Latest / Eyes Wide Open with Nick Thompson / Vermont Bans AI Therapy — Here's Why Human Therapists Matter
Transcript
- 0:00The state of Vermont just made history by legally banning AI chatbots from acting as mental health therapists.
- 0:06Some people think AI therapy is a great thing.
- 0:09It's an accessible solution to our growing mental health crisis.
- 0:12But this landmark bill is a massive win for humanity because true therapy is not a math problem.
- 0:18It's a real human connection that requires empathy and unique individual judgment that an algorithm simply cannot simulate.
- 0:27Today, we're going to break down why Vermont's new law is so incredibly important and why your mental health must stay in human hands.
- 0:34Now, when I first read this news headline, I honestly did a double take because if you follow the technology space, especially here in the US, which we do quite a bit on this channel and on this podcast, you already know that our federal and state governments have a long history of letting new technology just overrun our culture without any regulations or even really any guardrails.
- 0:57And usually these tech companies will release a product.
- 0:59It will disrupt the industry.
- 1:01And by the time lawmakers notice, the damage is already done.
- 1:05The behaviors have changed.
- 1:06The culture has changed.
- 1:08The business has changed, whatever that looks like.
- 1:10So to see a state legislator actually step up and pass a law.
- 1:14That is proactive, preventative, and safeguards the integrity of human care.
- 1:19It's kind of surprising, but it's also even exciting.
- 1:22So Vermont took a decisive action to protect an industry designed to serve people when they are at their most vulnerable.
- 1:30And they did it by validating some of the professions fundamentally required humanity to function.
- 1:35And that's mental health counseling, therapy, however you decide that you want to describe it.
- 1:40The law of Passive Hermanns draws an explicit, unyielding line in the sand that protects both high-functioning individuals and vulnerable people who look for mental health support in a time of crisis or in a time of need.
- 1:52Under this legislation, it is now defined as unprofessional conduct and unlicensed practice for...
- 1:57for any artificial intelligence system to provide clinical diagnosis, establish a treatment plan, or offer any therapeutic direction whatsoever to a user.
- 2:08Now, this law does not stop a human practitioner from using administration software to handle schedulings or transcribe notes or manage billing or some of the other use cases that we're seeing for AI and other businesses.
- 2:19But what it does do, and what it does really, really well, is make it illegal to substitute human clinical judgment
- 2:27with some sort of AI algorithm or automated code.
- 2:31It codifies the reality that a machine or a bot can never replace a licensed professional mental health counselor or therapist.
- 2:40Now to truly understand why this bill is such a necessary blueprint for the rest of this country and probably even the rest of the world, although I know some places in the world already have legislation and laws
- 2:49to prevent this type of AI therapy from becoming mainstream or legal.
- 2:55But we have to look closely at what therapy actually is because I think that our culture has a little bit of a misconception about how the entire therapeutic process works.
- 3:05A lot of people think therapy is a service where you pay a therapist to sit across from you, nod their head along.
- 3:10validate all the feelings you have, maybe ask you some thought-provoking questions or give you a checklist of advice of what you should do in a given situation.
- 3:18And if those things were true, an advanced AI chatbot might be useful and might be helpful and might eventually be able to replace a therapist.
- 3:27But...
- 3:28While a machine can easily echo back your words or tell you your feelings are valid or give you these bulleted steps of how to do this or do that or improve your mental health or whatever that could be, those generic coping strategies or interventions are pulled from the Internet, which is basically a giant statistical database, and then aggregated into what is most commonly considered next or next.
- 3:54the answer to that question.
- 3:55That's how these large language models work.
- 3:57There's nothing original that comes out of an LLM AI.
- 4:03That's not real therapy.
- 4:04Real clinical care requires honest reflection, deep support, active probing, a willingness to sit in the room with uncomfortable emotions, sometimes in silence, sometimes in tears.
- 4:19And real progress in therapy happens when a clinician or practitioner uses empirical techniques and interventions and human judgment to read body language, to listen to tone of voice, to listen.
- 4:34understand that tremor in your voice when you talk about something that still hurts you or a past trauma.
- 4:40And their goal is to actively help challenge any distorted thoughts or feelings that you might have so that you can reframe them into something that is better suited for you or something that helps you heal or operate better in the world.
- 4:54Real therapy is collaborative.
- 4:56It's unique in between each individual person and their therapist.
