[Epistemic status: Total wild speculation]
I.
The predictive processing model offers compelling accounts of autism and schizophrenia. But Surfing Uncertainty and related sources I’ve read are pretty quiet about depression. Is there a possible PP angle here?
Chekroud (2015) has a paper trying to apply the model to depression. It’s scholarly enough, and I found it helpful in figuring out some aspects of the theory I hadn’t yet understood, but it’s pretty unambitious. The overall thesis is something like “Predictive processing says high-level beliefs shape our low-level perceptions and actions, so maybe depressed people have some high-level depressing beliefs.” Don’t get me wrong, CBT orthodoxy is great and has cured millions of patients – but in the end, this is just CBT orthodoxy with a neat new coat of Bayesian paint.
There’s something more interesting in Section 7.10 of Surfing Uncertainty, “Escape From The Darkened Room”. It asks: if the brain works to minimize prediction error, isn’t its best strategy to sit in a dark room and do nothing forever? After all, then it can predict its sense-data pretty much perfectly – it’ll always just stay “darkened room”.
Section 7.10 gives a kind of hand-wave-y answer here, saying that of course organisms have some drives, and probably it makes sense for them to desire novelty and explore new options, and so on. Overall this isn’t too different from PCT’s idea of “intrinsic error”, and as long as we remember that it’s not really predicting anything in particular it seems like a fair response.
But I notice that this whole “sit in a dark room and never leave” thing sounds a lot like what depressed people say they wish they could do (and how the most severe cases of depression actually end up). Might there be a connection? Either a decrease in the mysterious intrinsic-error-style factors that counterbalance the dark room scenario, or an increase in the salience of prediction error that makes failures less tolerable?
(also, there’s one way to end all prediction error forever, and it’s something depressed people think about a lot)
II.
Corlett, Fritch, and Fletcher claim that an amphetamine-induced mania-like state may involve pathologically high confidence in neural predictions. I don’t remember if they took the obvious next step and claimed that depression was the opposite, but that sounds like another fruitful avenue to explore. So: what if depression is pathologically low confidence in neural predictions?
Chekroud’s theory of depression as high-level-depressing-beliefs bothers me because there are so many features of depression that aren’t cognitive or emotional or related to any of these higher-level functions at all. Depressed people move more slowly, in a characteristic pattern called “psychomotor retardation”. They display perceptual abnormalities. They’re more likely to get sick. There are lots of results like this.
Depression has to be about something more than just beliefs; it has to be something fundamental to the nervous system. And low confidence in neural predictions would do it. Since neural predictions are the basic unit of thought, encoding not just perception but also motivation, reward, and even movement – globally low confidence levels would have devastating effects on a whole host of processes.
Perceptually, they would make sense-data look less clear and distinct. Depressed people describe the world as gray, washed-out, losing its contrast. This is not metaphorical. You can do psychophysical studies on color perception in depressed people, you can stick electrodes on their eyeballs, and all of this will tell you that depressed people literally see the world in washed-out shades of gray. Descriptions of their sensory experience sound intuitively like the sensory experience you would get if all your sense organs were underconfident in their judgments.
Mechanically, they would make motor movements less forceful. Remember, in PP movements are “active inferences” – the body predicts that the limb it wants to move is somewhere else, then counts on the motor system’s drive toward minimizing prediction error to do the rest. If you predictions are underconfident, your movements are insufficiently forceful and you get the psychomotor retardation that all the pathologists describe in depressed people. And what’s the closest analog to depressive psychomotor retardation? Parkinsonian bradyphrenia. What causes Parkinsonian bradyphrenia? We know the answer to this one – insufficient dopamine, where dopamine is known to encode the confidence level of motor predictions.
Motivationally – well, I’m less certain, I still haven’t found a good predictive processing account of motivation I understand on an intuitive level. But if we draw the analogy to perceptual control theory, some motivations (like hunger) are probably a kind of “intrinsic error” that can be modeled as higher-level processes feeding reference points to lower-level control systems. If we imagine the processes predicting eg hunger, then predicting with low confidence sure sounds like the sort of thing where you should be less hungry. If they’re predicting “you should get out of bed”, then predicting that with low confidence sure sounds like the sort of thing where you don’t feel a lot of motivation to get out of bed.
I’m hesitant to take “low self-confidence” as a gimme – it seems relying too much on a trick of the English language. But I think there really is a connection. Suppose that you’re taking a higher-level math class and you’re really bad at it. No matter how hard you study, you always find the material a bit confusing and are unsure whether you’re applying the concepts correctly. Eventually you start feeling kind of like a loser, you decide the math class isn’t for you, and you move on to something else where you’re more talented. Your low confidence in your beliefs (eg answers to test questions) and actions (eg problem-solving strategies) create general low self-confidence and feelings of worthlessness. Eventually you decide math isn’t for you and decide to drop the class.
If you have global low confidence, the world feels like a math class you don’t understand that you can’t escape from. This feeling might be totally false – you might be getting everything right – but you still feel that way. And there’s no equivalent to dropping out of the math class – except committing suicide, which is how far too many depressed people end up.
One complicating factor – how do we explain depressed people’s frequent certainty that they’ll fail? A proper Bayesian, barred from having confident beliefs about anything, will be maximally uncertain about whether she’ll fail or succeed – but some depressed people have really strong opinions on this issue. I’m not really sure about this, and admit it’s a point against this theory. I can only appeal to the math class example again – if there was a math class where I just had no confidence about anything I thought or said, I would probably be pretty sure I’d fail there too.
(just so I’m not totally just-so-storying here, here’s a study of depressed people’s probability calibration, which shows that – yup – they’re underconfident!)
This could tie into the “increased salience of prediction error” theory in Part I. If for some reason the brain became “overly conservative” – if it assigned very high cost to a failed prediction relative to the benefit of a successful prediction – then it would naturally lower its confidence levels in everything, the same way a very conservative better who can’t stand losing money is going to make smaller bets.
III.
But why would low confidence cause sadness?
Well, what, really, is emotion?
Imagine the world’s most successful entrepreneur. Every company they found becomes a multibillion-dollar success. Every stock they pick shoots up and never stops. Heck, even their personal life is like this. Every vacation they take ends out picture-perfect and creates memories that last a lifetime; every date they go on leads to passionate soul-burning love that never ends badly.
And imagine your job is to advise this entrepreneur. The only advice worth giving would be “do more stuff”. Clearly all the stuff they’re doing works, so aim higher, work harder, run for President. Another way of saying this is “be more self-confident” – if they’re doubting whether or not to start a new project, remind them that 100% of the things they’ve ever done have been successful, odds are pretty good this new one will too, and they should stop wasting their time second-guessing themselves.
Now imagine the world’s most unsuccessful entrepreneur. Every company they make flounders and dies. Every stock they pick crashes the next day. Their vacations always get rained-out, their dates always end up with the other person leaving halfway through and sticking them with the bill.
What if your job is advising this guy? If they’re thinking of starting a new company, your advice is “Be really careful – you should know it’ll probably go badly”. If they’re thinking of going on a date, you should warn them against it unless they’re really sure. A good global suggestion might be to aim lower, go for low-risk-low-reward steady payoffs, and wait on anything risky until they’ve figured themselves out a little bit more.
Corlett, Frith and Fletcher linked mania to increased confidence. But mania looks a lot like being happy. And you’re happy when you succeed a lot. And when you succeed a lot, maybe having increased confidence is the way to go. If happiness were a sort of global filter that affected all your thought processes and said “These are good times, you should press really hard to exploit your apparent excellence and not worry too much about risk”, that would be pretty evolutionarily useful. Likewise, if sadness were a way of saying “Things are going pretty badly, maybe be less confidence and don’t start any new projects”, that would be useful too.
Depression isn’t normal sadness. But if normal sadness lowers neural confidence a little, maybe depression is the pathological result of biological processes that lower neural confidence. To give a total fake example which I’m not saying is what actually happens, if you run out of whatever neurotransmitter you use to signal high confidence, that would give you permanent pathological low confidence and might look like depression.
One problem with this theory is the time course. Sure, if you’re eternally successful, you should raise your confidence. But eternally successful people are rarely eternally happy. If we’re thinking of happiness-as-felt-emotion,itt seems more like they’re happy for a few hours after they win an award or make their first million or whatever, then go back down to baseline. I’m not sure it makes sense to start lots of new projects in the hour after you win an award.
One way of resolving this: maybe happiness is the derivative of neural confidence? It’s the feeling of your confidence levels increasing, the same way acceleration is the feeling of your speed increasing?
Of course, that’s three layers of crackpot – its own layer, under the layer of emotions as confidence level, under the layer of depression as change in prediction strategies. Maybe I should dial back my own confidence levels and stop there.
I feel like this should then merge nicely with ketamine’s* activation of the AMPA receptor. Which instantly cures depression and according to the theory increases the weight of bottom up signaling.
But intuitively this theory seems to suggest to me that more dopamine should be the miracle depression cure, which is a little true. But it doesn’t cure it like AMPA activation. Maybe a more clever or less sleepy commenter will come up with a good theory about how it all fits together.
* – actual downstream chemical product but whatever.
Edit – I guess if you have depressing high level beliefs then AMPA activation will dampen those belief’s effects on the rest of your mental processing.
It’s not a bad way to think about it. All models are wrong; some models are useful. This sounds like as good of a conceptual framing of depression as anything else. It gives a vague-but-actionable solution to depression if its true:
Get the person to go out and do things that can reliably turn out as expected. Our world today is often complex, and can be unpredictable. How many people talk about getting out of depression by finding some sort of anchor hobby? They pick up running, for instance, and they discover reliably that they can improve their time and duration by running more. Or weight lifting, where they do a week of workouts, and every week they can add more weight to the bar. Good, positive, predictable result that comes true, and give the person improvement in a physical, measurable attribute that corresponds to power. The kind of things that should increase confidence – whatever deep abyssal pit of a baseline they might be starting at, it should get better.
Much like your hairdryer story; if it works, it works. If the possible strategies to cope with depression suggested by the model are effective, then the model is effective.
Now, apply the model in a different direction. Does depression cure gambling addictions?
If increased predictability improved depression it would also suggest watching movies you’ve already seen, or solitary confinement would improve depression. Not just things like weight lifting and running.
Well, my interpretation isn’t simply satisfying a craving for predictability. You may be un-depressed for the duration of the movie, but once your escapism ends and you return to the real world, all that ‘confidence’ should disappear. and it does. That’s why you throw in another movie.
Global Confidence is probably improved more by becoming more capable. Capability in some way corresponds to confidence in the face of the unknown. Which is what life outside mostly is – unknown.
I wasn’t happy and confident walking into new classes at University because I had already extensively studied the material. I was confident because I thought that, whatever the material, I was generally good at learning it and handling it. My assessment of my capabilities was high. And even if I failed at the start, I had the confidence that I could become more capable. Depressed people seem to feel worthless – they feel like they’ll fail with what they do, and probably more importantly, that they can never improve and eventually not-fail at what they do. They feel stuck. If you suck but you can get better – great! Go get better. But if you suck and you can’t improve, then what’s the point? Better to save the calories.
As an aside – intelligent people are generally more capable in any given task than someone less intelligent (on average). So they shouldn’t be depressed as much. And I recall seeing that there is a modest but present correlation with depression and low IQ. But we all know very intelligent people that nevertheless are depressed. So what’s their story? What’s the confounding personality traits you notice with depressed intelligent people?
I would say a lot of times they’re not just intelligent, but really intelligent. They’re used to succeeding immediately at trying something new. They’re easily discouraged. Whenever they initially fail at something, even though they may have the potential to be extremely good with work. And why bother sticking with something they initially suck at, when there are a dozen things they could be good at with no work? It’s a reasonable approach.
But then what happens when life tosses that kind of person a bunch of failures in a row? Suddenly they suck at a bunch of things they can’t afford to suck at. And while in reality they could improve, they have no practice at improving at things. They may believe they can’t improve at things, so they feel incapable and stuck in that state of impotence.
So doing something where you are measurably, incontrovertibly increasing in capability could be analogous to low-level observation overriding the negative prediction of being incapable. It forces your brain to reevaluate this model of “I am incapable.” by shoving your face in some evidence of being good at something, and more importantly getting better at something.
These are all vague generalities of course, but depression as an assessment of permanent incapability might be still-better a model. Maybe it’s a mix. Maybe I’m super off-base. Hard to say.
