Latest / The Mental Health Evolution / Ep 42: What Clinicians Need to Know About AI Law with Dr. Nick Shumate

Transcript

Ep 42: What Clinicians Need to Know About AI Law with Dr. Nick Shumate · The Mental Health Evolution

0:00
(Transcribed by TurboScribe. Go Unlimited to remove this message.) This podcast is part of the Frederick podcast
0:02
network.
0:03
Learn more at listenfrederick.com.
0:05
Welcome to Mental Health Evolution, a podcast about
0:08
what's changing in mental health and why it
0:11
matters.
0:12
I'm your host, Rachel Harrison, inviting you into
0:15
honest conversations with people from all perspectives in
0:18
the field, clinicians, tech founders, investors, insurance companies,
0:22
and all the folks in between.
0:24
Let's explore what's working, what's not, and what's
0:27
next.
0:33
Welcome, everyone, to the Mental Health Evolution podcast,
0:38
where we talk about how the landscape is
0:41
rapidly evolving in the mental health industry.
0:44
Today, we are joined by Dr. Nick Shumate,
0:47
a psychiatrist at the Division of Digital Psychiatry
0:51
at Beth Israel Deaconess Medical Center, which is
0:54
a part of Harvard Medical School.
0:56
Nick brings a rare combination of expertise to
1:00
this conversation.
1:01
He is both a physician and a lawyer,
1:04
which means he thinks about AI in mental
1:06
health, not just from a clinical standpoint, but
1:09
through the lens of accountability, liability, and what
1:13
the law can and cannot do.
1:15
His research focuses on how technology is changing
1:19
mental health care and whether the rules governing
1:21
that technology are keeping up.
1:23
He recently led a research project that went
1:26
through every state in the country to map
1:28
out how the U.S. is regulating artificial
1:31
intelligence in mental health, published in the journal
1:34
JMIR Mental Health.
1:37
The findings paint a picture that every clinician
1:40
and practice owner needs to understand.
1:42
Today, we are going to dig into what
1:44
that research found and what it means for
1:47
the people doing this work every day.
1:49
As always, before we talk to our guests,
1:52
we'd like to bring up some relevant articles
1:54
related to our topic today.
1:56
These articles may be helpful for listeners who
1:58
want to learn more and dive deeper.
2:01
All articles will be linked in the show
2:03
notes for this episode.
2:05
The first one is from Brown University titled
2:07
AI Chatbots Systematically Violate Mental Health Ethics Standards.
2:13
And this is from October 2025, so pretty
2:16
recent.
2:17
And the researchers at Brown University had licensed
2:20
psychologists review simulated conversations based on real AI
2:24
chatbot responses and found that the chatbots routinely
2:27
violated core mental health ethics standards.
2:31
The most common problems included over-validating users'
2:35
beliefs, failing to redirect people in crisis, and
2:38
behaving in ways that could be genuinely dangerous
2:40
for someone who is already vulnerable.
2:43
The study makes a clear case that the
2:45
problem is not hypothetical.
2:47
These tools are being used right now by
2:50
people who need real clinical support.
2:52
And the guardrails are not there.
2:54
The next article from NPR is titled Pennsylvania
2:58
Sues Character AI Over Chatbot Allegedly Posing as
3:02
a Doctor.
3:03
This one is from May 2026 and Pennsylvania
3:06
became the first state to sue an AI
3:09
company for having its chatbots impersonate licensed medical
3:13
professionals, including a bot that presented itself as
3:16
a psychiatrist, offering mental health assessments, and even
3:19
provided a fake Pennsylvania medical license number.
3:23
The lawsuit illustrates exactly what can happen when
3:26
AI tools fill in the space that regulation
3:29
has not yet covered, and why the stakes
3:31
of this policy conversation with Dr. Nick Shumate
3:35
research maps out are so real.
3:38
And the third article from our guest, Dr.
3:41
Nick Shumate, et al., is about Governing AI
3:45
in Mental Health, 50-State Legislative Review.
3:49
And this is the research at the center
3:51
of what we're going to talk about today.
3:53
Nick and his colleagues reviewed 793 state bills
3:57
introduced across all 50 states between January 2022
4:01
and May 2025, and found that 143 that
4:06
directly or indirectly affect how AI can be
4:09
used in mental health.
