Latest / The Mental Health Evolution / Ep 5: Preparing for the AI Shift in Therapy with Dr. Jordan Harris

Transcript

Ep 5: Preparing for the AI Shift in Therapy with Dr. Jordan Harris · The Mental Health Evolution

0:00
(Transcribed by TurboScribe. Go Unlimited to remove this message.) This podcast is part of the Frederick Podcast Network.
0:03
Learn more at listenfrederick.com.
0:05
Welcome to Mental Health Evolution, a podcast about what's changing in mental
0:10
health and why it matters.
0:12
I'm your host, Rachel Harrison, inviting you into honest conversations with
0:16
people from all perspectives in the field, clinicians, tech founders,
0:20
investors, insurance companies, and all the folks in between.
0:24
Let's explore what's working, what's not, and what's next.
0:31
Welcome back, everyone, to the Mental Health Evolution podcast, where we are
0:38
talking about how the landscape is quickly evolving in the mental health
0:43
industry.
0:43
Today, we are taking a look at none other than AI.
0:48
It is a hot topic really everywhere, but especially in the mental
0:52
health world.
0:53
Our guest today, Dr. Jordan Harris, has researched and written about what
0:58
is happening with AI.
0:59
He is also a therapist, and he coaches therapy practices in their
1:03
growth and just figuring out what to do with artificial intelligence in
1:08
general.
1:09
As always, I want to give a little background here to set
1:12
the stage for our conversation so you have some information about the
1:16
topics we are addressing.
1:18
Here are a couple things that have been in the news.
1:21
First, from telehealth.org, we have the article about Amazon Health Services
1:26
and Talkspace Partner for Virtual Therapy.
1:30
So this is a corporate partnership between these two companies, and the
1:35
article outlines the elements of that partnership, as well as the implication
1:40
that these two companies plan to use recorded sessions and written interactions
1:45
with clients and therapists in order to train AI to take over
1:50
the therapist role.
1:51
They're not making this a secret.
1:53
That's kind of something that's happening in a lot of forums, and
1:56
this is just one of those.
1:58
Then I also want to reference an article from Fortune magazine.
2:03
This was a pretty well distributed article, and the title says, in
2:07
a showdown of psychotherapists versus chat GPT, the latter wins, new study
2:12
finds.
2:13
This study was interesting.
2:15
It set out to tell if people could tell the difference between
2:18
chat GPT, that therapist, and their therapist in which they preferred.
2:23
So they found that participants had difficulty identifying whether responses
2:28
were written by chat GPT or by therapists, and they correctly guessed
2:33
that the therapist was the author about 56 percent of the time,
2:38
and they correctly guessed that chat GPT was the author about 51
2:42
percent of the time.
2:43
So kind of a 50-50 split there, whether they could tell
2:47
if it was chat GPT or the therapist.
2:51
And then further, the article goes on to say that in most
2:54
cases that the clients preferred chat GPT's take on the matter at
3:00
hand, based on how they understood the speaker, showed empathy, what they
3:04
said was appropriate, and what a good therapist would say.
3:09
Of note, this was specifically couples therapy.
3:12
And this was written responses between either a therapist or chat GPT.
3:17
So not sort of the full AI presentation here.
3:21
And lastly, I want to reference the article that kind of was
3:25
the springboard for this episode of the podcast.
3:29
And that's Dr. Jordan's article.
3:31
And he wrote an article called AI is reshaping psychotherapy.
3:35
We have two years to get ready.
3:37
He outlines in this article that there will be four big players
3:42
in the AI therapy space.
3:44
Insurance companies, model-based apps, big box provider platforms and general
3:51
large language models.
3:53
He also talks through the idea of autonomous AI being a co
3:58
-pilot in therapy and potentially helping us with process coding, meaning that
4:03
we would receive real-time feedback to enhance attunement and effectiveness in
4:08
the therapy room.
4:09
His article is robust and there were a lot more pieces.
4:13
That's just me pulling out a couple of highlights there so that
4:17
our listeners kind of have some background with what it is that
4:21
we're referencing and what we're going to be talking about.
4:24
So with that, I would like to open up the conversation with
4:27
our guest today and talk through some of these events, as well
4:30
as looking at how to make meaning of all of this from
4:34
the perspective of clients, clinicians and businesses in this space.
4:39
So first, I want to say welcome, Jordan.
4:42
Thanks for being here.
4:43
Hi, thanks for having me.
4:44
Glad to be here.
4:45
Yeah.
4:46
So let's look at the overall patterns happening with AI in the
4:51
mental health space right now.
4:53
What are you seeing so far?
