Long-troubled actress Amanda Bynes has apparently been checked into a psychiatric facility to address the mental illness that has plagued her for years. She’s reported to have been wandering naked outdoors when she flagged down a car and asked for help, telling the driver that she was having a psychotic episode. People are naturally grateful that she was able to request help in that way, but I’m also seeing a lot of confusion. Which is understandable; a psychotic person being cognizant enough to request help is out of step with how most people conceive of serious mental illness. But psychosis is more complex and multivariate than most people think, and I hope this incident prompts people to broaden their understanding of the condition. The mind is complicated, and so are the ways in which it breaks down.
The story has stuck with me for days, including the story of Shina, one of the people who lives in a tent. She knows she has delusions, and tells the reporter—but also believes many of them sincerely. She also had practical survival skills and helped the others.
Liberals are always debating whether people like Shina should be permitted to live in a tent or forced into services. I admit I’m on team services, but it’s complicated when someone has delusional beliefs but is otherwise functioning.
“This may sound weird, but what can happen over time is that you get better at being psychotic.” I think this is an amazing opinion and accurate perspective! Obviously what happened to Amanda is tragic but I constantly hear rather cynical commentary on mental health conditions and that they “can’t be cured”. I love your article because it highlights that while psychosis is serious, especially when it persists as in Amanda’s case, maybe she’s getting a little better because she wasn’t hurt, didn’t hurt anyone else, and asked for help herself. I really think content like yours will help “end the stigma” around mental illnesses like psychosis so that people like Amanda can be treated with the respect and dignity the deserve and hopefully get better! 🫶
Great post. We often use the term insight in clinical practice. This can be multi-faceted - a person can have varying degrees of insight into their symptoms, their mental health diagnosis and need for treatment.
Obviously these are very different issues, but a lot of what you describe here lines up with my experiences with disordered eating (both my own and what I have observed in other people who struggle with them). I don't feel I have "healed" from my ED in the way that is often described in popular media, as much as I have learned how to live with and react differently to the "irrational mind" urges, thought patterns, habits and inner monologue that eventually lead to it.
This has been a double-edged sword for me, too -- on the one hand, I have gotten to a point where "learning to live with it" HAS allowed me to live a healthier and more fulfilling life, which feels an awful lot like healing/curing it. At the same time, "learning to live with it" can also look like "learning to hide it," "learning to go through the motions as a 'recovered person' while your mind continues to buzz under the surface," "learning to be just 'healed' enough to escape further treatment/scrutiny but not enough to fully move on," "learning how to game the system," "learning to parrot popular recovery mantras around people who you know scroll that corner of tiktok so you don't have to have a 100th conversation with them about HAES and diet culture" etc. (ok that last one is probably pretty specific to me...). I feel I have been all of these and more, sometimes at once, at different points in time, and while I'm pretty confident I'm headed in the right direction, it's hard to constantly wonder if I'm really on the "right path" or if I'm just doing further self harm. This is even more muddled thanks to modern information overload and access to too many opinions about the right and wrong ways to "be disordered" via the internet and social media.
Again I know these are different illnesses and am definitely not suggesting that my experiences are on the same level as those of someone who deals with psychosis. But this piece really resonated with me and I think this is a very useful framework for discussing mental illness and recovery.
This makes sense to me. I have never had to deal with mental illness, but it has always seemed that the idea that the mind was a unitary entity was not really that true.
I think people understand this when it comes to urges. Your "rational" mind knows you are on a diet, but part of you still wants a cookie and they fight it out. I tend to think that even your rational mind is subject to conflict. Part of you wants to be on a diet, part of you wants a cookie, and part of you is coming up with ways that you can eat a cookie while still believing you are on a diet.
Years ago, I read a book "Getting to Yes," which was about various negotiation techniques that can make the other side more open to accepting or sticking with you view. I never found it super helpful in negotiating, but a lot of the tricks, e.g. getting them announce a favorable position, have always struck me as a good way to negotiate among myself. I could see someone with a psychotic disorder learning to do the same thing to ensure that, when they are having issues, they still can get themself to get help.