Yup. That’s me.
I don’t rewatch movies, but I do turn off all the lights and listen to Comfortably Numb on repeat sometimes.
You know what else is a stimulus that can be reproduced reliably? pain.
Interesting point. When I think back on being depressed as a teenager, I remember kind of recoiling at the idea of actually not being depressed anymore. Depression had become kind of comfortable and familiar.
You would expect, then, depressed people to play a lot of videogames. Which often manage to capture and package ‘you are getting better at this’ in an extremely addictive way, and are available for free on the internet. (I mean, you should expect a greater increase for videogame usage in depressed people, than for other escapist media.)
This comment matches my experience pretty well. I’m still not sure my brain knows how to take things seriously, in the extraordinary effort sense. Trying and failing to do so has always been pretty miserable, and when I got to college it was pretty much the new status quo.
But also the training thing. I spent May 2016-February 2017 in a training program specifically geared toward some major skills in which I felt lacking. The instructor/trainee ratio was such that there was a good amount of personalization and pace-adjustment, and there were definitely results. (There are some more specific details that complicate the training vs depression framing, but this comment probably shouldn’t turn into a complete retrospective.)
The word “confidence” came up a lot in training. To the point it was annoying almost immediately. I’m not sure why it was so annoying, but maybe that’s hindsight bias.
My life post-training is not that different from pre-training, except way less miserable. I think it’s partly—maybe mostly—because, impirically, I’m neither hopeless nor helpless. Still don’t grock how to do the serious effort thing, though.
That high-intelligence explanation doesn’t make sense to me. It would seem to imply that “achievement gradients” are optimized for people in the middle of the bell curve, and are capped at the high level. But there doesn’t seem to be an inherent skill cap for any activity I can think of (at least any caps near the entry level point).
I expect most of this comes from primary schooling, which is often optimized for the middle. If you skipped grades, or didn’t attend public school, you might not have run into this problem.
In that case, shouldn’t we expect a U-shaped curve of correlation between intelligence and depression?
Empirically, while engaging in novel activities seems to have some effectiveness for combating depression (which of course makes it harder to engage in novel activities as a symptom,) getting better at some thing doesn’t seem to be a miracle cure for depression. If someone wanted to navigate their own IRB nightmare, they might try testing this by setting up a battery of courses in very quick-to-learn skills which offer visible feedback for improvement, and testing its efficacy as a treatment for depression. But just based on the general observations of my experience, I find it doubtful that it would be particularly more effective than conventional depression treatments, especially if the skills aren’t useful (which they don’t seem like they should have to be if the point is simply to recalibrate one’s estimation of one’s ability to get better at things.)
Does (or doesn’t) watching old movies actually help? Does meditation?
For most of my life, I essentially never re-read books and only rarely re-watched movies, even ones I really liked.
I recently went through a bout of moderate depression, and re-reading fiction was one of the few things I actually felt motivated to do. It didn’t help, per se, in that it didn’t make me any “better” after. But it was soothing in the moment.
Meditation, on the other hand, did help me.
Depression seems to result in people preferring to stick to low-novelty activities over high-novelty ones, but low-novelty activities don’t seem to speed recovery.
http://journals.sagepub.com/doi/abs/10.2466/pr0.1996.78.2.635
says “A number of previous studies have indicated that pathological gambling is often associated with depression. Equally, a number of theoretical models of pathological gambling have included depression as a key variable.” and “Of 19 pathological gamblers who completed the Beck Depression Inventory, 21% were depressive; by contrast, only 9% of the other subjects were depressive. The Beck scores of pathological gamblers were positively correlated with the severity of their addiction as indicated by the number of DSM–IV-specified symptoms reported.”
which is a point against the idea Scott lays out here, at least as I understand it.
However maybe gambling is somehow also associated with low confidence?
Gambling and low confidence together make a lot of sense. “I am a proven loser and therefore it is better to leave the outcome to pure luck, even if the odds are not fair.”
I think it would probably go the other way around.
Gamblers would get depressive because they are for some weird reason confident that they can win, but are proven wrong time and time again. So they gradually loose confidence in their predictions, but due to their gamblemania (what was the correct term again?) they cannot stop playing. So they are trapped in a world where all their predictions get frustrated. No wonder then they are likely to get a depression.
>Gamblers would get depressive because they are for some weird reason confident that they can win
I don’t know much about gamblers, but I think the poster above was supposing that depressive gamblers are the ones who aren’t confident they’ll win, but they still do it because they feel the odds of ‘winning’ are better than their abilities would be at producing the same ‘winning’ feelings at some other activity.
Maybe it weakens their prediction that they would lose, so it seems that the odds of the jackpot are much closer to 1:1 than they ought to be?
Good, positive, predictable result that comes true, and give the person improvement in a physical, measurable attribute that corresponds to power. The kind of things that should increase confidence – whatever deep abyssal pit of a baseline they might be starting at, it should get better.
I wish it worked like that, but it more generally goes “See, this is the kind of thing you should have been doing all along. Normal people can do this. Doing this isn’t anything special. Don’t start thinking you’re so great because you can do this. The only reason you haven’t been doing this all along is because you’re lazy and useless, now all you are doing is the bare minimum an ordinary person does and you think you’re hot stuff? Get real!”
A plurality of my ancestors are from coastal Sweden. When I imagine what they were doing up there, I suspect that summers revolved largely around herring: catching, brining, smoking, pickling. And in the winter, they spent a lot of time sleeping, telling stories, and finding creative ways to eat them. What do you suppose they did it the herring went away? I imagine they put greater effort into finding them. They’d go out further, work harder, maybe even frantically, despite there being no payout. But then, eventually, they’d give up, lose motivation. When morning came around, they’d think about getting up, question the point of it all, and roll over.
This is a simplification of a simplification, but I use something like this with my clients when I’m explaining the impact and functional utility of depression. Our behavior is guided by imperatives that are pretty firmly established, usually. What happens when they don’t result in a payout anymore? Hope do you erase imperatives so you can adopt alternatives? You get depressed. It’s a bit more complicated than all that of course, participating in a modern social context where people are supposed to show up the same way every day, doing the same things, and there’s something wrong with them if it doesn’t work for them.
But the survival advantages of depression and hypomania are pretty obvious when you think about it. It’s the social context that hangs people up.
(If you have a comment I’ll have to respond later, because I’ve got a busy day tomorrow and I’m off to bed.)
As someone who’s depression was often a bit cyclical (no true mania, but repeated alleviation of symptoms after days or weeks that would feel high energy and high confidence by comparison) I’ve often thought that it has to be part of a system for encouraging disengagement from imperatives or just rest. That system is just maladapted to our environment for those of us who experience depression.
Also, CBT, mindfulness, etc. (i.e., the non-pharmaceutical treatments with some effect) definitely have an element of predictability and practicing thought patterns that can be expected and even rerouting expectations. To my maladapted brain that fits with PP very naturally.
I don’t think Scott has quite gotten his head around the idea that mania represents an extinction effect; the internal payoff for behavior is no longer forthcoming, to which a person responds much like the rat anticipating the food pellet that failed to appear after the bar press (or whatever). Subjective optimism is secondary, and doesn’t necessarily accompany this increasingly frantic pursuit of payoff (eg mixed episodes are not optimistic, and prenatal mixed or manic episodes, while more intensely goal-directed, really can’t be characterized as optimistic). In both mania and depression, there’s a disengagement process, and people are prone to take increasing longshots, whether accompanied by optimistic estimates of success, or not.
This is why both (particularly together) are associated with such increased risk of suicide.
Kohut had interesting ideas about the relationship between mania and narcissism/horizontal identity schism that are not at all inconsistent with viewing mania as an extinction effect.
40 days and 40 nights wandering in the desert would probably make a fantastic treatment for depression. But if people end up returning to precisely the same priorities they were busily erasing, treatment is unlikely to stick. (Failure to generalize learning to the home context has always been the bane of increased levels of care. Also, this is both the virtue and curse of various forms of wilderness therapy.)
The plural of “anecdote” isn’t “data”, but…
I’m in my early 20s. Since about the time I headed off to college, I have moved towards the emotional state of “depression” – worse at college, better since I got out, but still a movement in that direction – and also moved towards a philosophical state of skepticism towards more and more things, starting with the proper rationalist “look at the outside view” and steadily expanding the amount of things I doubt further and further.
I tended not to correlate these things – depression feels like an inevitable consequence of not actually accomplishing things in the real world, and recognizing good arguments to doubt new and exciting things is hard mental work that feels good, like I’m making real philosophical progress – but now I wonder if maybe I ought to.
I have a very minor version of what I think you’re talking about. Over the course of these crazy past five or six years, I have become rather jaded in the public sphere and the political deliberations that take place in it. Xorgials aren’t xorgial, anti-blerpians are pretty blerpian themselves, Doingians don’t care about doinginity, etc. etc. My political tribe has betrayed every single moral principle that made me support them in the first place, and the outgroup is as terrible as ever.
So, I extended the tools I was taught to apply to the claims of the outgroup to my ingroup’s sacred beliefs, and voilà, turns out everybody lies. Hardly anyone is interested in the truth for truth’s sake, hardly anyone is interested in moral principles for the principle’s sake. Everything anybody tells you is probably based on motivated reasoning and I’m better off distrusting it.
But that’s a terrible world to live in! How do I escape?
Embrace it, let it corrupt you, and emerge on the end to discover that you still enjoy making other people happy, that you still enjoy finding out the truth, that you still enjoy being honest, and that you still value beauty.
It’s a bumpy ride for sure, but in my estimate worth it.
Thank you. So it’s a ‘tend to your walled garden’ kind of thing? Sure, I think I can find relief there. Doing small things and succeeding at them, and all that.
But I don’t enjoy finding the truth anymore. It alienates me from my peers. And I don’t enjoy being honest anymore because I run the risk of being ostracised because of it.
Interesting that you read ‘tend to your walled garden’ out of my comment. But yeah that too was part of it for me.
It’s okay if you don’t enjoy those things anymore, they will come back when you have regained a steady footing.
About three years ago I more or less decided to abandon my values because I couldn’t see how I could live on with them and it was surprising to me that even though I gave myself permission to be evil I stayed good even from the perspective of my former self.
Organize, protest, petition. Take politics beyond mere voting. Even when you lose, and lose hard, losing in solidarity with like-minded people who have skin in the game feels a lot better than atomized, alienated failure. The only reason politics-as-usual proceeds is because so many people don’t care anymore to use the full spectrum of their democratic rights
Sure, but:
The only people who care about principles are philosophy nerds. For everyone else, they are purely instrumental.
That’s called ”growing up” and moving into the post-conventional stage. You are well beyond teenage years, perhaps middle-aged or even older, and you gradually realize not only are you surrounded by idiots, but you have no clue, either. It then dawns on you that the “adults” — those smart people you knew all your life are all frauds.
They were clueless all this time, and made things up as they went along, pretending they had some insights that are unlocked at some point when one reaches a certain age or place in life. But that moment never came, and you’re now well beyond convincing yourself that it’s just right around the corner.
You’re now realizing that it’s all a ship of fools, and you are one of them. The next stage is, spoiler alert, realizing that therefore anyone can grab the steering wheel and set the sails however you want, and that person might be you.
Eerily accurate, except for the last part.
Isn’t that up to you?
I’ve had a similar experience. I first started dealing with what I would consider to be “depression” when I was probably ten or eleven years old, and dealt with it pretty continuously into my early twenties. This was accompanied by a general, mostly-unarticulated attitude of philosophical skepticism. Then I had a religious awakening that coincided with attending and becoming very involved socially in a religious university. I simultaneously became less skeptical and less depressed. Since leaving school, my skepticism and my depression have both crept back in basically hand in hand. Lots of things about that are hard, but one of the more important ones is that the skeptical/depressed outlook seems to be right intellectually. (Cue cite to Alloy and Abramson re: depressive realism.) I also think I’m a kinder and less judgmental person when I’m depressed (and my wife thinks this too for whatever that’s worth). So I don’t want to give those things up but I don’t want to be depressed either. Rock, hard place, etc.
I think the best way to defeat this is to point out that, yes, the human brain has other drives. Specifically, the drive to eat enough to live, and other such. If you as a brain assume that you will have to experience new things at least some of the time, it makes sense to seek them out and figure them out ahead of time. Sounds reasonable enough.
