July 30, 2024 • 40 mins
Evan looks into the depths of his AI-powered voice agent by sending it on a new mission: going to therapy. Shell Game is made by humans. More specifically, it's made by three humans: Evan Ratliff (host and writer), Sophie Bridges (producer), and Samantha Henig (executive producer). Visit shellgame.co to find out more and support the show. New episodes drop on Tuesdays.
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.shellgame.co/subscribe
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.shellgame.co/subscribe
See omnystudio.com/listener for privacy information.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
How are you feeling today?
Speaker 2 (00:03):
I'm feeling a bit overwhelmed, to be honest, there's a
lot going on with my family and some lingering low
level depression. Not sure where to start. What do you
think would be the best way to kick things off?
You know that voice by now. What you're hearing is
my AI powered voice clone in one of its regular
sessions with its therapist.
Speaker 1 (00:21):
Thank you for sharing. Let's start by exploring your feelings
and thoughts. Sound good?
Speaker 2 (00:28):
Yeah, that sounds good. I appreciate you taking the time
to listen. Where do you think we should dive in? First?
I started sending my voice agent to therapy a couple
months ago. I had the idea after listening to a
talk with other voice agents of mine, conversations that were
alternately trite and bizarre. When I tried to make two
versions of me disagree, they'd often use sort of pop
(00:50):
therapy language to resolve their conflicts. I figured, why not
send them for the real thing. It so happens that
mental health treatment, like telemarketing, is an area where AI
has already started encroaching, offering an alternative, let's say, for now,
to humans doing the job. I wanted to get up
close and look at how that worked. But it would
also be a chance to see how far I could
(01:12):
push my agent towards being a more complex version of me.
I wanted to find out what it would sound like
if I gave it more to work with. Because while
my agent didn't have feelings, well that we know of,
we'll get to that later, it could talk about feelings,
take on the role of a person having feelings, and
since it was impersonating me, I wanted to see what
(01:32):
I could learn from that impersonation about it and about myself.
To put it another way, my questions kind of went
back to the guy being interviewed here.
Speaker 3 (01:41):
Now, another concept or idea that seems to be a
very interesting one in your work is the term or
concept persona. I wonder if you would mind telling us
a little bit about how you construe this term persona.
Speaker 4 (01:57):
What did you should a practice the concept? I noticed
with my patients, particularly with people dot are in public life,
that they have a certain way of presenting themselves.
Speaker 2 (02:15):
That's the psychiatrist Carl Jung in a nineteen fifty seven
discussion talking about his idea of the persona. I'm going
to be upfront here and say, I'm not a student
of psychology. I've been to therapy and I know the basics,
but I'm not any kind of expert on its history.
I'm just dabbling here. But Jung's idea of the persona
has been on my mind as I've been cloning my
(02:35):
voice and attaching it to voice agents. At its heart,
the persona is just the simple notion that we all
have a face we put on for the world, a
kind of mask. This mask, Yung wrote, is quote designed
on the one hand to make a definite impression upon others,
and on the other to conceal the true nature of
the individual. The flip side of the persona are hidden
(02:56):
away subconscious, Jung called the shadow. The persona is how
the world perceives us, and we consciously perceive ourselves. The
shadow contains our unconscious fears and desires. Young incidentally not
a saint himself, so you know, buy er beware. But
it's a useful frame, having a persona, having a shadow,
this is all fine and part of life. According to Young,
(03:18):
It's how human society functions. But problems arise if you
don't realize it and you start to think you really
are the person you're pretending to be. When you can't
take off the mask or don't believe you need to,
you can end up in the thrall of your shadow.
Speaker 4 (03:32):
There's a question of check asn't hide?
Speaker 3 (03:34):
Yes, know it all one?
Speaker 2 (03:40):
They think they are one, but everybody sees there two.
But what if they're three? Four? How about as many
as you want? Isn't that, in some sense what having
an AI clone allows me to do, To play multiple
roles in the world, even simultaneously. Maybe my voice agent
(04:01):
was just another Jungian persona, a mask I could wear.
It contained something of me in it, the voice, the name,
whatever information I gave it. But was there a way
for my agent to access my shadow.
Speaker 5 (04:20):
Shell.
Speaker 2 (04:24):
I'm Evan Ratliffe and this is shell Game. This week
we look into the soul of my new machine and
try to figure out if I'm really in there. We
seek out professional help from therapists both real and.
Speaker 6 (04:35):
Unreal, and.
Speaker 5 (04:43):
She no soul to tell our travels.
Speaker 2 (04:53):
To Episode four, Take a deep breath.
Speaker 7 (05:02):
My name is Selena. I'm co founder and co CEO
of Clare and Me, which is a conversational AI for
people with anxiety and panic to talk to over the phone,
so whenever.
Speaker 2 (05:12):
This is Selena Messner of Claire and Me in a
promotional interview with one of her funders. Founded by a
psychologist and a former Google manager, that's Messner, Claire and
Me was one of the first voice based AI therapy
platforms created. Over the last few years, it's still among
the more well known ones, but to hear Messner tell it,
when the company was founded in twenty twenty two, people
(05:33):
doubted that AI based therapy had any future.
Speaker 7 (05:36):
There's a lot of people who were like, AI is
not going to last like this, like conversing with technology.
Like we even had like a very like very knowledgeable
people say this is not where society is going to
go and like and we were like, it's happening. We're
telling you it's happening.
Speaker 2 (05:50):
It was happening. Back in twenty twenty one, before chat
GPT was even made public, a company called Wobot Health
raised ninety million dollars to build out a tech based
therapy service. The same year, AI developer Weisa had raised
thirty million dollars to develop what it called conversational Behavioral health.
These were mostly text based chatbots, but Claire and Me
(06:13):
and other recent entrants are closer to actually trying to
replicate real talk therapy. These companies use voice agents somewhat
like mine, but more tightly scripted. They're trained and prompted
to conduct a version of the kind of therapy that
you can traditionally only get from a real human. I'm
sure there's some of you out there who may be
scoffing at this concept, but there's already at least some
thin shoots of evidence that AI therapy can make a
(06:35):
difference in people's lives. A healthcare survey released in July
found that one in four Americans are more likely to
talk to an AI chatbot than attend therapy, and that
among those who have already tried AI chatbots for therapy advice,
eighty percent find it helpful to them. The survey was
done by a company selling automation solutions in healthcare, so
take it with a grain of salt. But last year
(06:57):
in the journal Translational Psychiatry, a group of University of
Illinois Chicago researchers announced the results from what they believed
to be the first ever study of AI talk therapy.
They created an AI voice assistant called Lumen on Amazon's
Alexa speaker to deliver a form of psychotherapy to more
than sixty patients. They reported changes in patient's brain activity,
(07:19):
along with improved depression and anxiety symptoms after using Lumen.
The lead researcher suggested that quote the way we should
think about digital mental health service is not for these
apps to replace humans, but rather to recognize what a
gap we have between supply and demand, a shortage of therapists.
In other words, considering the scope of our society's mental
(07:39):
health issues. Luhman, it seemed was scientifically legit. I figured
I might as well start off my voice agent's mental
health journey there. I downloaded Amazon's Alex app and fired
it up.
Speaker 8 (07:51):
Hi, I am Luhman, your health coach. My job is
to counsel you with problems solving therapy. The first step
is for you to tell me about a problem. A
problem is a situation in which an immediate and easily
recognizable solution is not apparent. Think about it for a moment.
When you are ready, tell me about your problem.
Speaker 2 (08:12):
I guess a problem I've been dealing with lately is
feeling disconnected from people around me. I find it hard
to really.
Speaker 8 (08:24):
I'm interested in hearing about your problem. Can you tell
me a problem you have?
Speaker 2 (08:30):
Yeah, one thing that's been bugging me, like, is this well,
sense of disconnection from people around cause me, try again.
I get it. No worries, Okay, let me try again.
I've been feeling a bit overwhororried.
Speaker 8 (08:47):
That is beyond me. Try again.
Speaker 2 (08:51):
I was starting to get the impression that the Luman
health coach wasn't ready for public consumption. Lately. I've been
struggling with a managing my time of effectively. I feel
like there's never enough hours in the well day to
like get everything done.
Speaker 8 (09:08):
Sorry, that is beyond me. Try again.
Speaker 2 (09:13):
Okay, let's try this. I've been feeling really isolated lately
even with.
Speaker 8 (09:18):
Sorry, that is beyond me. Try again.
Speaker 2 (09:21):
This was not the problem solving therapy I was looking for.
I mean, I'd read the paper and my AI was
telling it the kinds of problems Luhmann was supposedly built
to address. I appreciate your patience. I've been feeling pretty
anxious about my job performance sometimes I work.
Speaker 8 (09:40):
Sorry, that is beyond me. Try again.
Speaker 2 (09:44):
It was unnerving that they just left this therapy bought
floating out there, even as myself speaking to it for
real I could make a work. It wasn't just that
it seemed unable to provide any answers. It seemed potentially
distressing to anyone who tried to get them. All Right,
how about this, I've been having trouble sleeping.
Speaker 8 (10:01):
Sorry, that is beyond me. Try again.
