ?)
I don't think "confidence in neural predictions" is part of the theory anymore (i.e. the theory espoused by this post).
Mathematically, I interpret "low confidence" as "the brain uses Bayes' Theorem correctly, but then pushes all probabilities toward uncertainty" (which, for a binary problem, means pushing all probabilities toward .5). This is *not* about having heavier weights on priors or on evidence.
The theory espoused in this post (I think) is that priors are weighted more heavily than they should be. So if the "correct" weighting is given by Bayes' theorem:
p(H|D) = p(H) * p(D|H)/p(D)
...then a depressed person is using something like:
p(H|D, depression) = p(H|D) + a(p(H|D) - p(H))
Note: this is the same as what dogiv wrote in their comment.
The difference is that "low confidence" as defined above would result in being less effective at everything, because all of your predictions would be biased towards uncertainty. Your parameter estimates are biased in expectation, meaning you don't converge on truth.
In contrast, a heavy weight for priors would result in requiring more data to learn (i.e. all priors update slower than they should). Your parameter estimates are unbiased in expectation (you do converge on truth).
A person with a heavy prior (meaning "highly-weighted prior") will make good predictions when their prior is close to objective truth. If I have a heavy prior that the world is bad, and the world *is* bad, then my predictions will be great. However, a "low confidence" person with a prior that the world is bad will still make somewhat-bad predictions, because their low confidence pushes them towards uncertainty about whether the world is good or bad.
(Lastly, a person with "light priors" would have high-variance, unbiased predictions. They converge on truth but their predictions are always swingy. Does this describe a known psychiatric abnormality?)
On a separate but related point, Scott writes:
"Looking over the date after you get home, you hyperfocus on the one really embarrassing thing you said, and any good parts of it mysteriously slip your mind."
Humans selectively attend to evidence (I imagine this is doubly true for manic/depressed humans). Scott doesn't explicitly say this, but it's important to call out, because he's assigning some significance to this fact here. This isn't about a neural implementation of Bayes' theorem that underweights evidence. This is about ignoring evidence altogether.
People look for evidence that fits their worldviews. This happens at a level of processing that we can be consciously aware of (or at least catch ourselves doing, if we reflect upon it). So if the "heavy priors" theory is true and explains the unconscious phenomena mentioned here, I think there are also higher-level things going on that contribute to someone being stuck at a certain set point for p(the world is bad). This would help explain why priors are not inevitably updated toward reality after enough data is gathered.
How do you differentiate "ignoring evidence" from "really low weight for evidence / really high weight for priors" ?
Thanks for asking that, because I'm not entirely sure! My thoughts:
The down-weighting of evidence that Scott talks about in this post is fundamental to the person's neural processing and has a variety of sensory, memory, and motor effects. AFAIK you can't "become aware of it" and thereby make it go away. You can't say "oh I'm just seeing the world in gray because of my silly brain down-weighting evidence. Now that I realize what's happening, I can see the world in color again."
The "ignoring evidence" that I was thinking of is at a higher level of cognition. It is something a person can consciously notice (e.g. "I'm interpreting this person's actions negatively because I am in a bad mood") and do something about ("I'm going to consciously think of positive interpretations in order to de-bias myself.") Down-weighting evidence is not fundamental to this person's neural processing. They probably see the world in color and have normal memory and ability to feel their body.
So the predictions differ:
A person with a depressed mood but who has otherwise normal neural processing should be able to regain a positive mood more easily than someone who downweights evidence, as long as they attend to evidence. This can be done using cognitive behavioral strategies.
These strategies would *also* help someone who down-weights evidence, but the improvement would be slower because even after consciously attending to evidence the person's brain is still down-weighting it at a level they don't have conscious access to.
Thoughts?
Conscious effort is known to be able to eventually, after a training period, affect subconscious behavior (as well as, with more or less effort, being able to override subconsciousness in the moment).
I wouldn't be so sure that ignoring evidence is happening on a purely (or even mostly!) conscious level, and that you can NOT train yourself to see the world more in color (see : 'enlightenment').
This also might explain why mania and creativity go together. If depression means you up-weight priors, then mania could mean that you down-weight priors. Your mind is more open.
I think the premise is too broad. If you start lumping things like memories and introspection and other cognition together with more basic sensory input such as vision or hearing, and call of it "senses", then it seems like the word "senses" might as well be synonymous with "anything your brain does".
To the extent that a prior can be described as a semistable attractor state in the energy landscape of the brain, there really isn’t any difference for changing the shape of that state between new data coming in from the nerves connected to your sensory organs, and the nerves elsewhere. The priors/evidence weighting balance simply describes how easily new data is able to dislodge that attractor state into a new configuration.
I am confused however how things such as memory and learning play into this. It seems like this framework would also predict depressed people to be poorer learners. Is that true? I’m not sure what it would say about memory.
“ Researchers in one 2013 study discovered that people with depression couldn’t identify objects on a screen that were identical or similar to an object they had seen previously. According to researchers, this suggests that memory can be diminished as a result of depression. Researchers in a 2015 studyTrusted Source came to a similar conclusion. They concluded that depression might cause short-term memory loss.”
The article: https://www.healthline.com/health/depression/depression-and-memory-loss#research
The studies:
2013:
http://www.sciencedirect.com/science/article/pii/S0166432813004920
2015:
http://www.ncbi.nlm.nih.gov/pubmed/25562416?dopt=Abstract
Not really? It might make it synonymous with "anything that you consciously experience", but your brain does a vast amount of things that you are never consciously aware of. And your senses just being anything that transmits information to the global workspace seems like a reasonable enough definition.
I recall Thomas Metzinger making the point that the distinction between introspection and non-introspection is somewhat artificial; in both cases you are experiencing some data that a subsystem has output into your consciousness, it just happens that some of the data is tagged as "internal" and some of it "external". But that's just a tag on the data, and while it does to some extent track a real difference, it can also easily get confused (e.g. projecting your own assumptions on other people is common).
In reality, what feels like raw sensory data is always mixed in with internally generated high-level models of what one expects to see, and what feels like purely internally generated thought always has a causal link to *some* past experience.
Well, to an extent it shouldn't bee surprising if everything your brain does share various important systems-level features, because it's all instantiated on the same computing substrate.
Like, 'everything a computer does' can be expressed in the language of ones and zeros, logic gates and cycles, etc.
Whether or not it's semantically useful to call it all 'senses' is another issue, but I don't think that being overly broad is a problem in and of itself - we should expect a broad theory, at some low-enough level of analysis.
Examined in detail, there effectively is no "more basic sensory input" – I think maybe smell and taste might be the closest. But _vision_ is incredibly sophisticated (and complicated) and is, in large part, actively _predicted_ by the brain and not some kind of 'raw data'.
This is fascinating. I have a child whose depression is exactly like this. But I’d like to suggest a different style of depression, one where the pt downplays or “forgets” positive priors, and focuses unduly on the (negative) now. Here, I think, the disconnect is more cognitive than sensory, and may respond to highly tactical approaches such as cbt.
Since you've used the term 'forget', you might be confusing priors and memory, which is NOT (directly) part of priors ?
Though I guess that *technically*, priors are a kind of memory ?
On the "date" hypothetical: not that my experience of depression is universal, but my experience of that scenario would go as follows:
I am at a -9. I go on a neutral date, which is a 0. Typically (though not always), I have a brief elation period--I feel the date went better than it likely did, say a +2 or a +3. And then abruptly my mood shifts back. I begin to feel anxious and/or paranoid: maybe this dude is a serial killer, and he was charming and into me because he wants to kill me, or maybe he's got a wife and he's a cheater, or maybe he's repulsed by me but really hard up and so he's gonna go along with it until he gets some. Sometimes the anxiety is only the realization that one good date means there will be *another* date, and it just seems like too much work to repeat that performance.