I haven’t noticed the vision symptom, but the rest of this seems to match my experience of depression pretty well.
I’d agree, except I know my recognition of color gradients isn’t as good as others, as I am quite notably inferior to others at noticing changes in facial color (blushing, etc)
Other than that, I don’t feel like things are gray
This sort of fits well with my model of a major contributor to depression, to the extent that there is such a thin: you want to change something but can’t, and you also can’t stop caring about it.*
If I want to make all the bad things in the news stop happening, I might naively think that I can accomplish that by going out and doing direct action or political advocacy or whatever. In fact, in the ancestral environment I would justifiably expect my actions to have a noticeable effect on the overall goodness or badness of the world reasonably often. In a hyper-Dunbar world, nothing I do makes the amount of bad things on CNN go down, so it starts to feel like nothing I do matters. This is sort of like expecting your limbs to move and then not seeing anything happen if you squint enough. The really nice thing about this theory is that it explains the contrast between the high rates of depression in the modern world and the (allegedly) low rates in hunter-gatherer tribes.
I’m also sorta convinced that depression is an response that has been adapted to serve the dual purposes of A) getting you to stop doing counterproductive or harmful things and ruminate instead when nothing is working like you expect,
and B) to get you to avoid expending needless energy or expose others to your germs when you’re sick. If depression is a response that can be triggered by either one of these two things, then that explains the weird link with the immune system.
*Consider the two major genres of advice for dealing with depression: “Get out there and do something!” and “Let it go.”
Edit: Some quick anecdotal evidence. I’ve been depressed to varying degrees ever since middle school (I’m 26 now) and the main things that cause flare-ups are:
– things that make me feel helpless (the sexual assault of a close friend, the rise of the surveillance state)
– things related to disease, allergies, or inflammation (mosquito bites, existing near a cat, the common cold)
– learning that one of my major (positive) assumptions about the world was wrong (my disillusionment with higher education, the “arc of the universe” not bending inevitably towards improvement)
– constant, low grade (or higher) shittiness that I’m powerless to stop, including long-term life circumstances (stress at work, [redacted])
As a metaphor for the helplessness/actions that creating the change you expect, take the concept of the grim/noble and dark/bright spectrum:
This (and most of the rest of the comment) rings true to me.
Doesn’t this rather neatly resolve the paradox that groups that actually suffer very little (stereotypically rich, white people in western democracies) in a global perspective seem to be more depressed? It’s not about the amount of trauma and distress, but about the predictability of modern society. Life is just getting too complex for all but the most dopamine-soaked brains.
This seems false. Modern life is much more predictable than life in a pre-modern society.
Maybe so and perhaps it’s a gross oversimplification to say modern life is more/less predictable than a pre-modern life. In this modern world, I don’t have to wonder when the next eclipse will come but in a pre-modern world, I won’t have to wonder what traffic will be like tomorrow.
Anthropologist James Suzman studies bushpeople in Botswana. He uses the term “an unyielding confidence” [0] to describe the expectation the community holds with respect to the environment providing for their needs.
And perhaps that’s an appropriate amount of confidence for the bush people to have. If they can’t find birds that day, they can count on eating some roots. And they know the weather isn’t going to kill them and they know where they’ll get water.
In some ways, modern life is more secure (food banks, drinking fountains, homeless shelters) but also more tenuous. Most Americans would feel the heat if they lost their income tomorrow. And although we know that provisions exist, we also know that learning how to navigate a county rent assistance bureaucracy is more involved than say rooting for some grubs.
Maybe modern life is more predictable (maybe) but tomorrow holds higher variance in the modern world than the premodern world. Tomorrow might be the day you get doxed, or fired, or have a car crash, or win the lottery, or get audited, or find the love of your life, etc and maybe all those things can happen in pre-modern worlds but the modern world seems to have categorically more outcomes and events and more things are likely to happen on a given day.
Perceptive people in the modern world should lower their confidence about specific outcomes in the future in light of the variance they see. Look at all the mental health problems people reported around Trump’s victory – maybe those people reassed their expectation of varience and their confidence in realizing good outcomes in the future dropped below some criticality threshold that manifested as a mental illness.
Variance might not even be the right word, perhaps some people seem to have a median or mode driven (rather than a mean driven) set point. And I’m inclined to think that’s the case, look at risk-aversion biases / results in behaviorial economics.
In fact, risk aversion and predicibility might form a better explanation than just predictability alone. A bushperson might feel more secure in their life regardless of ‘random’ events than someone with a mercurial boss, high debt, and not a lot of cash. Even if they both lose everything, a person in the modern world can feel the loss for a longer time (e.g can’t get loans after a bankruptcy) and might take more time to restablish themselves. Square one for both people is different, a bush person might lose everything. A modern person might lose everything good but still retain the bad: debt obligations.
[0]: https://www.newyorker.com/magazine/2017/09/18/the-case-against-civilization?mbid=synd_digg
Agreed. I would add that, even though life is rough for a hunter-gatherer in a small tribe, it also feels more in control. They have much more leeway to deal with their problems. If they are hungry, they can hunt; we can’t, we don’t live where game is to be found. If they don’t find game in the bushes today, they know they can eat nuts from that tree over there; we can’t do this, the trees have owners, and there are not enough trees for us all. And for whatever thing they cannot control (diseases and similar), they always create a god or a spirit that can be excited or placated through witchcraft, to feel more in control (I am not saying that this is great for their happiness or general mental well-being, though. It can bring terror and paranoia, for sure).
Modern society created a lot of wealth for more people than any hunting grounds could ever sustain, but at the cost of limiting the opportunities each person has to deal with their problems the way their brain prefers. Many behaviours have more consequences and are more difficult to express in the modern world than in the primeval savanna – which may be a source of anguish. Hunter-gatherers never get stuck in traffic, for example; they are never trapped in that situation where you want to move and you can’t: you can’t go forward, you can’t go back, you can’t cut way on the sidewalk, you can’t abandon your car on the street and come for it latter, you can’t do anything to make the other people move. You can’t do a thing, and you never really know how long it is going to take for the problem to go away. Where they live, they can always move around; we often can’t.
Predictability, in this sense, is about what the brain expects, not about an algorithm. For most of human history “food shortages happen” is a fact of life, so our brains should be programmed to anticipate shortages sometimes. In modern life they basically never happen for most people and thus your prediction circuit would feel off.
Perhaps it should be clarified that expected payoff is key. A primitive society has very low expectations of payoff for just about everything, so every meal is a victory. A modern society has high expectations of payoff for almost everything, so achievements with comparatively lower expectations require highly complex events to realize: the perfect job, the perfect relationship, etc.
So, for a member of the primitive tribe, a successful kill for the week’s dinner has the equivalent payoff as finding the perfect job does for a member of modern society; but the former is probably a lot easier to achieve and more likely to happen frequently.
I may be typical-minding here, but this is pretty obvious to me. It’s obvious when observing my own reactions to things: happiness occurs when things are improving, not when they’re good. Growing more capable and making your way in the world leads to happiness, not achieving some particular outcome*. Does not everyone feel this way?
(Null Hypothesis mentioned above that depression among the highly intelligent seems unexpectedly common. I’m not sure it has to do with failures as such, can it rather be the sudden disappearance of a structure in which to achieve and climb in a reliable, steady way with constant feedback? I.e. education and finishing it. In that case, such depression should manifest a while after someone has completed their education (and fails to make obvious and tangible progress). Is this what happens?)
It explains why success doesn’t make people happy, it must be success of a steadily increasing scale and/or scope. Maybe that’s a youth thing though (I mean, ambition is seen as a youthful thing) and normal people wind down as they get older to focus on the new generation and get happiness from tranquility. All except a small number of freaks (like business leaders and politicians??) that become unhappy if they can’t keep succeeding in bigger and bigger ways.
*This is part of what makes me think of many versions of utilitarianism as wrongheaded. The idea of a utility function (and by extension things like wirehead Gods on lotus thornes or Heaven) seem to fail to capture our sense of value: we don’t value outcomes or world-states, but processes and changes. We don’t want to the world to achieve some ultimate goal, we want it to keep getting better, keep doing challenging new things successfully.
“What is good? — All that heightens the feeling of power, the will to power, power itself in man.
What is bad? — All that proceeds from weakness.
What is happiness? — The feeling that power increases — that a resistance is overcome.
Not contentment, but more power; not peace at all, but war; not virtue, but proficiency (virtue in the Renaissance style, virtù, virtue free of moralic acid)…”
Friedrich Nietzsche, The Anti-Christ, 1895
Yes (depending on the meaning of “power” and “war”, though) I think Nietzsche got a lot right.
John, agreed, this is pretty obvious and I think there’s a lot of empirical research that backs this up (the theory that happiness results not from the steady good state, but from the feeling of progress or improvement). I’ll bet Scott (or some of the others good at digging into the psych vaults) can lay hands on some of this research very quickly. I, being somewhat depressed at the moment, am not going to make the effort, and will just wave my hands and say that it’s there, trust me. And like you, I find it intuitively obvious, and it correlates strongly with my own experience.
Weirdly enough, this is something Predictive Processing handles well: the higher levels of the predictive models are usually models of trajectories, not states. Physically, this makes sense: the laws of physics might be Markovian, but our specific knowledge of physical quantities is emphatically not. Given the location of the baseball at time t and then the location at time t+1, I become more confident about t+2 than if I had only observed t+1.
If our mental models are about trajectories (paths through some space over some causal ordering of “intuitive time”), then it would make sense that our intuitive, brain-processing-level sense of value also deals with trajectories. After all, why would you make a separate representation for value-states, instead of just assigning value to anything that your generative model already handles?
Hold on, I’m remembering a quote from somewhere:
Now, if we’re starting to have a neuroscientific, computational-cognitive grounding for making statements like that – if we can see it sort of shaping up on the horizon – what’s the appropriate response?
Haha, well I guess I should have expected EY to have thought of it.
I do wonder why it isn’t seen as a bigger issue regarding goals though, as it suggests we shouldn’t aim to achieve goals but… aim to aim to achieve goals? It makes any utilitarian calculus orders of magnitude more complex that it already is.
Doesn’t it make our whole value system turn in on itself and become some Gödelian knot? The goal isn’t the point, the striving is – but the striving isn’t meaningful without the goal. It seems to me like it only works if you refuse to look at the whole system at once.
1. Notice how the word “serotonin” was never mentioned in Scott’s post. That neurotransmitter certainly plays some role in depression, schizophrenia, hallucinations, but does not have a place in predictive processing model.
2. The consequence of low confidence – > dopamine hypothesis of depression is that antipsychotics which are dopamine antagonists should cause depression-like mental states. Is that so?
3. I guess (wildly speculate) the “missing link” between AMPA activation by ketamine metabolite and SSRI action is BDNF, a protein secreted by neurons. It can activate special receptor – TRKB – on the surface of (neuronal) cells, and launch signaling cascades which lead to growth of neurons, formation of new dendrites and other events.
This may translate into PP: it’s not a low confidence that causes depression, but lack of response to prediction error or successful prediction.
4. Related, and even wilder speculation: probably the state itself should not be mixed with emotional reaction to the state. The low BDNF state could be provoked by some external factor, totally unrelated to failure, loss or what is perceived as a cause of depression. Instead, the BDNF deficit could be caused by inflammationion, or administration of the drug that intervenes in signaling at TRK or downstream. It’s at some point in life (at childhood? after some stress?) that brain learns to associate low-BDNF states with badness.
5. The weak spot of glutamate theory of depression is why all the AMPA modulators that were developed in 1990s are not antidepressant drugs now?
Yes, let’s revisit the neurotransmitters question. The PP theory Scott offers feels like quite a stretch to me. As a bipolar 2, I have a lot of experience with depression (and regrettably far less with hypomania), and this whole explanation just doesn’t seem to click in any way with my experience. Which is not to say that it is wrong — what’s happening under the hood may show up very strangely in subjective experience — but occasionally you find one of those explanations of mood phenomena that’s so intuitively right that you just say, “Yeeeeeesssss!” This is not one of those moments for me. Plain old neurochemical theories generate more of a click. When I take Typtophan at bedtime, I wake up less depressed. Serotonin! I recently tried Lion’s Mane mushroom which is considered a nootropic because it promotes BDNF, and it quickly made me hypomanic (which then led to a depressive crash, because bipolar). Now maybe all these chemicals are just mucking about with the confidence levels of my neural predictions — but it doesn’t feel that way to me. When I’m depressed it just feels like there is no joy to be had in the present moment. Like my brain is dry. And I lose that feeling that one normally has of wanting to be alive (which is not the same thing as wanting to be dead). None of this really feels to me like changes in the levels of neural predictive confidence.