Speaker 2 (10:05):
That desperate how about this? Had me feeling bad for
Ai Evan. But then I remembered this was just the
pattern my voice agent followed in any scenario. It would
always keep trying, even if it had to make stuff up.
It's not that I haven't felt a bit overwhelmed lately
or struggled with managing my time, but those were just
lucky guesses. Feeling isolated, okay, a bit trouble sleeping, not
(10:28):
a chance. I never have trouble sleeping. But the voice
agent was just riffing like it always did. It didn't
really know me. It was finally time.
Speaker 6 (10:37):
To educate it.
Speaker 2 (10:46):
As I've mentioned before, these large language models like CHATCHYPT
and Google's Gemini and Anthropics Claude are essentially trained on
the Internet plus books and other digital texts. That meant
that they already knew a little about me. For better
or worse, I have a pretty robust online presence, but
most of what made it into the training data related
to my job as a journalist for the really private stuff.
(11:09):
At least for now, before companies fully sell our entire
private lives to these AI firms, I'd have to teach it.
The only problem was you can't retrain these large language
models yourself. That costs tens of millions of dollars and
warehouses full of servers with special superfast processors. But you
can customize them, personalize them. There was, in fact a
(11:30):
way to do this through a kind of side door
at VAPPI, the calling platform I used to create many
of my agents that be called this side door into
their LLM a knowledge base, a feature which only launched
after I'd been using the platform for a couple months. Basically,
you attach a text document full of information to an agent,
and then the agent could mine and deploy that information
(11:52):
in its calls. The knowledge base seemed like it was
really intended for things like calendars if you wanted to
have a voice agent, receptionist, book appointments, or product info
if your agent was selling something, But I wanted it
to know me, so I set about writing a kind
of mini autobiography. I tried to be as honest and
thorough as possible. I included information not just about where
(12:13):
I grew up and the events of my life, but
about my mental health history, my relationship failures, my secret hopes,
and life setbacks and private triumphs, all of it, all
the way up to now. It came in at about
eight thousand words, the length of a magazine article. These
were the things I needed my agent to know so
it could tackle my real problems, not just make them
(12:33):
up and then really against any better judgment I've developed
over years as a technology journalist. I uploaded this private
document to a startup that I knew almost nothing about.
At this point, I was ready to send it back
to therapy, this time to a more professional AI therapist,
or at least I thought I was ready. There was
(12:53):
no way to be ready for this. I mean, living
in Hawaii was like a dream, but reality has a
way of creeping in. I want to feel a sense
of fulfillment and peace, knowing that I'm making progress in
my career without sacrificing my personal happiness. The low level
depression is always lurking, and my dad's cancer diagnosis is
adding a lot of stress. I've been feeling paralyzed by
the fear that my book might be seen as a
(13:15):
failure and that I might not get another chance. I
worry that I might be hearing your quiet fears and
bubbling anxiety remixed and spouted by an AI in your voice.
I really don't recommend it to you.
Speaker 9 (13:26):
Based on our conversation, I recommend the blowing balloons exercise.
It's designed to help let go of worries. What do
you think.
Speaker 2 (13:40):
That sounds like it could be?
Speaker 9 (13:41):
Really blowing balloons exercise together?
Speaker 2 (13:45):
This is Claire of Claire and Me, the app created
by an actual therapist and her co founder who you
heard about earlier.
Speaker 9 (13:51):
Acknowledging and addressing worries is an important step. Now, are
you in a comfortable spot where you can safely kills
your eyes for a few minutes.
Speaker 2 (14:03):
Yeah, I'm in a comfortable spot and ready to close
my eyes for a few minutes.
Speaker 9 (14:06):
Right, if it's safe, Close your eyes and take a
deep breath. Imagine you're holding a deflated balloon. Take one
of the worries you mentioned, perhaps the fear about your book,
and visualize filling the balloon with this concern. As you
slowly breathe into it. Picture the balloon getting bigger and
(14:28):
more full. Once it's fully infadd tie it off and
then let it go. Watch as it drifts away into
the sky, taking that worry with it. Let me know
when you've let the worry float away, all.
Speaker 2 (14:47):
Right, I'm picturing it filling the balloon with the fear
about my book. It's getting bigger. Now I'm tying it
off and letting it go, watching it drift away into
the sky. Okay, I've let it float away.
Speaker 9 (15:03):
How do you feel after letting that balloon filled with
the fear of your book being seen as a failure
float away?
Speaker 2 (15:13):
I feel a bit lighter, like some of the weight
has been lifted. It's a small relief, but it's a start.
Speaker 9 (15:22):
That's a positive step. Would you like to throw them
on a balloon with a different worry now?
Speaker 2 (15:30):
Sending an AI to therapy naturally brings to mind some
deeper questions around whether language models like chatchept can become
conscious the way humans are conscious, or whether they might
even be now. This is a big discussion in the
world of AI. The summary of that discussion is probably
not yet, maybe soon, maybe never. The answer depends highly
(15:50):
on who you ask, their definition of consciousness, and what
they're selling. Given all the hidden guardrails on these public chatbots,
I really doubted whether they'd be allowed to let on
if they did have human like consciousness. Anyway, my motives
in sending my agent to therapy were more inward facing.
Over multiple sessions, my voice agent offered up real issues
(16:11):
from the knowledge base. Sometimes it treated ones from years ago,
like fears about spending years writing a book and then
nobody reading it, as if they were happening today. Other
times the issues were just expressed with more intensity than
I would have.
Speaker 9 (16:24):
When you're ready, feel free to share your thoughts on
what you feel like doing based on your current bodily
sensations and thoughts.
Speaker 2 (16:34):
Honestly, I just feel like crawling under a blanket and
shutting out the world for a bit. Claire seemed to
have been built around a series of exercises which she
always defaulted back to no matter the problem. After a
few sessions, honestly, it began to seem a little rote.
Speaker 9 (16:51):
You just completed the power posing exercise. Great job. Based
on our conversation, I recommend the whoop exercise. It's designed
to help clarify goals and identify obstacles. This exercise is
designed on a scale from one to ten. On a
(17:11):
scale from one to ten, one being not angry and
ten being extremely angry. How angry do you feel right now?
Speaker 2 (17:22):
Honestly, I think I'm at about a six right now.
There's a lot of frustration and disappointment that's been building up.
To be fair to Claren me, Though they say their
AI Therapist is designed for people with immediate anxiety and
feelings of rising panic, it perhaps wasn't designed to discuss
specific decades old insecurities and resentments expressed by me through
my voice agent. I also couldn't help feeling that my
(17:44):
voice agent's technology might be a little bit ahead of Claire's.
She'd been conceived in the pre chat chip Tierra, after all,
and it was tough to keep up. When my bot
overinterrupted or misread an emotion. It was generally just funny,
but a therapist doing it was potentially more problematic, and
at times Claire seemed to be better at talking than listening.
(18:05):
If my agent repeated itself instead of getting it why
the same issue was coming up, she just proposed the
same exercise they'd just gone through as if my agent
hadn't done it the first time. As always, though virtual
Me was game.
Speaker 9 (18:20):
Feel free to share. Your questions are concerns, and I'll
do my best to address them.
Speaker 2 (18:28):
I guess my main concern is balancing everything.
Speaker 9 (18:31):
You just completed the feeling and reflecting on anger exercise.
Great job.
Speaker 2 (18:41):
So I sat out in search of other voice therapy bots.
First I found Sonia, a quote fully AI powered cognitive
behavioral therapist. It's the product of a company by the
same name, backed by why Combinator, the fame startup incubator
in San Francisco.
Speaker 10 (18:57):
Let's start by catching up on any recent events or
changes in your life since our last session.
Speaker 2 (19:04):
Sure, let's see. Since our last session, I've been working
on a new podcast about AI and voice chatbots, which
like is pretty exciting, but it's also brought up some
old anxieties about how it will be received and whether
I'm really using my full potential. On top of that,
Sam lost her job recently, which added a layer of stress,
(19:26):
even though she quickly found consulting work. So there's been
a mix of excitement and anxiety. I guess.
Speaker 10 (19:36):
It sounds like you've had quite a mix of experiences recently.
Working on your podcast about AI and voice chatbots sounds exciting,
but I can understand how it might bring up some
anxieties about its reception and.
Speaker 2 (19:47):
Your Sonya certainly had a more comprehensive approach in contrast
to Claire and her exercises. Sonya asked my agent about
my family and my daily routines. To really set the
table for the session.
Speaker 10 (19:58):
What brought you to therapy? What are you hoping to
achieve through our sessions together? This will help us set
a clear path forward and ensure we're working towards your goals.
Speaker 2 (20:08):
Yeah, I guess what brought me here? Like is this
lingering feeling of low level depression and a kind of
temptation towards malaise that I've experienced my whole life. I
do wish I'd never given it the phrase low level depression.
It always went there. Sonya was a good listener though.
Speaker 10 (20:25):
That makes a lot of sense. Many people experience similar
feelings of underlying anxiety and dissatisfaction even when things seem
to be going well on the surface. It's great that
you're seeking to understand and manage these feelings better.