Then, and more importantly, I forget. I forget that the experience was good and can only recall the experience with difficulty. I cannot overstate how much my experience of depression involves forgetting even basic facts about my life. If I feel bad about not being married, I will *literally forget* the people I've dated, and other people/things must prompt me before I remember. If I feel like a failure, I will *literally forget* how much money I make, or how my bosses have given me glowing performance reviews. I don't know that it's anything to do with weighting--at least for me, it seems like the issue is that, outside an extremely short window, the new experience (and many old experiences) is failing to be considered at all. The prior cannot update.
Have you tried writing down a detailed list of good things to be reminded of in such cases? for example, good dates, good feedback, etc (especially ones you have a special reason to believe are honest) - and then go to that list whenever you feel like you might be forgetting or discounting evidence?
It's something i thought of doing myself, but haven't actually done since i rarely feel that way or have trouble coming up with those examples from memory. but i think if i had the problem of often forgetting about them and feeling bad i would make that list.
Yes, one of my strategies for managing my depression is a "scrapbook" of good things on my life
Yes. Scott doesn't seem to be picking up that people negatively downgrade things further and ignore positive evidence.
We know that people do this with beliefs all the time. I'm surprised that Scott treats beliefs as a naive comparison of belief and evidence when we know the vast majority of people do not work like that.
I think that this is, in part, an attempt to come up with a theory about how that works. Have a linear function for the prior in the example (ie -9 + actual sensory data) is probably wrong but is simple enough to illustrait a point about veteran reasoning.
Reality may be more in line with a discrete system where things fall in categories of very bad, bad, meh, ok, good etc. If the prior is more heavily weighted enough, then new information will fail to update the prior out of the discrete category is in.
Or it could be any number of other update functions. But the core idea is that there is a failure to weight the evidence vs expectation in such a way that expectation dominates and you get stuck in a depressive state.
Agreed. The way depressed people (generally) seem to think is much more like an attractor/quantum state/gravity well where you need energy to move to a new stable state, and anything less than that energy doesn't move you even slightly.
Move someone up 0.9 and it rounds down to 0. Move up a whole point and you're great - but you can slide back into a lower orbit again because the sucking void beneath you continues to exert its pull.
But isn't the attractor state coming from the self-reinforcing loops that Scott describes ?
(and the other way around for people who are naturally happy.)
Trying to test the theory: naively, we expect a prior in a negative direction to increase attention to evidence/sensory inputs in the OPPOSITE, namely positive direction, because that is the definition of surprise: you pay more attention to things you did not expect.
However, depressed people despite their supposedly negative prior seem to dwell much more on negative experiences than positive ones, and also have heightened pain and light sensitivity, so that would contradict our prediction. If you already expect bad things to happen to you AND your attention to sensoral inputs is lower, why would you be able to tolerate pain LESS? In some ways, depression seems to make people numb, like you explained, but in others more sensitive.
OTOH, I am not really sure whether my understanding of what a prior entails is correct: If you fear tigers, you will pay more attention to sounds that could be coming from tigers than other people, so it is probably more compicated.
Heightened sensitivity to light: https://pubmed.ncbi.nlm.nih.gov/2748876/ , and pain: https://pubmed.ncbi.nlm.nih.gov/26867484/
I don't think the pain thing is necessarily a contradiction. The model could explain it as your sense data reporting a neutral pressure or mild discomfort, and your higher-level cognition favoring a prior towards any given sensation being unpleasant.
I'm not sure how this squares with the other sensory data, though. If I'm right, I'd predict that the reduced olfactory capability reported in depressed patients wouldn't extend to unpleasant smells, but I don't know if the current olfactory tests cover this angle.
That's a good point about unpleasant smells! I'm very curious as to whether that would affect the results.
I think you might be on to something.
I looked up the tests and they seem to be these (or some combination, or something similar from the same company/organization):
- [Screening 12 Test - MediSense | Smelltest.eu](https://www.smelltest.eu/en/product/buy-burghart-sniffin-sticks-smelltest-pens/)
- [Sniffin' Sticks Identification Test 16, blue - MediSense | Smelltest.eu](https://www.smelltest.eu/en/product/burghart-sniffin-sticks-identification-test-16-blue/)
- [Sniffin Sticks Identification Test 16, purple - MediSense | Smelltest.eu](https://www.smelltest.eu/en/product/buy-sniffin-sticks-identification-test-16-purple/)
I couldn't find a list of smells for the last two tests but for the first one I did find this:
> The odors of these sticks are peppermint, fish, coffee, banana, orange, rose, lemon, pineapple, cinnamon, cloves, leather and liquorice.
For what it's worth, some types of depression I've had have made me tolerate pain less (mostly the "general malaise" sorts), and some types have made me tolerate pain more (mostly the "sitting motionless like a lump" sorts). It might also depend on the type of pain.
My depression definitely has a physical component. I don't know how I would score on any of those tests, but I know I feel incredibly lethargic and nothing is interesting. At all. Meditation and exercise don't help - one of my worst episodes came when I was going through a daily bootcamp.
Ketamine assisted therapy briefly worked for a few weeks but the effect faded. I never found any kind of trauma behind it - my therapist remarked that I was pretty unusual compared with his other patients.
Thanks also did some somatic therapy at a weekend workshop. Parts of it I liked but again, no lasting benefit.
I've had good luck with SSRIs but I have to change the drug every few years.
If you're looking for suggestions, you might try TMS (transcranial magnetic stimulation). There's no side effects that I've heard of, and it seemed to help. I got mine from a place that was also offering ketamine therapy, so you might try asking at your ketamine place and see if they do TMS, or know of anyone in the area doing it.
Ha, I've tried that too. Thanks for the suggestion, though.
I'm doing fine with SSRIs, I'd just like something that doesn't have side effects.
The only other official treatment I've tried is ECT, and hoo boy does that have side effects. :-( Quite effective, though.
Great find and impressive synthesis, thank you.
Briefly, for your third point under implications, that's been tried under the rubric "concreteness training" and appears at least in two studies to be fairly effective.
I am somewhat sceptical though. To me it falls under the Jordan Peterson "clean your room" category of advice that sounds good but doesn't really work in shifting the "attractor state" (to use your terms) - except maybe as a placebo. Both appear to be positive symptoms of a "deeper" shift that is not easily amenable to reverse engineering at the surface level, at least absent other important changes.
Is the default mode network relevant here? Or maybe that's just a higher level description of low weighting of sensory data?
I had the same question. Having read a tiny bit of Karl Friston (confusing) and recently Michael Pollan's "How to Change Your MInd" I'm guilty of trying to integrate this subject and other examples of brain activity into the models for DMN and Markov blankets. The VDBEA piece seems to mesh well with a model where DMN dominance is overruling the sensory updates being discussed. Not so much with the Markov model - if the sensory input contradicts the prediction generated by priors why isn't the model updated accordingly?
How does this relate (if at all) to the account of autism that sees it as an unusually bottom-up style of information processing? Naively it would seem like a case of opposite weights on priors vs sensory experience, although I don’t know of any indication that autistic individuals are less likely to be depressed, and it doesn’t seem like something I would guess anecdotally either.
Also: some of the descriptions here about depression and trauma having to do with their having very negative priors; others have to do with weighting priors over new evidence. These seem logically distinct and in principle you could have something where someone has a high weight on very positive priors. But as you note happiness is associated with eg color - we don’t seem to have a lot of cases of abnormally happy people describing the world in greys. (Maybe people having mystical experiences that make them more detached from earthly experience? I guess there is a lot of language of happiness that implies floating, flying, various forms of ungroundedness.)
+1 to the question about autism and depression, if anyone knows.