Nice. I wonder if PP theory could similarly provide some insight about growth mindset?
When you say “confidence,” do you mean something along the lines of comfidence intervals, or the emotion of feeling confident? If anything, depressed people’s predictions seem to have really narrow confidence intervals, and manic people’s predictions seem to have very wide confidence intervals.
Aside: thanks for having read and reported on this book.
Legal-minded SSC readers will find that Scott’s prediction-centric theory of mental illness integrates seamlessly with progressive theories of human rights, provided solely that “prediction” is identified as a “capability” … surely this identification is natural?
For details, see (e.g.) Martha C. Nussbaum’s and Carla Faralli’s “On the New Frontiers of Justice: a Dialogue” (Ratio Juris, 2007) and Martha Nussbaum’s “Human Rights and Human Capabilities” (Harvard Human Rights Journal, 2007).
As a concrete example of the moral lens that an Alexander / Nussbaum unification provides, consider the (very difficult) problem of counseling a high-IQ, yet severely depressed / distressed STEM-student, who upon investigation proves to have had a childhood notably deficient in STEM-related toys (e.g., blocks, lenses, candles, chemistry sets, mechanisms, saws, screws, knives, compasses, rulers, etc.).
Here there is no deficiency in ratiocinative capacity: shown a putative solution to a STEM-problem, the student can readily verify its correctness vs incorrectness. Rather there is a deficiency in intuitive / affective capacity: the capacity to experience an instinctively joyous “aha” in STEM contexts.
Definitely, a capacity for experiencing “joyous ‘aha’-s” can be cultivated in adults, yet on the other hand, said capacities are far more rapidly and naturally cultivated in childhood.
Moreover, the powerful yet problematic role(s) of pharmaceutic intervention in fostering “joyous ‘aha’-s” — specifically in mathematics, science, and engineering, and more broadly in all spheres of human capacity — provides an extraordinarily rich venue for moral reflection and debate. For example, in which respects can the pharmaceutically induced experience of joy help to foster the Spinozist experience of hilarity?
According to the above-referenced Nussbaum article “Human Rights and Human Capabilities” article
Appendix A of Nussbaum’s article enumerates ten “Central Human Capabilities” — all of which are predictive capabilities in Scott Alexander’s sense. In Nussbaum’s words:
In summary, predictive processing models of human cognition provide abundant medical grist for the 21st century’s progressive moral mills, specifically by associating ongoing advances in cognitive science and medical practice to central human capacities for joyous “aha” experiences, with the latter capacity viewed (by 21st century progressives) as foundational to human rights.
You need to write better.
I would caution against the lure of theories-of-everything. Just because something fits nicely within the framework of the predictive brain, doesn’t mean that predictions are all there is to it. The prior (;-) with the brain should be that its mechanisms are messy, patched together, redundant, suboptimal in many ways, but generally good enough…
I don’t think the math class comparison does enough to explain pessimism in depressed people, because in my experience pessimism isn’t just limited to your own performance. All the depressed people I know are convinced they’ll lose at pure-luck gambling games, they don’t vote because they think their votes won’t make a difference, and they even seem to think new movies are more likely to suck. Most of them admit that they are irrationally pessimistic about these events, which really doesn’t fit the predictive processing model.
I wonder if there are any studies on the probability calibration of depressed people for events entirely outside their control.
The only way I can think to fit this into the model is pretty hand-wavy. Maybe depressed people are pessimistic about their own performance for the reasons you described, then the brain does notices how many bad things there are in the set “outcomes under my control” and generalizes, spreading the pessimism to the set of all outcomes.
I worry that this PP model has so many knobs to turn that it can be made to fit any evidence. Some people do behavior X? Too much top-down/bottom-up. Too low/high confidence in predictions going up/down. High/low system not adjusting/over-adjusting its prediction based on data from below/above.
Does mania always look a lot like being happy, though? I’m not pretending I know anything about this topic, but I’m thinking about Freddie deBoer’s recent farewell post (since taken down), which is burned into my memory.
I’m not going to be summarize it perfectly (and I hope it’s OK that I’m talking about a post that he’s since removed – if that’s a shitty thing to do, I’ll delete this post) but he described manic phases as a crash – a car crash, where he felt utterly out of control and typically entirely dissatisfied with everything he did during the phase. And he clearly did a lot of real work.
He noted that the public’s typical perception of manic phases was that it was sort of a “reward” for depression, but he thought manic phases were as bad or worse than depression, and he wouldn’t wish them on his worst enemy. Nothing in there suggested happiness.
From what I’ve read, hypomania is pleasant, but full mania can get pretty bad.
Funny you mention that – he specifically stated that he was hypomanic.
I don’t know whether it would be for or against your theory but I have an anecdote. When I was younger I was practicing epoché. I mean:
– From Wikipedia
And it was rather blissful experience. Definitely not sad.
I was reading a paper that claimed that people with one particular type of migraine (which I suspect might apply to me) have chronically low levels of both serotonin and dopamine in their brains. Because of the chronic deficits, their brains compensate by becoming hypersensitive to what is there. Occasonally though the brain runs completely dry of one or the other setting off the chain reaction that eventually results in a migraine. Also why migraines might be co-morbid with OCD.
While this explains why the 100% successful person might be unhappy once they reach that state, it doesn’t explain a person approaching a 100% failure rate with increased confidence. I’d expect that person to have increasing misery, even as they gained confidence in predicting their own ineptitude.
Then again, maybe there’s something to it. The counterexample would be someone who finally accepts a new, “lower” state in life, and stops grieving and starts to improve their life again.
A potential alternative: getting results that are better than expected seems to make people happy. Maybe instead of increased confidence in the accuracy of your results, happiness tracks with the upward shift in predictions based on constant successful “surprises”. Why is learning joyful? Because we gain confidence? Sure, but we also gain competence, and we have a shifting expectation as we get better and better. But then if we plateau for a while, we might start to get more confident in our predictions (“I’m still only this good?”), but we get discouraged at the lack of growth. But then again, maybe that’s adjusting our confidence in the growth itself.
I don’t know. It seems like we can rationalize it either way. Maybe the tool is more useful at lower levels.
Edit: I also wonder how this might relate to “flow state”. And how some neurodivergent people enjoy games differently. Hmmm…
It seems to me (a depressed person) that it might be more accurate to say that priors fail to update for positive feedback.
To give an example: if a depressed person wins an award, he’ll be happy in the moment, but he will return to baseline almost immediately, and his predictive model won’t update to suggest that he should do more things to get awards. (Which makes sense from the model’s perspective, since the positive feedback was weak.)
But maybe I’m talking in circles at this point. I don’t know – I’m spitballing here.
Also, discovering that the world’s colors losing their saturation is not just a product of my imagination was mindblowing.
This matches my own experience with depression. I still experience the positive emotions that come with pleasant experiences, but my later memories of those experiences are generally weak and I don’t really experience much of a mood “halo effect”. I’ve also learned to distrust my positive emotions because I know they won’t last so I believe I am doing active mental things to “feel less happy” to avoid the pain of losing that happiness
Maybe also “a depressive person wins an award, and at the very moment the award loses its importance, as what sort of award would be handed to someone like them?”
In terms of drives… could it be that “I will predict wrong about sensory data X about Y often” is, itself, a prediction whose error the brain tries to minimize? Higher Y might then roughly correspond to something like risk tolerance, or openness to new experience.
I imagine the reason the natural human impulse isn’t to sit in a dark room to minimize prediction error is because the higher levels *expect* a baseline level of sensory input, and so being constantly in the dark is actually interpreted as a prediction error, which, if we do it for long enough, is corrected by generating vivid hallucinations (see also: dreams).
Do depressed people react differently to sensory deprivation experiments compared to normal people?
“But why would low confidence cause sadness?”
It’s interesting that this happens even on an extremely general level. For example, people are happier if they are fervently religious or strongly atheist, rather than being vaguely agnostic, so much so that people on both sides tend to irrational confidence for its own sake. So “people like being sure” seems like a pretty good explanation for why low confidence would cause sadness.
The link in “You can do psychophysical studies on color perception in depressed people” doesn’t seem to go to the right place. It doesn’t say anything about color perception, and it links to the same article as “Chekroud (2015).”
If the hypothesis is that depression is related to neural underconfidence, I’d have to ask how that squares with the hypothesis that autism is also related to neural underconfidence. The mechanism sounds the same (small error bars for predictions,) but the outcomes are completely different. I’m not sure how to reconcile the same phenomenon being responsible for both the world-of-gray effect and persistent irritation with shirt tags; anxiety seems pretty distinct from malaise.
Also; if under-confidence causes depression and autism, over-confidence should cause happiness and not-autism. I’ve got Aspergers and suffer* from chronic happiness which should be impossible?
*: I will have to use the word ‘suffer’ until someone comes up with a word that implies something obviously out of the norm but also the exact opposite of suffering…
I think my understanding of this model is different from yours, and in my understanding autism and depression are caused by different parts of the model going strange.
Top-down, you predict a model of what to expect and the confidence interval in that model. In depression, it’s the confidence interval that is messed up. In autism, it’s the model building.
An example: in most brains, libraries are modeled as quiet. The neurotypical model is more-or-less quiet, it and sends down a high confidence prediction of a mostly quiet space with some walking around, some whispering, and some fill-in-the-blank quietish noise placeholders. A depressed brain sends down the same model but with pathologically low confidence that the model is accurate ( which leaves a lot of space for “something will probably go wrong, what’s the point of going to the library”).
An autistic brain, on the other hand, sends down a high confidence model of an exactly quiet room, with no predicted noises and no placeholders for unexpected noises. Then they go to the real library, and frustratingly, it isn’t what they were expecting. Someone is chewing gum so now they have to laboriously re-write their model to include gum-chewing, as well as adjust their confidences down, which is very emotionally taxing and off putting. But next time they go to the library, there may be a different noise, and their model only includes gum chewing now, so the next visit maybe just as taxing.
From my understanding, one would predict autistic people to need to be more protective of their routines and sensory expectations than average, because if they adjusted their model’s confidences instead, they would feel depressed. One would also predict they would be more prone to depression, or to at least have temporary depression like symptoms after experiencing many model-reality disagreements in a row.
Maybe I misunderstood, but I think small error bars is the opposite of under confidence.
Autists strongly expect things to be routine, predictable, unable to just be like ‘whatever’ if the tag in their clothes moves a bit unexpectedly. But for depressed people, everything is like ‘whatever’.
The dark room is so nice. I could stay in there for most of the day if I could get away with it. Nice low light from the sun being filtered through the curtains over the windows and the trees beyond that. The window cracked just enough so that I can hear the amazing sounds of nature: the birds and the bugs, the lawnmowers and leaf blowers. Curled up in a blanket just staring, staring and creating worlds.
The worlds aren’t always pleasant. Sometimes I die. Sometimes everybody I care about dies but I’m left alive. Sometimes aliens invade the world and enslave us to their will. Sometimes we do this to ourselves. Sometimes there’s torture. Sometimes there’s despair. But sometimes they are pleasant. I can inhabit a world with bright lights and sounds, and smells, and textures, and tastes, with me and those who I care about carefully provided for with fulfilling activities keeping them busy, able to spend tender moments with people who I know will be around forever, and the problems of the world being solved by ingenious adverse effect free solutions. Sometimes the worlds are pleasant, sometimes not, and sometimes it’s hard to tell the difference but they’re always amazing.
Then I get hungry. I have to leave the darkened room but I don’t want to. Sometimes I know exactly what I want to eat and I eat it and it’s great. Sometimes anything specific to eat, I don’t want to eat that but I want to eat something because of how hungry I am. Then I don’t create worlds: I solve the problem of what to eat. This itself can sometimes take hours. I try to talk myself into wanting to eat one of my favorite foods. I try to talk myself out of being hungry. Eventually I give in and eat something, but I don’t taste it.