4:55
Yeah, I mean, there's sort of two things that I'm seeing.
4:59
One is, well, one is that no one knows what's happening.
5:01
I mean, this thing is moving so fast.
5:04
Every week there's a new thing happening, particularly in the mental health
5:08
space.
5:09
People are really trying to figure out, like, what are the limits
5:12
of what we should allow it to do?
5:15
So it can do a lot, but like, should we allow it
5:17
to record sessions?
5:19
Should we not?
5:20
Those are all questions that we're trying to figure out.
5:23
And so with that, a lot of what I'm also seeing is
5:25
the first thing that a lot of people who are testing out
5:29
is this extreme note taking, like, can this thing do my notes?
5:33
Beyond that, I mean, I think there's a few different ways that
5:35
the field is sort of unfolding, but we're trying to figure out
5:39
as a field and I think as a society, like, what is
5:42
this thing and what's it going to do?
5:43
Yeah.
5:44
Do you have a particular stance on this whole idea of AI
5:48
taking notes for us in session?
5:51
Yeah.
5:51
So, you know, a lot of people think that it's unethical, that
5:56
it violates privacy and confidentiality sort of things.
5:59
There's two things to that in my mind.
6:02
One is that I don't think that it does.
6:04
Right.
6:04
Like, you have a duty to do informed consent, but as long
6:09
as you're doing that, there's no reason why you couldn't have it
6:13
record your sessions unless there's some sort of law specific to your
6:17
state.
6:18
If you think about, like, for me, I went to college and
6:21
all of our sessions were recorded and we review them with our
6:24
supervisors.
6:24
Now, this is back in the olden days.
6:27
The old school.
6:27
We met in person.
6:28
We met in person and like, yeah, to me.
6:31
Yeah.
6:32
We didn't store it to the cloud.
6:33
We just destroyed it right on the, yeah.
6:36
Right.
6:36
So, like, but we had it in our informed consent.
6:39
People knew when I came to the university clinic where I was
6:41
learning that they were going to be recorded and watched and that's
6:44
what happened.
6:45
The other thing is, and this is, I think, the thing that's
6:47
important because the field is moving so fast.
6:50
I don't think regulatory bodies, which are usually like AAMFT or the
6:56
APA or the state boards, can move fast enough to tell people
7:01
what to do.
7:02
We might see some like standards of care type stuff, but like,
7:05
you know, my state board is struggling to get us updated onto
7:08
a digital platform for license renewal.
7:12
So I don't think they're going to be telling me about how
7:14
to use A.I. anytime soon.
7:15
It's just not going to happen.
7:17
No, I think you're right about that.
7:19
Yeah, but I, I am one of those people that I do
7:23
think that there are ethical risks to doing this because once that
7:28
are once that recording exists, we don't have control at that point.
7:33
I hear your client consent piece, but I also sort of wonder
7:37
if clients know fully what they're consenting to from the perspective of
7:42
I've talked to several different A.I. companies about like what happens,
7:47
right?
7:47
And they all will say, well, some of them destroy the recording,
7:52
but all of them data mine the recording.
7:55
So there's this whole nuance between, OK, maybe the actual recording is
8:00
destroyed, but what about the data mine?
8:02
And my additional concern is what if policies change and they decide
8:06
it doesn't benefit them as a company to destroy these anymore?
8:09
What happens if attorneys in a case get wind that there is
8:15
information they can access?
8:17
You know, there are all these protections maybe there to just think
8:21
about.
8:21
I think for me, rolling out anything like this, we need ethics
8:25
in our field.
8:27
Privacy and confidentiality is the whole reason why people are safe enough
8:31
to come and talk to us.
8:32
Right.
8:33
So that's just for me, that's just one piece that I wonder
8:36
about with A.I. notetaking.
8:38
When I'm thinking about ethics, I'm thinking about like what are the
8:41
actual ethical codes I have to adhere to and what are my
8:45
state legislators saying?
8:47
And so to to my mind, if I'm answering this question, it's
8:51
there's nothing in the ethical code that I have to adhere to
8:53
or the state legislator.
8:55
Now, like morally, what's maybe best for the practice for my clients?
8:58
I think your questions are totally valid.
9:00
The only thing I would say, the only thing I would say
9:03
is most of our clients are already giving away their data and
9:07
they don't care.
9:07
So if I have a client who says this wigs me out,
9:11
I don't want to do it, then I'm not going to do
9:13
it.
9:13
Right.
9:14
But, you know, like even in the article that I wrote, they
9:18
surveyed about 500 clients and like 98 percent of them are already
9:23
using chat GPT or some other sort of large language model.