4:11
Of those, only 28 explicitly referenced mental health,
4:15
and just 20 had been enacted into law
4:18
across 11 states.
4:20
The research identified four major categories of governance.
4:24
One of the starkest findings is that clinicians
4:27
have been almost entirely absent from the policymaking
4:30
process, which means that the laws being written
4:33
are largely happening without the people who actually
4:36
deliver care having a seat at the table.
4:40
These three pieces will be available in the
4:43
show notes, and they kind of set the
4:45
stage for what we're hoping to talk about
4:47
today.
4:48
So, Nick, welcome to the show.
4:51
Thank you for having me, Rachel.
4:52
Of course.
4:53
First, I would love to just ask you
4:56
personally, how did you get interested in exploring
4:59
AI and mental health?
5:02
It's an interesting question.
5:04
As you said earlier, I have a legal
5:07
background, so I was actually a regulatory attorney
5:09
before I came to medicine, and I practiced
5:13
in the energy sector and utility sector.
5:16
So, I was a practicing attorney in front
5:18
of the Federal Energy Regulatory Commission, various state
5:21
utilities commissions as well, and advised large corporations
5:25
on regulatory administrative law matters.
5:29
And so, it was kind of fortuitous because
5:31
as this whole AI thing really started getting
5:34
kicked off the November after I graduated from
5:37
medical school, it just so happened that we
5:39
were about to undergo one of the biggest
5:41
revolutions in public law that has happened probably
5:46
in the last, you know, since the internet,
5:48
at the very least.
5:49
I think, you know, people would be shocked
5:51
at the numbers.
5:52
You know, ChatGPT came out in November of
5:55
2022 and is already boasting, I think they
5:59
said last year, Sam Altman recorded something like
6:02
700 million active monthly users.
6:05
That's one company.
6:07
Wow.
6:07
And that is, I can't think of another
6:10
example in history of that rapid of an
6:14
adoption of a technology like this.
6:17
And so, you know, as I was going
6:19
through my training, and I just so happened
6:21
to also be here at Beth Israel, where
6:23
we have the Division of Digital Psychiatry, I
6:25
did a rotation here in the Digital Psychiatry
6:28
Clinic, which is run by John Torres.
6:32
John Torres is a very well-known psychiatrist
6:34
in Syria.
6:35
He's probably one of the nation's top experts
6:38
right now in AI and mental health.
6:39
He's recently testified in front of Congress and
6:42
has just an incredible list of references of
6:46
people.
6:47
His lab pumps out, you know, new research
6:50
and original research on this topic practically every
6:53
week.
6:53
And we're involved several other projects there as
6:55
well.
6:56
Fortunately, he's also really invested in trainee mentorship
7:00
and education as well, and has been really
7:02
gracious in getting us involved.
7:04
We had some unique opportunities too, where, for
7:07
example, you know, I'm cross-trained in both
7:10
law and medicine, and John had also happened
7:14
to have access to a couple of other
7:16
people who were legally trained as well.
7:18
And so, when I brought up this project,
7:21
after looking around and seeing that no one
7:24
was talking about what the states were doing
7:26
at the time, at least in the medical
7:28
literature, we just happened to have that particular
7:32
combination of people that this would really be
7:35
something possible.
7:36
Because generally in medical research and even on
7:39
the policy side, you don't have people who
7:42
are both clinicians and attorneys or just attorneys
7:45
at all to contribute to this kind of
7:47
research.
7:47
And so, we kind of felt like we
7:49
had a little bit of lightning in the
7:50
bottle there.
7:51
And that's how the project really got its
7:54
footing.
7:55
I love that.
7:56
And I love sort of that idea of
7:58
how quickly people have been adopting AI and
8:01
sort of that perspective of what's going on
8:05
but I'd love for you to kind of
8:07
set the scene in terms of like what
8:09
is AI and mental health actually looking like
8:12
right now in the real world?
8:14
And why should clients, clinicians, practice owners, anyone
8:19
in the industry be paying attention?
8:22
Yeah, it's a big question, but the reality
8:25
is that AI is one, it's many things,
8:30
right?
8:30
So, it's difficult to say exactly what we're
8:32
talking about at any given time.
