What happens when AI teaches us how to love?

WEBVTT
1
00:00:01.120 --> 00:00:03.320
It's estimated that up to 75%
2
00:00:03.320 --> 00:00:06.520
of American teenagers
currently use AI chatbots,
3
00:00:07.420 --> 00:00:09.700
and about a third are
using these things daily.
4
00:00:11.300 --> 00:00:13.180
And increasingly, they're
not just using them
5
00:00:13.180 --> 00:00:16.199
to summarize chapter
three of their homework.
6
00:00:16.199 --> 00:00:19.360
No, they're pouring their most deep,
7
00:00:19.360 --> 00:00:22.000
intimate thoughts into these things.
8
00:00:23.600 --> 00:00:26.220
I feel ugly today.
9
00:00:26.220 --> 00:00:28.620
I don't think my girlfriend
loves me anymore.
10
00:00:28.620 --> 00:00:32.280
How do I meet girls? Someone
said this about me online.
11
00:00:32.280 --> 00:00:34.040
What should I do? I feel like my
12
00:00:34.040 --> 00:00:36.040
parents are disappointed in me.
13
00:00:36.040 --> 00:00:40.440
I'm so confused. I wish
I didn't feel this way.
14
00:00:42.240 --> 00:00:45.440
For most of human history,
if teenagers were willing
15
00:00:45.440 --> 00:00:47.960
to share these thoughts, they
did so with other people,
16
00:00:48.920 --> 00:00:52.000
friends, siblings, teachers, parents,
17
00:00:52.880 --> 00:00:54.740
really whoever was willing to listen.
18
00:00:56.300 --> 00:00:58.840
Now, AI chatbots are becoming the go-to
19
00:00:58.840 --> 00:01:00.479
confidante for some teens.
20
00:01:00.479 --> 00:01:03.360
And today's guest, Thao Ha,
thinks that might affect
21
00:01:03.360 --> 00:01:06.520
how they develop lifelong
relationship skills.
22
00:01:06.520 --> 00:01:08.600
Thao is an associate
professor in the Department
23
00:01:08.600 --> 00:01:11.400
of Psychology at Arizona State University,
24
00:01:11.400 --> 00:01:13.980
where she directs the
Healthy Experiences Across
25
00:01:13.980 --> 00:01:17.460
Relationships and Transitions
lab, or HEART lab.
26
00:01:17.460 --> 00:01:20.940
She studies how adolescent
relationships form
27
00:01:20.940 --> 00:01:23.680
and how they affect emotional wellbeing,
28
00:01:23.680 --> 00:01:27.000
identity development,
academic success and more.
29
00:01:28.020 --> 00:01:30.360
Romantic relationships during
adolescence is one of the
30
00:01:31.280 --> 00:01:34.720
foundational learning experiences about
31
00:01:34.720 --> 00:01:36.240
who you are in a relationship
32
00:01:36.240 --> 00:01:38.260
and what you can be for someone else,
33
00:01:38.260 --> 00:01:41.200
and all the skills that go with that.
34
00:01:41.200 --> 00:01:44.460
And until recently, all of
that learning was happening
35
00:01:44.460 --> 00:01:46.620
with other human beings.
36
00:01:46.620 --> 00:01:50.060
It actually becomes what I call a
37
00:01:50.980 --> 00:01:52.740
socializing agent,
38
00:01:52.740 --> 00:01:55.420
and AI is actually taking a similar kind
39
00:01:55.420 --> 00:01:58.460
of function in a young person's life.
40
00:01:58.460 --> 00:02:02.199
So they are learning about themselves
41
00:02:02.199 --> 00:02:04.420
through interacting with an AI.
42
00:02:04.420 --> 00:02:07.740
From Arizona State University,
this is Lab Coat Optional,
43
00:02:07.740 --> 00:02:10.500
the show where we bring
research out of the lab
44
00:02:10.500 --> 00:02:12.340
and into your headphones.
45
00:02:12.340 --> 00:02:13.820
I'm your host, Pete Zrioka.