9:26
People, I mean, if if you're on social media, like if you're
9:29
on X or LinkedIn, like they're already mining your data.
9:32
So I don't think people I think we care a lot about
9:34
that for other people.
9:35
But for ourselves, most of us just give our data away anyway.
9:39
I mean, that's a that's a fair point.
9:42
I think that is happening everywhere all the time, but maybe not
9:46
with the content that people are sharing in therapy.
9:50
I mean, that's a that's a good point.
9:53
The other thing and this is my bias, you know, my my
9:56
background is really in clinical expertise.
9:59
How do we become a better therapist?
10:01
And one of the things I think we see across fields, there's
10:04
two big ways.
10:05
It's to become better at any field.
10:07
And one of those ways is through the data.
10:09
Like if you get better analytics, whether you're a pilot in an
10:13
airplane or a baseball player or even comedians, I've seen comedians like
10:17
track the amount of laughs they get.
10:19
But they just get better at the field.
10:20
So personally, I want to help people.
10:23
And again, if my client says that this is not for me,
10:25
I'm not I'm not going to force this on them at all.
10:27
But I will take what tools I can help.
10:29
I can get to help people.
10:30
Yeah, I hear that.
10:31
That's understandable.
10:32
I would love to talk about this whole idea.
10:35
I mentioned it in one of the articles about that AI or
10:38
chat GPT might be a better therapist to a human therapist.
10:42
And I would love to hear some of your thoughts about that.
10:47
Yeah.
10:47
So, I mean, this goes back two things to my mind.
10:51
So last year, I said my my whole push in my actual
10:55
academic career is expertise for clinical clinical expertise.
10:59
And so a buddy of mine who has written a ton of
11:02
books on clinical expertise recommended me to an article by two researchers,
11:08
Henny Westra and Alyssa DiBartolomeo about clinical expertise.
11:11
And basically what they found is they specialize in studying resistance.
11:16
And what they mean by resistance is any moment where the client
11:20
says this is not a good fit for me therapist.
11:22
What they basically found is that most therapists, not only do they
11:25
misrespond, they don't notice those moments are actually happening.
11:28
They just don't see them.
11:30
And so I actually, I mean, this is kind of extreme, but
11:32
I paid Alyssa to coach me for about a year so that
11:35
I could see this dynamic and it's everywhere.
11:38
It happens a lot.
11:39
So when we're talking about AI and therapy, one of the big
11:44
things that I see online is people saying things like, man, this
11:48
AI, it's sort of too nice to other people.
11:52
And on the one hand, if someone's like in a real sort
11:55
of delusion, right, if they're really in a place where they're really
11:59
at risk for themselves and for other people, like that's not good.
12:02
And I have friends who are working on this problem now of
12:04
how do we actually make these systems safe?
12:06
And a lot of systems, not not all models are safe, but
12:10
some of them are actually pretty safe already.
12:11
But the other side of this is, is that this thing that
12:14
AI does where it is consistently validating is actually the antidote to
12:20
a lot of resistance.
12:21
Now, you have to be careful that you don't validate poor behavior,
12:25
but validating positive intent, validating underlying emotions and being
12:29
relentless in it.
12:30
I mean, from the research that I've done and the coaching that
12:33
I've gotten, we don't do enough of that.
12:35
And I think that's one of the key reasons why this these
12:38
these large language models are so persuasive because they are consistently
12:44
validating.
12:45
I mean, even back to the old to me, this all is
12:48
like sort of a weird competition to have because like 60 years
12:51
ago, you had Eliza, which was the first chatbot, and all it
12:55
did was Rogerian reflection and people loved it.
12:59
Like they lost their minds about how amazing it was.
13:02
Right.
13:03
Yeah.
13:05
So now we have that.
13:07
But like a million X, like, yes, it's really good because when
13:11
you're working with people in these really tender spots, they need a
13:14
lot of support.
13:15
And so and especially when when most therapists don't really see they're
13:20
not giving their clients as much support as the client would really
13:22
like.
13:23
It's no surprise to me that these models are doing really, really
13:25
well.
13:26
Yeah.
13:26
But there I mean, I guess there's also the other piece of
13:29
that.
13:29
I've heard it called psychophantic bias, right?
13:32
That that chat GPT is a yes man, that AI is a
13:36
yes man.
13:36
That's kind of what you're saying.
13:37
Right.
13:37
And that if you ask it to if you say, I don't
13:42
like that response, give me a different one.
13:44
They'll give you a 180 response.
13:47
I think that's fascinating.
13:49
So we're almost there's I guess for me, it takes away that
13:53
human element in the room of psychotherapy where we are talking and
13:59
putting out who we are.