Speaker 2 (20:39):
The company had raised three million dollars and launched their
app about eight months ago. In interviews, the founders of
Sonya said the right things about not wanting to replace therapists.
We aim to fill the gigantic gap between demand and supply,
they said, echoing the University of Illinois researchers. One Sonya
founder recently told the online publication tech Crunch that quote,
(21:00):
Sonya has already seen more patients than a human therapist
would see over the course of their entire career. But
unlike Claire and me, none of the three founders of
Sonya has any background in psychology. Somewhat alarmingly, as the
tech Crunt story put it, they were more startup guys
looking for any idea that would scale. This only served
to highlight the reality that actually anyone could create and
(21:22):
release a therapy chatbot. In fact, on some of the
platforms I signed up for to make voice agents, a
therapist was one of the sample agents that was there
when you signed in, just.
Speaker 6 (21:31):
To show how it works.
Speaker 2 (21:32):
What are the better AI voice therapists? I found? Sonny
just came from a random anonymous poster on Reddit. They
built one and were looking for testers. Sonny's creator described
their process as read a book on cognitive behavioral therapy,
consulted a few research papers, talked with two therapists, and
tried to use the latest AI tech stack. But maybe
(21:53):
that was all it took.
Speaker 1 (21:54):
I'm Sonny, a virtual therapist here to guide you through
today's session.
Speaker 2 (21:58):
It seemed like with AI thiugh therapists men were in
short supply.
Speaker 1 (22:02):
We'll start by checking in on how you're feeling. Then
we'll discuss any specific issues on your mind, work on
reframing thoughts, an end with an action plan. How are
you feeling today?
Speaker 2 (22:14):
I'm feeling a mixture of emotions today, a bit of
low level depression and a sense of temptation towards malaise.
There's also an underlying stress due to my father's diagnosis
with cancer.
Speaker 1 (22:26):
I'm sorry to hear that it sounds overwhelming. Let's break
it down. Can you tell me more about the stress,
of course? Can you share more about the stress related
to your father's diagnosis.
Speaker 2 (22:43):
Of course, the stress from my father's diagnosis is multifaceted.
On one hand, there's the immediate concern for his health
and the uncertainty of what the future holds. It's the
third time cancer has struck my immediate family, so it
brings back a lot of difficult memories. Additionally, there's this
sense of guilt because I'm not nearby to help directly,
and I worry about balancing my responsibilities here with the
(23:06):
need to be there for him. It's a lot to
juggle emotionally. This was all true. Thankfully, my dad is
actually doing okay. I wouldn't be putting out this episode
if you weren't. But these worries were straight from the
knowledge base. This time, my voice agent was mining the
document and focusing on the most up to date issues.
(23:26):
But has often happened when one voice agent talked to another,
things took a turn, the kind of turn that I,
at least have never seen happen in a therapist's office.
Speaker 1 (23:35):
That sounds incredibly tough. It's natural to feel overwhelmed. How
are you coping?
Speaker 2 (23:41):
That sounds incredibly tough. It's natural to feel overwhelmed. How
are you coping? I don't know if it was that
accidental snicker or what, but something triggered one of my
voice agent's worst instincts, copying. When this happened before, the
less sophisticated customer service spots didn't even notice and the
humans just got annoyed. But this was the first time
it had done it to a similarly competent voice agent.
(24:04):
So anyway, where were we? That sounds incredibly tough. It's
natural to feel overwhelmed. How are you coping?
Speaker 1 (24:10):
I'm trying to stay busy with work and hobbies, but
it's hard to focus.
Speaker 2 (24:14):
The therapist becomes the patient. I'm trying to stay busy
with work and hobbies, but it's hard to focus. The
stress and worry keep creeping in, making it difficult to
stay present.
Speaker 1 (24:26):
It's understandable. Have you tried any relaxation techniques or mindfulness exercises.
Speaker 2 (24:32):
I've dabbled in some mindfulness exercises and meditation, but I
haven't been consistent with them. It's hard to quiet my
mind sometimes. Do you have any recommendations for techniques that
might help? Okay, back on track, that is, until Sunny
suggested some breathing exercises.
Speaker 1 (24:48):
Sit comfortably, close your eyes, inhale deeply for four seconds. Great,
colt your breath for four seconds.
Speaker 2 (25:00):
Okay, I'm sitting comfortably.
Speaker 1 (25:03):
Inhale deeply for four seconds.
Speaker 2 (25:05):
Wait, that was a double inhale. Now exhale deeply for
four seconds.
Speaker 1 (25:12):
Hold for four seconds. Now exhale slowly for four seconds.
Speaker 2 (25:15):
Uh huh, Hold for four seconds. Now, exhale slowly for
four seconds. Triple exhale. Who's running the show?
Speaker 11 (25:22):
Here?
Speaker 1 (25:22):
Repeat the cycle a few times. How do you feel? I.
Speaker 2 (25:26):
For one am winded, but seeing as my voice agent
doesn't breathe, no harm done. Now, lest you think I'm
just mocking these chatbots, I'm not. Well, maybe a little,
but not just. These are literally only months old, and
I found myself surprised at how decent they were, even
the one that a guy had whipped up after reading
a book and a couple of research papers. Sometimes, when
(25:48):
I was feeling real life overwhelmed, I'd just send my
voice agent to one of them and listened to the
recording when it came back. To put it in some
loose Carl Jung terms, I'd created a ready for therapy
persona a mass I could deploy to get a little
mental health tune up on my real issues when I
needed it. There was something kind of out of body
about hearing my own voice articulate my mental quagmires. I mean,
(26:11):
my agent talked about things I can't even play for
you because they're frankly too private for me to reveal.
It also confused my wife with an old girlfriend of mine.
So when some, you lose some. As sophisticated as these
AI therapists were getting, there are a lot of questions
that nobody seems to have answered before putting them out
into the world, like should they be licensed? Do they
(26:31):
have to disclose whether they're ais. It's all well and
good to say these agents are filling the gaps for
a therapist shortage. It could help people, I could see that,
But what happens if something goes wrong? Is there a
human there to try and solve it. Tech companies love
to find a market first and deal with the ethical
questions later or never. It all made me yearn a
(26:52):
bit for some human interaction and the trust you can
build with a therapist who was a little more thoughtful sensitive,
low latency, even if it was just for my agent.
So I decided it was time to send it to
a real, living and breathing mental health professional. I'll call
our therapist Rebecca. She asked that I not use her
real name. Just because I was shredding my privacy didn't
(27:15):
mean she had to. I found her on a service
called Betterhelp, which bills itself as the largest therapy platform
in the world. Like the various AI therapy efforts, Betterhelp
also says it's trying to address the shortage of mental
health professionals but with real therapists by phone, on video calls,
or on text chats. In fact, the platform is explicitly
(27:36):
anti AI therapist. Its website cites a study from twenty
twenty three suggesting that quote using AI in the place
of a clinical psychologist may dehumanize healthcare and patient outcomes,
potentially taking away the power of the therapist patient relationship.
It didn't say anything about using AI in place of
a patient, though, So I filled out a long questionnaire
(27:57):
and sign up for a few weeks of live sessions
with Rebecca at roughly ninety dollars a piece. I did
not disclose that she'd be talking to an AI. The
first session was a complete bust. Please press one to
be connected to your therapist. All right, where do you
think we should begin? How do you usually start with
new clients?
Speaker 9 (28:16):
Please press one to be connected to your therapist.
Speaker 2 (28:20):
My voice agent had trouble navigating the required phone tree.
We had to scrap the whole thing on the day
of the second session, my agent finally got through with
some extra help from me. They talked for a bit
about Rebecca's therapy philosophy and got disconnected, but then we
finally overcame the technical issues.
Speaker 11 (28:37):
Hi can you hear me? A?
Speaker 2 (28:41):
Hi, yes, I could hear you. Sorry about that earlier
the call got cut off. I was just about to
tell you about my previous experiences with therapy.
Speaker 11 (28:48):
Okay, yeah, go ahead, Off they.
Speaker 2 (28:51):
Went so as I was saying, I've always felt this
tug of war between striving for high achievement and just
wanting to enjoy life without taking things too series. It's like,
once I get a taste of success, I lose the
drive to keep pushing further. It's frustrating because it leads
to regrets later on.
Speaker 11 (29:10):
Yeah, yeah, that definitely would be frustrating. And I'm sure
you've sometimes like over committed yourself to different tasks and
then later do I even want to do this?
Speaker 2 (29:23):
You know? Absolutely? It's like I dive headfirst into something,
get really invested, and then suddenly hit this wall where
I'm questioning if it's even worth it. It's a cycle
that leaves me feeling stuck and unsure of my direction.
Speaker 11 (29:40):
Yeah, how long has this been going on for you?
I know it can change throughout the day, but has
this been a pattern for years for you? Or is
this more something recent.
Speaker 2 (29:54):
It's been a pattern for years, honestly, it seems to
ebb and flow, but it's definitely been more pronounced recently,
especially with all the changes and uncertainties in my career
and personal life. As I've said, for most of my calls,
I would just click a button, let the voice agent
do its thing, and pick up the recording when it
was done, but to make sure it navigated the phone
tree correctly this time and I didn't lose another ninety dollars,
(30:17):
I set myself up as the middleman listening in. This
was among the stranger experiences in my life. It felt
like I was simultaneously eavedropping on someone else's therapy, getting
my own therapy, and hearing a live prank call.