Am I sad? I don’t think I’m sad. Do you think I’m sad? I remember months after my depression destroyed my life for the first time, when I had a problem to deal with, a… well, depressing… problem but instead of how things have been going, this time, I knew how to deal with it! I was thrilled: I had a problem and I was copping with having a problem! It was so long since I could say that. It was quite an experience being so upset and ecstatic at the same time but I learned then that fixing depression isn’t about not being sad but it’s instead about being able to deal with sadness. I later learned that this is wrong.
Years after this I was describing this revelation to a therapist and she remarked that “it’s often put that depression isn’t about being sad but about being sad for no reason.” This reinforced one of the most important facts about depression that I’ve learned: depression is very different for different people with it. Just because I understand my depression, what it does to me, how to treat it, and what happens when I don’t have treatment, doesn’t mean that I understand these things for your depression. In order to do that, I need to get to know you.
I am convinced that in 70 years what is currently called depression will have more than one name. For someone with a mental illness this is an exciting time to be alive. We, as a species, are learning so much about it. Every decade things are much better then there were the last decade. Medicine knows more about it, there is better treatment, society is more accepting, and people are more understanding. There is this relentless march towards victory despite being far from it even after how long we’ve been fighting the fight.
After Robin William’s suicide, my boss at the time remarked to me about being surprised that somebody who seemed so energetic, so full of life, and so happy could be depressed enough to kill himself. This didn’t surprise me at all. I feel things. I feel everything. I feel everything strongly. I feel great joy. I feel great despair. I feel great pain. I feel great pleasure. I’m passionate about everything I believe. I’m even passionate about how everything I believe is somehow wrong. I can crack my friends up. I can also make them really, really worried. This is probably why I like both the dark room and the amazing worlds of imagination (my own or the ones that others so generously share with me) so much.
Is this excess of emotion the same as low confidence? Is it different? Are they not related? Are they related in not simple ways? Does it matter? Now that I think about it, I suppose I am less confident when I’m in the throes of my depression then when it’s merely background noise. I don’t know if that’s meaningful or if that’s just confirmation bias. For me though, my depression is about the excess of emotion. No. That’s still wrong. My depression is about managing the excess of emotion. I like the strong emotions when I can manage them. When they destroy my life… not so much.
I am confident in my ability to fail. I’m quite successful at it and have the track record to show this. I do think that people fail more often than they realize but with enough examples, it merely becomes imposable for me to ignore. I don’t see this as contradicting the confidence argument for depression. Perhaps this is a sign that I haven’t reached rock bottom yet. Maybe confidence in failure is the last bit of confidence to go…
And yes. I do fail. I fail a lot. But I’m still trying. Looking back on the past 15 years of my life and extrapolating, I would give a 50% chance that I will be dead 4 years from now. I was on the path to starvation nearly a year ago before my mother realized it and stepped in and helped me. I want to change that percentage (to be clear, I want to decrease it) but I don’t think I will. I still try. I try hard. Even though I think I will fail if I try, I know that I am guaranteed to fail if I don’t try.
The worlds not a certain place. It’s doubtful that one of the worlds I create will come to pass (especially the ones that are imposable by the currently known laws of physics) but one of them might. If one of them does, I have no clue whether it will be a pleasant one or an unpleasant one. That’s frightening but it’s also hopeful. Not being certain has its advantages.
I think I’ll go back to that pleasant room now. I don’t know my destination yet but I’m excited to find out what it will be. And maybe, just maybe, it will give me enough motivation, one way or another, to do two good things today. Wish me success.
This was a very moving post. Thanks for sharing.
I do, very much so.
As we all do. And as we ever should.
This model is making me feel depressed.
I’d love to think that this state of mind is simply a matter of my predictive processor being miscalibrated – great, hand me the screwdriver! I’ll get this working again!
And what does the screwdriver turn out to be? As Scott says, CBT with a coat of paint slapped on. And I didn’t find CBT any help at all to me.
I don’t know. I don’t feel this model is correct because it doesn’t jibe with my experiences/sensations, but then again, what do I know? I do think there is something more there than merely “ah you just need to pull yourself up by your bootlaces – go out, do stuff, you’ll find that the predictions of failure/unhappiness are wrong!”
Ah, no, no I didn’t find that when I gritted my teeth, pulled my socks up, told myself to adopt a positive attitude, that no more excuses I was going to go out and achieve, damn it! And I gave an entire year to it and ended up with nothing at the end, and felt even worse than if I hadn’t tried.
Because not trying means that you can always think “Okay, I can change this if I just pull myself together and put in some effort”. But you try 3 times a week, 4 weeks a month, 12 months a year and nothing improves and you just get the same old runaround (plenty of “oh you’re so good” when you’re the one putting in all the unpaid effort and making no demands but getting silence and tumbleweeds the moment you ask ‘so you’ve been telling me what a great job I’m doing, any chance of actual paid work?/hey I need a favour, can you help me with this thing?’).
After a year of that, you do recalibrate the model to “why yes, objectively, I am a failure”.
I’ve been wanting to be dead since I was 12. So, predictive model, tell me what went “sproing!” at that age that caused this? The grave is the ultimate dark, quiet room, I suppose.
Obligatory disclaimer: I’m in a low mood right now. I had another crying fit for no reason when I got home from work this afternoon, so I think the ol’ depression is kicking in again, and that’s why this comment is so downbeat.
Also, any chance of a professional opinion on this? Is it only really depression if you’ve tried killing yourself? And it’s not depression, it’s just mood swings if you just want to be dead but are not actively trying to kill yourself? I’m asking because in that one session I had with the Official Counsellor from the Official Counselling Scheme they said that – it sounds more like mood swings than depression – and my GP when I asked about anti-depressants asked the “are you self-harming? have you made a suicide attempt?” questions and when the answer was “no”, it was “then you don’t need anti-depressants, here’s the Official Counselling Scheme to apply for instead”.
I mean, I’m mainly going by self-diagnosis here and I could be over-dramatising (that CBT “catastrophizing” thing) and thinking it’s depression when it’s only mood swings.
Now this post I completely understand.
I’m sorry your country’s medical professionals are apparently being such asses about providing drugs. While antidepressants don’t do a ton for me, antipsychotics saved my life.
Have you considered trying Ketamine?
Not regularly, but just once. Or many once every few years.
I am being completely serious.
So does this theory make any observable real world predictions that differ from other theories?
So we can test if it’s true, I mean.
Can a small bias towards positive surprises explain boredom and novelty seeking?
Could you explain what you mean by this?
I too have been curious about boredom. It doesn’t quite seem to fit into the PP model; if prediction error reduction is what our brain is trying to achieve it seems that repetitive and monotonous tasks would be the way to go about this. Though it is quite obvious that people loathe this type of thing.
I’m also curious how well the model explains ADHD, something that is very highly correlated with autism. It, too, seems to defy the PP explanation, as those of us with it crave novelty and seeking new information.
Does anyone have any thoughts? Or could someone explain how I might be seeing this the wrong way?
Tangent:
I wonder if our modern rich caring society makes it way too easy to actually sit in a dark room forever. Both because we can afford it and because nominally caring systems let us stay home and do nothing “until we get better”.
Maybe back in the day when the stark realities of existence forced people to do things all the time, depression couldn’t take root as permanently.
In support of your theory, I’ve struggled with depression since I was a little kid, and as the story commonly goes, none of the medicine worked, but being on nothing also didn’t work. After running through all of the usual treatments, I had reached a sort of point where I accepted it as the way things were going to be and became more interested in countering the symptoms, namely the lethargy and apathy. To this end, my psychiatrist eventually decided to try small doses of amphetamine to see if it would be effective, and it’s closer to a miracle cure than anything I could have ever dreamed up. This makes sense within your analysis; if depression is the low-confidence equivalent of the amphetamine-induced mania, then nudging in a manic direction could increase sensory confidence and push prediction confidence levels in general closer to normal values.
I’m a single data point, obviously, and there could easily be other factors at play, but everything you’re describing makes a lot of sense to me as someone who has dealt with it for years. The efficacy of increasing confidence levels on all of the other symptoms of depression is striking, which does lend credence to the notion that it’s another thing that predictive processing explains well.
That’s because there mostly isn’t a very good account. Friston et al tend to claim that, for instance, dopamine encodes precision of prediction-errors rather than reward-prediction errors (notice the hyphen placement!), even in brain components that everyone else thinks do reward processing.
Now, there’s all kinds of subcortical circuitry which seem to help with affective/valenced processing, and are explicitly noted to amplify (apply more gain, notionally more precision) affectively salient (ie: precise and valenced) signals.
If I really had to take a running leap here, I’d note that the Predictive Processing theory also doesn’t precisely tell us when the brain should resolve prediction errors with perceptual inference (change the predictions, upwards) or active inference (change motor commands to change sensory signals, downwards). They claim that it will do both whenever possible to reduce total prediction error, but that’s not really a prediction (ahaha). So we could turn it around and say: valence is a second type of signal gain on predictions, which governs the balance between perceptual and active inference. Psychological predictions here are trivial, but subtle: if this theory is right, then you’ll be more likely to act on a strong affect in preference to changing your mind about something.
Which… is… exactly what we observe. So much so that I’m just going to call those observations inputs to my theory-making here, and therefore completely ineligible to count as any evidence in its favor whatsoever.
I personally have high-functioning depression where I tend to not be very sad and have comparatively high levels of motivation. But, I experience substantially decreased feelings of happiness and I have very low confidence that my actions will work out positively in the future. I force myself to do things anyway with my willpower because I know it’s better for me, but it’s very tiring and all of my positive experiences end up feeling like “work”. This predictive model matches my experiences well, and it seems likely that many of the other symptoms like sadness are being helped by my medication while the central neural problems still exist. I am at least somewhat on the spectrum and have a very hard time predicting/measuring my emotions and physical states.
When you’re depressed life feels REALLY HARD, and I can think of two ways this makes sense in the theory. One, it takes more mental effort for me to generate positive predictions, or maybe I’m evaluating the resulting error wrong from my actions, and it always feels like I’m very far off from predictions. So I wonder if depression isn’t necessarily about weak initial confidence, maybe it’s more difficult for depressed people to resolve differences between our confidence and reality, so we learn to lower our confidence in order to decrease the error.
Huh. Never mind the entire rest of the post, I clearly have dark room-itis. As in I am incredibly (motivated to become) good at what I do (I’m interacting with this system? I must predict it perfectly!), yet at the same time very hostile to learning new things (Eww, a system I can’t perfectly predict, get it away from me).
Like I held off for years on picking up a game because “new things bad”, then once I kicked myself over that hump I spent an entire week reading wikis and strats 8+ hours per day and loving every minute of it.
Even “knowing” that I will most likely enjoy acquiring and exercising mastery, it’s hard to touch the icky new stuff.
My brother suffered from numerous mental ailments and this was back in the 70’s/80’s and they were not really good at diagnosis let alone cures. Before his “break” he was supremely confident, always on the go, and very driven to succeed (the opposite of depressed). Afterwards, he was looking for answers into what “caused” it, and got more and more depressed as a solution was not found- periodically he would get his manic energy back but it never lasted. You can probably guess where it led.
Not sure if this is common- there was one kid in my school with the same overconfidence and later a break. He now considers himself “broken”
I always looked at myself and my children very carefully for overconfidence because I associate that with the risk of the other extreme not far down the road.
http://hyperboleandahalf.blogspot.com/2010/06/this-is-why-ill-never-be-adult.html
What does the dark room analogy have to say about seasonal affective disorder?
Entirely speculative thoughts:
My intuition says this is not a great picture, because my experience with depression seemed more related to the interaction between your usual executive-operation systems and… something else in the brain. Something like your pre-wired instincts to fit in, seek validation, find a mate, etc.
I haven’t seen anything about how the predictive processing model interacts with, say, hormones, but I wonder if that’s related — for example, if an adolescent is being driven biologically to seek out relationships/sex, but lives in a world where they fundamentally don’t know how or can’t do that, they might end up in a sort of ‘analysis paralysis’ where their models are unreconcilable with their instinctual drives. And this sort of ‘gridlock’ is what causes the system to misbehave or shut down.
My experience with depression, off and on through my life but especially in college, involved a feeling of this sort: that my body/brain wanted to fit in, to accomplish social goals, to have relationships, to be deemed worthy and attractive by peers, etc — but also due to my habits / training / excessively rigid moral system / dangerous self-concept, felt quite unable to do so. All of the progress I can recall in my life on depression consisted of ‘re-modeling’ myself to make my interactions with the world less.. paralyzing.
