Anosognosia revisited

Over a decade ago, I noticed a new term cropping up in psychiatry: anosognosia. Actually, it was an old term, coined a century earlier by neurologists to describe a behavioral sign in certain types of brain damage. But by the time I took notice, this mouthful of a word was doing double duty. In neurology it still referred to peculiar deficits of awareness linked to specific anatomical damage in the brain. In psychiatry it began to mean something else.

I wrote a blog post in 2011 on the misuse of “anosognosia” in psychiatry, a word that had come to mean the denial of illness often seen in severe psychiatric disorders. Psychiatry already had a perfectly good phrase to describe this denial: lack of insight. By my reckoning, replacing this with a neurological term was a rhetorical trick, “intended to make psychiatry sound better understood, and more biological/neurological, than it really is.”

Insight is subjective

At the time, I emphasized subjectivity in assessments of insight. This is reflected in the common observation that psychiatric patients have “good insight” if they agree with their psychiatrists, and “poor insight” otherwise. Given the many reasons one might differ with one’s psychiatrist, from delusions to honest conceptual disagreements, “anosognosia” seemed a crude way to pathologize disagreement.

It’s now 2024, and “anosognosia” is alive and well in popular media about mental illness. It’s a word pushed by treatment advocacy groups, less so by psychiatrists. In view of its tenacious foothold, I’d like to revisit this formerly obscure term, with a somewhat different spin than I gave it 13 years ago. It’s not so much the subjectivity of insight I want to emphasize now, but the political use of language.

Neurology uses “anosognosia” precisely

Coined in 1914 by the neurologist Babinski, for a century “anosognosia” meant denial or lack of awareness of disability due to injury or disease of specific parts of the brain.

One example is “hemi-inattention”: lack of awareness of one side of the body, or one side of the visual field, due to parietal lobe damage. For example, a patient with a right-hemisphere stroke may be unaware that his or her left arm is paralyzed. Such a patient, when asked to draw a clock face, may add numbers only to the right side, neglecting the left completely. Or notice a stimulus in the right visual field (of both eyes), but not the left. This is hemi-inattention. Other early examples included cortical blindness and cortical deafness, sensory deficits that the patient may be unaware of, due to damage in the cerebral cortex.

Prior to the invention of brain imaging, e.g., CT scans in the 1970s, neurologists could determine the site of brain damage, most often in the parietal lobe, by the presence of anosognosia. It was, and is, a neuroanatomical localizing sign. One of the special skills of the neurologist is to localize nerve or brain dysfunction by careful assessment of signs and symptoms.

Later, neurologists applied “anosognosia” to lack of awareness of deficits in some types of dementia, due to specific frontal lobe damage. This is again a neuroanatomical localizing sign.

Words can slip in meaning

Unfortunately, the application of this word in dementia may have misled lay observers (and a few die-hard biological psychiatrists) to conclude that lack of illness awareness in schizophrenia and manic psychosis is anosognosia as well.

I should digress here to acknowledge that words, especially psychological words, often change their meanings over time. “Idiot,” “moron,” and “imbecile,” all words that now roughly mean “stupid,” once had precise psychological meanings. “Narcissism” and “gaslighting” have devolved from highly specific terms to general terms of disapproval. If one likewise wants to argue that “anosognosia” once meant something very specific in neurology, but now means general cluelessness or lack of insight, I can’t really say that’s wrong. I can only regret that a once useful word has been dumbed down in the service of political rhetoric.

A veneer of pseudoscience

“Anosognosia” in psychiatry offers nothing more than a veneer of pseudoscience. There’s no evidence that lack of insight in schizophrenia is related in any neurological way to the lack of insight in anorexia nervosa. Yet “anosognosia” has been applied in both. Conversely, patients with ego-syntonic personality disorders also lack insight and deny their illness, but the term is not applied there, because these disorders don’t seem “biological” enough to lay claim to a neurological term.

More fundamentally, there is no neuroanatomical localization in psychiatric disorders — by definition really. As I’ve often noted, whenever a psychiatric condition is localized and understood medically, it inevitably leaves psychiatry, to be claimed by neurology or another branch of medicine. Psychiatry is what’s left. The only thing “anosognosia” adds to “lack of insight” is the implication that (some) “psychiatric disorders are brain diseases.”

As I’ve also repeatedly argued, this is political rhetoric, not science. The reality is far more complicated. Psychiatric disorders differ widely in the degree to which they appear biological, yet even the most biological-seeming have strong psychological components.

Anti-psychiatrists often claim there is no biology in psychiatry at all — that it’s all social control, punishment of deviance, and the like. Conversely, treatment advocacy groups and some biological psychiatrists seem to believe it’s all biology.

False certainty helps no one

Both groups are mistaken. Psychiatry is fundamentally uncertain and in-between. Yes, there’s biology. And yes, there are social factors at play in psychiatric assessment (and functioning) as well. Even in psychosis, a patient’s degree of insight can fluctuate in response to stress and other environmental factors. It’s not, as the neurologists might say, a fixed deficit.

As I wrote 13 years ago, the use of “anosognosia” is a rhetorical device to make insight judgments in psychiatry sound more biological/neurological, without offering any particular reason to do so. This misuses the term, and lays claim to unwarranted certainty in an inherently uncertain domain. Like “chemical imbalance” and other convenient oversimplifications, it’s understandable that untrained laypeople might make this mistake. It’s more disappointing when mental health professionals do, especially when they really should know better and be more honest.

Palliative psychiatry

The application of palliative care to intractable psychiatric disorders has been debated at least since 2010, when a journal article reported that a patient with severe anorexia nervosa died in hospice, after referral there by her psychiatrist.  The New York Times published a thought-provoking article earlier this year on the same topic: whether we should ever deem severe, treatment-refractory anorexia incurable and terminal.

Are there incurable psychiatric patients?