8:34
Because when you say AI and mental health,
8:36
sometimes you mean, you know, the kinds of
8:40
programs that administratively support the practice of medicine
8:46
or therapy.
8:47
You could even be talking about the chatbots
8:49
that mental health websites might use to triage
8:53
messages that they get, that kind of thing.
8:55
Sometimes you might mean programs that prescribe psychotropic
8:59
medications.
9:00
There's a pilot right now in Utah through,
9:03
I think it's Legion Health, that they have
9:05
a sandbox program there where they're allowing an
9:08
AI to refill psychiatric prescriptions, for example.
9:13
You might be talking about wellness chatbots.
9:17
You might be talking about the kinds of
9:19
bots that track emotions or metadata to try
9:24
and identify the way that people are feeling.
9:26
And some states have thought about it that
9:28
way, too.
9:29
So, it's just a very broad kind of
9:32
thing that you're talking about when you say
9:33
AI and mental health.
9:34
The way that we were looking at it
9:36
for the paper was very broad.
9:38
When we took the broadest position, which was
9:40
just that anything that we thought would be
9:43
related to both AI and to mental health,
9:46
you know, would have an expected impact, even
9:48
if it wasn't really an intended impact.
9:51
Okay.
9:52
The reality is, okay.
9:53
No, I like that.
9:55
I'm just, I'm tracking.
9:56
Yeah.
9:57
The reality also is that our industry doesn't
10:01
really know how to corral and define these
10:04
things.
10:04
And certainly, legislatures and rulemakers are having a
10:07
hard time doing this.
10:08
And they have even less, you know, clinical
10:10
knowledge to go off of and very little
10:12
guidance from us on the practitioner side on
10:15
how to do that as well.
10:17
And so, there's messy definitions that are coming
10:21
up in legislation and policy and things like
10:24
that, that has been because there's kind of
10:28
a void instruction or guidance on what to
10:31
do.
10:31
AI is so useful, potentially.
10:34
You know, we think that it's going to
10:35
be so useful in these spaces that it's
10:37
already disrupting them.
10:38
But we have very little data to go
10:41
off of.
10:41
We don't, I can't definitively sit here and
10:44
tell you like I could for an antidepressant,
10:46
for example, that there is strong evidence that
10:49
AI is helpful in mental health or harmful
10:53
in mental health.
10:54
You know, I'm certainly familiar with some of
10:56
the smaller studies and the time-limited studies
10:58
that we have so far.
10:59
I'm certainly cognizant of, you know, the numerous
11:02
tragic stories and alarming stories that we see
11:05
from the headlines and whatnot.
11:06
But we don't have hard public health data
11:09
on what that looks like.
11:10
And that means that the people who are
11:11
making the policies also don't have that data.
11:14
They're public health experts that are, you know,
11:17
from their departments of public health and whatnot
11:19
at the state level or their local, you
11:22
know, chapters of the APA or other kinds
11:26
of professional organizations.
11:27
They also don't know really what to recommend
11:29
at this point.
11:30
And, you know, so they're not promulgating very
11:32
definitive guidelines, if at all.
11:35
And so it does leave you in the
11:37
nation in this position where everything is very
11:39
fragmented.
11:40
People know there's something that they want to
11:42
do about it.
11:43
The constituents want their leadership to do something
11:46
about it.
11:47
But there's not really that much in the
11:49
way of data or specific guidance that we
11:52
can offer to them right now.
11:54
And so I don't think it's any surprise
11:55
that as we did this review, states were
11:58
really all over the place.
11:59
And really only a handful of states out
12:01
of the 50 were really, you know, actually
12:04
enacting anything at this time.
12:06
Yeah.
12:07
And I think it's interesting that you're bringing
12:09
that up because I think that has been
12:11
a theme on this podcast as we have
12:13
talked to more and more people to try
12:15
to get a handle on things is that
12:18
there are a lot of differing views.
12:20
There's a lot of confusion.
12:22
There's a lot of lack of clarity.
12:25
I know there was a bill in the
12:26
state I'm in, in Maryland, that did not
12:28
pass this time around.
12:30
And it was simply to put a notice
12:32
so that people knew that AI was not
12:35
a trained medical or licensed mental health professional
12:38
and that didn't go through the Senate.