My one experience with bad psychosis (and when I was diagnosed as bipolar) had me very delusional, with hallucinations, and hyper-real violent images playing in my head, which scared the shit out of me. Scared my young daughter too, when I told her I thought I was dying, in tears, and that I was sorry. Immediate access to quality care in these situations is crucial. I was able to see a good psychiatrist that day, because my wife, an MD, called him and said I wanted to go to the hospital. He said to come to his office, where he talked me through the worst of it, reassured me that the drugs would start working soon, and I ended up slowly, slowly recovering over the next couple months like it was a brain injury, without ever being admitted. Saw him once a week for a couple months, then less as time went on. I was definitely conscious and aware, just not at all hopeful that things would turn out well. To me, it felt like a delusional panic attack from hell.
I've had interesting conversations with my nephew who is a severe paranoid schizophrenic. He says the medications don't take away the crazy ideations but it allows him to use his rational mind to control his reaction to them. No. They're not cooking babies in that restaurant people would find out and shut them down.
The problem he's had is not taking his meds, or not enough, and the meds become part of the psychosis. The meds are poisoning him. He gets way too far out there to ask for help.
My niece says he's on a new medication and he's the best she's seen him. He's been institutionalized for a year and a half and is looking for a less restrictive place, but one that still provides some structure.
He'd probably be dead without the institutionalization.
Too much money and celebrity are the opposite of defenses - they massively increase the odds you'll suffer mental illness and/or have a terrible quality of life.
I wonder what, if anything, this implies about AI alignment (partly because I wonder that about everything these days). Yes, sure, there's no ironclad reason to presume that specific aspects of human minds will generalize onto weird superpowered programs. But in my view the most believable AI failure modes sound a lot like how you describe psychotic episodes: thoughts getting lost in themselves; obsessive itches taking over ever-increasingly from rational behavior as real-world feedback fades away; etc.
On the one hand, it's gratifying that minds tend to somewhat recognize when they're misfiring in this way, even as they continue to misfire; i.e. that "meta-rationality" can make sense to human-like minds when the rubber hits the road. On the other, it's worrisome that these tendencies may come and go, in unpredictable waves, and that, at least for humans, we have serious trouble debugging the root cause, even when we solve the specific crisis and offer general diagnoses, tips, resources, etc.
I tend to think this is the sort of failure mode that will characterize near-term worrisome AIs, rather than the issue being like paperclip maximizers. In other words, there seem to be reasons to believe that the most relevant issues will involve "inner misalignment" (learning the wrong heuristics to achieve the right goal), rather than "outer misalignment" (learning, with perfect clarity, the wrong goal). Think of it like how evolution taught us to pursue evolutionary fitness by liking sugar, and being horny, and so forth, because in our ancestral environment these were good and easy strategies to focus on, but now they probably harm our ability to propagate our genes (or at least easily end up getting hacked with activities which evolution would view as psychotic, like non-procreative sex).
In other words, I think the issue won't be that AIs don't learn to care about us -- "human persons" is a very natural abstraction to focus on in their training process -- just as the human species obviously learned to care about our relatives (especially our offspring) despite not having any understanding of DNA until quite recently. But think of all the ways in which parents can fail horribly by caring about their kids way too much (the failure modes I'm thinking of sound a lot like your description of psychosis, but obviously in a much less acute or medicalized way). Or think about how dogs can be optimized to just hijack our impulses to rear cute things, whereas kids have to be optimized to produce kids of their own (or else their genes stop showing up in the human story); so dogs can outcompete kids in this way, precisely as they become completely unfit on their own (think of e.g. how dependent pugs are), hence the fact that there are now more dogs than children in the US.
Here's an I think apropos quote (from https://cebk.substack.com/ [shameless self-promotion]) about how horribly wrong this can go; if the Bing chatbot's alarming hallucinations are any indication of what happens when AIs get unleashed on us, I think this could get bad fast:
You can evaluate whether some advice sounds good within a few seconds; however, you’ll need about a month—i.e. a few million seconds—to actually test out whether it works. Thus, if chatbots are trained by seeking reward, then those that focus on flattering our society’s vapid sensibilities will “evolve” about a million times faster than those that offer us harsh but useful truths (to the extent that reinforcement learning is like natural selection). That’s a huge advantage! If biological evolution had been sped up a million-fold, then dinosaurs would’ve first emerged around 1800, and would’ve gone extinct around 1960. Other forms of machine learning seem even more subject to this extreme social desirability bias: for instance, GPT-style transformer-based attention-only self-supervised prediction models only learn what the text corpus they’re trained on is likely to say about our world, not what’s actually true of it. And because language is already an incredibly effective compression of reality, language programs are likely to stay in control of AI progress.