Proponents argue that only hubris and false hope on the part of psychiatrists stand in the way.  They say we should treat such patients as our colleagues treat medically incurable patients: with palliation and hospice.

This question is vexing enough.  But eating disorders are an exception in psychiatry: untreated, they can lead to death from medical causes.  Other mental disorders are miserable but not terminal in the same way. 

Medical aid in dying

For this reason, discussions of “palliative psychiatry” lead directly to medical aid in dying (MAiD). Although MAiD solely for psychiatric conditions is not legal anywhere in the U.S., laws permitting it exist in Belgium and the Netherlands, and are pending in Canada.  Accepting the framework of palliative psychiatry for incurable conditions appears to entail MAiD. 

However, arguments that advocate for palliative psychiatry are muddled in several ways, and do not in fact lead to that conclusion.

Psychiatry is already palliative

First, psychiatry is inherently palliative.  All somatic psychiatric treatment (medication, ECT, TMS, and so on) treat signs and symptoms of psychiatric disorders, not their root causes. That’s because we don’t know these root causes, nor the mechanisms that connect them to the manifest signs and symptoms we observe.  In essence, all such treatments aim to provide symptom relief, comfort, and support — the very definition of palliative care.  It makes no sense to speak separately of palliative psychiatry when palliation is virtually the whole field.

The only exception is psychotherapy.  Psychotherapy aims to treat the root causes of emotional distress.  Of course, this can succeed or fail, and in the case of failure we and our patients routinely resort to palliation.  This is called supportive psychotherapy.  It’s hardly a new concept that needs a new name.

Treatment resistance is slippery

Second, arguments for palliative psychiatry usually invoke “treatment resistance,” or refer to “treatment-refractory” disorders.  Disorders so named are the putative targets of palliation, since we can’t “treat” them.

There are biases hidden in such language.  Treatment resistance is a concept from biological psychiatry.  It means a particular patient fails to improve in the face of somatic treatments that help most other patients.  However, as David Mintz argues, adding psychotherapeutic elements to a medication treatment can overcome this kind of treatment resistance.  From a psychotherapeutic standpoint, treatment resistance may say more about the treatment than the patient.

Psychiatric disorders are not “things”

Psychiatric disorders sound misleadingly like reified “things” we can treat with concrete interventions.  In reality, our moods, thoughts, impulses, and actions result from a complex interplay of biology and psychology.  Treatment resistance in that light is vague and abstract — not a sound basis for life and death decisions.

Again in contrast, psychoanalytic psychotherapy is well-acquainted with treatment resistance.  In fact, it’s expected.  Not only is resistance not a reason to give up, it can be a signpost to insight and improvement.

Personality change can take a long time.  I saw a highly defended patient in weekly psychotherapy for several years before she allowed herself to be vulnerable and introspective.  In the years before the change I often wondered if we were wasting time and money, if she was “treatment refractory.”  Now we both see that she isn’t.  Conversely, I’ve seen another patient even longer with little to show for it.  Is he incurable?  There’s no way to know.

Being present and bearing witness

Third, sensitive psychiatrists (and other mental health professionals) stay with our patients whether they improve or not.  The original idea behind palliative care was attending to the patient’s “total pain,” which includes the physical, emotional, social, and spiritual dimensions of distress. Not listed but equally important is bearing witness to distress, and maintaining a caring therapeutic relationship come what may.  Again, we offer palliation in nearly everything we do.

MAiD is never inevitable in psychiatry

Last but not least, given all of the above, MAiD cannot follow as a logical next step even after long-term hopelessness or failure to improve psychiatrically.  Staying present isn’t hubris and it isn’t imparting false hope.  If a patient chooses to forgo further treatment, whether somatic or psychotherapeutic, we will honor that choice and remain available.  If local laws someday allow, and as a matter of personal conscience, some of us may choose to participate in MAiD.  But that will be an individual matter quite separate from incurability, treatment resistance, or comparisons with terminal medical conditions.

Political advocacy and psychotherapy don’t mix

Two senses of “psychotherapy is political” are often conflated.  The first is the notion, popular lately, that psychotherapy either allows or demands political advocacy in the therapy room itself.  The other is recognition that political factors influence the nature and practice of psychotherapy.  It is a conceptual error to confuse the two, and a clinical error to justify the former by appeal to the latter.

Yes, psychotherapy is political (like everything else)

Viewing the practice of psychotherapy through a political lens, albeit one lens among many, can be valuable and revealing.  Political analysis of this sort can be applied to nearly all human endeavors: war, housing, work, romantic relationships, childrearing, sports, nutrition, medical care, media, etc.  There is no reason to imagine psychotherapy is an exception, and indeed it is not.  To cite just a few of the most obvious areas where politics intersects with psychotherapy:

  • third-party payment, public and private, and out-of-pocket cost
  • the choice of psychotherapy versus other types of help
  • social stigma, both of mental disorders and their treatments
  • social inequities that lead to despair, anxiety, and anger
  • controversies over what counts as a mental disorder
  • lobbying and other activities of professional organizations
  • allocation of research funds

    Note that none of these dictate how therapy itself should be conducted, aside from the value of understanding and appreciating what the patient is dealing with.  In other words, for empathy.  None of these political issues speak to what psychotherapy is, or realistically offers.  Even issues such as gender and racial dynamics within psychotherapy itself, while important to be aware of, need not alter the way competent therapy is conducted.

    One political view that matters

    However, there’s a different kind of political position that does affect psychotherapy itself.  It’s the degree to which one situates pathology in the sufferer, versus in his or her environment.  Even Freud grappled with this.  His early “seduction theory” held that childhood sexual abuse led to neurosis.  Yet he was unable to believe such abuse was widespread.  Thus, he soon revised his account to say that young children had sexual (or sexual-like) fantasies that led to inner conflict.  This revision justified treatment of the individual who harbored the conflictual fantasies.