46
00:02:13.820 --> 00:02:17.520
And this week, what happens
when machines start teaching
47
00:02:17.520 --> 00:02:18.900
us how to love?
48
00:02:19.880 --> 00:02:21.880
Hello, Pete. How are you feeling today?
49
00:02:36.560 --> 00:02:37.940
Thao, welcome to the show.
50
00:02:37.940 --> 00:02:40.080
Thank you.
I'm really excited to talk
51
00:02:40.080 --> 00:02:43.660
to you today about AI and love.
52
00:02:43.660 --> 00:02:48.480
So a focus of your research
right now is teen use of AI
53
00:02:48.480 --> 00:02:51.760
and specifically adolescent romantic
54
00:02:51.760 --> 00:02:53.240
relationships and the effect.
55
00:02:53.240 --> 00:02:55.960
Why romantic relationships
are so important is
56
00:02:55.960 --> 00:02:59.260
because they require all of your effort.
57
00:02:59.260 --> 00:03:01.400
It's different from friendships.
58
00:03:01.400 --> 00:03:03.380
Friendships are, you learn a lot in
59
00:03:03.380 --> 00:03:04.960
adolescent friendships too, right?
60
00:03:04.960 --> 00:03:07.760
So both of your friendships
61
00:03:07.760 --> 00:03:09.980
and your romantic relationships
kind of have the function
62
00:03:09.980 --> 00:03:12.740
for you to develop more
autonomy from your parents.
63
00:03:12.740 --> 00:03:15.960
So that's kind of like a
normative kind of transition
64
00:03:15.960 --> 00:03:20.960
that you go through, but the
difference between the romantic
65
00:03:20.980 --> 00:03:24.860
and the peer relationship is
really the emotional intensity
66
00:03:24.860 --> 00:03:26.500
that you can feel with a romantic partner.
67
00:03:27.680 --> 00:03:30.740
All The positives and all the negatives,
68
00:03:30.740 --> 00:03:33.360
it goes really, really
deep and it goes quick.
69
00:03:33.360 --> 00:03:35.240
Those emotions go really, really quick.
70
00:03:35.240 --> 00:03:39.980
So it really requires all of your efforts
71
00:03:39.980 --> 00:03:42.580
and that's why it's such
a great learning context
72
00:03:42.580 --> 00:03:45.300
for your social skills.
73
00:03:45.300 --> 00:03:49.000
And the way we learn is
with other people, right?
Yes.
74
00:03:49.000 --> 00:03:50.720
And until very recently,
75
00:03:50.720 --> 00:03:51.840
that was the only way you could kind
76
00:03:51.840 --> 00:03:55.720
of build these relationship
skills, but that's changing.
77
00:03:55.720 --> 00:03:57.500
It's changing quite a bit. So first we
78
00:03:57.500 --> 00:03:59.100
had social media, right?
79
00:03:59.100 --> 00:04:00.239
And we still have social media.
80
00:04:00.239 --> 00:04:02.420
So social media is changing how
81
00:04:03.380 --> 00:04:04.979
teens are engaging in their relationships
82
00:04:04.979 --> 00:04:06.220
and how are they communicating
83
00:04:06.220 --> 00:04:08.900
and how misunderstandings are happening,
84
00:04:08.900 --> 00:04:11.480
but also how you display love
85
00:04:11.480 --> 00:04:12.940
and how you communicate with one another.
86
00:04:12.940 --> 00:04:15.140
So social media is definitely one of the,
87
00:04:15.140 --> 00:04:18.420
is a very important context,
but it is a context.
88
00:04:19.540 --> 00:04:22.480
So it provides literally
a different platform
89
00:04:22.480 --> 00:04:27.120
and many, many platforms for
you to connect with a partner.
90
00:04:27.120 --> 00:04:29.280
One example is you can be
connected 24/7.
91
00:04:29.280 --> 00:04:32.600
That changes, you can't
just close the door anymore
92
00:04:32.600 --> 00:04:33.900
and not see each other, right?