A powerfully illuminating and empathetic piece that serendipitously allows me to discuss with an actor in a play this potential for a character's consciousness that he is having a psychotic episode.
"As I’ve mentioned before, the technical term is anosognosia - the inability of the sick to understand that they’re sick. Anosognosia is indeed very common among people with schizophrenia and schizoaffective disorder, but not universal, and while many bipolar patients experience the condition, most estimates I’ve seen suggest that a majority don’t".
My best friend will, most of the time, acknowledge that he has schizophrenia and that he has previously experienced hallucinations and paranoia, but for him, the main voice he hears is an exception this this, and //exploits// the fact that he has diagnosed schizophrenia to hoodwink family, friends and medical staff while continuing to torment him.
This is really illuminating. I live in a place with, unfortunately, a lot of mentally ill people living on the streets, and it’s not uncommon to run into several of them just walking across the small downtown. Something I’ve noticed is that most of them are lucid to some degree – the man who frantically, compulsively pushes a broom around the sidewalk while muttering to himself will move out of the way without breaking his stride if you say “excuse me!” to him. The woman having a furious growled conversation with the air while she tries to navigate her shopping cart over a curb breaks midsentence to say “I’ve got it, thanks!“ in a perfectly normal voice if you ask if you can help, then goes back to talking to the air. Once a guy apologized for being too loud and wished me a nice day as I walked by, because he accidentally startled me by yelling while he was in the middle of a heated conversation with Jesus.
Of course the nature of psychosis is that it’s unpredictable. But it seems that for many people who have been living with it for a long time, even dysfunctionally, it’s not that they’re wholly incapable of engaging with reality – I’m precisely as real to them as the unreal things they’re seeing and hearing. It just sucks, because while that means we can sometimes have a brief, fairly normal interaction when we’re both trying to navigate the same physical space, it also means that I can’t see the unreal thing that may be standing between them and seeking help like Bynes did.
Thanks for writing this. We have found the same is true with a sister of mine. She has "learned" over the years what is happening, and so have we. So we can all deal with it as an illness instead of a momentary catastrophe. I wish you and all others in this situation all the best.
Evaluating insight into their condition when patients try to leave against medical advice can be very interesting. All they really have to say is that they appreciate the doctor thinks they're at risk by leaving but they're willing to take that chance. If they're not organized enough to say something like that, they have to stay.
Great piece. I think this is very hard for people to understand if they’ve never experienced psychosis, as I haven’t.
Did anyone else see the story in the NYT over the weekend about 2 sandwich shop owners overwhelmed by an encampment? https://www.nytimes.com/2023/03/19/us/phoenix-businesses-homelessness.html
The story has stuck with me for days, including the story of Shina, one of the people who lives in a tent. She knows she has delusions, and tells the reporter—but also believes many of them sincerely. She also had practical survival skills and helped the others.
Liberals are always debating whether people like Shina should be permitted to live in a tent or forced into services. I admit I’m on team services, but it’s complicated when someone has delusional beliefs but is otherwise functioning.
“This may sound weird, but what can happen over time is that you get better at being psychotic.” I think this is an amazing opinion and accurate perspective! Obviously what happened to Amanda is tragic but I constantly hear rather cynical commentary on mental health conditions and that they “can’t be cured”. I love your article because it highlights that while psychosis is serious, especially when it persists as in Amanda’s case, maybe she’s getting a little better because she wasn’t hurt, didn’t hurt anyone else, and asked for help herself. I really think content like yours will help “end the stigma” around mental illnesses like psychosis so that people like Amanda can be treated with the respect and dignity the deserve and hopefully get better! 🫶
Great post. We often use the term insight in clinical practice. This can be multi-faceted - a person can have varying degrees of insight into their symptoms, their mental health diagnosis and need for treatment.
Obviously these are very different issues, but a lot of what you describe here lines up with my experiences with disordered eating (both my own and what I have observed in other people who struggle with them). I don't feel I have "healed" from my ED in the way that is often described in popular media, as much as I have learned how to live with and react differently to the "irrational mind" urges, thought patterns, habits and inner monologue that eventually lead to it.