I can offer a few speculations.
About the "bottom-up processing": I'd describe autism from PP perspective as a very rigid internal model structure that is optimized for high precision and is slow to update (contrast this with a very fluid worldview on the opposite side, related to say schizotypals, see the SSC post on the diametrical model of schizophrenia and autism for context). Byproducts of the rigid/highprecision structure are (again, my lax interpretation):
- you're probably very good in areas where high-precision reasoning is required (say mathematics) and very uncomfortable in areas where a high-precision model is not possible (say social interactions)
- as your model is slow to update, this gets better over time as you slowly incorporate more complex rules (afaik moderate autism, e.g. Aspergers, smoothes out over time as people get older and are maybe slowly building internal rules of coping strategies / picking places and situations they can handle / ...)
So from this model, "low certainty" in internal predictions should not be an issue if you are on the Autism spectrum. BUT: this comes at the price of being hit really hard in case of unexpected situations, e.g. where your predictions fail. Two elaborations:
- I know a small child that is somewhere close to the autism spectrum (no medical diagnosis and this might all smooth out over time) and if things don't go as expected, THIS IS LIKE VERY BAD PHYSICAL PAIN for her. This matches my experience with many other people (most of which are not autism-diagnosed and - if at all - only losely connected to the spectrum!) where i've experienced many times how unforseen events (e.g. a small mishap or unexpected error) quickly spirals out of control in their head and they end up tormenting themselves mentally. If we think of autism as an extreme, this is a potential cause for a very bad mental pain spiral (which is potentially very difficult to escape).
- somewhere in the comments (or Scotts post?), there was a reference that "high prior, low update rate" not only refers to exterior senses but also your internal feeling. I could list a few anecdotal cases (the child mentioned above, a friend diagnosed with Aspergers) where this slowness / neglect of internal sense data lead to followup injuries (the friend mentioned just sprained his arm while "doing normal movements" and I could easily imagine that there were some subtle warning signs coming from his body earlier he missed). Combine this with the point above and you have another recipe for potential mental blaming cascades.
Speculation on trauma and PP:
If we think of hyperpriors as the general structure of our internal model of the world (how the hierarchy is connected / unconnected, how much update happens to the model structure in case of failed predictions, ..., resulting e.g. in the individual personality traits that are pretty robust over lifetime), we can think of trauma as "an event that shreds the whole model structure into pieces". This would align quite nicely with personality traits being changed by traumatic events and would explain why trauma therapy looks like "slowly rebuilding your internal model of the world". Also if event memories are tied to an operational internal model of the world, this would explain both why we all have no toddler memories and memory failures after trauma.
I'm more confused about depression. The link to low confidence sounds very reasonable, but as far as I can tell from the PP framework it's not really clear how high/low motivation works. I need to reread the "Guyenet on motivation" and "chamber of guf"-posts on SSC, but to me it looks as if this is mostly a personality trait / evolutionary preset parameter stored somewhere in the hyperprior (maybe not as a constant but as a function that reacts more or less to internal parameters... really, I'm confused).
I wonder if much of the positive effect, and positive affect, of exercise are mediated by just being more in your body. Exercise pushes you to feel your body, for sure. (Who is that in the back row saying "torture"?)
...and more out of your head.
I have been reading about feedback control systems lately. This makes me think of a PID controller - would strongly weighting priors be akin to a more prominent integral function?
I think it's more complicated than a PID controller. Think of a system that generates a hierarchy of controllers on different levels and then chooses on the fly which one to use, discard or update. The hyperprior is the general hierarchical structure and set of update rules (eg if something goes wrong, how much adjustment should you do?)
Dante might have been commenting on this in Inferno Canto 10. The damned are chronologically far-sighted; they can remember things from the past and have a sense of what will happen in the future, but the present is a haze. This turns into a merciful feature once judgment day comes and the future ends, because they'll just sort of reach a steady state. That also sounds a lot like "curl into a ball" depression, where you just want to stay under the covers forever and shut out all stimulus.
It seems that you can see, if I hear rightly,
Beforehand whatsoe'er time brings with it,
And in the present have another mode."
"We see, like those who have imperfect sight,
The things," he said, "that distant are from us;
So much still shines on us the Sovereign Ruler.
When they draw near, or are, is wholly vain
Our intellect, and if none brings it to us,
Not anything know we of your human state.
Hence thou canst understand, that wholly dead
Will be our knowledge from the moment when
The portal of the future shall be closed."
https://www.gutenberg.org/files/1001/1001-h/1001-h.htm
A very intriguing connection!
One quibble though: I’m not sure the final haze is meant to be a merciful feature of Hell. Aren’t we meant to think that the denial of all knowledge will be a further punishment? The other sinners seem to be punished more after the world ends: consider how the heretics will have their tombs covered after the final judgment. Dante admired Aristotelian scholasticism, and one of Aristotle’s most famous doctrines is that all men by nature desire to know. On that score, it seems difficult to think that Dante thinks God is being merciful in this situation.
To the issue of why thinking about the trauma does not make it go away on its own. Stress seems to really ingrain memories in a way that normal memories are not. So equally as powerful is needed to dislodge the bad memory and EMDR is somehow simulating a high stress environment.
This makes me wonder about aphantasia and depression. I see some claims that aphantasia can become a thing after a depressive or psychotic episode or brain injury. But I wonder if it goes the other way too.
Also,
> your brain artificially down-weights their signal, marks them as suspicious and untrustworthy.
I don't know about suspicious, but you talked about "maggots" in your Muggeridge review, and that's how I generally feel in the mornings: whatever I happen to be thinking or feeling is getting stamped with big DISGUST labels.
A recent breakthrough for me was noticing that I feel like I have a highly tuned empathic sense that notices all the suffering and imperfection in the world, and noticing that it's odd that it doesn't also detect all the joy and triumph. It was quite overwhelming (in a good way) to even begin to contemplate that. So I've been paying more attention to the "Humans, Fuck Yeah!" genre, eg., https://space-australians.tumblr.com/
Interesting. I have always been depressed and just realized I have aphantasia. I am going to have to read up on that.
> So I've been paying more attention to the "Humans, Fuck Yeah!" genre
I've (weakly) found this to be helpful for me too – and MDMA.
"That is, the first time you bungee jump you're probably pretty scared. But you bungee jump a few times, nothing bad happens to you, you learn on a preconscious level that it's not dangerous, and then you're fine with it." Adam Savage would disagree: in this Q&A video (https://youtu.be/-AvhYMPvPR4), he describes that it actually gets *worse* the second time, at least if you do multiple jumps in a row (ex. to shoot a TV show).
Huh, it looks like the link made it duplicate the text. I've seen duped-text glitches before here, but not in that way.
It's so scary that you mostly remember being scared the second time ?
P.S.: Scared of the unknown vs the known. The various weights resulting in being *less* scared from the 3rd time on.
I can’t quite reconcile this argument with certain examples of over-weighting the sensory data going badly. I’m thinking of the stereotypical hypochondriac obsessing over any minor bodily sensation, or anxious people who live in California and are hypervigilant about any tremor that could be an earthquake. It seems that in those cases there’s both an incorrect negative prior and too much attention paid to sensory data. Perhaps because in these examples there’s no way to “disprove” the prior (there’s no sensation that proves one is completely healthy apart from maybe the passage of time without any other sensations) so the two directions are not in conflict with each other, and it’s another possible stable state?
Is that necessarily overweighting the sensory data? It seems like you could be hyperaware of every minor bodily sensation or tremor without attributing it to sickness or an earthquake. If you were updating on information correctly, what should happen is that you'd notice that you're picking up on all such minor signs all the time without them actually being associated with anything bad, so you'd continue to be aware of them but they wouldn't cause you to freak out.