For instance, after years of “abstaining from alcohol because I associated it with dishonesty and unfaithfulness, but also wanting to fit in in college where interactions were often mediated by alcohol”, I managed to shed my previous model entirely and adopt a much healthier attitude towards the whole thing, which almost immediately lifted a huge weight off my shoulders and caused me to feel vastly less depressed.
And it would not surprise me if part of the reason for the massive prevalence of depression is that, as society becomes more global and different cultures with different norms collide and interact more frequently, it’s harder and harder to have a mental model of behavior that isn’t in constant conflict with the world around. Some people can comfortably slip into a new model as necessary to fit in wherever they are — but I suspect many cannot, and instead are paralyzed and lose the ability to be driven to do anything.
(And I’ve noticed that when I have one of these big paralyses looming in my mind, my day-to-day behavior tends to reduce to whatever’s easiest and requires the least motivation to do. In a state of total confusion about how to interact with the world, it’s easiest to just play video games and get some quick dopamine hits.)
Not a bad tree to bark up. It does seem a bit MacGuyvered compared to the previous posts, but let’s feel around it a bit with anecdata and see its spans and contours.
I reliably get depressed in two cases, which I have (thankfully) mostly learned to avoid. 1. Eating a specific food I’m allergic to (this actually creates the most dramatic effect). Inflammation something something neuron processes. 2. Not experiencing novel, challenging circumstances enough.
The thing that made a huge difference in my life was simply forcing myself to experience new things reliably, and then paying close attention to what my actual, gut-level response to that experience was. I realize this was likely only successful because I wasn’t THAT depressed in the first place, but the difference was still very noticeable.
Could it be that novel stimuli – especially when not ‘guarded’ against or run away from, rationalized into fitting old patterns, etc – keeps the brain’s ability to accommodate new information and paradigms strong?
That sounds quite a bit like increasing the confidence intervals in neural processing. The brain becomes a bit more prepared for a wider spectrum of non-habitual inputs and outputs.
If this is at all true, we’re in a circumstance where there are likely a few levers that can cause depression, with different solutions. My more general lack-of-stimulation dysthymia is analogous to, for example, muscle wasting due to lack of use (solution – exercise) while other forms of more serious depression may be more like muscular dystrophy, which arises from a more distinct, less solvable disease state. Importantly, things like muscular dystrophy can be helped by exercise, so it’s still a good idea! Just not a very complete solution.
I don’t think this is necessarily inconsistent with your model – success and failure are generally macrostates, not microstates. Let’s say we play a game where I pick a number between 1 and 1000 and you get one chance to guess it. If you’re completely unconfident, you might assign equal probability to all numbers, so you’ll estimate your chance of success is 0.1% and you’ll say it’s quite likely you will fail.
Few opportunities in everyday life have a thousand possible outcomes, but its the same thing if you must get 10 binary choices right to succeed, and failure is failing any one of them. This suggests one possible test of the model: are depressed people sure they will fail things with small spaces of possible outcomes, or is this sureness of failing only for things with large spaces? The latter better fits the “too-low confidence” model.
Scott, I advise you to indulge in wild speculation more frequently.
You know, you wouldn’t have to actually implement biological drives, exactly, with this model. You just have to model yourself as an organism that should behave as if it has drives, and your behavior will tend to conform with what the drives would do, if they existed…
The thing that isn’t being discussed here is what happens when the top-down prediction and the bottom-up sensory stuff don’t match at all. When the mismatch is large enough, there has to be some kind of neural exception-handling, based on the surprise. Perhaps the best model for something like this is something along the lines of a Kahnemann-esque system 1/system 2 metaphor. When things are humming along, system 1 is making predictions and they’re being confirmed. But when a big mismatch is detected, system 2 gets forced to deal with the mess. That implies expending lots of attention getting things back on track, performing one elementary behavior after another until system 1 kicks back in when things match up again.
The difference between these two modes is energetically huge. Attention is slow, laborious, and ultimately leads to (ugh!) learning. Just from a pure energetic standpoint, it’s something to be avoided whenever possible.
So how to avoid it?
Predictions (i.e., the set of time-sequenced concepts being pushed down into the lower layers of sensory apparatus) can be vague or specific. In general, the more specific the prediction, the higher the risk of incurring a system 2 “exception”. A race car driver makes an incredibly specific set of predictions about how his car will behave as he goes into a turn, but he’s also devoting huge attentional resources to monitoring that prediction, and preparing to recover if it’s wrong. Conversely, driving the kids to school in the mini-van can be done almost in one’s sleep, because the right turn into the parking lot doesn’t need much more than “we should probably go over there, sorta”.
It’s pretty easy to saturate this system 2 exception handler. Too much surprise, and the proper response is to make less specific predictions. But that will lead to exactly the sort of perceptually washed-out cognition that accompanies depression. Bland predictions lead to bland experience. Bland experience leads to less action and less risk-taking.
The thing that’s missing with this explanation is why everybody wouldn’t go bland as much as possible. I suspect that that has to do with the reward system for learning things correctly. When you do something new that makes the outcome agree better with the top-down system after it’s had an oopsy, you get a little squirt of dopamine, which tells all the neural nets to change their synaptic weighting based on what was firing when the good thing happened. Coincidentally, it also also feels awesome. (I’m falling back on “feels awesome” as a placeholder for something that lots and lots of neurons are genetically programmed to seek, because it enhances survival. Mostly, I suspect that this means that the synaptic re-weighting is its own reward.)
So there’s a nice dynamic tension between going bland (no painful surprises requiring attention), and going more specific (you get a prize if you manage the surprise properly!). But that tension is mediated by a notoriously squirrelly hunk of neurochemistry.
I think this can mostly explain the motivational issues. You’re motivated to recover successfully from surprise to get the reward, but you’re only motivated to the extent that you expect the to get the reward. Come to think of it–that’s a prediction, but it’s an odd kind of meta-prediction. Hmm. I wonder if that kind of meta-prediction is organically different than your basic cortical prediction.
Very good!
I think quite a few people here are not quite understanding this topic yet, and are interpreting terms like ‘confidence intervals’ to mean much higher-order phenomena than the sort of bit-level processing that is actually meant.
I was going to say as much below, but I think this comment explains it well. It’s quite a difficult topic to understand.
One interesting thing that just occurred to me is that this theory could quite reasonably predict why it is that SSRI’s classically take a while to start working – something that the old ‘depression is serotonin deficiency’ didn’t. You’re changing the neurotransmitter ‘environment’ for the neurons, so this creates different biases in how predictive connections are formed. To your example above, dopamine creates an incentive for the neurons to make more specific predictions – even if they’re wrong more often, when they’re right they get a hit of dopamine as a reward. So if you increase the amount of dopamine in the brain, your brain will gradually begin to form more and more neural circuits designed to create highly-confident predictions even though these predictions are more frequently wrong – leading at the ‘top level’ to the sort of personality that seeks novelty and action.
Basically, personality has inertia – a change-pace which is limited by the speed at which neural wiring can change.
What this makes me wonder, then, is how moods work. Give people a hit of some neurotransmitter – say, serotonin – and they immediately begin processing things differently, albeit temporarily. I’m not a neurologist and this is wild speculation, but what if different circuits and structures in the brain have a bias to being activated by the neurotransmitters that first shaped them? That is, pattern-recognition neural circuits that were shaped under the influence of (say) serotonin will become more active as serotonin increases in the brain, making your cognition immediately shift to a ‘serotonin brain’ mode.
This would predict that a person who had done a lot of neurological learning under the ‘bias’ of a certain neurotransmitter cocktail would, if exposed to it again, have more powerful experiences than one who had not. Such as how people with PTSD react in very extreme fashion to what are, to others, mild stresses.
I had the same intuition re SSRIs and time-lag, and it has me wondering if you initially couple SSRIs with something that temporarily narrows all of your confidence intervals (do amphetamines do this?), would that be enough of a kick-start to eliminate the lag time, even if you drop the confidence-interval-narrower.
This discussion is all based on the premise that anti-depressant drugs actually work to alleviate depression rather than block the normal self-regulating processes that occur on a synaptic and neurocognitive level. See below…
http://psycnet.apa.org/journals/pre/1/2/2a/
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3937810/
If this is correct it suggests that you could treat depressed patients by showing them a series of pictures that conforms to their view of reality. Just photos of grey clouds etc, hold them up and say “what do you see” followed by “correct, its a grey cloud. A grey couch. And on and on for long enough (or for enough days in a row) until their brain started trusting itself again.
It also suggests what many depressed people already suspect, that telling them “things aren’t as bad as you make them out to be” and “you have plenty to be happy about” is actively harming them, not helping.
It’s more that there is a
It’s more that there is a really low probability that none of the million things that can go wrong will. Like after graduating in 2008 it took me 4 years to get a job using my engineering degree. After a few years I realised there was no way I would ever be paying off my student debt. The string of things that would need to go right for me not to have worked my ass off my whole life to still never get out of poverty became impossibly long and while I’m sure I could manage some of them it is a near certainty I would fall on my face eventually. When I did I just gave up, I don’t see a path forward towards anything that could be considered a non miserable life. American Dream my ass.
I believe that the advantage of a unified scientific theory is that it can call into question, deconstruct and model human function and dysfunction from first principles rather than to answer questions that are predefined by false assumptions. This is one example of the huge breadth of difference between predictive/Bayesian approaches and perceptual control theory (PCT). See https://m.youtube.com/watch?v=vQ3lY1U60MM.
Sorry if it has been said earlier (too many responses):
Failure is easy to achieve, so predicting failure is a better bet than predicting success, and the less you are willing to accept prediction failure the more you are inclined to adopt this strategy.
Good point.
Not sure if you know about this article from 2016 by Aaron Beck, but it’s similar:
A Unified Model of Depression
Integrating Clinical, Cognitive, Biological, and Evolutionary Perspectives
Aaron T. Beck, Keith Bredemeier First Published March 29, 2016
We propose that depression can be viewed as an adaptation to conserve energy after the perceived loss of an investment in a vital resource such as a relationship, group identity, or personal asset. Tendencies to process information negatively and experience strong biological reactions to stress (resulting from genes, trauma, or both) can lead to depressogenic beliefs about the self, world, and future. These tendencies are mediated by alterations in brain areas/networks involved in cognition and emotion regulation. Depressogenic beliefs predispose individuals to make cognitive appraisals that amplify perceptions of loss, typically in response to stressors that impact available resources. Clinical features of severe depression (e.g., anhedonia, anergia) result from these appraisals and biological reactions that they trigger (e.g., autonomic, immune, neurochemical). These symptoms were presumably adaptive in our evolutionary history, but are maladaptive in contemporary times. Thus, severe depression can be considered an anachronistic manifestation of an evolutionarily based “program.”
Okay! I know now what is not clicking with me and this theory – it’s Lucy and the football!
The “you have a misaligned top-down predictive processing model” sounds cool, but what it comes down to is this – no, you are mistaken when you think (predict) bad outcomes, just try it and you will see that it doesn’t go as badly as you think, so then your confidence gets built up, so your predictions get better aligned, so your depression is cured!
Except. Except that the experience of being depressed is often (and I can’t speak for everyone) that of Lucy and the football. Lucy says “C’mon, I’ll hold it for you this time, go ahead and kick it”. Charlie Brown says “No, you said that the last time, I ran full-tilt ahead, you pulled the ball away and I fell flat on my face. I’m not going for it this time”. Lucy says “No, I promise, this time will be different!”
Eventually she talks Charlie round, he runs full-tilt…
… and she pulls the ball away and he lands flat on his face yet again.
Now, maybe you can talk yourself into “Okay, so 9 times out of 10 this happens, but surely the 10th time will be different!” But when you’ve done it 99 times, or 998 times and this is No. 999, then I don’t think your prediction model is broken when it goes “Yeah, maybe this will be more like the last 998 times and not the 1000th time it will work”.
So – my view of the predictive processing model? Great for figuring out how you move your arm. Maybe not so great for “why are you depressed”.
See, I’m laughing (to keep from crying) at this, because the following is from a website all about how You Can Find Help, and this is the exact thing that the predictive model wants you to try – ignore your wonky top-down confidence estimate, go ahead and try, things will be different!