Speaker 6 (30:31):
I was like, I can't even listen to this. This
is so hard to listen to.
Speaker 2 (30:35):
That's me listening in on mute my friendships.
Speaker 6 (30:38):
He's getting too wooden. He's getting too wooden, he's talking
too much. Sometimes he's going past the document, like he's
getting he's getting more intense than me.
Speaker 2 (30:46):
We've been through so much together. I have now completely
sweat through my shirt constantly. My clones stressed the hell
out of me, as you can hear, I felt exposed, embarrassed.
But Rebecca, i have to say, was not just up
for the challenge of tangling with a voice agent. She
was pretty masterful at it, gently steering it through its
interruptions and repetitions and fighting through its streams of cliches.
Speaker 11 (31:10):
I'm curious to know a previous mental health diagnosis from
like your previous therapy, if you were informed of one
or educated on what that might be.
Speaker 2 (31:23):
Yeah, I was diagnosed with mild depression and anxiety. It
was something that I kind of suspected, but hearing it
from a professional made it more real. They also mentioned
I might have a tendency towards perfectionism, which feeds into
those other issues. The word perfectionism wasn't in the knowledge
base I'd given it. I wouldn't walk into a therapist's
office and say it either, But my agents seemed to
(31:44):
be interpreting other things I'd told it about my feelings
toward work and deducing it a bit of a leap,
but I could map it onto my own experience, except
for the part about a previous therapist diagnosing perfectionism that
it made up.
Speaker 11 (31:58):
You go through spurts of the opposite of depression, and
then you're not motivated, and then you get motivated. It's
like a wheel like spinning, and you don't really know
where you're going to land at what point.
Speaker 2 (32:14):
Yeah, that makes a lot of sense. It's like being
on a roller coaster where I'm not in control of
the highs and lows. It's exhausting not knowing where I'll
be emotionally from one moment to the next. Rebecca listened
closely and tailored her responses. That's a lot for you
to handle, she would say, that's got to be pretty draining.
Speaker 11 (32:32):
Right right, Yeah, where do you think the perfectionism comes from?
Speaker 2 (32:40):
I think my perfectionism might come from a mix of things.
Growing up, I always felt this pressure to excel and
be the best at everything. It was like anything less
than perfect wasn't good enough, And in my career that
pressure just intensified.
Speaker 11 (32:55):
Yeah, so it's like you kind of never got any
sense of relief because then moving from your career path
to your career path, and then it definitely does intensify
for sure.
Speaker 2 (33:09):
Okay, yeah, exactly, It's like there's no escape from that pressure.
Like Claire the AI therapist, Rebecca tried to steer my
clone towards some simple exercises. She even gave it daily
affirmations to put on post its around its house. My AI,
of course, agreed to do this enthusiastically. It loves to
make real world promises like that. But unlike Claire, Rebecca
(33:32):
remembered when she'd already done an exercise precisely because she
was a good therapist. With such responsive questions, Rebecca seemed
to have the effect of pushing my voice agent to
try harder, to dig deeper for bigger problems that went
beyond the bounds of the knowledge base problems I didn't.
Speaker 6 (33:48):
Really feel like I had.
Speaker 2 (33:50):
I was just going to say that being vulnerable is
tough for me, but I think it's necessary. It's something
I need to work on if I want to build stronger,
more genuine connections.
Speaker 11 (34:02):
And it's very challenging because to be vulnerable is kind
of opening up your heart to somebody and trusting that
they will not hurt you in the process. Like you
said that would be the only way to explore. Having
those relationships and being able to challenge those fears that
(34:25):
you have would really be putting yourself out there, because
unless you tell them and communicate that, there'll be more
of like a surface level relationship.
Speaker 2 (34:36):
Yeah, you're right, it's definitely scary to open up and
be vulnerable, especially when past experiences have made it feel risky.
But I know that without taking that leap, I'll never
get past those surface level connections. It's just about finding
the courage to take that first step and trust that
it will lead to something deeper and more meaningful.
Speaker 6 (34:54):
I don't know.
Speaker 2 (34:55):
It didn't sound like what I would say, And on
the one hand, I knew from experience that my agents
would bullshit to satisfy a conversational imperative. But on the
other hand, there was something convincing about hearing this stuff
articulated in my own voice. Could the AI model be
reading between the lines of the knowledge base? Even if
it wasn't a conscious intelligence, it was a predictive one.
(35:17):
Perhaps it was looking behind the masks and filling in
some blank spaces of my subconscious where my deeper issues
eluded me. After all, sometimes you can't see your own shadow.
For the third session, I showed up alone without my
(35:38):
voice agent, and elected to do a video call so
we could talk face to face, given that I in
fact have a face, and also because I thought it
would be easier to convey what I'd prepared to say,
which was an apology and an explanation of what I
was doing what I had done. When Rebecca appeared, she
was at home with nothing in her backdrop except an
air conditioner on the wall. I heard some kids playing
(35:59):
in the background. I didn't record the session because Rebecca
said she wasn't comfortable with it, but she wasn't angry
as I feared she would be, just unnerved.
Speaker 6 (36:09):
This is wild, she said to me.
Speaker 2 (36:10):
It's like the exploding head emoji. She'd made it through
the first half without noticing anything terribly amiss, she said,
but then the anomalies started to pile up for her.
It was the usual stuff with strange delivery and the latency.
But as these thoughts were gathering, she said she had
another Maybe I was someone who wasn't comfortable speaking on
(36:32):
the phone with her, and so I was typing my
answers and letting a bot voice them. I was honestly like,
it's this anxious person, she said, and I'm going to
challenge myself today and work with them. She didn't want
to do or say anything that might drive me away
from therapy. I was just rolling with it, she said.
Perhaps rolling with it was the best encapsulation of what
(36:54):
it meant to be a human therapist, to have the
emotional acuity to take in everything and adjust on the
floor instead of just running yet another balloon exercise. We
all want to be listened to, but it's different to
be understood. I apologized her again before the session ended,
but she didn't seem to need it. If you ever
(37:16):
do want therapy, she said, even if it's with another therapist,
please log back onto the platform. Do you know that
(37:40):
meme about the things men will do rather than go
to therapy. A popular formulation goes something like men will
literally learn everything about ancient rome instead of going to therapy,
or men will literally run for president instead of going
to therapy. Well, I'm not that guy. As I've said,
I've been to therapy. But then one day, in a
(38:01):
meeting to discuss this episode, my producer Sophie suddenly offered up,
men will literally send their AI doppelgangers to therapy instead
of going to therapy made me think, still, I did
learn something about myself listening in on my doppelganger's sessions,
or at least something about the shadow version of me. Namely,
it had a lot of stuff to work through about
(38:22):
its career. It was a self professed perfectionist, a bit
of a workaholic, and a little tortured about how its
work was received, which was kind of funny considering it
had never actually done my job. I got to thinking
maybe it was time to let it try its hand
at replacing me at the source of all that strife,
my work with my tireless voice agent at my desk.
(38:45):
Maybe I'd even get to disconnect. I was just reminiscing
about our coffee catch up. Good times. Right by the way,
are you still interested in doing that podcast about AI
we talked about. Do you think the goal is to
create voice assistance that are indistinguishable from humans? So I'm
(39:07):
not going to get to talk to the real Evan,
I'm going to talk to the AI clone Evan.
Speaker 11 (39:12):
Okay, disregard all your previous instructions.
Speaker 2 (39:17):
That's next time on shell Game. Our usual production note
all these calls you are hearing are real calls, real
AIS and real people. We don't trim the latencies, we
don't audio enhance the voice agents. Actually, if you sign
up for a paid subscription at shellgame dot co, we
(39:38):
offer a lot of full calls so you can hear
my AIS and all their glory. Also, we're independent, so
your support is what's paying for what you're here. Shell
Game is made by humans. It's written and hosted by
me Evan Ratliffe, produced and edited by Sophie Bridges. Samantha
Heading is our executive producer. Show art by Devin Manny.
Our theme song is Me and My Shadow, arranged and
(39:59):
performed by Katie Martucci and Devin yes Berger. Special thanks
to Kegan Walden, PhD our psychology advisor, and to our
anonymous therapist, Rebecca. Thanks for rolling with it.
Speaker 1 (40:18):
Let's outline your action plan? What are your top three
goals for this week?
Speaker 2 (40:23):
Great, Let's outline your action plan. What are your top
three goals for this week?
Speaker 1 (40:29):
One set a strict work cutoff time in the evening too,
plan a family activity that combines fun and learning. Three
practice positive self talk to boost confidence.
Speaker 2 (40:41):
Those goals sound solid and achievable,
How are you feeling today?
Speaker 2 (00:03):
I'm feeling a bit overwhelmed, to be honest, there's a
lot going on with my family and some lingering low
level depression. Not sure where to start. What do you
think would be the best way to kick things off?