I think what you're describing is actually a case of having a high prior on there being an earthquake or illness and that prior then "looking for" anything that would be compatible with that story, and then not updating on the sensory evidence that repeatedly says "no illness" or "no earthquake".
Wonderful food for thought, thanks a lot. So far, my prejudice about psychology is that they mostly correlate things with other things. And most of the time, questionnaires with other questionnaires. So it is refreshing to see people developing proper theories that might lead to improved therapy
Wonderful food for thought, even if the theory turns out to be wrong or incomplete. So far my prejudice about academic psychology (not psychotherapists!) is that they correlate stupid questionnaires most of the time. It's so refreshing to see that people actually think and develop theories and models that may even improve therapeutic work.
Ups, this works better than expected. One can even respond twice :-)
>> manic people say the world seems oversaturated, even brighter and more colorful than usual
As a person with unipolar mania, this effect is *awesome* (and overwhelming). I get so stimulated walking or biking about and seeing little bits of color pop out at me, like flowers or plants that I've never noticed before on a common route. It's like a low-level psychedelic trip.
(Also, food often tastes better.)
Albert Hofmann, the Swiss chemist who first synthesized LSD, wrote a book called "LSD: My Problem Child". He recounts an incident where, on a hike as a child, the world suddenly appeared much more vibrant and alive, a religious experience, for reasons that were mysterious to him. He compared his experiences on LSD to that moment. The Magic Theater scene in "Steppenwolf", by Herrmann Hesse (written many years before Hofmann discovered LSD) is the best description I've ever run across of an LSD trip. Something about those German Swiss, I suppose.
The author Daniel Quinn also described a similar incident to what you described Hofmann experiencing.
I think you meant NMDA antagonists rather than NDMA antagonists. Though I can see why you'd have MDMA on the mind when typing about trauma; by all accounts MDMA-assisted psychotherapy is incredibly effective for PTSD.
Speaking as someone with PTSD...
I'd rephrase that bungee jumping example. Let's say I went bungee jumping *without proper safety procedures*. I was scared, but everything went OK. I did it a few more times, I learned on a preconscious level that bungee jumping without proper safety procedures is not dangerous, and I was fine with it. So was everyone else, so I felt quite normal. Then something bad happened, something that impacted me directly, something rare enough that most people haven't seen it, let alone had it happen to them or near them. All of a sudden I'm no longer fine with bungee jumping without proper safety procedures. I have a different idea in my head of what proper safety procedures are, a different idea than pretty much everyone else out there has. It's hard for me to relate to them: they all think I'm paranoid and worrying too much, and I think they're all ignoring truths that are inconvenient to them at the moment. "Sweet summer child" is a good start, but combine that with one of Lovecraft's protagonists who just barely survived and encounter with Something, and that gets closer to how I feel. There's only one friend who I can really be open with; she was one of the people who cut ties with me as I started going downhill, but a bit later she had an eye-opening experience of her own, and now we can actually have serious discussions about stuff, even though we don't always agree on everything. That more than anything else has made me feel as though I'm not completely insane.
To put it another way, what was Julius Caesar's prior for "being stabbed by a friend" 1 day before his death, and 1 millisecond before his death? Which was more accurate? Could he have saved his life if he'd changed his prior 1 day earlier? But would it also have had a negative impact on his ability to form friendships afterwards? If he saw a therapist 1 year later, would the therapist conclude that he worried too much, and should trust more easily? Would he have lived a happier, healthier life if he'd somehow recognized what would happen, taken steps to neutralize it, and then lived the rest of his life without that event having any impact on his personality?
All that said, EMDR is amazing, although I feel like it's borderline witchcraft. My operating theory is that, given how we apparently rewrite memory every time we access it (last I heard, anyway), EMDR is a way to access and rewrite PTSD memories without rewriting the full emotional power, leading toward a more normal emotional level associated with the PTSD memories. But I try not to think too much about it, in case it makes the EMDR stop working. It feels a bit sacrilegious to consciously avoid thinking about something, but this is by far the best chance I've seen at my ever having a relatively normal life again, and I don't want to mess it up. :-/ My main complaint is that it's best done in-person, and I'd just started to make progress when covid-19 hit. So no more EMDR for me for a while, and even if my therapist gets trained for online EMDR, it doesn't seem to work as well for cases like mine.
Also, something you didn't mention, and something I hadn't heard of: part of my EMDR has involved my therapist asking me to pay attention to where I'm feeling stress in my body, after I revisit traumatic memories. And it does happen, I usally feel tense and sore somewhere, like in my shoulders, or neck, or back, or head, usually spots around 1" in diameter. So that may connect with the somatic therapy angle.
I also wonder a bit about the back-and-forth eye tracking of the fingers, and whether that has any relation to the pendulum used in classic hypnosis. Not that I feel particularly hypnotized, but would I know? :-)
And I realize here that I'm conflating some things. There's the "flashback"-like dump of emotions that I get in response to certain stimulus, which causes several seconds of highly distorted thinking, and then has a diminishing effect on my mind for at least half an hour. And then there's the set of beliefs about the world which I think follow fairly logically from events, but which are also reinforced by the "flashbacks", and which are closely-enough associated with the stimuli for the "flashbacks" that it's hard to examine them too closely without triggering more "flashbacks".
Perhaps a comparison would be between a vegetarian who doesn't eat meat, and feels a bit sad when meat-eating is going on, and will share their philosophy when asked, and a vegetarian who gets a rush of intense negative emotions every time they're around meat, to the point where they can only shop for food in vegetarian-only grocery stores, and hearing co-workers talk about a barbecue can make them unproductive for an entire day. Maybe if the PTSD is fixed (to whatever extent possible), the second one will stop being a vegetarian and start enjoying bacon again. Or maybe they'll just be like the first one.
"Magic is the science and art of causing change to occur in conformity with the will." – occultists. So, yes, technically it *is* 'witchcraft' !
Well, yeah, but that definition seems overbroad... It also includes eating food if I'm hungry and want to not be hungry, right? :-)
IIRC there's also a less broad definition that restricts it to "changes in consciousness" ?
This is interesting to compare with this account of what it's like to be enlightened: https://www.youtube.com/watch?v=cMep4wZyRr4 .
In that account, Daniel Ingram gives the example of someone walking on a street where they're completely safe, when they suddenly recall the memory of something upsetting. The memory then keeps making them feel more and more upset, as it causes stress and that stress response keeps making the upsetting memory more salient, and so on.
Whereas for someone enlightened, they retain such a strong sense of awareness of their current environment that memories of past events just can't get any hold. One can still remember them and think about them, but one is also constantly aware of the fact that they are just memories and not a sign of something in the environment that's actually threatening, so they cannot go low-grade loopy and keep retriggering the same upset. In your terms, the sensory data keeps being high-precision enough to outweigh the priors in past memories.
But then Scott puts memories in the sensory data bin rather than the priors bin ?
I think the way to square this circle is by considering 'priors' to be some kind of global statistic that gets updated and "memories" as an active recall, i.e. indirectly experiencing them through something like 'conscious awareness'. The priors aren't memories – they're a 'cached' calculation, even if, in a very real sense, both they and 'memories' (actively recalled) are derived from the same original 'sensory data'.
I wonder if neurologically, priors aren't more about the brain's chemical balance ? (And maybe somewhat about large-scale neuronal architecture ?)
While memories are about the low-scale physical architecture of neuronal links, and memory recall is electronic, but obviously affected by chemical balance.
Maybe the 'weight' given to priors is encoded in "the brain's chemical balance" but my naïve understanding is that there's no way to encode priors in it exclusively.
I haven't read up on the details of this kind of stuff in a long time so I don't really know that much (or can't recall any details I once 'knew').