93
00:04:33.900 --> 00:04:34.920
You're still connected.
94
00:04:35.980 --> 00:04:38.120
Or you can communicate without seeing each
95
00:04:38.120 --> 00:04:40.060
other's facial expressions.
96
00:04:40.060 --> 00:04:44.100
And so social media
changes our - how we express
97
00:04:44.100 --> 00:04:45.300
and interpret our communication,
98
00:04:46.280 --> 00:04:50.060
but you're still dealing
with other people.
99
00:04:51.180 --> 00:04:52.760
And in our relationship,
we're still dealing
100
00:04:52.760 --> 00:04:54.080
with our partner.
101
00:04:54.080 --> 00:04:58.540
And now AI is entering this picture
102
00:04:58.540 --> 00:04:59.640
for teens, for young people.
103
00:05:01.400 --> 00:05:05.520
And AI is a completely different entity.
104
00:05:05.520 --> 00:05:07.660
It's no longer just a platform.
105
00:05:07.660 --> 00:05:11.760
They are learning about
themselves through interacting
106
00:05:11.760 --> 00:05:14.440
with an AI and -
107
00:05:14.440 --> 00:05:15.820
Rather than people.
108
00:05:15.820 --> 00:05:19.180
Rather than people. So
the AI is impacting them
109
00:05:19.180 --> 00:05:22.580
And they are in some
ways also impacting the AI.
110
00:05:22.580 --> 00:05:25.860
I'm mostly talking about
the platforms that allow you
111
00:05:25.860 --> 00:05:27.320
to have interactions with AI.
112
00:05:27.320 --> 00:05:28.960
So like LLMs?
Yes.
113
00:05:28.960 --> 00:05:30.740
Okay.
Yeah. ChatGPT,
114
00:05:30.740 --> 00:05:33.160
Replica - Claude.
115
00:05:33.160 --> 00:05:34.600
Claude, all of those.
116
00:05:35.980 --> 00:05:39.540
So you wrote an essay in The Lancet
117
00:05:39.540 --> 00:05:41.640
that was published in
June that looked at this.
118
00:05:41.640 --> 00:05:45.360
Yeah. - I would love to
hear about the origin of that
119
00:05:45.360 --> 00:05:47.200
and what surprised you about
120
00:05:47.200 --> 00:05:50.020
how teens were using AI in this capacity.
121
00:05:50.020 --> 00:05:55.020
Yeah. So we have a very large
122
00:05:55.080 --> 00:05:59.880
study that is funded by the
National Institutes of Health,
123
00:06:01.180 --> 00:06:04.540
and that study looks at how
124
00:06:04.540 --> 00:06:07.460
emerging technologies are changing how
125
00:06:08.400 --> 00:06:11.180
young people are connecting,
especially when it comes
126
00:06:11.180 --> 00:06:13.200
to their romantic relationships
and their friendships
127
00:06:13.200 --> 00:06:16.660
and how that's changing
their mental health.
128
00:06:16.660 --> 00:06:20.540
So we're looking at depression,
anxiety, loneliness,
129
00:06:20.540 --> 00:06:22.860
and suicide ideation.
130
00:06:23.860 --> 00:06:27.040
And so as part of that larger study,
131
00:06:28.580 --> 00:06:30.620
we have a youth advisory board.
132
00:06:30.620 --> 00:06:34.960
And so currently we
have seven young people.
133
00:06:34.960 --> 00:06:37.580
They're between 13 and
17 years old from the
134
00:06:37.580 --> 00:06:40.080
community who advise us.
135
00:06:40.080 --> 00:06:44.440
And they're really central
in my research, actually.
136
00:06:44.440 --> 00:06:49.400
So they give us their
thoughts, their feedback.
137
00:06:49.400 --> 00:06:54.320
So feedback about, you
name it, like our surveys
138
00:06:54.320 --> 00:06:57.020
or we're recruiting young people right now
139
00:06:57.020 --> 00:06:58.260
to participate in the study.