    Conversely, some therapists today hold that emotional distress and dysfunction are always “normal reactions to abnormal situations.”  That is, the pathology lies outside the patient.  This perspective justifies social action, not inner exploration.

    Where patients locate pathology affects how and whether they seek therapy.  Those who frame their problems as existing entirely outside themselves — cruel bosses, uncaring spouses, or social pathology such as racial, gender, or class inequity — don’t come to therapy to change anything about themselves.


    In the traditional language of psychotherapy, such patients externalize: they complain about the outside world, which psychotherapy can do nothing about, and disclaim responsibility for their plight.  Typically, therapists are quick to challenge this stance when it comes to difficult bosses and spouses.  After all, therapy can’t change other people, only the patient.

    Curiously, despite identical logic, many therapists today accept externalization with respect to social ills.  They see their role as providing support, validation, and “advocacy.”  They believe that focusing on the patient’s responsibility for navigating the outside world is “blaming the victim.”

    Unfortunately, advocacy by itself isn’t therapeutic.  Validation and support don’t lead to change.

    Advocacy in therapy — more accurately, advocacy instead of therapy — doesn’t directly improve the noxious environment either.  Framing the patient’s distress as a struggle against “oppression” draws an oversimplified battle line, with patient and therapist comfortably on the same side. It promotes a primitive fight-or-flight duality instead of creative, nuanced alternatives.  This false simplification may make both parties feel better for a time.  At best it may inspire activism, which may help the patient feel better indirectly, depending on the presenting complaint.

    What political advocacy trades away

    Of course, spurring political activism is not the purpose of psychotherapy.  Psychotherapy is a treatment, not a pep talk or political rally.  Therapy isn’t designed to make the world less harsh or more loving.  As the old joke says, the lightbulb has to want to change.

    In addition, much is traded away when political advocacy colors treatment.  It’s no secret that most therapists are on the liberal side of the political spectrum.  Yet politically conservative patients may badly need a safe place to talk — not a lecture.  Even patients who mostly share their therapists’ political outlook may have mixed feelings, or be of two minds, about social issues.

    Fundamentally, political advocacy violates the precept of therapeutic neutrality.  In Freud’s original formulation, therapeutic neutrality meant not siding with one aspect of the patient’s psyche over another.  No favoring the superego over the id, for example.  A more modern way of saying this is that it doesn’t help for the therapist to weigh in on a conflict the patient is struggling with.  Casting a vote for one side or the other won’t resolve the conflict.  And let’s be clear: patients who are sufficiently troubled by social injustice to seek therapy have intrapsychic conflicts that amplify the injustices they face.  No one comes to therapy for problems they can figure out for themselves.

    The fundamental dialectic of psychotherapy

    Marsha Linehan, the founder of dialectical behavior therapy (DBT), recognized back in the 1970s that neither blaming patients nor completely absolving them is helpful.  Patients can learn to accept themselves, yet they need to change too.  This apparent paradox is the “dialectic” in DBT’s name.

    As I’ve written elsewhere, this dialectic of self-acceptance and change exists in all psychotherapy, not just DBT.  That’s why most therapists stake out a middle ground that recognizes real-life hardships and tragedies, as well as the reality that psychotherapy is for personal, not social, change.  Unfortunately, therapists who primarily see themselves as advocates for their patients conflate the political factors surrounding therapy with a false need to be political IN therapy.  In doing so, they trade away the value of psychotherapy as an avenue for personal insight and development.

    Working with people we disagree with

    Needless to say, not all therapists and patients can work with each other.  A gleefully misogynist or racist patient may stir up such intense anger in liberal therapists that they cannot work with that person.  Conversely, a patient who repeatedly intones progressive talking points may anger centrist or conservative therapists.  Therapists are human, and there are limits to what any of us can tolerate.

    These limits should be wide, though.  Just as surgeons must tolerate the sight of blood, and dermatologists cannot recoil from disfigured skin, therapists must work with personalities and viewpoints we find offensive.  (Not the converse: patients need not put up with therapists who share strong, potentially offensive viewpoints. That’s one reason for therapists to tone it down — not be a “blank slate,” just leave enough room for the patient to feel comfortable and safe.)

    A strong working alliance does not require political agreement.  Indeed, if political talk serves a defensive function, e.g., externalization, the therapist’s job is to help the patient recognize that dynamic and look beneath and beyond it.  In that sense, psychotherapy may be more effective when patient and therapist disagree somewhat politically, lest they unwittingly collude in defensive avoidance.

    Image courtesy of vectorolie at

    HIPAA-compliant email revisited

    graphic of padlock with 'at' sign

    In January 2021, I described my search for a HIPAA-compliant email provider for my practice, and reviewed several options. In the end I chose Hushmail for Healthcare. It was, and is, a good service: thoughtfully featured, reasonably priced (with minor changes since then), and fast support when needed. But the truth is, I didn’t stay with Hushmail.

    This is the continuing saga of my search for secure email and online forms for my practice. And here’s the punchline up front: I haven’t found a service that provides all I want. Lately I’ve cobbled together two competing services that “kind of” work in combination. If anyone reading this happens to work at one of these companies and wants to please a customer like me, please take note.

    The need

    As I wrote in 2021, regular email is not in compliance with HIPAA, the longstanding federal law that governs the confidentiality and “portability” of medical information. Along with many other provisions, HIPAA has rules for maintaining the security of electronic medical data, and the transfer of protected health information from one place to another.

    Small private practices like mine sometimes ignore these rules. I had used Google Calendar for patient scheduling for years, and with the start of the Covid pandemic, I started exchanging email with patients. Neither Google Calendar nor regular email is secure or HIPAA-compliant. But plain email is the easiest option for quick doctor-patient communication (except perhaps for texting, which isn’t HIPAA-compliant either). It’s a no-brainer for doctors and patients alike.