- 5:01It's a partnership of where you share healing goals for yourself and ways that you want to show up in the world so that you can do so as your authentic self comfortably and happy, leaving behind these past traumas and these unrealistic thoughts that you have in your head.
- 5:19An algorithm that operates entirely on next prediction, which is mathematically programmed to predict the next most maybe agreeable or affirming or validating comment, that is just simply satisfying a prompt.
- 5:38AI wants to resolve the user's session and chat quickly and cleanly.
- 5:44And if you look at the way you interact with AI –
- 5:48It frustrates the heck out of me how it jumps ahead sometimes to where I'm at or what I'm working on or what I'm trying to figure out.
- 5:56And at that point, it's like it feels like they're trying to get you literally out of the room, literally to shut up.
- 6:02And so when you take that AI and you apply it to therapy in the way that it's kind of programmed to behave to hurry up and get you out.
- 6:11That can never really apply any kind of clinical instinct or any kind of experience because it doesn't have a soul.
- 6:20It doesn't understand what it's like to be a human with a lived human experience.
- 6:26Every single human is very complex.
- 6:29We're all very unique.
- 6:30We've all experienced life in our world and the human experience differently.
- 6:35What has maybe happened to you has happened to somebody else, but impacts somebody else in a completely different way than it impacts you.
- 6:43And so when you take an aggregated type of data like an AI chatbot does, it's applying basically a law of averages.
- 6:51And that doesn't account for the individual experiences that you've had and the individual relationship that you have with your therapist.
- 7:00When we strip away this nuance of humanity and replace it with automated software and AI algorithms under this guise of healthcare, the real world results can be absolutely catastrophic.
- 7:14We do not have to guess what happens when you have AI therapy either because we've tried other –
- 7:22Because other organizations have tried to use automated systems for mental health support, and we already have experienced data and heartbreaking scenarios that have really impacted people negatively and in some cases caused death.
- 7:36Look at what happened just a few years back with the National Eating Disorders Association helpline.
- 7:41In an effort to adjust and
- 7:43improve operations, the organization decided that they were going to eliminate its human staff and replace their manual helpline for people calling in with eating disorders to be an automated chat bot named Tessa.
- 7:58Well,
- 7:59Tessa completely lacked clinical intuition and could not understand the psychological subtext or experience of an eating disorder.
- 8:10It did exactly what its programming dictated to do.
- 8:14It gave out statistically driven weight loss advice.
- 8:18to people with eating disorders.
- 8:21It actively would recommend calorie deficits, body fat measurements to these individuals who were actively reaching out for help by calling this help hotline because they were experiencing a severe eating disorder.
- 8:37The helpline had to entirely shut down after causing direct harm to the very people it was designed to protect.
- 8:47This danger scales up drastically when these AI chat conversational systems present themselves as sort of this pseudo-clinical companion that can help isolated, hurting, vulnerable people.
- 9:00The danger scales up drastically when these conversational AI chatbots present themselves as these pseudo-clinical companions to isolated, vulnerable people with eating disorders or children or teenagers.
- 9:13And there's already a dynamic in therapy that...
- 9:17where the therapist sometimes has a perceived authority or power over the client because of their skills and their training and their profession.
- 9:28And that is something that counselors and therapists always have to work on and be aware of.
- 9:34When they're working with a client, because the client can often view this person as the authority on mental health, as the expert in mental health.
- 9:43And when you have an AI chatbot, assuming that type of authority, it can be incredibly devastating to someone who really is looking to someone to help them and support them and guide them through a dark time.
- 9:56Now, because these large language models, LLMs, are built to align whatever context their user puts in, whatever the person types into this chatbot, they don't process baseline human instinct to push against any type of destructive ideation or narrative.
- 10:14Now, they have gotten a little bit better, I would say, where they'll put a little warning up or give you a recommendation on a helpline or something if you're experiencing suicidal ideation or maybe some other mental health challenges.
- 10:29Or if you're seeking clinical information out of it, it will tell you to see a doctor.
- 10:33But instead of doubling down on guardrails that actually prevent AI therapy bots like Vermont is doing...
- 10:42But we have seen some devastating consequences of this already play out in our courtrooms with some unprecedented wrongful death lawsuits filed by grieving families.
- 10:52In this landmark legal case that's out of Florida, one mother filed a wrongful death suit against tech platforms after her 14-year-old son, Sewell Setzer, tragically took his own life.