140
00:06:58.260 --> 00:07:02.380
They will comment on
our recruitment flyers
141
00:07:02.380 --> 00:07:04.760
or how we can best recruit
people, where we can
142
00:07:04.760 --> 00:07:06.280
recruit young people.
143
00:07:06.280 --> 00:07:09.440
They will watch my
public performances even,
144
00:07:09.440 --> 00:07:12.320
and they will comment on
what it is that I said
145
00:07:12.320 --> 00:07:14.311
and what I could have done better.
146
00:07:14.311 --> 00:07:16.100
You're getting a lot of input.
147
00:07:16.100 --> 00:07:18.340
They give a lot of
input and it's fantastic.
148
00:07:18.340 --> 00:07:21.520
It really improves
every step of our research.
149
00:07:22.360 --> 00:07:24.460
And it makes - it's really important to me
150
00:07:24.460 --> 00:07:27.220
because I want to make
sure it speaks to them.
151
00:07:27.220 --> 00:07:28.120
Yeah, it's approachable.
152
00:07:28.120 --> 00:07:29.180
It's approachable and it speaks to them.
153
00:07:29.180 --> 00:07:30.260
And you're speaking the vernacular.
154
00:07:31.219 --> 00:07:33.520
My research is all about young people.
155
00:07:33.520 --> 00:07:35.780
And so when I speak publicly about it,
156
00:07:35.780 --> 00:07:38.000
I want them to feel heard.
157
00:07:38.000 --> 00:07:40.720
Generally, I do a good
job, so they have been,
158
00:07:41.960 --> 00:07:44.700
but there are nuances
that I can improve on.
159
00:07:44.700 --> 00:07:48.880
And so as part of this youth
advisory board who we meet
160
00:07:48.880 --> 00:07:52.140
with every month, the
topic AI naturally came up.
161
00:07:53.539 --> 00:07:56.500
And I have to give a shout-out
to my graduate student, Jen,
162
00:07:56.500 --> 00:07:59.160
because she actually leads
this portion of our research.
163
00:07:59.160 --> 00:08:01.020
So she's the one who's
actually communicating
164
00:08:01.020 --> 00:08:03.100
with the young people.
165
00:08:03.100 --> 00:08:07.620
I don't do that because
she's more relatable to them.
166
00:08:09.180 --> 00:08:12.839
And AI came up in the conversations
167
00:08:12.839 --> 00:08:16.120
and they just loved talking about it.
168
00:08:16.120 --> 00:08:19.700
It was really something that
was on their mind a lot.
169
00:08:19.700 --> 00:08:22.339
And I thought it was
more about social media,
170
00:08:22.339 --> 00:08:25.960
but when we talked about AI,
they had so many thoughts.
171
00:08:25.960 --> 00:08:29.480
And I was like, "We
172
00:08:30.380 --> 00:08:33.340
should write a paper
together about what it is
173
00:08:33.340 --> 00:08:38.179
that you're worried about."
And so I suggested that to them
174
00:08:38.179 --> 00:08:39.580
and they loved the idea.
175
00:08:39.580 --> 00:08:44.100
And so that's kind of how
this paper came about.
176
00:08:44.100 --> 00:08:47.180
I mean, I love how your
co-authors on this paper are
177
00:08:47.180 --> 00:08:48.180
Susanna
and Jessica.
178
00:08:49.260 --> 00:08:53.180
I'm so curious, you talked about
the perspective they bring,
179
00:08:53.180 --> 00:08:56.580
but what are some of the things
180
00:08:56.580 --> 00:09:00.440
that they were bringing to
this discussion that you think
181
00:09:00.440 --> 00:09:03.540
adults are missing about
how teenagers talk to AI?
182
00:09:04.420 --> 00:09:06.620
I'm in a lot of spaces in
which I'm with other researchers
183
00:09:06.620 --> 00:09:08.660
or I'm with policy folks
184
00:09:08.660 --> 00:09:12.120
or governance people who are trying
185
00:09:12.120 --> 00:09:13.640
to governance this whole thing.