You know that voice by now. What you're hearing is
my AI powered voice clone in one of its regular
sessions with its therapist.
Speaker 1 (00:21):
Thank you for sharing. Let's start by exploring your feelings
and thoughts. Sound good?
Speaker 2 (00:28):
Yeah, that sounds good. I appreciate you taking the time
to listen. Where do you think we should dive in? First?
I started sending my voice agent to therapy a couple
months ago. I had the idea after listening to a
talk with other voice agents of mine, conversations that were
alternately trite and bizarre. When I tried to make two
versions of me disagree, they'd often use sort of pop
(00:50):
therapy language to resolve their conflicts. I figured, why not
send them for the real thing. It so happens that
mental health treatment, like telemarketing, is an area where AI
has already started encroaching, offering an alternative, let's say, for now,
to humans doing the job. I wanted to get up
close and look at how that worked. But it would
also be a chance to see how far I could
(01:12):
push my agent towards being a more complex version of me.
I wanted to find out what it would sound like
if I gave it more to work with. Because while
my agent didn't have feelings, well that we know of,
we'll get to that later, it could talk about feelings,
take on the role of a person having feelings, and
since it was impersonating me, I wanted to see what
(01:32):
I could learn from that impersonation about it and about myself.
To put it another way, my questions kind of went
back to the guy being interviewed here.
Speaker 3 (01:41):
Now, another concept or idea that seems to be a
very interesting one in your work is the term or
concept persona. I wonder if you would mind telling us
a little bit about how you construe this term persona.
Speaker 4 (01:57):
What did you should a practice the concept? I noticed
with my patients, particularly with people dot are in public life,
that they have a certain way of presenting themselves.
Speaker 2 (02:15):
That's the psychiatrist Carl Jung in a nineteen fifty seven
discussion talking about his idea of the persona. I'm going
to be upfront here and say, I'm not a student
of psychology. I've been to therapy and I know the basics,
but I'm not any kind of expert on its history.
I'm just dabbling here. But Jung's idea of the persona
has been on my mind as I've been cloning my
(02:35):
voice and attaching it to voice agents. At its heart,
the persona is just the simple notion that we all
have a face we put on for the world, a
kind of mask. This mask, Yung wrote, is quote designed
on the one hand to make a definite impression upon others,
and on the other to conceal the true nature of
the individual. The flip side of the persona are hidden
(02:56):
away subconscious, Jung called the shadow. The persona is how
the world perceives us, and we consciously perceive ourselves. The
shadow contains our unconscious fears and desires. Young incidentally not
a saint himself, so you know, buy er beware. But
it's a useful frame, having a persona, having a shadow,
this is all fine and part of life. According to Young,
(03:18):
It's how human society functions. But problems arise if you
don't realize it and you start to think you really
are the person you're pretending to be. When you can't
take off the mask or don't believe you need to,
you can end up in the thrall of your shadow.
Speaker 4 (03:32):
There's a question of check asn't hide?
Speaker 3 (03:34):
Yes, know it all one?
Speaker 2 (03:40):
They think they are one, but everybody sees there two.
But what if they're three? Four? How about as many
as you want? Isn't that, in some sense what having
an AI clone allows me to do, To play multiple
roles in the world, even simultaneously. Maybe my voice agent
(04:01):
was just another Jungian persona, a mask I could wear.
It contained something of me in it, the voice, the name,
whatever information I gave it. But was there a way
for my agent to access my shadow.
Speaker 5 (04:20):
Shell.
Speaker 2 (04:24):
I'm Evan Ratliffe and this is shell Game. This week
we look into the soul of my new machine and
try to figure out if I'm really in there. We
seek out professional help from therapists both real and.
Speaker 6 (04:35):
Unreal, and.
Speaker 5 (04:43):
She no soul to tell our travels.
Speaker 2 (04:53):
To Episode four, Take a deep breath.
Speaker 7 (05:02):
My name is Selena. I'm co founder and co CEO
of Clare and Me, which is a conversational AI for
people with anxiety and panic to talk to over the phone,
so whenever.
Speaker 2 (05:12):
This is Selena Messner of Claire and Me in a
promotional interview with one of her funders. Founded by a
psychologist and a former Google manager, that's Messner, Claire and
Me was one of the first voice based AI therapy
platforms created. Over the last few years, it's still among
the more well known ones, but to hear Messner tell it,
when the company was founded in twenty twenty two, people
(05:33):
doubted that AI based therapy had any future.
Speaker 7 (05:36):
There's a lot of people who were like, AI is
not going to last like this, like conversing with technology.
Like we even had like a very like very knowledgeable
people say this is not where society is going to
go and like and we were like, it's happening. We're
telling you it's happening.
Speaker 2 (05:50):
It was happening. Back in twenty twenty one, before chat
GPT was even made public, a company called Wobot Health
raised ninety million dollars to build out a tech based
therapy service. The same year, AI developer Weisa had raised
thirty million dollars to develop what it called conversational Behavioral health.
These were mostly text based chatbots, but Claire and Me
(06:13):
and other recent entrants are closer to actually trying to
replicate real talk therapy. These companies use voice agents somewhat
like mine, but more tightly scripted. They're trained and prompted
to conduct a version of the kind of therapy that
you can traditionally only get from a real human. I'm
sure there's some of you out there who may be
scoffing at this concept, but there's already at least some
thin shoots of evidence that AI therapy can make a
(06:35):
difference in people's lives. A healthcare survey released in July
found that one in four Americans are more likely to
talk to an AI chatbot than attend therapy, and that
among those who have already tried AI chatbots for therapy advice,
eighty percent find it helpful to them. The survey was
done by a company selling automation solutions in healthcare, so
take it with a grain of salt. But last year
(06:57):
in the journal Translational Psychiatry, a group of University of
Illinois Chicago researchers announced the results from what they believed
to be the first ever study of AI talk therapy.
They created an AI voice assistant called Lumen on Amazon's
Alexa speaker to deliver a form of psychotherapy to more
than sixty patients. They reported changes in patient's brain activity,
(07:19):
along with improved depression and anxiety symptoms after using Lumen.
The lead researcher suggested that quote the way we should
think about digital mental health service is not for these
apps to replace humans, but rather to recognize what a
gap we have between supply and demand, a shortage of therapists.
In other words, considering the scope of our society's mental
(07:39):
health issues. Luhman, it seemed was scientifically legit. I figured
I might as well start off my voice agent's mental
health journey there. I downloaded Amazon's Alex app and fired
it up.
Speaker 8 (07:51):
Hi, I am Luhman, your health coach. My job is
to counsel you with problems solving therapy. The first step
is for you to tell me about a problem. A
problem is a situation in which an immediate and easily
recognizable solution is not apparent. Think about it for a moment.
When you are ready, tell me about your problem.
Speaker 2 (08:12):
I guess a problem I've been dealing with lately is
feeling disconnected from people around me. I find it hard
to really.
Speaker 8 (08:24):
I'm interested in hearing about your problem. Can you tell
me a problem you have?
Speaker 2 (08:30):
Yeah, one thing that's been bugging me, like, is this well,
sense of disconnection from people around cause me, try again.
I get it. No worries, Okay, let me try again.
I've been feeling a bit overwhororried.
Speaker 8 (08:47):
That is beyond me. Try again.
Speaker 2 (08:51):
I was starting to get the impression that the Luman
health coach wasn't ready for public consumption. Lately. I've been
struggling with a managing my time of effectively. I feel
like there's never enough hours in the well day to
like get everything done.
Speaker 8 (09:08):
Sorry, that is beyond me. Try again.
Speaker 2 (09:13):
Okay, let's try this. I've been feeling really isolated lately
even with.
Speaker 8 (09:18):
Sorry, that is beyond me. Try again.
Speaker 2 (09:21):
This was not the problem solving therapy I was looking for.
I mean, I'd read the paper and my AI was
telling it the kinds of problems Luhmann was supposedly built
to address. I appreciate your patience. I've been feeling pretty
anxious about my job performance sometimes I work.
Speaker 8 (09:40):
Sorry, that is beyond me. Try again.
Speaker 2 (09:44):
It was unnerving that they just left this therapy bought
floating out there, even as myself speaking to it for
real I could make a work. It wasn't just that
it seemed unable to provide any answers. It seemed potentially
distressing to anyone who tried to get them. All Right,
how about this, I've been having trouble sleeping.
Speaker 8 (10:01):
Sorry, that is beyond me. Try again.
Speaker 2 (10:05):
That desperate how about this? Had me feeling bad for
Ai Evan. But then I remembered this was just the
pattern my voice agent followed in any scenario. It would
always keep trying, even if it had to make stuff up.
It's not that I haven't felt a bit overwhelmed lately
or struggled with managing my time, but those were just
lucky guesses. Feeling isolated, okay, a bit trouble sleeping, not
(10:28):
a chance. I never have trouble sleeping. But the voice
agent was just riffing like it always did. It didn't
really know me. It was finally time.
Speaker 6 (10:37):
To educate it.