In a general sense, both priors and 'memory' are both memory in the sense of information. But just based on introspection and other's peoples reports about what we're calling here 'priors' and 'memory', accessing priors must be, effectively, a much cheaper computation to perform, i.e. 'cached' calculations. 'Memory' seems to, often, involve much more involved 'active recall', which seems (to me, and apparently many other people) something more like 'replaying sensory data'. I _believe_ that there's strong evidence that that's a fairly accurate summary of what happens neurologically, e.g. the same parts of the brain that process sensory data in 'real time' are also used during 'memory recall'. We very much do in a sense 'relive' our memories.
P.S.: We should be wary about applying computer metaphors to something as different as the brain.
Like all metaphors, they are useful up to a point to get understanding, but are only hindering after that point.
For a previous example of such a metaphor, see the modern model of living beings as literal clockwork automatons.
That's a good warning to repeat and I try to remain cognizant of it in these kinds of discussions!
But I generally use _computation_ metaphors, which, admittedly, have the weakness that they're not always easily distinguishable from 'computer metaphors', given that they share many of the same technical terms. You're right that brains are _very_ different than the architecture of what most people understand 'computers' to mean.
But from both a mathematical and physics perspective, the brain is just another 'computer', i.e. it must have some mechanism for storing information, retrieving it, and performing computations on that information. From that perspective (especially the mathematical side), the 'implementation details' are 'trivial'. Obviously, the implementation details are actually interesting and important, but they also don't fundamentally change the relevant 'computer science' of how the brain _has_ to work, albeit in an extremely abstract (and impractical) sense.
Did anyone else laugh out loud at the depiction of Scott reinventing meditation?
Was funny, but it made me notice that I don't think I'd ever thought of the 'awareness' usually associated with meditation and some drugs as literal sensory awareness before. Apparently I'd biased that word towards awareness of not-actually-physically-there sensations (mystical, hallucinatory, or metaphysical).
I wonder if that insight might help one to approach it in a more effective way.
(And then I remember I'm prone to things like health anxiety, making the prospect of trying to feel things *more* seem analogous to reaching out to touch a stove that might be very hot...)
It can be both. Extreme awareness of what you are looking at can also help you see small distortions in it. Think about looking through a screened window casually versus looking through a screened window very precisely at a pattern on the other side. At a casual glance, the screen just becomes a grayish blur over the image, but a precise glance lets you see the shape and size of the screen cells.
AFAIK there are two 'kinds' of meditation :
- (Somewhat improperly called ?) 'Mindfulness meditation' where you're trying to be as aware of your 'corporeal' senses as possible, suppressing everything else.
- More 'traditional meditation', where, on the contrary, you're trying to suppress your 'corporeal' senses, but *also* your memories (which Scott puts into the 'senses' bin), and even your thoughts, so I guess what is left to focus on are the processes by which the brain itself works ?
I first learned of the Sniffin' Sticks Test when I saw it in a paper investigating anosmia in COVID-19 patients. I thought it was just something they made up for the study but no, apparently it's a common benchmark of olfaction.
Also explains why my wife, who grew up in an abusive household, needs the TV on to sleep. Sensory deprivation lets her priors run wild.
Anyone have the actual best references for the color perception - depression link? This post just links to an older post, which links to a paper by Chekroud, a much more general review. I ask because this seems like the sort of embodiment study not to survive the replication crisis, and I'd be interested to see what sort of control groups and measures they used.
Relatedly, I looked into the attention bias literature (anxiety, OCD) a while back, and it's not very solid (much p hacking, little evidence that biases play a causal role).
Typo: NDMA. I've made that same typo so many times I am traumatized by it and pay a ton of attention to it.
Quite interesting though i wouldn't lean on EMDR too much: not just because all psychotherapies are kind of equally effective (wampold, 2015) but it's been shown that EMDR doesn't require eye movements at all, and it's not particularly extra-effective in controlled trials. In my opinion, is enough with any kind of attention splitting process so you get old fashioned exposure plus interference with your emotional arousal. In any case, in my opinion any psychotherapy involves some degree of exposure (+ acceptance?) of felt/sensed disturbance plus s certain reprocessing of it (and priors as well) towards new narratives. Just an idea.
Congrats from Spain on extremely amazing blog and smart commentators.
How do you reconcile this with the fact that the most effective and best supported trauma therapies (Prolonged Exposure and Cognitive Processing Therapy) both involve revisiting traumatic memories over and over in excruciating detail?
The literature to date suggests EMDR works independent of the eye movements, i.e. it is the exposure piece that is doing the work. It is probably a mistake to spend too much time trying to reconcile theories with van der Kolk's stuff; the years have not been kind in finding supporting evidence for it and the trauma researchers do not take it very seriously. At the very least, his work is not some well-established, gold standard, obviously true framework.
These therapies, by taking place in the safest possible environment, would be either slowly pushing the prior to be more favorable and/or give more weight to the immediate senses (rather than memory senses) ?
Interesting. I wonder if this is part of why people with migraines are at such high risk for major depression. The risk goes up about 3.2x the general population once you get a migraine diagnosis:
https://headachejournal.onlinelibrary.wiley.com/doi/pdf/10.1111/j.1526-4610.1994.hed3407387.x
I have chronic migraines that cause a lot of sensory/nervous system hyperactivity, so I literally can't trust most of my senses most of the time. Obviously dealing with the pain and disability that comes with migraines can't help here, but it seems like over time have actually unreliable sensory data would take it's toll. Interestingly, the new CGRP blockers are the only medications I've taken that get rid of the weird sensory issues I get when I get a migraine. All the other medications just address the pain part. Massage helped, though I could never quite figure out why. I think it helped put me back in touch with my body more effectively.
Scott, could you please replace https://web.archive.org/web/*/https://slatestarcodex.com/* links with https://slatestarcodex.com/* in the https://astralcodexten.substack.com/about page? The archive loads way slower than the blog.
(I would have sent an email but couldn't find any. Some kind of "Contact" button would be nice)
This discussion may be adjacent to some of the discussions surrounding predictive coding and fatigue in exercise.
This discussion paper frames fatigue in exercise as a response to a scenario where sensory information cannot be trusted: https://psyarxiv.com/nqxz2/
The argument is essentially that, when exercising at high intensity, the allostatic mechanisms that regulate various systems up and down the stack starting making bad predictions and fail to keep various parameters (heart rate, blood CO2 levels, contractile potential, etc.) within acceptable bounds.
So, the organism experiences "fatigue" as a way of returning to a situation in which sensory information can be trusted — aka "not exercising really hard."
In the case of depression, it seems that the low trust of sensory information could create a similar state where the organism looks to "shelter and rest" in order to return to a situation in which sensory information regains trustworthiness.
As mentioned, it seems like the depression attractor state could potentially be found through some combination of all of these phenomena.
t seems to me that depression starts with a lack of control in the quality of thoughts and then establishes itself as a phenomenon in brain chemistry. Thus, after a certain stage, the neurotransmitters are compromised and chemical treatment becomes necessary. However, if we have the maturity to realize that everything starts with the quality of thoughts and we are able to maintain good mental hygiene, depressive states can be preventively prevented
We just recently tested whether the tendency to downweigh sensory evidence relative to priors varies with peoples trait anxiety in perceptual decision making: https://www.sciencedirect.com/science/article/abs/pii/S0010027720302936
To the question of perceived -6 event still being better than an expected -9 event and yet not resulting in proper updating: It’s possible that being depressed imposes an additional -3 penalty on every experience simply because being depressed is itself actively unpleasant. Sort of like how my brain doesn’t automatically subtract away the pain of a migraine when judging whether I enjoyed a movie while suffering through one; the migraine was part of the experience, and so the movie seems worse. Depression is painlike, in addition to all the other stuff. In this case, the depression is part of the experience, and I think it sucks even worse than the sensory dulling effect accounts for. When you add up the sensory dulling, the update downweighting, the intrinsic suckiness of your experiential frame, and the low prior, it’s possible that you could still manage to undershoot your unrealistically low expectations.