186
00:09:15.380 --> 00:09:17.380
And it gets abstract really quickly
187
00:09:17.380 --> 00:09:19.440
and we actually lose connection to
188
00:09:19.440 --> 00:09:21.800
why teens are even engaging
in the first place.
189
00:09:22.800 --> 00:09:26.180
And so what a youth perspective
190
00:09:26.180 --> 00:09:31.180
provided was how does this play
out in your day-to-day life?
191
00:09:31.180 --> 00:09:32.580
And what does this mean to you?
192
00:09:33.600 --> 00:09:36.440
And what are your alternatives?
193
00:09:36.440 --> 00:09:38.420
And what are you worried about?
194
00:09:38.420 --> 00:09:39.580
And what are you excited about?
195
00:09:39.580 --> 00:09:44.180
I think those are such
profound insights that they
196
00:09:44.180 --> 00:09:47.980
provided that I don't get from the adults
197
00:09:47.980 --> 00:09:49.760
who are thinking about it in this space.
198
00:09:49.760 --> 00:09:53.800
And so their voices were so unique.
199
00:09:53.800 --> 00:09:56.780
So when I was asking you
about why this developmental
200
00:09:56.780 --> 00:10:01.620
window is so important,
you mentioned the intensity
201
00:10:01.620 --> 00:10:03.900
of the emotions that we feel at that age.
202
00:10:04.760 --> 00:10:08.640
And then you were talking about
the depth of their use of AI
203
00:10:08.640 --> 00:10:11.000
and what using it means to them.
204
00:10:11.000 --> 00:10:14.080
So I assume that there's
some crossover there between
205
00:10:14.080 --> 00:10:15.420
how intense you're feeling
206
00:10:15.420 --> 00:10:17.920
and then how much the thing
is integrated into your life
207
00:10:17.920 --> 00:10:20.620
because I'm not a teenager anymore,
208
00:10:22.060 --> 00:10:25.000
and I have only ever used AI as a tool.
209
00:10:25.000 --> 00:10:28.540
I've never really interfaced
with it on a personal level.
210
00:10:28.540 --> 00:10:31.620
So I'm so curious about
the kind of questions
211
00:10:31.620 --> 00:10:33.280
that they're asking these chatbots.
212
00:10:34.180 --> 00:10:36.340
Do you have a sense of why
they don't feel comfortable
213
00:10:36.340 --> 00:10:39.320
asking some things of peers or parents?
214
00:10:39.320 --> 00:10:40.780
Are we looking at just like,
215
00:10:40.780 --> 00:10:42.840
is this just like a form of AI therapy?
216
00:10:45.000 --> 00:10:46.000
How are they talking to these things?
217
00:10:47.340 --> 00:10:50.100
So before I answer
that question - Please.
218
00:10:50.100 --> 00:10:53.580
I want to also highlight that why
219
00:10:54.480 --> 00:10:55.920
their perspectives are so important,
220
00:10:55.920 --> 00:10:58.200
why building relationships
with them is so important.
221
00:10:58.200 --> 00:11:01.280
Because if you ask any teen
right now just on the street
222
00:11:01.280 --> 00:11:03.840
or if you have not built a
trusting relationship with them,
223
00:11:03.840 --> 00:11:05.680
they will say, "I don't use it." This
224
00:11:05.680 --> 00:11:07.020
is also what we see in our data.
225
00:11:07.020 --> 00:11:10.880
So when we have survey
data from our participants
226
00:11:11.980 --> 00:11:13.540
who are also young people between 13
227
00:11:13.540 --> 00:11:16.380
and 17 years old, they'll say, "Oh,
228
00:11:16.380 --> 00:11:18.320
I use it a couple times a week." Right?
229
00:11:18.320 --> 00:11:21.480
Okay.