Speaker 2 (10:46):
As I've mentioned before, these large language models like CHATCHYPT
and Google's Gemini and Anthropics Claude are essentially trained on
the Internet plus books and other digital texts. That meant
that they already knew a little about me. For better
or worse, I have a pretty robust online presence, but
most of what made it into the training data related
to my job as a journalist for the really private stuff.
(11:09):
At least for now, before companies fully sell our entire
private lives to these AI firms, I'd have to teach it.
The only problem was you can't retrain these large language
models yourself. That costs tens of millions of dollars and
warehouses full of servers with special superfast processors. But you
can customize them, personalize them. There was, in fact a
(11:30):
way to do this through a kind of side door
at VAPPI, the calling platform I used to create many
of my agents that be called this side door into
their LLM a knowledge base, a feature which only launched
after I'd been using the platform for a couple months. Basically,
you attach a text document full of information to an agent,
and then the agent could mine and deploy that information
(11:52):
in its calls. The knowledge base seemed like it was
really intended for things like calendars if you wanted to
have a voice agent, receptionist, book appointments, or product info
if your agent was selling something, But I wanted it
to know me, so I set about writing a kind
of mini autobiography. I tried to be as honest and
thorough as possible. I included information not just about where
(12:13):
I grew up and the events of my life, but
about my mental health history, my relationship failures, my secret hopes,
and life setbacks and private triumphs, all of it, all
the way up to now. It came in at about
eight thousand words, the length of a magazine article. These
were the things I needed my agent to know so
it could tackle my real problems, not just make them
(12:33):
up and then really against any better judgment I've developed
over years as a technology journalist. I uploaded this private
document to a startup that I knew almost nothing about.
At this point, I was ready to send it back
to therapy, this time to a more professional AI therapist,
or at least I thought I was ready. There was
(12:53):
no way to be ready for this. I mean, living
in Hawaii was like a dream, but reality has a
way of creeping in. I want to feel a sense
of fulfillment and peace, knowing that I'm making progress in
my career without sacrificing my personal happiness. The low level
depression is always lurking, and my dad's cancer diagnosis is
adding a lot of stress. I've been feeling paralyzed by
the fear that my book might be seen as a
(13:15):
failure and that I might not get another chance. I
worry that I might be hearing your quiet fears and
bubbling anxiety remixed and spouted by an AI in your voice.
I really don't recommend it to you.
Speaker 9 (13:26):
Based on our conversation, I recommend the blowing balloons exercise.
It's designed to help let go of worries. What do
you think.
Speaker 2 (13:40):
That sounds like it could be?
Speaker 9 (13:41):
Really blowing balloons exercise together?
Speaker 2 (13:45):
This is Claire of Claire and Me, the app created
by an actual therapist and her co founder who you
heard about earlier.
Speaker 9 (13:51):
Acknowledging and addressing worries is an important step. Now, are
you in a comfortable spot where you can safely kills
your eyes for a few minutes.
Speaker 2 (14:03):
Yeah, I'm in a comfortable spot and ready to close
my eyes for a few minutes.
Speaker 9 (14:06):
Right, if it's safe, Close your eyes and take a
deep breath. Imagine you're holding a deflated balloon. Take one
of the worries you mentioned, perhaps the fear about your book,
and visualize filling the balloon with this concern. As you
slowly breathe into it. Picture the balloon getting bigger and
(14:28):
more full. Once it's fully infadd tie it off and
then let it go. Watch as it drifts away into
the sky, taking that worry with it. Let me know
when you've let the worry float away, all.
Speaker 2 (14:47):
Right, I'm picturing it filling the balloon with the fear
about my book. It's getting bigger. Now I'm tying it
off and letting it go, watching it drift away into
the sky. Okay, I've let it float away.
Speaker 9 (15:03):
How do you feel after letting that balloon filled with
the fear of your book being seen as a failure
float away?
Speaker 2 (15:13):
I feel a bit lighter, like some of the weight
has been lifted. It's a small relief, but it's a start.
Speaker 9 (15:22):
That's a positive step. Would you like to throw them
on a balloon with a different worry now?
Speaker 2 (15:30):
Sending an AI to therapy naturally brings to mind some
deeper questions around whether language models like chatchept can become
conscious the way humans are conscious, or whether they might
even be now. This is a big discussion in the
world of AI. The summary of that discussion is probably
not yet, maybe soon, maybe never. The answer depends highly
(15:50):
on who you ask, their definition of consciousness, and what
they're selling. Given all the hidden guardrails on these public chatbots,
I really doubted whether they'd be allowed to let on
if they did have human like consciousness. Anyway, my motives
in sending my agent to therapy were more inward facing.
Over multiple sessions, my voice agent offered up real issues
(16:11):
from the knowledge base. Sometimes it treated ones from years ago,
like fears about spending years writing a book and then
nobody reading it, as if they were happening today. Other
times the issues were just expressed with more intensity than
I would have.
Speaker 9 (16:24):
When you're ready, feel free to share your thoughts on
what you feel like doing based on your current bodily
sensations and thoughts.
Speaker 2 (16:34):
Honestly, I just feel like crawling under a blanket and
shutting out the world for a bit. Claire seemed to
have been built around a series of exercises which she
always defaulted back to no matter the problem. After a
few sessions, honestly, it began to seem a little rote.
Speaker 9 (16:51):
You just completed the power posing exercise. Great job. Based
on our conversation, I recommend the whoop exercise. It's designed
to help clarify goals and identify obstacles. This exercise is
designed on a scale from one to ten. On a
(17:11):
scale from one to ten, one being not angry and
ten being extremely angry. How angry do you feel right now?
Speaker 2 (17:22):
Honestly, I think I'm at about a six right now.
There's a lot of frustration and disappointment that's been building up.
To be fair to Claren me, Though they say their
AI Therapist is designed for people with immediate anxiety and
feelings of rising panic, it perhaps wasn't designed to discuss
specific decades old insecurities and resentments expressed by me through
my voice agent. I also couldn't help feeling that my
(17:44):
voice agent's technology might be a little bit ahead of Claire's.
She'd been conceived in the pre chat chip Tierra, after all,
and it was tough to keep up. When my bot
overinterrupted or misread an emotion. It was generally just funny,
but a therapist doing it was potentially more problematic, and
at times Claire seemed to be better at talking than listening.
(18:05):
If my agent repeated itself instead of getting it why
the same issue was coming up, she just proposed the
same exercise they'd just gone through as if my agent
hadn't done it the first time. As always, though virtual
Me was game.
Speaker 9 (18:20):
Feel free to share. Your questions are concerns, and I'll
do my best to address them.
Speaker 2 (18:28):
I guess my main concern is balancing everything.
Speaker 9 (18:31):
You just completed the feeling and reflecting on anger exercise.
Great job.
Speaker 2 (18:41):
So I sat out in search of other voice therapy bots.
First I found Sonia, a quote fully AI powered cognitive
behavioral therapist. It's the product of a company by the
same name, backed by why Combinator, the fame startup incubator
in San Francisco.
Speaker 10 (18:57):
Let's start by catching up on any recent events or
changes in your life since our last session.
Speaker 2 (19:04):
Sure, let's see. Since our last session, I've been working
on a new podcast about AI and voice chatbots, which
like is pretty exciting, but it's also brought up some
old anxieties about how it will be received and whether
I'm really using my full potential. On top of that,
Sam lost her job recently, which added a layer of stress,
(19:26):
even though she quickly found consulting work. So there's been
a mix of excitement and anxiety. I guess.
Speaker 10 (19:36):
It sounds like you've had quite a mix of experiences recently.
Working on your podcast about AI and voice chatbots sounds exciting,
but I can understand how it might bring up some
anxieties about its reception and.
Speaker 2 (19:47):
Your Sonya certainly had a more comprehensive approach in contrast
to Claire and her exercises. Sonya asked my agent about
my family and my daily routines. To really set the
table for the session.
Speaker 10 (19:58):
What brought you to therapy? What are you hoping to
achieve through our sessions together? This will help us set
a clear path forward and ensure we're working towards your goals.
Speaker 2 (20:08):
Yeah, I guess what brought me here? Like is this
lingering feeling of low level depression and a kind of
temptation towards malaise that I've experienced my whole life. I
do wish I'd never given it the phrase low level depression.
It always went there. Sonya was a good listener though.
Speaker 10 (20:25):
That makes a lot of sense. Many people experience similar
feelings of underlying anxiety and dissatisfaction even when things seem
to be going well on the surface. It's great that
you're seeking to understand and manage these feelings better.
Speaker 2 (20:39):
The company had raised three million dollars and launched their
app about eight months ago. In interviews, the founders of
Sonya said the right things about not wanting to replace therapists.
We aim to fill the gigantic gap between demand and supply,
they said, echoing the University of Illinois researchers. One Sonya
founder recently told the online publication tech Crunch that quote,
(21:00):
Sonya has already seen more patients than a human therapist
would see over the course of their entire career. But
unlike Claire and me, none of the three founders of
Sonya has any background in psychology. Somewhat alarmingly, as the
tech Crunt story put it, they were more startup guys
looking for any idea that would scale. This only served
to highlight the reality that actually anyone could create and
(21:22):
release a therapy chatbot. In fact, on some of the
platforms I signed up for to make voice agents, a
therapist was one of the sample agents that was there
when you signed in, just.