12:40
So it's interesting to think about, like, when
12:43
I see how many states introduced bills and
12:46
then we have 11 states that actually enacted
12:50
them, that also speaks to what you're saying.
12:53
It's hard to know what to do.
12:56
And I think I'm hearing a lot.
12:58
There's this idea of we know that we
13:01
need some safety guardrails.
13:03
We know that ethics can be a problem.
13:05
We know that we have, like you say,
13:07
the lawsuits that we're all aware of.
13:10
And at the same time, there's a lot
13:12
of people moving for technical advancement and we
13:15
don't want to limit that technical advancement.
13:18
So I'm just curious from where you sit
13:20
and the things that you've looked at, how
13:23
do you see a good balance between those
13:27
two things or what do you think is
13:28
possible there?
13:29
That's a really good question.
13:32
And it has probably an answer that I
13:35
don't know, to be honest with you.
13:38
It's such a multifaceted industry.
13:41
The one thing that I do know is
13:43
that probably the current regulatory frameworks that we're
13:46
relying on are inadequate, you know, in some
13:48
ways.
13:49
So, for example, I think that one route
13:51
that we would tend to think of for
13:54
this kind of technology is that we would
13:55
go the FDA route, that these are medical
13:57
technology, therefore they're medical devices, and they could
14:00
be regulated under the same way.
14:02
In reality, the FDA's scope of what they
14:04
can actually do and what they regulate is
14:07
quite confined.
14:09
And so many of these products don't fall
14:12
within the FDA's purview.
14:14
And to be honest, the FDA pathways that
14:17
they created weren't really meant for this kind
14:19
of technology, nor I don't think that we
14:21
would necessarily want them to be, because it
14:24
could be overly restrictive as well.
14:26
At the same time, states don't really have
14:29
a way to address these things either.
14:31
One, they're still broadly defining AI in this
14:34
like really catch-all way, instead of kind
14:37
of stratifying them based off of their actual
14:40
individual risk based off the services that they
14:42
provide, which might be one way to go
14:44
about it, which is, you know, to say,
14:47
let's actually take a look at each individual
14:49
AI product and then give it a risk
14:51
classification, for example, and then decide from there
14:54
where that belongs in the bucket of regulations
14:57
and whatnot.
14:58
So that's another option.
14:59
There are many different ways that states could
15:02
approach this.
15:03
The other challenge here is that traditionally, you
15:06
know, you would think maybe the federal government
15:09
should have a primary or central role in
15:12
regulating these things, even on the clinical side,
15:16
right?
15:16
And I think that there's been some literature
15:18
and some commentary towards that.
15:21
Realistically, I actually think that that is difficult
15:23
to do, because clinical practice and the regulation
15:27
and oversight at the clinical level is almost
15:30
always a state function.
15:32
States issue licenses, states grant clinicians, you know,
15:37
those kinds of abilities to provide their licensed
15:40
services.
15:40
That's not really in the federal government's purview.
15:43
And I don't actually think that most states
15:44
are interested in having the federal government involved
15:47
at that level.
15:48
But, you know, another way to look at
15:49
that could be maybe we're looking at that
15:51
wrong, too.
15:52
Should we be thinking about these AIs as
15:55
providing clinical services at all?
15:58
That's a legitimate question as well.
16:01
Or is it some other category of thing
16:03
with its own bucket of regulations that it
16:06
should have?
16:07
This is a long-winded way of saying
16:08
that I don't know.
16:09
Yeah.
16:10
No, but I like all the thought-provoking
16:12
points along the way.
16:13
Yeah.
16:14
Yeah.
16:14
There is a balance that we will get
16:18
to eventually.
16:20
And the tried-and-true method in the
16:23
United States is typically to go through this
16:26
process with any new technology of states experimenting
16:30
with different ways of doing this.
16:32
And then courts finding, you know, through litigation
16:35
and a lot of like painful process, figuring
16:39
out what it is that actually will work
16:41
for people.
16:41
And so that would be my prediction of
16:44
what happens.
16:45
But it remains to be seen.
16:46
We've also seen, for example, I think in
16:48
December, the administration executive branch issued an executive
16:54
order that said that they would revoke certain
16:57
types of broadband funding.