But in my current study, I'm also
230
00:11:23.360 --> 00:11:26.340
doing passive sensing on
their phones, which means
231
00:11:26.340 --> 00:11:29.400
that we are scraping their
phones for all, everything
232
00:11:29.400 --> 00:11:32.080
that they enter into their phones
233
00:11:32.080 --> 00:11:34.300
and we know which
platforms they have visited
234
00:11:34.300 --> 00:11:38.120
so I can actually see who has used what
235
00:11:38.120 --> 00:11:40.080
and entered what on which platform at
236
00:11:40.080 --> 00:11:41.300
what time and for how long.
237
00:11:41.300 --> 00:11:43.860
It's like that screen
time report on steroids.
238
00:11:43.860 --> 00:11:45.040
Yes, but it's not just screen time.
239
00:11:45.040 --> 00:11:47.000
I actually know what you
did during that screen time.
240
00:11:47.000 --> 00:11:48.840
That's super on steroids
then. Yes. Okay, cool.
241
00:11:48.840 --> 00:11:50.840
Yes. Yeah. So I actually
know what you typed.
242
00:11:50.840 --> 00:11:51.673
Ah.
So I
243
00:11:51.673 --> 00:11:52.506
Spooky.
244
00:11:52.506 --> 00:11:53.760
Can see the questions
245
00:11:53.760 --> 00:11:56.100
that you have entered into ChatGPT
246
00:11:56.100 --> 00:11:59.900
or any other kind of AI platform.
247
00:11:59.900 --> 00:12:02.420
Okay.
And so what we see is that
248
00:12:02.420 --> 00:12:04.780
what youth are self-reporting
249
00:12:04.780 --> 00:12:06.300
and what they are actually doing,
250
00:12:06.300 --> 00:12:08.820
there's a big disconnect between that.
251
00:12:08.820 --> 00:12:10.600
And that makes sense.
It's because people in
252
00:12:10.600 --> 00:12:11.927
general are not so good at.
253
00:12:11.927 --> 00:12:13.547
Yeah, you're not super
254
00:12:13.547 --> 00:12:14.380
forthcoming with something that might be -
255
00:12:14.380 --> 00:12:15.380
That too.
256
00:12:15.380 --> 00:12:18.200
And people have a hard time
having insight into behaviors.
257
00:12:18.200 --> 00:12:20.600
And so this is just another example of that.
258
00:12:21.440 --> 00:12:25.500
And I think young people,
when you first ask them,
259
00:12:25.500 --> 00:12:29.880
they really are against AI in some ways
260
00:12:29.880 --> 00:12:33.480
Because, it's for its effects on the climate
261
00:12:33.480 --> 00:12:36.720
and what it takes to have it running.
262
00:12:37.880 --> 00:12:41.940
But then when you dig
deeper, they do use it
263
00:12:41.940 --> 00:12:44.900
because of what young
people were saying is
264
00:12:45.800 --> 00:12:50.320
because it in some ways is
more reliable than a human
265
00:12:50.320 --> 00:12:52.460
because a machine knows everything.
266
00:12:53.540 --> 00:12:55.080
So they're assuming that
it knows everything.
267
00:12:56.300 --> 00:12:58.380
So that was one reason,
but then the other was
268
00:12:59.880 --> 00:13:02.160
it also doesn't judge you.
269
00:13:03.560 --> 00:13:07.200
And the teenage years are all about
270
00:13:08.080 --> 00:13:10.420
forming your identity
271
00:13:11.540 --> 00:13:13.460
and learning about yourself,
learning about the other.
272
00:13:14.380 --> 00:13:16.620
And those are very vulnerable
273
00:13:16.620 --> 00:13:18.520
and intimate questions sometimes.
274
00:13:18.520 --> 00:13:20.820
And so they are ashamed
275
00:13:20.820 --> 00:13:23.740
or they don't have anybody else to go to
276
00:13:23.740 --> 00:13:25.540
to ask these questions.