This has been a double-edged sword for me, too -- on the one hand, I have gotten to a point where "learning to live with it" HAS allowed me to live a healthier and more fulfilling life, which feels an awful lot like healing/curing it. At the same time, "learning to live with it" can also look like "learning to hide it," "learning to go through the motions as a 'recovered person' while your mind continues to buzz under the surface," "learning to be just 'healed' enough to escape further treatment/scrutiny but not enough to fully move on," "learning how to game the system," "learning to parrot popular recovery mantras around people who you know scroll that corner of tiktok so you don't have to have a 100th conversation with them about HAES and diet culture" etc. (ok that last one is probably pretty specific to me...). I feel I have been all of these and more, sometimes at once, at different points in time, and while I'm pretty confident I'm headed in the right direction, it's hard to constantly wonder if I'm really on the "right path" or if I'm just doing further self harm. This is even more muddled thanks to modern information overload and access to too many opinions about the right and wrong ways to "be disordered" via the internet and social media.
Again I know these are different illnesses and am definitely not suggesting that my experiences are on the same level as those of someone who deals with psychosis. But this piece really resonated with me and I think this is a very useful framework for discussing mental illness and recovery.
This makes sense to me. I have never had to deal with mental illness, but it has always seemed that the idea that the mind was a unitary entity was not really that true.
I think people understand this when it comes to urges. Your "rational" mind knows you are on a diet, but part of you still wants a cookie and they fight it out. I tend to think that even your rational mind is subject to conflict. Part of you wants to be on a diet, part of you wants a cookie, and part of you is coming up with ways that you can eat a cookie while still believing you are on a diet.
Years ago, I read a book "Getting to Yes," which was about various negotiation techniques that can make the other side more open to accepting or sticking with you view. I never found it super helpful in negotiating, but a lot of the tricks, e.g. getting them announce a favorable position, have always struck me as a good way to negotiate among myself. I could see someone with a psychotic disorder learning to do the same thing to ensure that, when they are having issues, they still can get themself to get help.
My one experience with bad psychosis (and when I was diagnosed as bipolar) had me very delusional, with hallucinations, and hyper-real violent images playing in my head, which scared the shit out of me. Scared my young daughter too, when I told her I thought I was dying, in tears, and that I was sorry. Immediate access to quality care in these situations is crucial. I was able to see a good psychiatrist that day, because my wife, an MD, called him and said I wanted to go to the hospital. He said to come to his office, where he talked me through the worst of it, reassured me that the drugs would start working soon, and I ended up slowly, slowly recovering over the next couple months like it was a brain injury, without ever being admitted. Saw him once a week for a couple months, then less as time went on. I was definitely conscious and aware, just not at all hopeful that things would turn out well. To me, it felt like a delusional panic attack from hell.
I've had interesting conversations with my nephew who is a severe paranoid schizophrenic. He says the medications don't take away the crazy ideations but it allows him to use his rational mind to control his reaction to them. No. They're not cooking babies in that restaurant people would find out and shut them down.
The problem he's had is not taking his meds, or not enough, and the meds become part of the psychosis. The meds are poisoning him. He gets way too far out there to ask for help.
My niece says he's on a new medication and he's the best she's seen him. He's been institutionalized for a year and a half and is looking for a less restrictive place, but one that still provides some structure.
He'd probably be dead without the institutionalization.
Another great one, Freddie.
Too much money and celebrity are the opposite of defenses - they massively increase the odds you'll suffer mental illness and/or have a terrible quality of life.
I wonder what, if anything, this implies about AI alignment (partly because I wonder that about everything these days). Yes, sure, there's no ironclad reason to presume that specific aspects of human minds will generalize onto weird superpowered programs. But in my view the most believable AI failure modes sound a lot like how you describe psychotic episodes: thoughts getting lost in themselves; obsessive itches taking over ever-increasingly from rational behavior as real-world feedback fades away; etc.
On the one hand, it's gratifying that minds tend to somewhat recognize when they're misfiring in this way, even as they continue to misfire; i.e. that "meta-rationality" can make sense to human-like minds when the rubber hits the road. On the other, it's worrisome that these tendencies may come and go, in unpredictable waves, and that, at least for humans, we have serious trouble debugging the root cause, even when we solve the specific crisis and offer general diagnoses, tips, resources, etc.
I tend to think this is the sort of failure mode that will characterize near-term worrisome AIs, rather than the issue being like paperclip maximizers. In other words, there seem to be reasons to believe that the most relevant issues will involve "inner misalignment" (learning the wrong heuristics to achieve the right goal), rather than "outer misalignment" (learning, with perfect clarity, the wrong goal). Think of it like how evolution taught us to pursue evolutionary fitness by liking sugar, and being horny, and so forth, because in our ancestral environment these were good and easy strategies to focus on, but now they probably harm our ability to propagate our genes (or at least easily end up getting hacked with activities which evolution would view as psychotic, like non-procreative sex).