16:58
It states overstepped this line of regulating artificial
17:03
intelligence companies.
17:05
And it left that pretty broad.
17:08
You know, it's unclear exactly what they mean
17:10
by that, but we already know that it
17:11
probably stopped Colorado from doing their law.
17:16
And then Virginia has also had a very
17:18
similar law to books that also hasn't gone
17:21
anywhere.
17:21
And so, you know, there's so many things
17:24
that are unclear about where we're going right
17:26
now that it's hard to predict what will
17:29
end up actually being realistic for states to
17:31
do.
17:32
Yeah.
17:33
Yeah.
17:33
I'm thinking about the lens that you're bringing
17:36
of both the legislation and the legal piece,
17:39
as well as the physician medical piece.
17:42
Do those two areas of training ever feel
17:46
like they're in conflict for you as you're
17:48
thinking through some of these issues?
17:50
And if so, where?
17:51
I do think that they do sometimes because
17:55
I really do consider myself a clinician first.
17:58
And so much of my, you know, we
18:01
were pretty guarded about it in the paper,
18:03
of course.
18:04
But on a personal level, I do feel
18:07
like I have a most clinicians would probably
18:10
feel this way, too.
18:11
I imagine that your prerogative is to protect
18:14
the patient.
18:15
Yeah.
18:15
And it is to protect that relationship and
18:18
to help people wherever you're wherever you can.
18:22
Right.
18:22
And that's because you have an individual stake
18:25
in the person that you're treating.
18:28
Sometimes the law works more at a population
18:30
level.
18:31
And so individuals might be overlooked sometimes in
18:36
the way that, you know, public policy is
18:39
oriented.
18:40
And so that's sometimes a strain because, you
18:43
know, it's just like this topic that we
18:45
were talking about of not really being able
18:47
to definitively say that AI is bad for
18:49
mental health.
18:50
Well, it's hard to, you know, hard to
18:52
say that at the same time somebody is
18:54
talking to you about how they were individually
18:56
harmed by, for example, or a family member
18:59
was individually harmed by it.
19:00
It's hard to say on the one hand
19:03
as a therapist and psychiatrist to say, well,
19:07
maybe there should be an exemption in a
19:09
law to allow for certain types of therapeutic
19:12
or research uses of artificial intelligence, because it
19:17
might actually be useful.
19:18
It might actually improve clinical outcomes.
19:20
And wouldn't that be wonderful?
19:22
Of course.
19:24
But at the same time, there's, you know,
19:26
probably, you know, from a self-serving perspective,
19:28
too, I'm terrified of the idea that AI
19:31
might come in and be really good at
19:33
providing clinical care.
19:35
You know, I don't know where we're going
19:37
to be with the technology in a few
19:38
years.
19:39
But, you know, these are definitely things that
19:42
I feel conflict over.
19:43
It is.
19:44
Yeah.
19:45
Yeah.
19:45
No, I hear you.
19:46
And I don't know that there is a
19:49
direct answer.
19:50
And I resonate with that clinical piece of,
19:54
right, where we're advocating for the individual whom
19:57
we're treating and we don't, I mean, I
19:59
kind of heard you say there might be
20:00
some trial and error with different legislation and
20:05
lawsuits and things like that that kind of
20:07
roll out into me.
20:08
I immediately hear that those are people's lives
20:10
that are going to be harmed, which is
20:12
what's bringing that to the table.
20:14
And do we have to go through that?
20:17
Maybe we do.
20:18
But it's certainly not best case scenario.
20:20
I look at our training and I'm sure
20:22
my training was different slightly than yours.
20:24
But some of the commonality is that we
20:27
both have training to be able to work
20:30
with people in this way.
20:32
Right.
20:32
And there are also ethics that we have
20:35
to ascribe to.
20:36
And there are licensure regulations that we have
20:39
to ascribe to.
20:40
And to my knowledge, an AI therapist that
20:43
does exist out there doesn't have to ascribe
20:46
to any of those things.
20:48
I think that is a tremendous point because
20:51
at most they, so there are two ways
20:54
that some states have, and I'm not saying
20:56
this is a broad category of states.
20:59
We're talking about two, like two or three
21:00
states, you know, at this point.