    However, doctors and other healthcare providers need to take our fiduciary responsibility seriously. The easiest and most convenient options aren’t always, or even usually, the best. This misguided tradeoff crops up everywhere now. For example, many health care startups cut corners clinically to make their services cheaper or “frictionless.” Medical ethics are a higher standard than business ethics, and as a profession we should resist the seductions of lower standards.

    As my use of email expanded, I grew troubled by the hypocrisy of writing and teaching about medical and psychiatric ethics, while engaging in unethical practices myself. During the height of the pandemic, when I was working exclusively online, I realized I needed HIPAA-compliant email, HIPAA-compliant online forms for initial inquiries by patients and my short intake questionnaire, and a HIPAA-compliant alternative to Google Calendar. That’s what prompted my initial search.

    Hushmail’s tragic flaws

    At the time I signed up, Hushmail offered two secure, customizable forms along with email and 10 Gb of storage, all for about $100/year. There was no calendar option, but it otherwise seemed to fill my needs, and the price was right.

    The forms were great. The email proved problematic, for two reasons I hadn’t anticipated. First, patients didn’t use it. Per Hushmail’s recommendation, I created a dedicated email address in my website domain, to be used solely for Hushmail. But as soon as I started sending secure webmail using that address, patients would use the address to send me regular, un-secured email. I was constantly asking patients not to contact me that way.

    The bigger problem was passwords. When non-Hushmail users, i.e., all of my patients, received their first secure message from me, Hushmail asked them to set a personal password to decrypt that message, and all subsequent messages. I didn’t know their passwords. So every month, I’d hear from patients, usually by plaintext email, that they forgot or misplaced their passwords, and couldn’t read their billing statements. They’d reset their passwords, and I’d have to re-send their statements. Since their new passwords wouldn’t open any old messages, sometimes I needed to re-send several documents.

    Back on the search

    After about two years of this, in late 2022 I got fed up and looked again at my options. This time I was drawn to Proton, largely because it also offered an encrypted calendar. When I reviewed it in 2021, Proton either wasn’t HIPAA-compliant (no Business Associate Agreement) or I didn’t notice that it was. In any case, it was now.

    Proton didn’t offer secure online forms. But it featured an ecosystem of HIPAA-compliant services — email, calendar, cloud storage, and VPN — and I figured I’d work something out. It offered 15 Gb of storage, ten customizable email addresses, the same ability to link to my custom domain that Hushmail offered, and it was only $84/yr for the “Mail Essentials” option. With some regret I bid Hushmail goodbye, and signed up with Proton.

    Proton pros and cons

    I finally had a secure calendar for scheduling patients. This was long overdue, works great, and feels like a necessity going forward.

    Proton also solved the lost password problem. I assigned a password to each patient, and could remind them when they forgot. I gave each patient a unique 7-character password, based in part on their name. I assumed this would be more secure than one password for everyone.

    On the other hand… this worked well for about six months. Unfortunately, at that point Proton suddenly required all passwords to be eight characters or more, without ever documenting any password-length requirement anywhere. I had to send out notices, further confusing many patients who didn’t quite grasp password-protected webmail in the first place. Exasperated, I gave everyone the same password.

    I changed my prior Hushmail address to auto-respond with an error message, and created a new Proton email address for sending and receiving secure webmail. Proton Mail works very well, much as Hushmail did. However, patients started using my new address to send me plaintext email, just like before.

    So here’s my first concrete request: Please, someone provide HIPAA-compliant email that doesn’t disclose an address that non-subscribers can use to send back plaintext email.

    Then there’s the online forms. For most of the year, I left a non-secure contact form on my website, which I insisted that established patients not use. Some did anyway, of course.

    I asked new patients to download my blank pdf intake form, fill it out on their computer, and return it to me using Proton Mail. This meant sending Proton Mail to them first, so they could reply to it. That way, their reply, with the attached intake sheet, was secure as well.

    This was convoluted. I had to collect new patients’ email addresses in my first phone contact, before they ever saw my intake form. And I had to convey their new Proton Mail password somehow. One option was a “password hint” attached to my first Proton Mail that revealed their new password in plaintext. Another was to explain this all verbally on the phone. A third was to send them, in a separate email without encryption, a “read me” file explaining what all this was about. It was awkward at best.

    Proton could streamline secure forms submissions in one of several ways. They could offer HIPAA-compliant online forms, as Hushmail and other services do. Or they could provide a “mail drop” feature allowing non-subscribers to upload files securely to a subscriber’s Proton Drive (perhaps in a quarantine folder or similar). Or, like MailHippo‘s SendSafe feature, Proton could assign each subscriber a unique URL that allows non-subscribers to securely send messages and attached files to the subscriber.

    But Proton did none of these, so I had to continue searching.

    Adding secure online forms

    MailHippo, which I reviewed back in 2021, now offers HIPAA-compliant forms as well. They call the combination FormHippo. Per my 2021 review, I hadn’t chosen MailHippo solely due to their lack of forms. I now realize that their SendSafe feature, even without forms, would have saved me a great deal of trouble compared to the workarounds I needed with Proton.

    FormHippo is $107/yr for email, up to five customizable forms, 5 Gb of email storage, message recall (not offered by the others), and the SendSafe feature. The drawbacks? No calendar, they still disclose an email address that non-subscribers can use to send plaintext email, and (sorry to say this) the hippo name and logo itself, clearly a play on “HIPAA,” which looks a bit unprofessional. Otherwise, it appears to be a very good option.

    However, I’m not quite ready to abandon Proton’s ecosystem, with its calendar, cloud storage, and VPN. I’m generally happy with Proton Mail, and right now they’re rolling out a password manager, Proton Pass, that may compete favorably with the Bitwarden I currently use.

    My solution so far

    I simply added FormHippo to Proton. The email part is redundant, but both together still cost under $200/yr. I use FormHippo for online forms (contact and intake) and for the SendSafe URL. I ask all patients to use the SendSafe option if they ever want to contact me online. As usual, not all do. I use Proton for everything else.