14:01
And there is another human being taking that in and feeding that
14:05
back in some way.
14:07
Right.
14:07
Basic attachment, kind of human interaction.
14:10
And maybe we lose that if we only get exactly what we're
14:14
wanting in response.
14:17
You're hitting on, I think, a few different points that I think
14:19
are really important.
14:20
The first is, will the models of tomorrow be as agreeable?
14:24
Right.
14:24
So like on on the level of reflecting, validating your intentions and
14:30
your emotions, I think that most clients actually need a lot more
14:33
of that than we're actually getting them in terms of like the
14:36
behaviors.
14:36
No, not every behavior that you have is valid.
14:38
And my sense is, is that the models that are trained to
14:42
be therapy models won't have that problem.
14:44
Like that would be part of the difference.
14:46
And so that's my first sense.
14:49
The second thing is, and I think this is a really important
14:51
question, I think that we're going to find out is I 100
14:54
percent agree with you.
14:56
Like there's something that we lose in the person to person interaction,
14:59
but I'm not sure that most people will actually care.
15:03
My model for this is the old Harry Harlow studies.
15:07
Oh, yeah.
15:08
The monkey studies.
15:09
The monkey studies basically showed us that mammals are cue based.
15:14
And so if this if this if this wire monkey doesn't have
15:18
the cues that we need for softness, right, we don't perceive it
15:21
as being warm.
15:22
But if it's terrycloth mother does, we perceive it as being warm
15:25
and we'll attach that thing.
15:27
And so that's my and I think that this really goes back
15:29
to your question of what do we lose?
15:31
And that's like the question.
15:33
But will most people be like, man, this terrycloth mother feels really
15:37
good.
15:37
People will think a lot of people will, you know, and I
15:40
saw something today that like 20 percent of people who use a
15:43
are using it already feel like romantic partners.
15:46
So I have seen that.
15:48
I know.
15:50
I know.
15:52
Yeah, that's I mean, that's a whole other topic.
15:55
We could go all the time.
15:57
It's all the topic.
15:59
But let me dive into another thing that you say.
16:03
Even in your title.
16:04
So I feel like it's an important piece.
16:06
But you kind of frame the next two years as a critical
16:09
window for adaptation.
16:11
So I'm curious from your thoughts what needs to happen during the
16:16
next two years and kind of what happens if we miss it.
16:20
That two year window comes from a few different things.
16:22
One, it's from a interview between Ezra Klein and the chief sort
16:28
of A.I. officer under Joe Biden.
16:31
And in the interview, they basically said that Biden was well aware
16:35
that within the next two years we'd have autonomous A.I. And
16:39
then everything flipped because Trump was sort of like the president and
16:42
they didn't see that coming.
16:43
So they're like, it's coming in the next two years.
16:44
The other thing is that this report out called A.I. Twenty
16:47
twenty seven.
16:48
I don't know if you've if you've read this report.
16:50
It's really fascinating.
16:51
OK.
16:52
And basically they're also saying two years from now, we're going to
16:55
have A.G. Right.
16:56
Super A.I., which I don't know if that's going to happen
16:59
in two years.
17:00
I don't know if anybody knows, but I do think the next
17:02
year is going to be like a crazy period.
17:04
And part of what that happens is we're in a Cold War
17:07
who gets A.I. first.
17:08
So we're getting a lot of funding.
17:09
And I mean, as soon as Trump was elected, he called all
17:12
the tech leaders and like literally behind him, right, like up on
17:15
them, up on the at the inauguration.
17:18
And he gave some moment like five hundred billion dollars.
17:21
Something some I don't remember the number of some astronomical amount that
17:24
he partnered with.
17:25
So we're in a Cold War with the government basically saying, give
17:29
as much money as you can toward this.
17:30
And the first thing that all the A.I. companies are trying
17:33
to do is use A.I. to code.
17:36
They're trying to use it in a cursive process to make it
17:39
better at making more A.I. So that's where all the energy
17:41
is going.
17:42
So that's why I think the next years are like we're going
17:44
to go through a time warp.
17:45
In my opinion, I think I think what this means for us
17:48
is and this goes back to the business side.
17:50
Right.
17:50
So as as much as I'm academically interested in this question of
17:55
expertise, I and I have a practice where I do all these
17:57
things.
17:58
I also have a business where I help people launch the practice
18:00
and run practices.
18:02
And the biggest issue in most practices is retention.
18:06
I saw a study last week.
18:08
I posted this on my on my website that went through and
18:11
categorized different types of therapists based on outcomes.