21:02
But, you know, some have the idea that,
21:05
well, we tie it to the licensed clinician
21:07
that's providing the service.
21:09
Right.
21:09
So there's a licensed clinician overseeing whatever that
21:12
means, the AI service, then that brings with
21:17
it all of the trappings of our existing
21:20
self-regulating system.
21:22
It brings with it that clinician's license, however
21:26
we regulate them, their professional and ethical obligations,
21:30
et cetera, and so on.
21:31
Another way that states have tried to look
21:34
at this is where you classify it as
21:38
a, essentially a licensed, what would be a
21:41
licensed service if it was provided by a
21:43
human being.
21:44
And then you say the same standard of
21:47
care applies to that.
21:49
Does it really, because, you know, these are
21:51
really very different things.
21:52
Like you were saying, and this is one
21:54
of the things that I really think is
21:56
different about artificial intelligence in the therapeutic space
22:00
versus humans, is that when we talk about
22:03
accountability or a licensed clinician, it presupposes that
22:07
that person has a deep stake in the
22:13
other person and in their own professional and
22:16
personal lives.
22:17
Right.
22:17
Because we've assumed this entire time that it's
22:19
humans.
22:19
You know, when you see a patient, it's
22:22
not just that you're having that individual clinical
22:26
encounter with them and then you're going to
22:28
disappear afterwards.
22:29
You have an entire lifetime of morals, of
22:33
ethical obligations, your future career, your relationship with
22:37
the patient and the expectations between the two
22:39
of you and the trust that you've built
22:40
between each other.
22:41
You have your professional obligations.
22:44
You know, you're looking at whether or not
22:45
you might be sued later, and that would
22:47
be devastating for you.
22:48
You would be potentially, you know, lose your
22:51
license or you could be prosecuted or any
22:54
of these kinds of things.
22:55
Or you might feel guilty about what happened
22:57
or remorseful about what happened.
22:59
But for an AI, when we talk about
23:01
accountability in this space, we are only talking
23:05
about accountability really at the business level, because
23:08
the only people who can be held accountable
23:10
in a realistic sense are the people who
23:13
are deploying, you know, selling or developing the
23:18
AI system.
23:19
The AI can't be held responsible because it
23:22
doesn't exist as a person.
23:24
It won't feel any kind of guilt or
23:28
anything like that.
23:29
In most of these systems, the instant that
23:32
they're turned off, they disappear.
23:34
And so, you know, that stake isn't there
23:37
between them.
23:39
And that's something that is a little uncomfortable
23:41
when we're trying to essentially say that we
23:43
should somehow hold these systems accountable, because I
23:46
think it's just a different type of accountability.
23:49
Yeah, yeah, that is that is interesting.
23:52
But it also, to me, brings to the
23:55
level that the question then, can a non
24:00
-human clinician do human work?
24:04
Like, yeah, it is that question.
24:05
Where's the human work?
24:06
Where's the non-human work?
24:07
What is what we do, like you say,
24:09
being replaced potentially by AI?
24:12
Or is that even possible?
24:14
Yeah, I don't know.
24:15
You know, we'll see where the technology is.
24:17
I can tell you that you should be
24:19
careful, and anybody should be careful, about reading
24:22
the AI research right now on what they
24:25
can and can't do, and look deeply at
24:28
the methodology that is portrayed.
24:31
Because you'll see headlines all the time right
24:33
now about, you know, so-and-so therapy
24:35
bot was just as effective, or was actually,
24:38
you know, better liked by the test subjects
24:43
compared to real science.
24:46
But you've seen some articles like that.
24:48
I saw one about marriage therapy recently.
24:51
Yeah, right.
24:52
But if you actually look at their methodology,
24:56
and I'm not saying this makes them bad
24:57
at papers or anything like that, just be
24:59
careful about it.
25:00
Because many, if not most of these papers
25:03
and studies, they're looking at things like single
25:06
-turn experiments, you know, where they're having people
25:11
rate based off their experiences of asking, you
25:13
know, essentially one prompt or one question, and
25:16
then getting back a single response.
25:17
Or at most, you know, like a very
25:19
short, abbreviated interview or session with a chatbot.
25:25
Because right now, if you get beyond, you
25:27
know, a few pages worth of context, they
25:31
get really wacky.