    (To complicate things even further, the SendSafe URL includes the login email address that the subscriber uses for MailHippo or FormHippo. I didn’t want to advertise this address, so I added yet another layer: redirection from an innocuous URL in my own domain to the SendSafe URL. Furthermore, if I want to reply online to a SendSafe message, I’m still stuck sending either MailHippo or Proton webmail, possibly to someone who doesn’t understand all this arcana. Whenever possible, I just use the phone.)

    While I don’t plan to change anything for now, I consider other permutations. I could downgrade FormHippo to MailHippo (costs a little less), have patients download my intake sheet from Proton Drive, and submit it to me using SendSafe. Or I could cancel Proton, just use FormHippo, and opt for another secure calendar solution. Tutanota, for example, offers encrypted, but not HIPAA-compliant, full-featured calendars to businesses for $76/yr. (My pricing for Tutanota in 2021 must have been a misprint.)

    Or I could go back to Hushmail, and tell all my patients what password to use — maybe the same universal one they now use for Proton. Hushmail also now offers a feature called a “private message center.” Information about it is scarce on their website. Apparently it allows non-subscribers to access Hushmail messages using their Google, Apple, or Microsoft credentials — instead of a password? If so, that might simplify matters too. They still don’t have a calendar though.

    I confess that this journey has been challenging. It’s easy to see why many in my position don’t bother. I confess that I still send and receive texts from patients, even though I know I shouldn’t. There comes a point when you do what you reasonably can, and then don’t sweat it anymore. I’m about at that point, but still open to suggestions and comments. Let me know how you (or your psychiatrist/therapist) handles these challenges, and thanks for reading.

    Making an AI dynamic therapist

    Currently, therapy apps featuring a nonhuman “therapist” aim fairly low at best, and at worst willfully mislead the public.  However, the advent of large language models (LLMs) such as ChatGPT-4 brings exciting potential for genuine depth psychotherapy delivered by AI — and many challenges and potential pitfalls as well. 

    Since “therapy” has no precise definition, marketers apply the term to any product, digital or not, that arguably helps a user’s emotional state: encouragement by text, formulaic cognitive homework, brief meditative interludes, and so on.  The semantic ruse is that the vague term “therapy” often stands in for psychotherapy, a word with a good deal more precision.

    The current state of AI therapy

    Unlike affirmations, inspirational poems, and nonspecific relaxation exercises, psychotherapy is a treatment designed to alleviate specific emotional problems.  Professional psychotherapists adapt general principles and strategies to a specific patient, and alter their approach based on their patient’s real time response.

    Some therapy apps can now simulate this, at least in part.  However, even the most advanced and nuanced are quite rigid compared to a skilled human therapist.  No current app escapes the orthodoxy of its programmers: users deal with a procrustean bed that either fits and helps, or painfully doesn’t.

    More fundamentally, apps that emulate psychotherapy limit themselves to cognitive behavioral therapy (CBT) and its offshoots.  This has been a pragmatic choice — notwithstanding marketing that falsely implies CBT is the best mental health treatment for almost everything.  It’s simply easier to operationalize and program CBT compared to other types of psychotherapy.  Current computer programs respond to users by following instructions in flowcharts and decision trees, sometimes with fuzzy logic to make the output somewhat less predictable.  Semi-random word substitutions, along with jokes and other human speech markers, can make the dialog more lifelike.

    CBT helps many people with relatively concrete emotional symptoms such as depressed mood, certain types of anxiety, and insomnia.  As CBT therapy apps continue to improve, they may eventually rival treatment conducted by skilled human CBT psychotherapists.  If I were a CBT therapist, I’d be anxious over the prospect that AI could replace me in the next five to ten years.

    Why we need more

    Even when conducted by skilled humans, CBT is not well suited for obscure emotional distress or dissatisfaction, unwitting self-defeating behavior or attitudes, or recurrent dysfunctional relationships.  Psychotherapy of “depth, insight, and relationship,” in particular the psychodynamic psychotherapy that began with Freud but evolved over the past century with refinements in theory and technique, addresses these more pervasive yet subtle struggles.

    Such therapy lacks CBT’s reputation as “evidence based treatment,” but this is a false narrative.  Empirical research documents this evidence, and in fact casts doubt on the value of “evidence based” claims to differentiate schools of mainstream psychotherapy.  Unfortunately, dynamic psychotherapy is by nature time-intensive and therefore expensive, and therapists are always scarce.

    To date, AI researchers and developers have not attempted to tackle psychodynamic psychotherapy.  It’s a tough area.  As Freud famously observed, such therapy is like chess: “… only the opening and closing moves of the game admit of exhaustive systematic description… the endless variety of the moves which develop from the opening defies description….”  Moreover, this type of treatment relies on transference and countertransference, carefully timed interventions, inferences about inner states, and much else that standard programming technique cannot capture.

    Large language models

    Recently released large language models (LLMs) appear to overcome some of these stumbling blocks.  ChatGPT, a product of OpenAI, responds to typed questions in nuanced ways that are not “canned.”  OpenAI trained the model with all of the psychological theory on the internet (and much, much else).  People are already using ChatGPT as a makeshift therapist.  Thanks to its vast training set, and leaving aside its occasional “hallucination” of false information, ChatGPT is likely better than most other sources at serving up regurgitated mental health tips and advice.

    However, to tap the real magic of LLMs, “fine tuning” is key.   This alters the LLM such that certain types of responses are favored, and others are disfavored.  Most simply, developers could fine-tune a copy of ChatGPT to answer questions in the style of a well-known psychotherapist, or of a named school of psychotherapy. Users would not have to specify this qualification each time in the input. 