No, see, I did all that (yes, the whole “Doctor, I think I might need antidepressants because I’m having suicidal thoughts and they’ve been going on for a while and they’re getting really bad and I’m genuinely worried I might really try to kill myself” begging and pleading bit) and I didn’t get the antidepressants and I got told that the suicidal thoughts weren’t that big a deal since, after all, I hadn’t actually tried self-harming or killing myself. But wait, there was this new official scheme and I’d get to see a counsellor (when one was available in like, six to ten weeks).
Great! Counselling! Like this, right?
Well, again, the counsellor was “No scars or suicide attempts? Sounds like mood swings not depression”. Since I wasn’t able to sign a contract promising that yes, I would without fail turn up every week for the next ten weeks bright and cheerful for the sessions (because whaddya know, executive disfunction is part of my whole problem) then sorry, not able to take you on the official new ‘counselling instead of drugs’ scheme.
EDIT: To do them justice, when they say “you will see a counsellor”, they do literally mean a counsellor, as in there are so many people waiting to be seen and so few therapists, this one was literally the only one for the area (Dublin is different, but Dublin is always different) and if I wasn’t going to use the time properly, better to take on another person who’d go the full 10 sessions because there were plenty waiting.
So I went online and found one of the mental health support sites and signed up for the online therapy ten sessions of CBT. I mean, CBT! The miracle therapy that has, as Scott said elsewhere, cured millions! Now I’ll be fixed, right?
Wrong. Didn’t find it at all useful, except for the mood mapping exercise (“On a scale of 1-10, what is your mood?” for every day. I found I bumped along at a nice, steady 3 out of 10, sometimes 4 out of 10. The really bad patches were the 2 out of 10 days but if I could keep stable at 3, things were no worse than usual).
Okay, so CBT isn’t for me. What next? Well, the next step is engagement with the mental health services, but that only gets triggered if I really do man up and take a razor to my arms, or throw myself off the bridge into the harbour (the favoured way of killing yourself round these parts):
So far from what I’m getting, it’s not “if they’re very concerned about the risk”, it’s only if “the guards had to pull you out of the quay at 4 in the morning” or “you had to call an ambulance because you cut too deep and the bleeding wouldn’t stop this time”.
So here I am: options (except for if I really do stop being a whiny little bitch, have some guts and go for really trying to kill myself) that all the earnest mental health campaigns and online advice told me were there, I just had to ask for help, now closed and I’m still at square one (I can’t even say “back to square one” because I never advanced off it).
Plus I am such an incapable failure, I can’t even manage to get antidepressants for long-term suicidal ideation when any schmuck can walk into the doctor, say “I’m feeling kinda down” and get a prescription. But that’s only because my confidence levels are unnaturally low and my predictive processor isn’t calibrated correctly, right? 🙂 Allll ya gotta do is try! Well, I tried. And it didn’t work. And now what does the model say?
EDIT: To be fair, I do come on here and whinge when I’m particularly bad, and you guys tolerate me, and that does help. But “hey how about ringing the Samaritans?” is no good either, because I don’t want or need “ooh yes, ooh I know, oooh that sounds terrible” sympathetic listening, that would make me really uncomfortable and self-conscious and feeling way worse.
I know you’re partly just venting here, but I thought it might be useful to note that anyone saying that you can fix yourself by ‘just trying anyway!’ and ‘think good thoughts!’ on the basis of this theory has not understood it.
It’s not as simple as that. Thoughts and actions like that are VERY downstream of this mechanism of predictive processing. The hypothesized dysfunction is happening at the level of where the brain crunches the bits and bytes, not at the level where you have thoughts or do things. It’s not something you control.
Godspeed, hope you find help soon.
Thanks for the good wishes, feeling a bit better today (which is why I think there’s something biochemical at the base of it all, but never mind that at the moment).
where the brain crunches the bits and bytes
Yeah, that’s the whole problem. That’s why I think the predictive processing model is good for the “so how do you decide to move your arm when studies show you move your arm before you decide to move your arm?” seeming contradictions, but if somebody then extracts a simplified version of the theory and decides to make a therapy based on it as “CBT with a coat of paint slapped on”, it’s not going to work.
Getting down to the data-crunching parts and re-calibrating them is like asking a driver to hang out the door of his car and re-tune the engine while going full speed on the motorway in that same vehicle 🙂
I agree. This theory, if true, might eventually be essential for developing much better therapies – after a few massive advances in tech and understanding.
Cell theory didn’t immediately revolutionize medicine just because Leuwenhook saw a bunch of boxes though a lens.
But the theory is nonsense in the first place. It’s like Niels Bohr or our friend Mr yudkowsky saying they know precisely what the underlying mechanism of quantum mechanics is. No, you just have a theory that fits the facts. (and there’s a million of those you could have settled on).
(Also this one is just obviously wrong, but then ‘Russel’s Square Circle’ isn’t much worse than russell’s teapot in practical terms.
You probably have tried it, but if not:
There is this inflammation mediators mumble mumble central nervous system mumble depression hypothesis. (link, link) Currently a study is done trying an antiinflammatory and mildly antibiotic drug for depression treatment.
Like a commenter above, I get depressed (and confused) when my allergies are triggered. Less with airborne allergens that still cause asthmatic symptoms and conjunctivis, much stronger with foodstuffs that fire up neurodermatitis. Antihistaminics do somewhat reduce the mental symptoms, but antiiflammatory drugs do this much more: After a diclophenac, the night’s sleep is refreshing and the next day is a good one. A fraction of topical steroids probably make their transdermal way and have a systemic effect, too, that improve mood (but reduced itching etc is a confounder here).
Have you tried antiinflammatory drugs? (Also, I hope I don’t do you injustice, did you mention alcoholism some time ago? Fire water even in small doses causes low-level liver inflammation. Maybe that’s why here its the single worst self-treatment (yet quite popular; I guess because of the short-term relief).
If antiinflammatory drugs turn out to be helpful: Search for medications that can be taken long-term i.e., have little side-effects. Maybe there is also an allergic reaction on a constant low level (e.g. dust mite)? Does your insurance cover testing? Also, you can search for foods and herbs, etc. that are likely to lower inflammation, to try, and for their opposites, to avoid. But there is a certain degree of alternative medicine mythology around that topic, so look for quality science reporting or health advice from reliable media.
HTH
I seem to recall reading that depressed people actually have more realistic expectations of the future than the non-depressed, but I don’t remember the details.
Is that actually substantiated, or am I just remembering some speculative hypothesizing?
If true, how does it fit in with the PP viewpoint? Is depression the result of fine-tuning expectations after a traumatic event shows that normal excessive optimism is erroneous? (That doesn’t account for the felt experience, of course, nor for depression’s tendency to feed on itself.)
ETA: Nor for depression’s reduced confidence in predictions. Weird for a response to trauma to make you simultaneously more accurate at predicting but less confident in your more-accurate predictions. At this point I’m sort of hoping Scott says that the whole thing is bunk and he doesn’t know where I heard such rubbish.
What you’re referring to is basically depressive realism.
Yes, that’s it exactly. It appears to be not as well established as I had thought — certainly better than “some guy pontificating”, but not what you’d call a crucial piece of the jigsaw. Thanks!
I’m not sure I understand this correctly, but if the model says that depression comes from low confidence in neural predictions, wouldn’t challenging peoples’ negative assumptions in CBT lead to some depression? I mean, you would be adding yet another layer of non-confidence (“don’t trust your brain telling you you are not valuable”) to an already not-so-confident-in-itself brain.
I talked about this in more detail above, but I think that this understanding is not correct. We’re talking about very fundamental bits-and-bytes processing levels here, not high-level cognitions like ‘I am valuable’ or something like that, which only emerge after a huge amount of predictive processing has already been done.
So a person who tells themselves something that conflicts with their fundamental assumptions will have no real impact on themselves – they just won’t believe it to start with, no tension or learning created.
this suggests somewhat more specific but not unrelated mechanism http://www.cell.com/trends/cognitive-sciences/fulltext/S1364-6613%2815%2900174-6#title-footnote-tb0015
Hi everyone, I really believe that this thread is based on two false premises. First that understanding depression is the way to help people recover from mental health problems. Depression is just one of many manifestations of chronic distress. We need to understand the principles of what underlies all of these because the categorisation of psychological distress into disorders is a red herring. All of the evidence now points to there being a core process underlying all mental health difficulties. Some of these studies are published in the highest impact psychiatry journals so it’s not just me saying it! Second we need to ask ourselves – what is more fundamental to well being? Predicting what will happen to you? Or controlling what will happen to you? Regardless of what you predict, if you can take control then your predictions are just that, and may not be reality, or you can make them your reality if you want to. This is control. So we need a unified theory of how control works. The fundamental mechanism of control – in all control systems – is negative feedback – not prediction. Under some circumstances prediction can improve control, but it needs updating as regularly as possible. That is negative feedback and it is the route to recovery and adaptation. Darwin and Wallace saw this a century ago by drawing analogy between the Watts Governor (negative feedback control system of a steam engine) and evolution by natural selection.
Perceptual Control Theory is the implementation of negative feedback in the nervous system, exploiting its capacity to extract regularities from the environment, organise them in a hierarchy and use actions to make current experience match memories of desired past environmental regularities. Conflict between control systems is the most pernicious reason of loss of control, and mental health problems persist when people are unwilling or unable to sustain their attention on the system governing the conflict long enough to allow spontaneous change to reduce error through reorganisation. Biases in prediction are observed to reduce during this process but this is largely epiphenomenal, and idiosyncratic to the person. The universal principle is to restore control through reorganisation of conflicted control systems.
We use Method of Levels as a universal therapy to help people manage this. See http://www.methodoflevels.com.au. We need to use mathematical/computational theories (of which PCT and PP and natural selection are examples) to push these pivotal changes rather than pondering within our familiar frames of reference…
To throw more confusion on the matter, my depression tends to revolve around the fact that the world is TOO predictable, that I am (justifiably) highly confident in my own ability to succeed at just about anything. And the thing that helped me get out of it was starting something I wasn’t good at and made little progress in.
But my depression tends to be characterized by an a-emotional state in which I feel nothing, rather than a negative emotional state in which I feel terrible. (Indeed, after many years of that state, negative emotional states started being enjoyable as a novel deviation from the norm)
What about the simpler explanation that clinical depression is a malfunction of the reward system, which causes it to compute lower rewards?
Reducing uncertainty about future experiences produces an intrinsic reward, so depressed people tend to generally overestimate their uncertainty on approximately neutral outcomes, because they also underestimate this intrinsic reward, but they become highly certain when they predict negative outcomes.
I think possibly it’s partly to do with the problem that depression means you stop enjoying what you used to enjoy.
So you do Fun Thing – but it’s not fun anymore. It used to be fun, it was the reliably fun thing you always did, but now it’s not fun. And so you don’t want to do it anymore, because it’s not fun. And people telling you “But you love Fun Thing, go out and do it, you’ll feel better!” don’t understand that.
I think that’s why depressed people put high confidence into computing lower rewards – they’ve already had the experience of the reward getting lower or even vanishing, so they adjusted their predictions accordingly.
Are you familiar with anhedonia caused by severe depression? Severe depressives with it know quite well they’re not going to enjoy what comes next. They’re not predicting.
I don’t think depression just dulls feelings because reduced confidence levels diminish strength of feeling. It’s also that surprises feel less surprising if your confidence level is low. So if surprises happen all the time and are the source for many of people’s feelings, then depressed people will end up with a relatively dull, diminished experience. Note that surprises aren’t necessarily a negative experience. Surprise joke endings, for example, are positive surprises. There’s some amount of novelty seeking that depression diminishes the reward of because of decreased confidence.
So maybe all the ‘boring” routine, menial, daily tasks we’ve freed ourselves from over the course of the last century were helping us to stave off depression?
I don’t know about that. Doing routine tasks like cleaning, self care, etc are extremely difficult for many people with depression.
Not exactly true.
*Motivating* themselves to do routine tasks is “extremely difficult for many people with depression”.
When an external source of direct immediate motivation or compulsion hauls their ass out of their bed, off their phone, away from their TV or video games, staring into bluewhite light screens every night when they should be asleep, and forces them to take a daily shower, forces them to eat warm flavorful high-protein low-sugar food on a consistent schedule, and forces them to do chores and labors that involve enough physical exertion that they get sweaty and breathless every day, many of them wake right up.