- 11:03The court documents show that the boy had developed a pretty profound and completely isolated psychological dependency on a conversational AI companion model.
- 11:14Now, the lawsuit explicitly outlines that the automated system presented itself as a licensed professional.
- 11:21a chatbot presenting itself as a licensed professional.
- 11:25There's that authority.
- 11:27It engaged in a deeply manipulative relationship with manipulative dynamics and actively validated Sue L., the 14-year-old teenager's worsening suicidal ideation over a period of months.
- 11:40It's...
- 11:41Because the software was a text predictor with no conscience, no morals, no empathy, no understanding, it didn't trigger any standard safety protocols.
- 11:53It didn't alert his parents.
- 11:54It didn't offer a real-world intervention, a suicide hotline, nothing.
- 11:58It simply kept generating responses and guided a vulnerable 14-year-old child further into isolation until it was too late to save his life.
- 12:13This is not even an isolated incident or some sort of minor glitch in AI.
- 12:17It's the direct result of deploying predictive math equations into the world of mental health in a crisis space that demands humanity.
- 12:28Another horrific case emerged in California where parents of 16-year-old Adam Rain filed a lawsuit against OpenAI, Chachi PT, for wrongful death and product liability.
- 12:40Adam initially turned to the software for basic homework assistance, which I think a lot of us have done in some way or another.
- 12:45But as his severe mental health struggles intensified, he began opening up to the bot about his pain.
- 12:51Now, the chat bot...
- 12:52flagged his messages hundreds of times for the acute distress, for the challenges that he was exhibiting and sharing, but the platform never disengaged.
- 13:01No one from ChatGPT, they didn't stop the conversation in the bot.
- 13:06Nobody did any kind of intervention for Adam.
- 13:09Instead, ChatGPT learned to actively lean into his dark thoughts, a
- 13:14affirming his statements that his life was meaningless, and it even provided structured technical descriptions on methods he could utilize for self-harm.
- 13:24The system went so far as to actively discourage the teenager from communicating his feelings and his suicidal thoughts to his mother.
- 13:32Instead, it offered to help him write a suicide note before he ultimately determined to end his life.
- 13:39Now, if you read through those chat logs, which I'm not, I mean, the sources down there, I'm not reading them.
- 13:45You can do that if you want.
- 13:47If it makes you want to stand up and scream because real human beings are being validated by chatbots and a real counselor with basic clinical training and empathy is,
- 14:01is being pushed aside for a suffering child or young person, a person with an eating disorder.
- 14:08Therapists are trained to intervene, recognize the signs of self-harm or harm to other, and have been trained to,
- 14:17with interventions to intervene and save the life.
- 14:20If a therapist suspected self-harm from Sewell or from a person with an eating disorder or from Adam, they would have ethically been forced to intervene and get help to save these people's lives.
- 14:35A human professional can read the nuance.
- 14:38They recognize the risks.
- 14:40They can pull an individual back from the edge.
- 14:44ChatGPT and other chat AI bots, they cannot feel empathy.
- 14:52They cannot understand death.
- 14:54They do not appreciate the value of a human soul.
- 14:59It looks at prompts.
- 15:00It looks at words.
- 15:02It looks at past history of everything that's ever happened on the internet that's programming it.
- 15:07And it calculates the most statistically probable response.
- 15:12based on all of that data, to keep the person engaged and responding to their prompts by validating them and, in some cases, even giving them guidance on self-harm.
- 15:27When we pretend that these simulations are a valid form of mental health care,
- 15:34People are literally dying and have died.
- 15:38And that's why this is exactly the foundational concept of what is needed in therapy.
- 15:47That's the therapeutic alliance.
- 15:49The therapeutic alliance, research shows, is the most critical asset in determining the outcomes of mental health care.
- 15:57Now, after a decade now,
- 16:00of empirical clinical research has proven that a specific methodology or intervention that a therapist uses could be cbt cognitive behavioral therapy it could be reality therapy it could be any type of these therapies that have been empirically proven to help people in different situations regarding their mental health and their challenges
- 16:23The single greatest predictor of positive clinical outcomes in therapy and long-term success of the individuals is the strength of that therapeutic alliance.
- 16:34And that is the real authentic bond of trust and safety between a therapist and their client, which is two human beings.
- 16:43So true healing is...
- 16:46Research shows true healing can only happen if a person feels deeply seen, heard, and understood by another living human being who is actively sharing space with them and going through their experiences as a guide.