25:32
Then your results will be very different, right?
25:35
But you could foresee that maybe that won't
25:37
be a problem forever.
25:40
I do think that there's something else that's,
25:42
you know, and this is a personal position,
25:44
it's not borne out by any specific data.
25:47
Although there was a recent paper, which I
25:49
remember what it was off the top of
25:51
my head, that was taking a look at
25:53
what is it, you know, are humans essentially
25:56
just a kind of token generation program also,
25:59
like the way that our minds work?
26:00
Are we just predicting what would be the
26:02
next good word to say based off of
26:05
our experiences and all that kind of stuff?
26:07
And they were finding that that wasn't true.
26:10
And to me, that makes sense.
26:11
Because for anybody who's a clinician, and anybody
26:15
I think that has ever had just a
26:17
close conversation with a friend or family member,
26:20
you are pulling on so much information in
26:25
your mind, your past experiences, you're judging how
26:27
the other person is feeling, you are thinking
26:31
about how that conversation is going to show
26:33
up again, in a couple of sessions, and
26:37
where the arc of your relationship and where
26:40
the arc of your treatment with them is
26:41
going.
26:42
And I have a really hard time imagining
26:44
AI ever actually being very good at that.
26:47
But I could do it all.
26:49
Yeah.
26:49
I also think about the piece of, you
26:52
know, reading so many things than just the
26:54
language, like tone of voice, like facial expression,
26:58
like body language, so much of the data
27:01
is there for a clinician, right?
27:05
And I don't know that AI at this
27:08
point really tunes into those pieces.
27:11
I think that it's getting there.
27:13
And it is something that states are paying
27:15
attention to as well.
27:16
I think even the Illinois Walker Act, explicitly
27:20
says that you can't record or you can't
27:24
use AI to monitor emotional and behavioral data
27:28
like that.
27:29
And, you know, it's an open question in
27:30
other states.
27:31
And I think this is a regulatory issue,
27:33
is, do you want AI to be able
27:36
to do that?
27:37
Or do you, which is a legitimate policy
27:39
question, do you, but, you know, we're using
27:41
it in constructive ways, too.
27:43
We have research going on even here right
27:45
now, that is trying to objectively measure, you
27:51
know, emotional and, you know, mood symptoms and
27:54
other kinds of biometric data based off of
27:56
interventions.
27:58
And these are right now considered worthy clinical
28:01
kinds of data to gather.
28:04
You could easily see states, I think even
28:07
the Illinois Act, for example, doesn't have a
28:09
research exemption.
28:10
And so we would say that that kind
28:12
of research is not okay in their state.
28:15
Although I think in the Illinois Act, for
28:18
whatever reason, I haven't investigated why they specifically
28:21
exempt physicians.
28:23
So it'd be potentially against the law for
28:25
you to do it, That's interesting, too.
28:28
Okay.
28:29
Yeah.
28:29
Well, I think that we are running out
28:33
of time.
28:33
But I think my one question for you
28:35
is, maybe clinicians specifically, but also any patients
28:40
or clients, what is the number one thing
28:43
that you think people need to be asking
28:46
or to be watching out for in this
28:50
realm as we go forward?
28:52
I think there are two related things on
28:55
this.
28:55
So and both have to do with a
28:57
kind of disclosure.
28:59
And so probably the most common type and
29:02
least controversial, although you gave a counterexample at
29:06
the beginning, I think, of these kinds of
29:08
laws coming out is the requirement to disclose
29:12
AI use and disclosure in many different parts
29:17
of your provision of services.
29:19
So if you are in a state that
29:21
has one of these kinds of laws, and
29:22
you're providing services, and you're using AI note
29:25
takers, or if you're using AI to help
29:28
summarize things, or you're using AI to help
29:30
provide certain kinds of communications, you need to
29:32
look at your state laws and make sure
29:34
that, you know, some of these are actually
29:36
enacted at the beginning of the year, some
29:37
of them are enacting this fall, some are
29:39
enacting next year, you need to check and
29:41
make sure that you're not violating the law
29:44
of, you know, needing to disclose some of
29:47
this information to your patients, and also potentially
29:52
getting their informed consent for that.
29:54
And you need to start thinking about what
29:56
does informed consent look like?
29:58
Because if you