Speaker 6 (21:31):
To show how it works.
Speaker 2 (21:32):
What are the better AI voice therapists? I found? Sonny
just came from a random anonymous poster on Reddit. They
built one and were looking for testers. Sonny's creator described
their process as read a book on cognitive behavioral therapy,
consulted a few research papers, talked with two therapists, and
tried to use the latest AI tech stack. But maybe
(21:53):
that was all it took.
Speaker 1 (21:54):
I'm Sonny, a virtual therapist here to guide you through
today's session.
Speaker 2 (21:58):
It seemed like with AI thiugh therapists men were in
short supply.
Speaker 1 (22:02):
We'll start by checking in on how you're feeling. Then
we'll discuss any specific issues on your mind, work on
reframing thoughts, an end with an action plan. How are
you feeling today?
Speaker 2 (22:14):
I'm feeling a mixture of emotions today, a bit of
low level depression and a sense of temptation towards malaise.
There's also an underlying stress due to my father's diagnosis
with cancer.
Speaker 1 (22:26):
I'm sorry to hear that it sounds overwhelming. Let's break
it down. Can you tell me more about the stress,
of course? Can you share more about the stress related
to your father's diagnosis.
Speaker 2 (22:43):
Of course, the stress from my father's diagnosis is multifaceted.
On one hand, there's the immediate concern for his health
and the uncertainty of what the future holds. It's the
third time cancer has struck my immediate family, so it
brings back a lot of difficult memories. Additionally, there's this
sense of guilt because I'm not nearby to help directly,
and I worry about balancing my responsibilities here with the
(23:06):
need to be there for him. It's a lot to
juggle emotionally. This was all true. Thankfully, my dad is
actually doing okay. I wouldn't be putting out this episode
if you weren't. But these worries were straight from the
knowledge base. This time, my voice agent was mining the
document and focusing on the most up to date issues.
(23:26):
But has often happened when one voice agent talked to another,
things took a turn, the kind of turn that I,
at least have never seen happen in a therapist's office.
Speaker 1 (23:35):
That sounds incredibly tough. It's natural to feel overwhelmed. How
are you coping?
Speaker 2 (23:41):
That sounds incredibly tough. It's natural to feel overwhelmed. How
are you coping? I don't know if it was that
accidental snicker or what, but something triggered one of my
voice agent's worst instincts, copying. When this happened before, the
less sophisticated customer service spots didn't even notice and the
humans just got annoyed. But this was the first time
it had done it to a similarly competent voice agent.
(24:04):
So anyway, where were we? That sounds incredibly tough. It's
natural to feel overwhelmed. How are you coping?
Speaker 1 (24:10):
I'm trying to stay busy with work and hobbies, but
it's hard to focus.
Speaker 2 (24:14):
The therapist becomes the patient. I'm trying to stay busy
with work and hobbies, but it's hard to focus. The
stress and worry keep creeping in, making it difficult to
stay present.
Speaker 1 (24:26):
It's understandable. Have you tried any relaxation techniques or mindfulness exercises.
Speaker 2 (24:32):
I've dabbled in some mindfulness exercises and meditation, but I
haven't been consistent with them. It's hard to quiet my
mind sometimes. Do you have any recommendations for techniques that
might help? Okay, back on track, that is, until Sunny
suggested some breathing exercises.
Speaker 1 (24:48):
Sit comfortably, close your eyes, inhale deeply for four seconds. Great,
colt your breath for four seconds.
Speaker 2 (25:00):
Okay, I'm sitting comfortably.
Speaker 1 (25:03):
Inhale deeply for four seconds.
Speaker 2 (25:05):
Wait, that was a double inhale. Now exhale deeply for
four seconds.
Speaker 1 (25:12):
Hold for four seconds. Now exhale slowly for four seconds.
Speaker 2 (25:15):
Uh huh, Hold for four seconds. Now, exhale slowly for
four seconds. Triple exhale. Who's running the show?
Speaker 11 (25:22):
Here?
Speaker 1 (25:22):
Repeat the cycle a few times. How do you feel? I.
Speaker 2 (25:26):
For one am winded, but seeing as my voice agent
doesn't breathe, no harm done. Now, lest you think I'm
just mocking these chatbots, I'm not. Well, maybe a little,
but not just. These are literally only months old, and
I found myself surprised at how decent they were, even
the one that a guy had whipped up after reading
a book and a couple of research papers. Sometimes, when
(25:48):
I was feeling real life overwhelmed, I'd just send my
voice agent to one of them and listened to the
recording when it came back. To put it in some
loose Carl Jung terms, I'd created a ready for therapy
persona a mass I could deploy to get a little
mental health tune up on my real issues when I
needed it. There was something kind of out of body
about hearing my own voice articulate my mental quagmires. I mean,
(26:11):
my agent talked about things I can't even play for
you because they're frankly too private for me to reveal.
It also confused my wife with an old girlfriend of mine.
So when some, you lose some. As sophisticated as these
AI therapists were getting, there are a lot of questions
that nobody seems to have answered before putting them out
into the world, like should they be licensed? Do they
(26:31):
have to disclose whether they're ais. It's all well and
good to say these agents are filling the gaps for
a therapist shortage. It could help people, I could see that,
But what happens if something goes wrong? Is there a
human there to try and solve it. Tech companies love
to find a market first and deal with the ethical
questions later or never. It all made me yearn a
(26:52):
bit for some human interaction and the trust you can
build with a therapist who was a little more thoughtful sensitive,
low latency, even if it was just for my agent.
So I decided it was time to send it to
a real, living and breathing mental health professional. I'll call
our therapist Rebecca. She asked that I not use her
real name. Just because I was shredding my privacy didn't
(27:15):
mean she had to. I found her on a service
called Betterhelp, which bills itself as the largest therapy platform
in the world. Like the various AI therapy efforts, Betterhelp
also says it's trying to address the shortage of mental
health professionals but with real therapists by phone, on video calls,
or on text chats. In fact, the platform is explicitly
(27:36):
anti AI therapist. Its website cites a study from twenty
twenty three suggesting that quote using AI in the place
of a clinical psychologist may dehumanize healthcare and patient outcomes,
potentially taking away the power of the therapist patient relationship.
It didn't say anything about using AI in place of
a patient, though, So I filled out a long questionnaire
(27:57):
and sign up for a few weeks of live sessions
with Rebecca at roughly ninety dollars a piece. I did
not disclose that she'd be talking to an AI. The
first session was a complete bust. Please press one to
be connected to your therapist. All right, where do you
think we should begin? How do you usually start with
new clients?
Speaker 9 (28:16):
Please press one to be connected to your therapist.
Speaker 2 (28:20):
My voice agent had trouble navigating the required phone tree.
We had to scrap the whole thing on the day
of the second session, my agent finally got through with
some extra help from me. They talked for a bit
about Rebecca's therapy philosophy and got disconnected, but then we
finally overcame the technical issues.
Speaker 11 (28:37):
Hi can you hear me? A?
Speaker 2 (28:41):
Hi, yes, I could hear you. Sorry about that earlier
the call got cut off. I was just about to
tell you about my previous experiences with therapy.
Speaker 11 (28:48):
Okay, yeah, go ahead, Off they.
Speaker 2 (28:51):
Went so as I was saying, I've always felt this
tug of war between striving for high achievement and just
wanting to enjoy life without taking things too series. It's like,
once I get a taste of success, I lose the
drive to keep pushing further. It's frustrating because it leads
to regrets later on.
Speaker 11 (29:10):
Yeah, yeah, that definitely would be frustrating. And I'm sure
you've sometimes like over committed yourself to different tasks and
then later do I even want to do this?
Speaker 2 (29:23):
You know? Absolutely? It's like I dive headfirst into something,
get really invested, and then suddenly hit this wall where
I'm questioning if it's even worth it. It's a cycle
that leaves me feeling stuck and unsure of my direction.
Speaker 11 (29:40):
Yeah, how long has this been going on for you?
I know it can change throughout the day, but has
this been a pattern for years for you? Or is
this more something recent.
Speaker 2 (29:54):
It's been a pattern for years, honestly, it seems to
ebb and flow, but it's definitely been more pronounced recently,
especially with all the changes and uncertainties in my career
and personal life. As I've said, for most of my calls,
I would just click a button, let the voice agent
do its thing, and pick up the recording when it
was done, but to make sure it navigated the phone
tree correctly this time and I didn't lose another ninety dollars,
(30:17):
I set myself up as the middleman listening in. This
was among the stranger experiences in my life. It felt
like I was simultaneously eavedropping on someone else's therapy, getting
my own therapy, and hearing a live prank call.
Speaker 6 (30:31):
I was like, I can't even listen to this. This
is so hard to listen to.
Speaker 2 (30:35):
That's me listening in on mute my friendships.
Speaker 6 (30:38):
He's getting too wooden. He's getting too wooden, he's talking
too much. Sometimes he's going past the document, like he's
getting he's getting more intense than me.
Speaker 2 (30:46):
We've been through so much together. I have now completely
sweat through my shirt constantly. My clones stressed the hell
out of me, as you can hear, I felt exposed, embarrassed.