I am not sure we can separate 'priors' and 'evidence' so neatly. Perception is actively constructed by the brain so our here-and-now perception ('evidence') is shaped by priors as well. Ambiguous cues will be interpreted according to our priors. Occasionally, the data will be both less ambiguous and incongruous with our priors but misperceptions are fairly common experiences, so it is not surprising that when our experience is incongruous with our expectations, we are at least sceptical that things are what they appear to be. This scepticism is probably fairly adaptive a lot of the time: "if the deal seems too good to be true, then it probably is"-kind-of-thing.
Haven’t read the paper yet to know how this might work with positive emotion (i.e. mania), but it seems like there might be a handy explanation here for how meditation leads to awakening/enlightenment/etc. I mean, if meditation in those with depression or PTSD can help correct the underweighting of the sensory channel, couldn’t it cause overweighting in normal people? And, as I assume the paper suggests with mania, wouldn’t overweighting the sensory channel make you feel pretty good assuming your life isn’t absolute garbage? I.e that decent coffee you have every day tastes really good (which is something “enlightened” people have actually described to me). And might those experiences be self reinforcing in the same way as negative ones are with depression? And, if you’re overweighting the sensory channel, couldn’t you convince yourself of some pretty odd beliefs that might correspond with the bare facts of the sensory channel, i.e. that’s there’s really no concrete sense of self to be found and thus the Buddhist conception of noself?
Thank you. You article made me recall & reassess my personal experience of (mild) depression. Hope it will add on to your observations. What you are describing, seems familiar at least.
I experienced sensor issues, as in - less colour, less taste in food, less enjoynment of life.
I experienced negative perception in work, communications with people, or simply reading or watching films.
Overall, it feels like all cells in your body tell you "i want to die" instead "I want to live".
What I understood over time, and what it took to get back to healthy:
1. Self-perception. I am a very happy and vivid person overall, so it took me months to realize smth is not okay, as my self-perception and reality do not match anymore. Tiredness, eager to be upset over anything, conflicts, irritability etc.
It`s hard to accept that it`s a problem. You keep thinking, its tiredness, and it will pass soon. I started having more walks, eating fruits, taking D3 etc after a while, which helped too. But those phisical habits are hard to get back to. It`s not that you do not understand it`s important. You just don`t have any energy, and you have to retrain yourself back to it for months. This is I think one of the reasons why depressed people get so angry over "take a walk & eat chocolate" advice.
2. Psychodinamics, hormons, neural activity changes drastically. You constantly feel everything is not working, you gonna loose, you gonna fail. It takes a huge effort to set up a self-feedback loop, to separate what you can reasonably assess is true, and your impressions about it.
Eventually I ended up spending 1-2h a day every morning for 1 year, just talking to myself, remembering my usual life perception & logic, reactions to similar situations before, reasoning over scenarios & things to do, and distancing within my mind from the constant negative "we all gonna die" feeling. It`s what I think you describe on ranks. I reranked myself basically.
3. Broken external feedback loop. When depressed, you react differently, and you provoke slightly different reactions from people or even attract similar more often. You keep getting negative signals not just within, but from outside, and at certain point it entangles so much you do not know who you are and if you ever going to be okay and if you can trust people.
Same as before - I tried to distant myself from my state, so to speak, to see where my reactions are more or less reasonable, and where it`s a negative spark. And then go out to talk to people, do some work etc - while contiously holding this framework, that I might overreact or take things more negative tnan they are. And read feedback from that perspective. Plus openly ask questions - to the point of being blunt, "did you mean to offend me or i misunderstood you entirely?". Again, retraining myself.
4. Dependance. You grow dependant on emotions, cos they are rare. Whatever makes you feel is important (90% not valid or true). It`s extremely unhealthy and shakes the psycho. It takes huge effort & discipline to stop reacting, to play down emotions, concentrate on how you feel, and think, and reason. This part of the equation is a major destabilizer, and less control you have over yourself, longer you can`t handle the whole thing. That was one of the reasons I stayed away from any farma; transfering my dependancy to chemicals didnt seem like a good idea.
5. Being stupid and in the i do not know are the most wonderous states of mind, but they are the hardest to get back to. Experience and knowledge becomes a problem & limitation. Nomally you learn and re-learn easily, and it`s a part of - i think - just being curious and open. In depression you grab ahold to what you know, and try to apply it to a changed situation. Basically, everything that you know is limiting you from getting better, while kinda protecting from being hurt further. Reconstructing, shifting the assemblage point seems like an evident way through, but it`s pretty much impossible to pull off. I attempted that after about a year of just reframing myself towards less negative view of the world.
6. Self-destructive & destructive relationships. Really hard to fix while in depression. Feedback loop makes all worse many times over and toxic relations become unrecognizable. I eventually chose solitude, and got back to being around pretty much anyone except close family after a few months or so. Cleaned out about a half of my networks afterwards too.
Strangely, afterwards I realized that I had a sense of happiness within depression, tho somewhat hidden. At the very same time I remember a feeling, that I will always be that sad and gray, and my life won`t go back to mornal - which I somehow accepted.
I started to see people who are not doing okay - I could now recognize face muscles, changed mimics, changed reactions, slight differences in behaviour - things I did not see or recognize before. It really helped to connect to friends and family better, care of them better.
Whole experience was powerfull, cos it teaches to take in all bad and all good, to be perceptive over a wide range of feelings and emotions, but not be ruled by those as much. Take life in, but not be destroyed by it. Be able to sit in the dark and not be scared of it. Be present for people around. Well, those kinds of things.
Thanks for your articles, btw. I genuinely enjoy the way you think.
I have two questions.
First, if "quieter" sensory input leads a pessimist into ever greater depression via the proposed feedback loop, shouldn't it also lead an optimist to euphoria? How would we know if there were lots of people made super-happy people by the same sensory deficit?
Second, if a preference for abstract thought over sense experience caused depression--without the balancing euphoria I just proposed--wouldn't we find that, historically, rationalists (say, philosophers and theologians) had a strong tendency to be depressed, while empiricists (say, modern scientists) had very little depression?
I'm still pretty confused by the numerical dating example. I kept trying to think of solutions, but they all came back to "the brain uses coarse-grained binning for good/neutral/bad events when reinforcing hyperpriors". Anybody has something better than that?
My math here suggests some ideas:
https://astralcodexten.substack.com/p/the-precision-of-sensory-evidence#comment-1267958
Thanks, that really helped to clarify the situation!
Glad it helped; but remember my comment is just math, it doesn't mean that the depressed brain really updates that way!
Sapolsky seems highly relevant here
If stress from subordinated SEC status degrades the frontal cortex, then your ability to reassess your priors is damaged. Feeling able to appreciate the beauty or what-not, again, comes with not being shat-upon on a daily basis. Seems obvious to me what the limitations of this “epiphany” are.
"subordinated SEC status" ?
Is part of the problem a failure of actually update priors? People in general aren't actually very good Bayesians - and there are perhaps good evolutionary reasons. No matter how many times the thing you saw out of the corner of your eye turns out to be a tree stump and not a snake, you still want to hold a decent prior that it might be a snake. So if your date was a -6 not a -9 as per your prior, you might not actually update your prior, because, you never know, snakes!
People aren't good Bayesians on the conscious level... any way to determine how good we are on a subconscious level ?