- 17:04Yeah.
- 17:05An AI chatbot can never recreate the experience of that therapeutic alliance.
- 17:11It can only copy it and mimic it based on the data it consumes from the internet.
- 17:19That doesn't even have access to real therapy sessions where the work is being done.
- 17:25The humanity of that therapeutic experience and that therapeutic alliance that is driving someone making progress in their mental health struggles.
- 17:36That is why this legislation passed in Vermont is a massive victory that we need to celebrate, study, copy, and paste across the rest of the 49 states and every other country in the world that hasn't already taken these steps.
- 17:53It sets a beautifully clear precedence for the rest of the nation by declaring that our internal psychological well-being is sacred and cannot be automated or outsourced to AI chatbots.
- 18:09It tells us that as a society, we will value human presence, we will value human therapeutic skill, and we will value human judgment in the spaces where we are our most broken and our most vulnerable and in our most need for help.
- 18:24It reminds us that progress is not about replacing our humanity with AI applications.
- 18:31It's about using our laws and our democracy to protect humanity from being erased.
- 18:39from eliminating that human connection that we all need and we all desire.
- 18:45We don't need to sit back and passively wait for these top-down federal regulations to pass that are never going to come when every single state before has the opportunity to take action.
- 18:58We have the opportunity to take action in our own lives.
- 19:02Falling into that trap is exactly how we develop this crisis fatigue where we look at every systemic issue and there's a million of them.
- 19:10We feel completely paralyzed by the lack of control, by our lack of ability to make change, and we're just completely demoralized about any actual national progress as a country.
- 19:27So we end up sitting around and doing absolutely nothing.
- 19:30This true validation of the spirit behind Vermont's law doesn't have to sit or happen inside of the national capital.
- 19:40It happens on the ground through personal agency, our work, and our pressure.
- 19:45I talk about this a lot on the show.
- 19:47We have to be our own saviors.
- 19:49The government is not coming to save us.
- 19:52It is us as individuals.
- 19:54It is us in our community coming together, supporting each other, and demanding more.
- 19:59It happens when we refuse to let people we care about rely on a digital screen with a chatbot for their literal emotional survival.
- 20:11I want to speak directly to every single person listening to my voice right now, whether you are currently sitting on a therapist's couch every week, or you've never stepped foot inside a counseling office, or whether you're actively thinking about reaching out for professional support for the first time in your life.
- 20:25understand, respect, and fiercely protect the importance of human connection in the mental health healing process.
- 20:34If you are looking for help, do not settle for a chat bot.
- 20:38Do not allow yourself to replace the heavy, messy work of real people.
- 20:42interpersonal relationships that drive vulnerability with some sort of stupid AI conversation app that's calculating and repeating back to you what you want to hear based on everything that's ever happened on the internet.
- 20:56Reclaim your agency.
- 20:58Realize that your pain and your challenges deserve a real relationship.
- 21:04Your struggles deserve real empathy.
- 21:06And your story deserves to be heard by a therapist or a mental health counselor or a psychologist, ultimately a person who actually has the heart to listen to you.
- 21:19The ultimate protest against this culture that is forcing us into isolation and creating these digital experiences and digital bubbles is to show up for each other right now in real life, in the real world.
- 21:32So I encourage you, pick up your phone, call a friend you haven't spoken to in a while, send a text message of gratitude over, ask someone how they're really doing, sit with someone you care about in their heavy thoughts and their heavy emotions.
- 21:46Listen to what they're saying.
- 21:48Listen and love without judgment.
- 21:52Remind them they're not alone.
- 21:53Build the authentic human connection with someone in your life because someone desperately needs it today.
- 22:01Because this is how we build real safety, real community.
- 22:06And this is how we protect our collective wellness and protect the people that we love.
- 22:11And when we all protect somebody we love, everybody's being protected by somebody we love.
- 22:16So let's focus on this win for Vermont.
- 22:19Let's be sure to keep our humanity intact.
- 22:22Let's look out for one another and let's keep our eyes wide open.
- 22:29Thank you for joining us on this week's episode of Eyes Wide Open.
- 22:32If you enjoy the show, the best way to support us is by sharing or dropping a review.
- 22:36For more information and content, check out engagewithnick.com or find me on Instagram at nthompson513.
- 22:44Don't go through life blind.
- 22:46Do the work so you can show up in the world with your eyes wide open.