In other words, I think the issue won't be that AIs don't learn to care about us -- "human persons" is a very natural abstraction to focus on in their training process -- just as the human species obviously learned to care about our relatives (especially our offspring) despite not having any understanding of DNA until quite recently. But think of all the ways in which parents can fail horribly by caring about their kids way too much (the failure modes I'm thinking of sound a lot like your description of psychosis, but obviously in a much less acute or medicalized way). Or think about how dogs can be optimized to just hijack our impulses to rear cute things, whereas kids have to be optimized to produce kids of their own (or else their genes stop showing up in the human story); so dogs can outcompete kids in this way, precisely as they become completely unfit on their own (think of e.g. how dependent pugs are), hence the fact that there are now more dogs than children in the US.
Here's an I think apropos quote (from https://cebk.substack.com/ [shameless self-promotion]) about how horribly wrong this can go; if the Bing chatbot's alarming hallucinations are any indication of what happens when AIs get unleashed on us, I think this could get bad fast:
You can evaluate whether some advice sounds good within a few seconds; however, you’ll need about a month—i.e. a few million seconds—to actually test out whether it works. Thus, if chatbots are trained by seeking reward, then those that focus on flattering our society’s vapid sensibilities will “evolve” about a million times faster than those that offer us harsh but useful truths (to the extent that reinforcement learning is like natural selection). That’s a huge advantage! If biological evolution had been sped up a million-fold, then dinosaurs would’ve first emerged around 1800, and would’ve gone extinct around 1960. Other forms of machine learning seem even more subject to this extreme social desirability bias: for instance, GPT-style transformer-based attention-only self-supervised prediction models only learn what the text corpus they’re trained on is likely to say about our world, not what’s actually true of it. And because language is already an incredibly effective compression of reality, language programs are likely to stay in control of AI progress.
A powerfully illuminating and empathetic piece that serendipitously allows me to discuss with an actor in a play this potential for a character's consciousness that he is having a psychotic episode.
Thank you for writing this, Freddie.
"As I’ve mentioned before, the technical term is anosognosia - the inability of the sick to understand that they’re sick. Anosognosia is indeed very common among people with schizophrenia and schizoaffective disorder, but not universal, and while many bipolar patients experience the condition, most estimates I’ve seen suggest that a majority don’t".
My best friend will, most of the time, acknowledge that he has schizophrenia and that he has previously experienced hallucinations and paranoia, but for him, the main voice he hears is an exception this this, and //exploits// the fact that he has diagnosed schizophrenia to hoodwink family, friends and medical staff while continuing to torment him.
This is really illuminating. I live in a place with, unfortunately, a lot of mentally ill people living on the streets, and it’s not uncommon to run into several of them just walking across the small downtown. Something I’ve noticed is that most of them are lucid to some degree – the man who frantically, compulsively pushes a broom around the sidewalk while muttering to himself will move out of the way without breaking his stride if you say “excuse me!” to him. The woman having a furious growled conversation with the air while she tries to navigate her shopping cart over a curb breaks midsentence to say “I’ve got it, thanks!“ in a perfectly normal voice if you ask if you can help, then goes back to talking to the air. Once a guy apologized for being too loud and wished me a nice day as I walked by, because he accidentally startled me by yelling while he was in the middle of a heated conversation with Jesus.
Of course the nature of psychosis is that it’s unpredictable. But it seems that for many people who have been living with it for a long time, even dysfunctionally, it’s not that they’re wholly incapable of engaging with reality – I’m precisely as real to them as the unreal things they’re seeing and hearing. It just sucks, because while that means we can sometimes have a brief, fairly normal interaction when we’re both trying to navigate the same physical space, it also means that I can’t see the unreal thing that may be standing between them and seeking help like Bynes did.
No matter what happens, Amanda Bynes is always the man! I hope and pray for her recovery!
Thanks for writing this. We have found the same is true with a sister of mine. She has "learned" over the years what is happening, and so have we. So we can all deal with it as an illness instead of a momentary catastrophe. I wish you and all others in this situation all the best.
Evaluating insight into their condition when patients try to leave against medical advice can be very interesting. All they really have to say is that they appreciate the doctor thinks they're at risk by leaving but they're willing to take that chance. If they're not organized enough to say something like that, they have to stay.