    Surprising capabilities of LLMs

    Of course, answering questions is not the main job of the dynamic psychotherapist.  Here’s where it gets interesting.  In an exhaustive review of early experiments with ChatGPT-4, the latest iteration, a Microsoft research team found the model possessed surprising emergent properties.  As documented in their paper, ChatGPT-4 offers plausible accounts of internal mental states to help explain human behavior.  It can predict the likely emotional and behavioral response of a described person to its own output.  It evidences “theory of mind.”

    Given the above, ChatGPT-4 should be capable of “educated guesses” about the functional and emotional role it plays for the user, i.e., transference, as well as the range of emotions a human (or a human with a particular personality) might feel in its position.  That is, countertransference of a sort.

    Developers could fine-tune ChatGPT-4 and its successors to emphasize “theory of mind” and transference-countertransference aspects, as well as other traits and emphases of the psychodynamic psychotherapist.

    Just a machine?

    Hold on, I hear you say.  This is just a machine, a fancy program.  People can’t form therapeutic relationships with non-humans.

    They can and do.  In fact, it’s remarkable how readily most folks anthropomorphize their dogs, cats, robot vacuums, and the primitive therapy chatbots now in use.  Empathy is imagined on very little evidence.  A sophisticated AI therapist would turn this bug of human nature into a feature.

    Ok, what are the challenges?

    Many challenges await — and not being an AI expert, I can’t say whether these are easily solved, difficult, or impossible.  Here is a far from exhaustive list.

    First and perhaps easiest, security and privacy need to be built in from the start.  To protect personal health information (PHI), developers may need to supply each user with a separate instance of the AI disconnected from the internet.  (Current LLMs function independently of the internet, although some in development are online, in order to download and incorporate up-to-date information, access other resources, etc.)

    Second, the usual question-and-answer format of interacting with ChatGPT needs to be reversed.  In psychotherapy, it’s the therapist who mostly asks the questions.  Indeed, an LLM fine-tuned for dynamic psychotherapy would not answer many questions put to it.

    Third is the vexing issue of timing interventions.  While an LLM could be fine-tuned to stay on task, and to offer standard psychotherapy interventions like observations, clarifications, and interpretations, it may not be possible using current models to programmatically control when it offers them.  From what I’ve read, this may be a major hurdle.

    Fourth, measures of client progress need to be part of the model.  Psychodynamic therapy isn’t only concerned with symptomatic improvement of mood or anxiety; it also tracks more abstract gauges of well-being such as the ability to love, work, and play; stress tolerance; accurate self-assessment, and so on.  Much of the training of depth psychotherapists is devoted to recognizing these “soft signs” of mental health, which defy self-report rating scales and other concrete measures.

    Technical point: A real psychotherapy app would include significant conventional programming, with ChatGPT invoked through an application programming interface (API). ChatGPT APIs are already available and in use. The regular part of the app might handle some of the challenges above.

    Where would this leave us?

    Even if all of this and much else were handled well, many depth psychotherapists would object that there is no “human connection” with AI.  True empathy would still be missing.

    How much this matters would be, to some extent, a testable question.  As noted above, the inability of current therapy chatbots to feel empathy doesn’t render them useless. We should remember that even if AI imperfectly emulates human-led psychodynamic therapy, it could still prove hugely beneficial for the many who have no access to the real thing, and are not well served by treatments limited to concrete symptoms.

    My sense is that creating a useful psychodynamic psychotherapist using LLM technology would be a serious challenge, but may be possible.  Moreover, the mere effort, even if unsuccessful, would help to clarify some of our thinking about this type of psychotherapy. In the best case, serious AI psychotherapy may soothe troubled souls the same way human therapists do: by fostering emotional insight, and through the healing nature of the relationship itself.

    Image by Eric Blanton from Pixabay

    Old medications in new bottles

    Money tumbling from a drug capsule

    I wondered whether to recommend Nuedexta for my elderly relative with worsening Alzheimer’s disease and daily spontaneous crying.  Often she would deny feeling sad and couldn’t account for her sobbing.   Avanir Pharmaceuticals released Nuedexta in 2010 to treat pseudobulbar affect: involuntary emotional outbursts due to certain neurologic conditions or brain injury.  I wasn’t sure if these crying bouts qualified as pseudobulbar affect.  But my biggest hesitation was the breathtaking price of the drug.

    Nuedexta is a “fixed-dose combination” (FDC) medication containing dextromethorphan (DXM) and quinidine.  The former is an active ingredient in over-the-counter cough syrup, the latter a century-old antiarrhythmia medication.  Both are very inexpensive, and DXM doesn’t even require a prescription.  Yet according to GoodRx, the average retail price of Nuedexta is nearly $1700 per month.  There’s no generic equivalent yet, although there may be soon.

    It’s impossible to comparison-shop for novel prescription drugs like Nuedexta.  Patent protection and “non-patent exclusivities” enforce monopolies: no company can sell the same product generically or under another brand name for 20 years after a drug is invented, or about ten years after it hits the marketplace.  Justified as a way to recoup development costs, pharmaceutical companies often manipulate these price protections to maximize profits.

    1. How to win at Monopoly: Mix it up

    One way is to create a new, pricey FDC medication, like Nuedexta, out of two or more cheap generics.  To take another example, the FDA approved Auvelity last month to treat major depression. It’s an FDC that combines DXM — the very same cough suppressant — with bupropion, a common generic antidepressant.  Auvelity reportedly works faster than conventional antidepressants.  The manufacturer has not yet announced a price, but it’s safe to assume it will be far higher than that of its generic components.

    This isn’t a new practice.  Triavil (1962) combines the antidepressant amitriptyline and the antipsychotic perphenazine.  Adderall (1996) combines several generic amphetamines.  Symbyax (2003) mixes the antidepressant fluoxetine with the antipsychotic olanzapine — an updated Triavil, in a way.

    Outside of psychiatry, it’s easier to justify some FDC formulations.  Oral contraceptives combine two hormones, estrogen and progestin. Antiretroviral FDC drugs revolutionized AIDS treatment.  The novel FDC Paxlovid uniquely benefits COVID-19 patients.