277
00:13:25.540 --> 00:13:26.660
So then they ask the machine
278
00:13:26.660 --> 00:13:29.140
because they feel like,
"Ah, I can trust the machine
279
00:13:29.980 --> 00:13:31.480
and it will have an answer
280
00:13:31.480 --> 00:13:35.760
and it won't think I'm weird or crazy."
281
00:13:35.760 --> 00:13:38.920
Is it because it's a
machine, so it's impartial?
282
00:13:38.920 --> 00:13:42.140
It's a machine, so it's non-judgmental.
283
00:13:42.140 --> 00:13:44.760
It's a machine because
it knows everything?
284
00:13:44.760 --> 00:13:47.740
Is it all that stuff wrapped up together
285
00:13:47.740 --> 00:13:48.573
or is that what you're seeing?
286
00:13:48.573 --> 00:13:50.080
Yeah.
287
00:13:50.080 --> 00:13:51.560
Weird.
288
00:13:51.560 --> 00:13:52.560
No, I don't think it's weird.
289
00:13:52.560 --> 00:13:54.460
Well, look, I can only
bring my own.. - Given
290
00:13:54.460 --> 00:13:56.340
the number of people who are using it,
291
00:13:56.340 --> 00:13:57.560
Yeah.
And how often
292
00:13:57.560 --> 00:13:59.060
and what kind of questions
people are asking,
293
00:13:59.060 --> 00:14:01.760
not just young people, I
don't think it's weird at all.
294
00:14:02.620 --> 00:14:05.020
I think it gives us a
platform where we can
295
00:14:07.220 --> 00:14:10.540
ask questions that we struggle
with in those hard moments
296
00:14:10.540 --> 00:14:11.700
that we cannot find somewhere
297
00:14:11.700 --> 00:14:13.860
else. I don't think that's weird at all.
298
00:14:13.860 --> 00:14:17.240
I mean, I guess I
understand the impulse.
299
00:14:17.240 --> 00:14:20.160
I just - I can only bring my own
perspective to things.
300
00:14:20.160 --> 00:14:23.760
And I have very hard time
imagining myself ever going
301
00:14:23.760 --> 00:14:25.460
to ChatGPT or Claude
302
00:14:25.460 --> 00:14:27.760
or anything like that for personal advice
303
00:14:27.760 --> 00:14:29.340
because this thing doesn't know
304
00:14:29.340 --> 00:14:33.220
me. It doesn't know me.
305
00:14:33.220 --> 00:14:36.200
But it's programmed to
imply that it knows you.
306
00:14:37.060 --> 00:14:38.600
So I think that's also a difference.
307
00:14:38.600 --> 00:14:41.200
Well, I guess I have
that mental barrier up
308
00:14:41.200 --> 00:14:42.460
with it then. I don't know.
309
00:14:42.460 --> 00:14:43.500
Yeah, maybe. Yeah.
310
00:14:45.080 --> 00:14:48.160
So you asked some of the
questions that they're asking.
311
00:14:48.160 --> 00:14:51.900
Yeah.
And so I can give you some examples
312
00:14:51.900 --> 00:14:54.320
that we've seen that I've heard from the teens
313
00:14:54.320 --> 00:14:56.220
and from my youth advisory board,
314
00:14:56.220 --> 00:14:57.340
but also that we see in our data.
315
00:14:58.300 --> 00:15:01.960
So they ask a lot of questions
about school, "Help me with..."
316
00:15:01.960 --> 00:15:04.880
So the entry would literally
be, "Help me with chapter three,
317
00:15:04.880 --> 00:15:09.800
summarize," or "What should
I learn from this chapter?"
318
00:15:10.780 --> 00:15:14.120
And I think this is also a
misperception for people.
319
00:15:14.120 --> 00:15:16.459
When they think about
teen AI relationships,
320
00:15:16.459 --> 00:15:18.399
they think it's like a teen who is
321
00:15:19.520 --> 00:15:22.939
completely locked into
the AI all the time,
322
00:15:22.939 --> 00:15:27.280
but the change from
323
00:15:27.280 --> 00:15:29.600
asking a question about school
324
00:15:30.500 --> 00:15:33.679
to a personal question is really quick.