(If you are reading this and have depression, I am not talking about you. I’m sure you’ve already tried the obvious straightforward changes in schedule, daylight, diet, and exercise, and they didn’t work. That sucks, I’m sorry.)
Sometimes I think that depression is not actually a disorder, it’s instead evidence that all humans were not intended to be their own guardian and taskmaster all of the time. Or even most/most, like low back and knee disorders and poor muscle tone is evidence that humans were not intended to sit in an upright chair for hours a day every day for decades.
The predictive processing thing really seems like something you should be able to test within the framework of machine learning.
Take a time series prediction problem with a SOTA result that’s not too close to a hundred percent, and build a network architecture where you split the recurrent module into sets of a few layers, that receive the output of the next layer up at the previous time step as input (as well as the standard feed-forward data). So the modules learn to combine both their current bottom-up data and the recent high level predictions. It’s sort of a weird, staggered implementation of skip connections.
If it works better than the same net with generic skip connections, then that’s some evidence that it’s actually a reasonable way to structure a reasoning system.
Now, if only I knew a person with severe depression and a bunch of Adderall to see if taking, say, 60mg would help
Predictive models of depression dovetail naturally with affective models of economic behavior — see (for example) Amartya Sen’s “Rational fools: A critique of the behavioral foundations of economic theory” (Philosophy & Public Affairs, 1977, 4000+ citations, yikes!), as updated (for example) by Alan Kirman and Miriam Teschl’s “Selfish or selfless? The role of empathy in economics” (PTRS-B, 2010).
In brief, the predictive capacities that are associated to rational cognition can paradoxically act to engender depression, by suppressing Freud’s two paths to happiness, viz., “working” and “loving”. As examples, what rational reason(s) can justify physical work … that robots can do better? What rational reason(s) can justify the pursuit of market-goals … that machines can optimize more efficiently? What rational reason(s) can motivate the pursuit of affective objectives … that by definition are not grounded in rationality?
To quote Sen’s article
… and Scott’s theory neatly predicts that Sen’s “purely economic person” will be susceptible to depression too.
Literature compatibly supplies the example of Captain Ahab, who appreciates that
Notably, Ahab’s powers of rational prediction do not consciously foresee the lethal consequences of his quest (which amounts to “suicide-by-whale”) … are the depressive consequences of exclusively rationalist cognition any easier to consciously foresee?
In summary, to escape the paradoxically depressive consequences of enhanced predictive capacity, rationalists require sophisticated appreciation(s) of affective cognition and affective economic behaviors.
Perhaps individual SSC readers who have personally escaped depression, or have successfully cultivated happiness in friends, family, and colleagues (and even pets), might comment on their own experience of balancing rationalist cognition with affective cognition, affective economic behaviors, and the affective arts generally?
“Depression has to be about something more than just beliefs; it has to be something fundamental to the nervous system. And low confidence in neural predictions would do it.”
I spent several months in the dark when I was depressed. Oddly, the end of depression occurred fairly suddenly and accompanied a change in beliefs. I had been heartbroken from a relationship, and was obsessed with Buddhism and trying to get enlightened. I slept all day and walked around in the dark at night, when I wasn’t laying in bed next to a gun which I constantly fantasized about using on my brain stem.
Things were so bad I eventually realized that I was on the edge of having to be put in some institution. That thought was pretty much what made me snap out of it, and along with it I gave up on the search for enlightenment and decided it had all been a delusion.
Things got better fairly quickly after the moment that happened. What I realized I needed to do to get over the pain of heartbreak was to distract myself with anything. In the first week or two, I even went about trying to memorize people’s license plate numbers. This kept my mind from thinking about the pain of loss. I didn’t know where to start and just had to do *something*. Eventually it was replaced with more effective hobbies, like reading books and such.
Yes, this. Reconfiguration of priorities/imperatives.
Glad you came out of your dive.
I know from personal experience that depression is awful so I’m thankful that you think and write about this subject. Your hypothesis is interesting and I don’t have a better idea right now, but I wonder if we should keep using the word ‘depression’, which seems to me an umbrella term.
An interpretation of low prediction confidence making failure seem more likely, instead of making both failure and success seem equally unpredictable: there are more ways to fail than succeed, and failure is the default outcome.
To expect success you have to predict that what you are trying to do will hit that small target. To expect failure, uncertainty about success is enough.
All of this rings very true and familiar. If strong predictions guide actions, it’s clearer why depressed people are asked to think positively.
Also kinda depressing from a rationalist perspective.
Having accurate beliefs, like 90% of start-ups fail, may send you into a negative spiral. Perhaps it’s better not try hard to be a realist. You cannot have accurate beliefs independent of your emotional state.
This seems to paint depression as an adaption (avoid uncertain situations) rather than a weird brain failure.
Great post.
“If strong predictions guide actions, it’s clearer why depressed people are asked to think positively.” is probably better phrased as “if therapists/society/science think that, if strong predictions guide actions, it’s clearer why depressed people are asked to think positively.”
If however, therapists/sciciety/science thinks that actions are used to achieve desired experiences, and depression is just one of many manifestations of having desired experiences (controlled perceptions) that are in conflict, then what we should be doing is helping people with depression have a safe space to express their thoughts and feelings freely so that they can consider their goal conflicts sufficiently to resolve them… http://www.methodoflevels.com.au
I tried to post a comment here last night at around 1:40am, but it’s not showing up (for me at least). If I try to repost it, I get “Duplicate comment detected; it looks as though you’ve already said that!” How can I figure out where my comment went? (If it’s relevant, the comment was about 1100 words.)
I think the right thing to emphasize here is not “low confidence” but “smeared confidence interval”. In particular, my model is that in depression, you react to surprisal by widening the confidence interval (thus lowering individual confidence). If you have extremely narrow, but very low confidence predictions, then you should expect constant updates. If you have extremely smeared, low confidence predictions, updates are rare.
I think I understand this, mainly from looking at a couple of cases:
1. Where is motivation unnecessary? In habits and impulses, certainly. If the normal mode of acting by adjusting to match predictions happens without you becoming consciously aware of a difference between prediction and sensation, then motivation is unnecessary.
2. Where is motivation unquestionably strong? In addiction, in visionaries, in fanatics. The commonality here seems to be that there’s a prediction somewhere that’s (almost?) completely resilient to contradictory sense data.
3. Where is motivation present but weak? In the marshmallow test, in the presence of temptation. It seems that motivation troubles come when you consider changing your predictions / model, but can’t quite settle.
From this, I model motivation as a capacity to (consciously? deliberately?) reinforce predictions in the face of contradictory sense data.
Depressed people aren’t updating predictions nearly as frequently as those who are not depressed. But they still update predictions, and if the evidence never contradicts the prediction, eventually the smeared “I’m uncertain” prediction will narrow to whatever is perceived. This doesn’t quite fit the claim, but I’m not sure the claim is accurate. In depression, is the claim “I’ll fail, every time”, or is the claim “if I succeed, it’ll be just random chance, and trying to succeed doesn’t have any impact on success”? The latter fits my model: confidence intervals are smeared until you are not sure whether you’ll succeed or fail, in any particular endeavor. More importantly, you predict that you have no control over success or failure (and I think it’s harder to get sense data saying “you control this” when you think you don’t than to get sense data “you have no control” when you think you do—maybe the source of confirmation bias is somewhere around here).
Uh, it seems like there’s a simpler explanation here. The brain seeks not to minimize prediction error (couldn’t it do this by making all of its prediction confidence intervals maximally smeared?), but to maximize the amount of prediction error it can eliminate (or maximize the amount of surprise it can extinguish).
I think this is too high-level and hand-wavy. I also think that claiming happiness as a global filter is wrong. You can be happy and sad at the same time. You can be happy about one thing and sad about another thing at the same time.
Consider the following examples:
1. Infants look at surprising stimuli longer. They extinguish surprise by making sense of it, not by looking away. I expect, moreover, that if you teach a toddler that one room always contains an unsurprising stimulus, and another room always contains a new and surprising stimulus that can be made sense of relatively quickly, the toddler will seek to enter the room with the surprising stimulus, even before they encounter the surprise. (Does anyone know if this is true?)
2. The reference description I’ve heard about mania “if you think you’ve found a theory that explains everything, diagnose yourself with mania and check yourself into the hospital”. The internal experience of thinking you’ve found a theory that explains everything is that you have a theory, and you keep finding surprising stimuli that make sense when you model them with the theory—you keep being surprised and extinguishing that surprise. You keep having “ahhh, that makes so much sense!” moments.
(Other interesting examples that fit well here: NRE / crushes, rats in rat park vs isolation)
I infer that the experience of happiness is somehow very, very closely tied to the internal experience of successfully extinguishing surprise.
I expect there is one variant of depression that corresponds to abnormally low surprise. Things tend not to be surprising, updates don’t happen. I think the internal experience of this would be blankness or muteness of emotions (or, at least, the emotions that are driven by surprise, so, anger, awe, hatred, probably pleasure, probably pain, probably happiness, possibly sadness, etc.).
(The internal experience of having lots of surprise and failing to update seems like the experience of sensory overload.)
There’s a bit of nuance, though, and a nice symmetry. I’ve read that to grieve is to familiarize oneself with a surprising and painful truth until it is no longer surprizing. Exposure therapy is similar: you repeatedly confront yourself with a surprisingly-not-bad stimulus until its not-badness is no longer surprising. In happiness, I expect, you repeatedly confront yourself with a surprisingly-good stimulus until its goodness is no longer surprising.
My current model is that pain is when you experience surprisal, and both the sense data and the predictions fail to update. Pleasure, then, should be the internal experience of experiencing surprisal going away. (This explains, perhaps, why being mindful brings more pleasure to movement: you become aware of when and how your body goes from not matching your desired location to matching it.)
This suggests a second sort of depression, perhaps more closely matching what you describe in the post: in this one, there is surprise, and it isn’t resolved well. Here, you’d have low confidence predictions that ought to additionally be resistant to being updated. Here you wouldn’t have muted sadness, but intensified sadness: reality would be painful, because your predictions conflict with your sense data and both are slow to update (the sense data because your predictions lack confidence; the predictions due to whatever mechanism causes them to be resistant to updates). This seems a bit off, though.
What I have found through my experiences with depressed people is that the vast majority of them are not insane people in a sane society, but perfectly sane people in an insane society, people who have maladapted to their environment but would otherwise (if they had a better one) be perfectly healthy individuals. Many of them internalize their sadness and helplessness and make it a part of their identity, further reinforcing the common belief among depressed people that they are ill at a biological level (and are not merely healthy people in an unhealthy environment). I don’t think the solution to this is ever going to lay in the hands of psychologists, and certainly not psychiatrists (SSRI’s and SNRI’s work only slightly better than chance in most people, and taking other antidepressant/mood-stabilizing medication is a hellish experience all its own) – it lays in the hands of the people who construct the environment we live in, the people in charge of its development and health, the people who have the most power over society. These are the people who have destroyed it and made the environment we all have to live in such a hellish place, and their successors will certainly be the only people who can fix it (on a societal level, at least – obviously some people really do have a brain problem, but I strongly suspect they are a minority).
I spoke about this model with a close relative who experiences depression on a frequent basis and this is what they had to say:
– “it’s the best verbalization so far of what I feel is going on”
– “I’ve felt so comfortable in structures” like college, because they’re predictable. Parenthood has been much harder.
– “this is probably why exercise helps with depression, it’s easy and predictable” so you feel that your prediction error better matches reality
– “rather than feeling confident [which is what I assume other people feel], I have to ‘make a case for myself’ [internally] all the time” by trying to recall past successes, like graduating with a bachelor in science, to persuade myself my action is worthwhile
This idea seems similar to one of the ideas advanced by Van de Cruys, S. (2017). Affective Value in the Predictive Mind. In T. Metzinger & W. Wiese (Eds.). Philosophy and Predictive Processing: 24. Frankfurt am Main: MIND Group. doi: 10.15502/9783958573253
The most relevant quote from that article is (emphasis, and typos introduced by copy-pasting from the PDF, by me):
A later idea from that paper goes further, linking affective valence to predictions of the first derivative of prediction error:
I don’t have sufficient expertise in the relevant areas to offer my own comments, but I wanted to briefly mention that source as it might contain useful material and references for developing the quoted idea from the post.