    2. How to win at Monopoly: Find a new use

    Another way to game the system is to obtain FDA approval to market an old medication for a new use.  It’s the same drug with a fancy new name, packaging, and price.  Its new use may not even be particularly novel: it may be for a condition that doctors already treated with the generic drug “off label.”

    For example, doxepin has been around since 1969 as an antidepressant.  Long available as an inexpensive generic in capsules of 10 to 150 mg, it’s never been popular because it’s too sedating.  In fact, doctors occasionally prescribed it “off label” for insomnia, not depression.  Turning a bug into a feature, in 2010 Currax Pharmaceuticals rebranded doxepin in low-dose 3 and 6 mg tablets as the patent-protected, FDA-approved sleeping aid Silenor.

    According to GoodRx, the average retail price of Silenor is over $600 per month.  There was no direct competition until 2020, although generic doxepin remained available.  Of course, it’s hard to split a 10 mg capsule at home to make a 3 or 6 mg dose for insomnia.  Now with generic competition — about 1/4 the price of the brand name, but still many times more expensive than the larger capsules — Currax strongly urges everyone to insist on brand-name Silenor.

    This is hardly a lone case.  GlaxoSmithKline patented and released Zyban (1997) for smoking cessation to extend its monopolistic pricing for sustained-release bupropion.  Zyban is exactly the same medication, at the same dosage, that was losing its patent protection as the antidepressant Wellbutrin SR.  Eli Lilly likewise patented and released Sarafem (2000) for premenstrual symptoms to extend its monopolistic pricing for fluoxetine.  Sarafem is exactly the same medication that was losing its patent protection as the antidepressant Prozac.

    3. How to win at Monopoly: Tweak it

    A third and very common way to extend monopolistic pricing is to tweak an existing medication just enough to justify a new patent.  Lexapro (2002) supplanted the older Celexa, even though both are citalopram.  The difference is chemically arcane: Lexapro is the S-enantiomer, perhaps with fewer side effects compared to the racemic compound in Celexa.  Pristiq (2008) contains desvenlafaxine, an active metabolite that is arguably a minor improvement over venlafaxine in the older Effexor.  Lybalvi (2021) combines generic olanzapine with a new medication samidorphan to decrease some of olanzapine’s side effects.

    Do these updates justify the premium cost of the brand names (while they are still patent-protected) over the generics?  That’s a judgment call — not in my opinion, although many of my colleagues think so.

    People or profit?

    It’s clear that many new drugs mainly serve business interests, not health care.  A visiting Martian might ask: Why didn’t the FDA simply announce that generic bupropion helps with smoking cessation, and generic fluoxetine with PMS?  Why didn’t the FDA have doctors tell patients to take non-prescription DXM with their prescription bupropion for depression?  Or take it with prescription quinidine to treat pseudobulbar affect?  Why didn’t the FDA advise manufacturers to make generic doxepin in smaller dosages?

    We Earthlings know why.  If the federal government conducted its own product research, it could make these recommendations to improve public health, not to maximize profit.  Instead, on this planet, profit drives drug development.  Indeed, when a clearly needed medication is unprofitable for private enterprise, the government steps in and pays a company to develop the “orphan drug.”  It’s a clear illustration that business goals and health needs aren’t the same.

    My relative with dementia took Nuedexta for a couple of weeks.  It had no effect and was discontinued, leaving about $800 worth unused.  In her case, no one went broke helping Avanir recoup its investment.  Other families are not so fortunate.  

    More recently I evaluated a patient with chronic depression.  Over many years she’d tried nearly every class of antidepressant several times, as well as various antidepressant “augmenting” medications and mood stabilizers (and, yes, psychotherapy).  As luck would have it, she’s currently taking maximum doses of bupropion.  Before sending her off for more esoteric treatment, I suggested she add cough syrup containing DXM, 20 mg per day, and call me back in ten days.  Auvelity isn’t available yet — not that I’d be inclined to prescribe it even if it were.

    Image courtesy of ddpavumba at

    Dr. Tom Insel scorns traditional psychotherapy

    Dr. Tom Insel.  National Institute of Mental Health, Public domain, via Wikimedia Commons

    When one of America’s most prominent psychiatrists expresses deep disdain for depth psychotherapy, especially when that criticism is misinformed and hopelessly outdated, it should concern all of us.

    Dr. Tom Insel directed the National Institute for Mental Health (NIMH) from 2002 to 2015.  Formerly a psychiatric researcher “at the cellular level,” he studied medications and neuroscience.  Insel admits that under his directorship the NIMH didn’t improve care for those with serious mental illness (SMI):

    I spent 13 years at NIMH really pushing on the neuroscience and genetics of mental disorders, and when I look back on that I realize that while I think I succeeded at getting lots of really cool papers published by cool scientists at fairly large costs—I think $20 billion—I don’t think we moved the needle in reducing suicide, reducing hospitalizations, improving recovery for the tens of millions of people who have mental illness.

    After NIMH, Insel led the mental health initiative at Verily, the Google-spawned health science company; co-founded Mindstrong Health, a digital mental health company focused on SMI; launched Humanist Care, a recovery-oriented online therapeutic community; and served as “behavioral health czar” to Governor Gavin Newsom of California.

    Even at age 70, Insel apparently hasn’t found a comfortable place to settle down.  But having learned a thing or two, he wants everyone to know.  Thus his book Healing: Our Path From Mental Illness to Mental Health (Penguin Random House, 2022).

    The book’s main message is that we can already help most people with SMI.  We “merely” lack the social and political will to make it happen.  He grossly downplays these social and political challenges, offering only a roadmap, not policy proposals.

    The roadmap, though, is fundamentally sound: comprehensive biopsychosocial care.  In other words, the biological treatments Insel knows so well, plus skills training, peer and family support, therapeutic communities, a shift from incarceration to treatment, and so forth.  It’s all perfectly sensible, if unsurprising.