I'm not a scientist but anecdotally speaking ketamine therapy saved my life. I can't say I was ever clinically depressed but I was incredibly anxious, prone to panic attacks, and had horribly low self esteem due to some negative childhood experiences. Even though I was professionally successful and had a vibrant social life I felt worthless, lonely, and unlovable. After a few months of ketamine infusions it was like my brain could suddenly let go of those negative childhood memories and accept the evidence that, at this present moment, I am a decent person with many people who care about me. I honestly wouldn't have believed it until it happened to me but it does feel like the ketamine rewired my brain to stop dwelling on prior negative experiences to the exclusion of sensory evidence. I also had a strong fear of driving due to a friend being in a terrible car accident when I was a teenager - I didn't drive a car for over 10 years. The idea of driving was inconceivable. Even riding in a car as a passenger would give me a panic attack 50% of the time. A few months after completing the ketamine theory something shifted in me and suddenly driving didn't seem so scary. Now I own a car again and drive on a daily basis and it's like my entire world has opened up. I have no idea how this happened except that I suddenly stopped caring about that one horrible memory that made driving seem impossibly dangerous and started letting myself be swayed by the fact that most people I know drive on a daily basis and are perfectly fine.
Anyway, fascinating stuff. I'm a true believer in ketamine therapy now and I look forward to seeing where it goes in the future. It feels like my brain was a train stuck hurtling down a very old, narrow, and broken track and ketamine turned it into a car capable of changing lanes, making sharp turns, and backing up when necessary.
How does this theory deal with the overconfident, happy/motivated type of person?
They would also weigh priors ("I'm going to succeed.") higher than evidence (like having multiple setbacks), but are not depressed.
Exactly the way that you suggest : their strong, positive priors make them dismiss most setbacks ?
The people I know who tend toward depression also have a narrowed bandwidth of memory of prior experience, too. But I really would characterize it as narrowed, not merely negative. They don't remember very positive experiences but they also don't seem to remember very negative, either. It's like everything gets compressed into this one narrow stratum of negative experience and anticipating and feeling at a certain level X on the emotional dial, and everything above or below gets folded into that channel. Like the brain is doing an FFT back and forth with only 64 points.
This is pretty much what Scott suggests ?
This accords well with my experiences with meditation. I find that, occasionally, I'm able to experience my thoughts not as an inner monologue, but as a phenomenon that I'm experiencing, independent of my consciousness. Those rare intervals coincide with the largest improvements in my mood and ability to deal with depression and anxiety.
>You update your prior some bit of the way between -9 and -6. Do this enough times, and you should be able to get your prior all the way back to 0, ie eliminate your depression. Right?
I can't comment on the medicinal aspects of this, but here are some mathematical thoughts.
The odd-ratio version of Bayes theorem is, for B meaning "the world is bad", ¬B meaning "not B" (ie the world is good), and G being the evidence of a good date (or a good anything):
P(B|G)/P(¬B|G) = P(G|B)/P(G|¬B) . P(B)/P(¬B).
What does it mean to have sticky priors? The prior odds are P(B)/P(¬B); the updated odds are P(B|G)/P(¬B|G). The term that updates is thus P(G|B)/P(G|¬B). Sticky priors means that that ratio is closer to 1 than it should be.
So what is P(G|B)/P(G|¬B) ? It's the ratio of P(G|B) (how likely is a good date given a bad world?) to P(G|¬B) (how likely is a good date given a not-bad world?)
So even if the depressed person feels "that kinda-good date is unlikely given a bad world", this will not update their priors if they also feel "that kinda-good date is unlikely given a not-bad world".
So there are two phenomena that can be in play: P(G|B) can be high (kinda-good dates are likely, even in a bad world - here we get all the stuff that says the date is not that good, or that good dates will lead to heartbreak anyway, and so on).
Or P(G|¬B) can be low - kinda-good dates are not likely, even in a good world. This could be unreasonable expectations; "if the world was good, this date would be much better".
The same result holds in the standard Bayes theorem, of course (the odds-ratio version and the standard version are equivalent). But I think it's easier to see in the odd-ratio version.
On a person note, as a less-depressed person than once I was, I can anecdotally confirm the poor sense of smell and the reduced autobiographical memory.
Anyway, don't know
>Anyway, don't know
Sorry, this was the remnants of the previous version of the comment.
So I think you're saying that people could have biased estimations of the likelihood ratio, such that all likelihood ratios are closer to 1 than they should be?
I feel like this might be barking up the wrong tree, compared to simply calculating p(H|D) as usual and then weighting the evidence more lightly (see i.e. https://astralcodexten.substack.com/p/the-precision-of-sensory-evidence#comment-1325734)
My reasoning: if depressed people have a high posterior for p(the world is bad), which they would if they had a high prior for it and were slow to update, then the depressed person will have a negative view about the future. What are my chances of having a good date, given how bad the world is? p(good date | bad world) = low.
Right? I feel like this is highly related to the premise of the post, although the post focuses more on unconscious predictions that underlie all brain computation.
So, the premise is that the brain *accurately* predicts likelihood (the technical term, not the layman's term). In the above example, the brain predicts that p(G|B) is low.
But you're suggesting that the brain can't accurately predict likelihood (or at least, likelihood ratios). If the brain doesn't accurately predict likelihood, the depressed person wouldn't have a problem. They would think "the world is bad but it's *totally possible* to have great dates anyway". And that's... not what happens.
Or, if you think the brain can predict likelihood but specifically gets likelihood *ratios* wrong... I dunno, do we think the brain models straight to the ratio when doing bayesian updates? Or does it do the equivalent of modeling the two likelihoods and dividing them?
...I agree that the former would have the same result as the math I linked to earlier - unbiased parameter estimates that converge slowly.
>So I think you're saying that people could have biased estimations of the likelihood ratio, such that all likelihood ratios are closer to 1 than they should be?
Maybe, maybe not. I was playing with mathematics to find a reason why the update would not happen. The math here suggests one possibility; that can translate into something in psychology that might be worth investigating. Since it comes directly from the math, it would be fascinating... if true.
There are many other reasons the update maybe wouldn't happen; this is just one possibility among many.
I fit this model, and it sounds quite useful. There is a difficulty in assessing evidence and being able to control the perceived danger. The danger is the thrashing of internal demons. Since being wrong on the assessment creates incredible emotional pain, the tendency is to not take anything of value from the experience.
When Scott is describing how the brain supposedly assesses an event based on priors and evidence he is assuming that the event is assessed once with priors and evidence acting as weighted multipliers.
There's another way this could happen. The brain could be performing an additive operation every time either a prior or sensory evidence enters consciousness. The level of consciousness is what assigns lower weights for sensory evidence.
I think his objection is somewhat silly on the part of the priors. Why would a prior be assessed only once? And only after the date? If the priors are only assessed in expectation of something they should be separate from the sensory evidence of the actual date.
If we see priors and sensory evidence acting on emotionality on this additive, separate way, we can explain the detrimental effects of rumination. Scott's proposal of a multiplier would always have rumination ending up at level 0 emotipnality.
But priors, by definition, do not 'enter consciousness' ?
So how does the memory loss fit into all this? If depression causes one to overweigh priors vs senses, why does this result in the experience that formed those priors being erased?
What makes you think that it has been erased, compared to has never been strongly fixed in the first place or being just hard to access at this point in time ?
Hi Scott, have you come across the Reconsolidation of Traumatic Memories (RTM) technique from Richard Gray for curing PTSD? If the website is to be believed, what research that has been done - three trials with @30 patients - they average 85-95 diagnosis elimination, far exceeding the efficacy of emdr etc.
Website here: http://researchandrecognition.org/research.html
Recent podcast: "life after PTSD" episode 36
> ...focus on trying to get the brain to pay unusually precise attention, under unusually safe conditions, to something it usually represses...