18:14
And in this study, 25 percent of therapists were inefficient and ineffective
18:20
on surveys of, hey, how much do you like to counsel clients
18:24
that I don't like them very much?
18:26
And then on outcome surveys of how much better do you feel
18:28
this?
18:29
So I don't really feel better.
18:30
Another 40 ish percent were good for some people, but for the
18:33
rest of their clients, for whatever reason, the clients dropped out really
18:36
early in therapy.
18:37
What this means is we have somewhere between probably 25 to 60
18:42
percent of counselors who struggle significantly, at least with some portion of
18:46
their of their caseload.
18:47
So here's what I think we have to do.
18:49
I think we kind of got a button up, right?
18:52
Like like I am.
18:53
I'm very pro people.
18:55
I don't think like I am very much of the belief that
18:59
other people might choose to use AI as a therapist.
19:02
I don't think it's a field that we should be.
19:05
I don't think that that's best.
19:06
I believe in the human connection, but I don't want to deny
19:09
what I think other people will do.
19:11
And so we want to really be people who are standing in
19:14
the gap for our patients, for our clients.
19:16
We have to really, I think, up level in our skills that
19:19
we, you know, if a quarter of therapists are not doing great
19:23
in their practice, like we got to help them, people help those
19:26
people to really meet clients where they are, so that clients can
19:28
actually get what they need from a person.
19:32
So that's the first big thing is we got to upscale.
19:35
And then the second thing is we have to be using these
19:37
tools now so that we're familiar with them.
19:39
And I mean, they change weekly.
19:41
So pay the twenty five bucks, get the pro subscription and like
19:44
start playing with it, start using it.
19:47
Maybe not for your notes, maybe for your marketing materials.
19:50
Right.
19:51
Maybe for your logos, maybe for your newsletters, like maybe for your
19:55
blogs, like use it to develop those.
19:58
And then I think the third thing that we need to be
19:59
doing is not fighting it, but really pushing for any way we
20:05
can to use it.
20:06
Right.
20:06
I think if we fight against it, then the house is against
20:09
us.
20:10
If we really join this thing and say, hey, let's use this
20:13
as a as a copilot.
20:14
Let's use this as a librarian.
20:16
Let's use this as a assistant to us.
20:18
We can use it and can use it as a superpower to
20:20
really help people.
20:21
So that's those are my like three big things that we have
20:23
to do.
20:23
I like that.
20:24
I think for that last one, to me, there's also this maybe
20:28
this duty for us to ethically use it ethically, right?
20:32
Figure out totally how we can make sure it does no harm
20:35
to clients.
20:37
We know that's already happened.
20:38
And use it in a way that can be of the greatest
20:42
benefit.
20:43
Yeah.
20:43
To people like that, I will say.
20:46
And, you know, so I also work at adjunct and do some
20:50
consultation with a university called Centio.
20:52
And they're in the forefront of this stuff.
20:53
So we're getting training on this in our faculty meetings on how
20:56
to use it ethically.
20:57
So maybe I'm less afraid of that because I'm getting training in
21:00
it.
21:00
And they have a class that they're putting out in September, I
21:04
think it is.
21:04
And I don't I mean, I'm not getting any sort of kickback
21:06
from this, but I think it's helpful.
21:07
They're getting a certification on working with using A.I. in your
21:10
in your practice.
21:12
So I'm very much a fan of what you're talking about.
21:15
And maybe I'm a little bit less worried because I'm getting training
21:18
in it.
21:18
Maybe people other people aren't.
21:20
Yeah, I think that's a unique thing.
21:22
I have not heard of another class like that.
21:24
Not that they're not out there.
21:25
Just hasn't crossed my path.
21:27
So that's very cool.
21:29
I like that.
21:30
All right.
21:30
Well, with that, we are about out of time.
21:33
Jordan, thank you so much for being here and digging into this
21:37
discussion.
21:38
I feel like there were even more things that we could have
21:41
gotten into.
21:42
And I encourage everyone listening to read his article and to dig
21:47
in more and more to what's happening so that we can be
21:50
aware, so that we can be smart.
21:52
Thanks, everybody, for listening.
21:54
And I hope you will join us next time for another look,
21:58
another view at the changes in the mental health industry.
22:04
Thanks for listening to the Mental Health Evolution podcast.
22:08
Join the conversation.
22:10
Find us on social.
22:11
Send us your thoughts and tell us what you are seeing in
22:14
your corner of the mental health world.
22:16
Until next time, stay curious and keep evolving.
22:25
Live radio, local podcasts, the Frederick Podcast Network and ListenFrederick
22:29
.com.