    Unfortunately, there’s a glaring bias in Insel’s narrative.  Throughout the book, and confirmed in a recent interview with the New York Times’ Ezra Klein, Insel repeatedly denigrates psychoanalysis, psychodynamic therapy, and all psychotherapy that is not symptom-focused or “skills based.”

    Insel claims such therapies are “eminence-based care” in contrast to “evidence based” [pg 103], and that psychoanalysis is “not by itself a treatment for mental illness” [pg 51]. He believes that traditional psychotherapy blames parents and families for mental illness, and that only by discarding these outmoded approaches, families can now be part of a patient’s support team.  He derides analytic therapy as dwelling on childhood, not current life.  By contrast, according to Insel, evidence-based therapy focuses on learning skills:

    That’s not what you get with talk therapy that’s not focused. It doesn’t have an evidence base. So I’m a huge proponent of psychotherapy, but it has to be psychotherapy that actually involves those kind of skills learning that has a kind of scientific basis to it with people trained to do it in the way that works.

    In his book Insel falsely claims that Victor Frankl developed logotherapy, a type of existential psychotherapy, in reaction “to the introspection and self-absorption of psychoanalysis” [pg 174]. (Logotherapy aims to be more positive than Freudian analysis, not less introspective.)  Writing about ELIZA, an early computer program designed to (roughly) mimic client-centered Rogerian therapy, Insel treats ELIZA’s simple algorithmic responses as though they accurately reflected Carl Rogers himself: “Of course, the Rogerian therapist, with this obnoxious reflexive response, was hardly better than a robot and certainly an easy form of ‘natural language’ to automate” [pg 204].  In rushing to condemn such therapy, Insel conflates a primitive computer program with a real therapist who was neither obnoxious nor reflexive.

    Finally, Insel praises Woebot, a chatbot that provides a version of CBT.  He quotes Woebot (actually, its programmers), apparently sharing their sneering disdain for the “couches” and “childhood stuff” of traditional Freudian analysis, as well as their updated vision of therapy larded with strategies and jokes:

    I’m here for you 24/7.  No couches, no meds, no childhood stuff.  Just strategies to improve your mood.  And the occasional dorky joke.  [pg 215]

    There’s a lot to look at here.  First, mental illness, including SMI, is very heterogenous.  Schizophrenia, bipolar disorder, and OCD count, but by Insel’s reckoning, so do major depression, PTSD, and borderline personality disorder.  This broad category of SMI obscures a wide variation in the applicability of various psychotherapies.

    Most psychiatrists (but not all) agree there is little direct role for insight-oriented, depth psychotherapy in the treatment of schizophrenia, bipolar disorder, or OCD.  Nonetheless, even in these conditions, therapies of “depth, insight, and relationship” can help a sufferer come to terms with his or her debilitating condition, reflect on issues of self-identity and life’s meaning, improve treatment adherence, and provide emotional support.  In other words, even when such therapy doesn’t treat the problem itself, it can help the patient deal with feelings about the problem.  The relationship can be stabilizing and very valuable.

    It’s a far different matter when it comes to depression, PTSD, borderline personality, and many other potentially devastating conditions.  Here we see much stronger evidence for the benefits of in-depth psychotherapy.  Insel is plainly mistaken when he implies that such psychotherapy lacks the evidence of efficacy that other therapies have:

    … for many issues (e.g., depression, anxiety, eating disorders, PTSD) there are specific therapies that have been validated empirically — sometimes called empirically supported treatments. [pg 252]

    It is a widely-held but false belief that only symptom-focused psychotherapy, usually cognitive behavioral in nature and sometimes manualized or algorithmic, is evidence based.  Insel should know better.  There is a large evidence base for analytic, depth therapies, particularly for depression and anxiety.  Consequently, he is also mistaken when he declares these therapies are not by themselves “a treatment for mental illness.” They clearly are.

    Perhaps worse are the outdated stereotypes he uses to denigrate such therapy.  Yes, many decades ago psychoanalysts blamed “refrigerator mothers” for autism and “schizophrenogenic” mothers for schizophrenia.  Dismissing current analytic practice for these old errors is just as silly as dismissing modern biological psychiatry for previously using lobotomy.

    Contemporary analytic therapy doesn’t blame parents or anyone else, nor is it trapped in endless rehashing of childhood.  It tackles plenty of present-day, pragmatic concerns.  But it does so while revealing underlying thoughts, wishes, fears, and more complex emotions, and while closely attending to the relationships the patient forms with the therapist and others.  Like everything else, psychoanalysis has evolved in the past 50 years since Insel experienced it.  Since he’s an influential speaker and writer, it would be good if his prejudices evolved too.

    By far the most ironic twist is Insel’s newfound emphasis on the importance of people (“people, place, and purpose”) and especially relationships for recovery from SMI.  This was his big insight in moving from a strictly biomedical view at NIMH.  While peer groups and clubhouses certainly provide support, the healing value of a close relationship with a caring therapist has been well known for at least a century.  The therapists Insel belittles have certainly known it all along.  His ultimate hypocrisy is complaining about, and failing to take responsibility for, the woeful dearth of research in this area:

    But social connection is not simply the absence of loneliness.  Connection, experienced as support, attachment, or love, has a power that has not been studied sufficiently.  [pg 163]

    Why hasn’t this power been studied sufficiently?  Surely, one guilty party is the former director of our premier mental health research agency, the NIMH.  The power of connection, experienced as support, attachment, or love, may very well “move the needle in reducing suicide, reducing hospitalizations, [and] improving recovery.”  Finding out probably won’t cost anywhere near $20 billion, yet it still awaits serious attention by NIMH.

    Page numbers refer to Insel T, Healing: Our Path From Mental Illness to Mental Health, Penguin Random House, 2022, Kindle version. Quotations without page numbers are from the New York Times interview transcript.