This is psychoanalysis! The free associative technique produces repressed contents, and the presence of the analyst makes the analysand feel safe, plus their desire to "please" the analyst (transference), beyond a mere desire to "get well", produces more motivation to actually uncover repressed contents.
And Wilhelm Reich in the 1920s tried to tie this practice into the body, as per the second idea, with, uh, questionable results.
Happy to have stumbled upon this article! correlates with my revisiting Bayes theorem yesterday (evidence does not determine beliefs; evidence updates beliefs, i.e. the evidence and hypothesis are weighed against memory or current beliefs) as well as SEP's entry on Karl Popper. Neither of these exactly relate to trauma, but their critical approaches: precise results (Bayes), and real, falsifiable theories (Popper) may contribute something to the problem of dealing with trauma.
I quote the relevant passage *"[Popper] argues that the Baconian/Newtonian insistence on the primacy of ‘pure’ observation, as the initial step in the formation of theories, is completely misguided: all observation is selective and theory-laden—there are no pure or theory-free observations." If this -- that there must be a 'selective and theory-laden' approach to observation -- is applied to trauma, would the problem not expand/explode one hundred fold? I ask the question on the assumption that theories of human nature are wanting in precision, repeatability...
Also, I disagree with Sniffboy. The link between low confidence and negative prior seems to me about as strong as the one between high confidence and positive prior.
Thank you for opening discussion.
*I only cite Popper because I don't yet fully understand Bayes theorem.
You might also want to check out Kuhn :
https://samzdat.com/2018/05/31/science-cannot-count-to-red-thats-probably-fine/
On the part you were a little bit resistant about, could you posit that the brain doesn't actually ever compare the experience to the prior in the step where it updates its new priors? Then you could have: Prior of -9, 'objectively' neutral experience: 0, but weighted between 0 and -9, let's say -3. In the step of updating priors, instead of comparing the -9 expectation with the -3 reality, it might simply record '-3' : bad experience, revise priors downwards. This would awkwardly approximate Bayesian reasoning, which might be the best our brains can mange. It would also be consistent with human brains being generally quite innumerate.
I really like Scotts take on borderline personality disorder. How does ”to loose priors” in that context align with this model of ”to rigod priors” in this context?
*rigid priors...
If most cases of depression can resolve themselves by averaging out the difference between overly pessimistic priors and data (your -9 expectation + 0 experience = -6 perception), then it follows the cure for most depressions is to have as many experiences as possible to dilute the bad priors. Go out and ovewhelm the system with new inputs that force an update.
Incidentally, that's precisely what depression prevents you from doing.
Huh.
One simplistic summary: depression, anxiety, and ptsd represent an over-reliance on priors in their constant sense-making negotiation with evidence. The antidote is to bolster our sensitivity to evidence, and decrease the power of the priors.
Framed in that way, this sounds mighty close to the overall project of rationalism. Because the techniques here are relational and embodied, they celebrate forms of evidence that rationalism doesn't usually focus on, but the project is the same. Are there clues here to what a more embodied rationalism might look like?
In contrast, a dogmatic community will over-rely on their collective priors and assign unusually low precision to empirical evidence. Can we interpret that as a kind of collective trauma/depression/anxiety, even if the individuals themselves don't have those conditions?
"Are there clues here to what a more embodied rationalism might look like?"
'Mindfulness' meditation ?
Late to the party but wonder how (if?) this relates to sensory deprivation tanks... some kind of reset or a variable removal which can help...? Or maybe the whole thing is bullshit :)
Very interesting! This seems to be missing an important piece, though. I would expect universal low sensory precision to translate to impulsivity or hyperactivity, whereas in fact in depression people seem to be more lethargic and it's harder rather than easier to accomplish movement.
Why ? Movement is sometimes related to prediction, so bad prediction would make for harder movement ?
(Also lower precision : https://www.pnas.org/content/111/42/15220 )
As per https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/ IV.5, movement is based on prediction of action. When your prediction mismatches the world, either action happens to bring the world into accordance with prediction, or prediction updates to match the world. Low precision sensory data would mean that it gets thrown out more often in favor of the prediction, i.e., it biases the system towards acting on any impulse/prediction/whim that arises. This also explains hyperactivity in fruit flies with ADHD (https://www.ted.com/talks/david_anderson_your_brain_is_more_than_a_bag_of_chemicals?language=en), albeit with a different mechanism (decreased signaling of surprisal rather than lower sensory precision).
(Also, the theory put forth by that link might explain low-functioning autism, but at best is exactly backwards for high-functioning autism. As described in https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/, the model of autism as overly narrow confidence intervals on predictions fits much, much better, IMO, than the model of generic failure of prediction.)
Also:
> so the system that should be habituating to the trauma is actually "learning" to become more and more afraid of it.
I'm suspicious of this. IIRC, Body Keeps the Score says that traumatic memories of war vets with PTSD were virtually unchanged ~40 years later, whereas the same memories of war vets without PTSD underwent the same sort of change and drift that normal memories do. I think in fact the "this is scary" response to re-experiencing the memory counts as sensory experience and is thus underweighted in the same way that normal experience is---the memory remains unchanged, rather than becoming either more or less scary.
Chatting with my friend Rivka, another component came up that seems quite related, though I'm not yet sure how it fits in. She told me that people with trauma tend to feel less in control of the world and their bodies. Since, IIRC, learning is best done interactively, when you can both observe the world and see the effects of the attempts you make to control it, it seems like half of this process is missing in trauma, where you feel like you can't control the world. It seems like this should somehow connect with the trouble updating on sensory experience, though I don't yet see the gears of how it fits in.
To control the world you have to predict it ?
"This is the most exciting paper I've read so far this year and an important addition to my understanding of predictive processing and psychiatry in general."
Writing this as the *last sentence* is some pretty serious lede burying, Scott. Maybe put this sort of thing in the first paragraph.
Are there known depression cases in primitive societies? If there are, what percentage compare to modern?
"In reality they average 0, ie completely neutral. You're depressed, so you have a prior that they'll average -9, ie quite bad indeed. And you assess things based 66% on your prior, 33% on evidence. You go on an average date, ie 0. Given your prior of -9 and your weighting, you perceive it as having a valence of -6. You expected things to be -9 bad, but they were really only -6 bad; it seems you were overly pessimistic. You update your prior some bit of the way between -9 and -6. Do this enough times, and you should be able to get your prior all the way back to 0, ie eliminate your depression. Right?"
Maybe it is, but how many people who predict -9 dates go on dates? When my depression is at its worst, I definitely predict terrible outcomes to things that normally are at least a little positive. Or if I intellectually know they're usually positive, I'll hyperfocus on some small part that could be hard or negative, and be convinced that that part will overshadow any good things. Whether or not people tend to push through (and thus get re-calibrated) could be a major factor for which attractor states they tend towards.
What makes this theory particularly unintuitive (whether coming in with knowledge of Bayes or not) and hard is that 'memory' gets placed in the 'new experience' bin, rather than the 'prior experience' bin !
But I guess that these 'priors' are still a very specific kind of memory, in the 'architecture of the brain' sense ?
We're getting here in the hard (but fuzzy, it's *philosophically* hard) frontier between 'mind' and 'body'...
"In theory you should be able to do the same with traumatic memories: you think about the traumatic event, but since the trauma's over and thoughts alone can't harm you, you're fine and you habituate to the trauma and don't mind thinking about it anymore."
Anecdotally, when an IED hit my truck in Afghanistan this was exactly what I did whenever there was down time, starting a day or two after medevac. I would relive the explosion, the adrenaline hits, after the memory-experience was over I would check my surroundings and verify to myself that I was in a hospital bed and everything was fine, and then back in for another round. I repeated the process until I didn't get the adrenaline spikes; might have taken a hundred iterations over the span of a week.