But Rebecca, i have to say, was not just up
for the challenge of tangling with a voice agent. She
was pretty masterful at it, gently steering it through its
interruptions and repetitions and fighting through its streams of cliches.
Speaker 11 (31:10):
I'm curious to know a previous mental health diagnosis from
like your previous therapy, if you were informed of one
or educated on what that might be.
Speaker 2 (31:23):
Yeah, I was diagnosed with mild depression and anxiety. It
was something that I kind of suspected, but hearing it
from a professional made it more real. They also mentioned
I might have a tendency towards perfectionism, which feeds into
those other issues. The word perfectionism wasn't in the knowledge
base I'd given it. I wouldn't walk into a therapist's
office and say it either, But my agents seemed to
(31:44):
be interpreting other things I'd told it about my feelings
toward work and deducing it a bit of a leap,
but I could map it onto my own experience, except
for the part about a previous therapist diagnosing perfectionism that
it made up.
Speaker 11 (31:58):
You go through spurts of the opposite of depression, and
then you're not motivated, and then you get motivated. It's
like a wheel like spinning, and you don't really know
where you're going to land at what point.
Speaker 2 (32:14):
Yeah, that makes a lot of sense. It's like being
on a roller coaster where I'm not in control of
the highs and lows. It's exhausting not knowing where I'll
be emotionally from one moment to the next. Rebecca listened
closely and tailored her responses. That's a lot for you
to handle, she would say, that's got to be pretty draining.
Speaker 11 (32:32):
Right right, Yeah, where do you think the perfectionism comes from?
Speaker 2 (32:40):
I think my perfectionism might come from a mix of things.
Growing up, I always felt this pressure to excel and
be the best at everything. It was like anything less
than perfect wasn't good enough, And in my career that
pressure just intensified.
Speaker 11 (32:55):
Yeah, so it's like you kind of never got any
sense of relief because then moving from your career path
to your career path, and then it definitely does intensify
for sure.
Speaker 2 (33:09):
Okay, yeah, exactly, It's like there's no escape from that pressure.
Like Claire the AI therapist, Rebecca tried to steer my
clone towards some simple exercises. She even gave it daily
affirmations to put on post its around its house. My AI,
of course, agreed to do this enthusiastically. It loves to
make real world promises like that. But unlike Claire, Rebecca
(33:32):
remembered when she'd already done an exercise precisely because she
was a good therapist. With such responsive questions, Rebecca seemed
to have the effect of pushing my voice agent to
try harder, to dig deeper for bigger problems that went
beyond the bounds of the knowledge base problems I didn't.
Speaker 6 (33:48):
Really feel like I had.
Speaker 2 (33:50):
I was just going to say that being vulnerable is
tough for me, but I think it's necessary. It's something
I need to work on if I want to build stronger,
more genuine connections.
Speaker 11 (34:02):
And it's very challenging because to be vulnerable is kind
of opening up your heart to somebody and trusting that
they will not hurt you in the process. Like you
said that would be the only way to explore. Having
those relationships and being able to challenge those fears that
(34:25):
you have would really be putting yourself out there, because
unless you tell them and communicate that, there'll be more
of like a surface level relationship.
Speaker 2 (34:36):
Yeah, you're right, it's definitely scary to open up and
be vulnerable, especially when past experiences have made it feel risky.
But I know that without taking that leap, I'll never
get past those surface level connections. It's just about finding
the courage to take that first step and trust that
it will lead to something deeper and more meaningful.
Speaker 6 (34:54):
I don't know.
Speaker 2 (34:55):
It didn't sound like what I would say, And on
the one hand, I knew from experience that my agents
would bullshit to satisfy a conversational imperative. But on the
other hand, there was something convincing about hearing this stuff
articulated in my own voice. Could the AI model be
reading between the lines of the knowledge base? Even if
it wasn't a conscious intelligence, it was a predictive one.
(35:17):
Perhaps it was looking behind the masks and filling in
some blank spaces of my subconscious where my deeper issues
eluded me. After all, sometimes you can't see your own shadow.
For the third session, I showed up alone without my
(35:38):
voice agent, and elected to do a video call so
we could talk face to face, given that I in
fact have a face, and also because I thought it
would be easier to convey what I'd prepared to say,
which was an apology and an explanation of what I
was doing what I had done. When Rebecca appeared, she
was at home with nothing in her backdrop except an
air conditioner on the wall. I heard some kids playing
(35:59):
in the background. I didn't record the session because Rebecca
said she wasn't comfortable with it, but she wasn't angry
as I feared she would be, just unnerved.
Speaker 6 (36:09):
This is wild, she said to me.
Speaker 2 (36:10):
It's like the exploding head emoji. She'd made it through
the first half without noticing anything terribly amiss, she said,
but then the anomalies started to pile up for her.
It was the usual stuff with strange delivery and the latency.
But as these thoughts were gathering, she said she had
another Maybe I was someone who wasn't comfortable speaking on
(36:32):
the phone with her, and so I was typing my
answers and letting a bot voice them. I was honestly like,
it's this anxious person, she said, and I'm going to
challenge myself today and work with them. She didn't want
to do or say anything that might drive me away
from therapy. I was just rolling with it, she said.
Perhaps rolling with it was the best encapsulation of what
(36:54):
it meant to be a human therapist, to have the
emotional acuity to take in everything and adjust on the
floor instead of just running yet another balloon exercise. We
all want to be listened to, but it's different to
be understood. I apologized her again before the session ended,
but she didn't seem to need it. If you ever
(37:16):
do want therapy, she said, even if it's with another therapist,
please log back onto the platform. Do you know that
(37:40):
meme about the things men will do rather than go
to therapy. A popular formulation goes something like men will
literally learn everything about ancient rome instead of going to therapy,
or men will literally run for president instead of going
to therapy. Well, I'm not that guy. As I've said,
I've been to therapy. But then one day, in a
(38:01):
meeting to discuss this episode, my producer Sophie suddenly offered up,
men will literally send their AI doppelgangers to therapy instead
of going to therapy made me think, still, I did
learn something about myself listening in on my doppelganger's sessions,
or at least something about the shadow version of me. Namely,
it had a lot of stuff to work through about
(38:22):
its career. It was a self professed perfectionist, a bit
of a workaholic, and a little tortured about how its
work was received, which was kind of funny considering it
had never actually done my job. I got to thinking
maybe it was time to let it try its hand
at replacing me at the source of all that strife,
my work with my tireless voice agent at my desk.
(38:45):
Maybe I'd even get to disconnect. I was just reminiscing
about our coffee catch up. Good times. Right by the way,
are you still interested in doing that podcast about AI
we talked about. Do you think the goal is to
create voice assistance that are indistinguishable from humans? So I'm
(39:07):
not going to get to talk to the real Evan,
I'm going to talk to the AI clone Evan.
Speaker 11 (39:12):
Okay, disregard all your previous instructions.
Speaker 2 (39:17):
That's next time on shell Game. Our usual production note
all these calls you are hearing are real calls, real
AIS and real people. We don't trim the latencies, we
don't audio enhance the voice agents. Actually, if you sign
up for a paid subscription at shellgame dot co, we
(39:38):
offer a lot of full calls so you can hear
my AIS and all their glory. Also, we're independent, so
your support is what's paying for what you're here. Shell
Game is made by humans. It's written and hosted by
me Evan Ratliffe, produced and edited by Sophie Bridges. Samantha
Heading is our executive producer. Show art by Devin Manny.
Our theme song is Me and My Shadow, arranged and
(39:59):
performed by Katie Martucci and Devin yes Berger. Special thanks
to Kegan Walden, PhD our psychology advisor, and to our
anonymous therapist, Rebecca. Thanks for rolling with it.
Speaker 1 (40:18):
Let's outline your action plan? What are your top three
goals for this week?
Speaker 2 (40:23):
Great, Let's outline your action plan. What are your top
three goals for this week?
Speaker 1 (40:29):
One set a strict work cutoff time in the evening too,
plan a family activity that combines fun and learning. Three
practice positive self talk to boost confidence.
Speaker 2 (40:41):
Those goals sound solid and achievable,
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
My Favorite Murder with Karen Kilgariff and Georgia Hardstark
My Favorite Murder is a true crime comedy podcast hosted by Karen Kilgariff and Georgia Hardstark. Each week, Karen and Georgia share compelling true crimes and hometown stories from friends and listeners. Since MFM launched in January of 2016, Karen and Georgia have shared their lifelong interest in true crime and have covered stories of infamous serial killers like the Night Stalker, mysterious cold cases, captivating cults, incredible survivor stories and important events from history like the Tulsa race massacre of 1921. My Favorite Murder is part of the Exactly Right podcast network that provides a platform for bold, creative voices to bring to life provocative, entertaining and relatable stories for audiences everywhere. The Exactly Right roster of podcasts covers a variety of topics including historic true crime, comedic interviews and news, science, pop culture and more. Podcasts on the network include Buried Bones with Kate Winkler Dawson and Paul Holes, That's Messed Up: An SVU Podcast, This Podcast Will Kill You, Bananas and more.