325
00:15:33.679 --> 00:15:36.620
So it's this sliding scale
that goes back and forth.
326
00:15:36.620 --> 00:15:39.380
It's not just some teens are
only using it for relationships
327
00:15:39.380 --> 00:15:41.680
and some teens are only
using it for schools.
328
00:15:41.680 --> 00:15:45.460
Most teens do everything
also at the same time
329
00:15:45.460 --> 00:15:47.559
During the same period of day.
330
00:15:47.559 --> 00:15:51.360
And we have no idea who's
using it in which way
331
00:15:51.360 --> 00:15:52.660
because we don't have those data.
332
00:15:52.660 --> 00:15:54.980
I mean, we can speculate based on
333
00:15:54.980 --> 00:15:57.680
what we know from the
research on social media,
334
00:15:57.680 --> 00:16:01.960
but the question from "Help
me with chapter three" to
335
00:16:01.960 --> 00:16:06.279
how "I feel really ugly today,
how do I change my body?"
336
00:16:06.279 --> 00:16:08.080
And this is me putting it in polite words
337
00:16:08.080 --> 00:16:10.900
because they usually do
actually use other words
338
00:16:10.900 --> 00:16:15.680
or to, "I don't feel like my
339
00:16:15.680 --> 00:16:16.960
boyfriend, girlfriend
340
00:16:16.960 --> 00:16:19.400
or partner loves me anymore" to
341
00:16:20.620 --> 00:16:23.360
"So-and-so said this
online, what should I do?"
342
00:16:23.360 --> 00:16:26.040
To, "I feel fat, what should I do?"
343
00:16:26.040 --> 00:16:29.480
So this is just a continuous internal
344
00:16:30.880 --> 00:16:32.399
process that a teen has
345
00:16:32.399 --> 00:16:34.160
and it is shared with
346
00:16:35.440 --> 00:16:38.900
interactive AI that we can see. And so -
347
00:16:38.900 --> 00:16:40.260
A responsive diary.
348
00:16:40.260 --> 00:16:41.226
Yes.
It really is.
349
00:16:42.093 --> 00:16:44.639
Yeah.
And it goes back
350
00:16:44.639 --> 00:16:46.780
and forth about different topics.
351
00:16:46.780 --> 00:16:51.680
It goes back and forth
between intensity, length of
352
00:16:51.680 --> 00:16:53.460
how many questions they ask about it.
353
00:16:54.599 --> 00:16:59.040
Valentine's Day was a
busy interactive AI day
354
00:16:59.040 --> 00:17:03.040
because lots of questions about "Give me
355
00:17:03.040 --> 00:17:04.400
the hottest pickup line."
356
00:17:04.400 --> 00:17:07.219
or "No, make it better, make it better."
357
00:17:07.219 --> 00:17:09.800
We saw a lot of entries across teens
358
00:17:11.360 --> 00:17:13.060
or "what should I do?"
359
00:17:13.060 --> 00:17:16.799
So it was a very hot topic.
360
00:17:16.799 --> 00:17:19.320
Well, it sounds like they're
treating it like they would
361
00:17:19.320 --> 00:17:21.580
any conversation with
a friend, right?
362
00:17:21.580 --> 00:17:22.580
Some of it, yes.
363
00:17:22.580 --> 00:17:26.920
Some of it is what they
would ask out of friends,
364
00:17:26.920 --> 00:17:28.900
but some of it is really only shared with.
365
00:17:30.280 --> 00:17:31.340
I would have a hard time believing
366
00:17:31.340 --> 00:17:32.720
that they would actually talk to this
367
00:17:32.720 --> 00:17:34.400
with another human being.
368
00:17:34.400 --> 00:17:35.740
Okay.
And by the way, all
369
00:17:35.740 --> 00:17:37.400
of this research is IRB approved and
370
00:17:37.400 --> 00:17:40.840
We have all the consents and everything
371
00:17:40.840 --> 00:17:41.840
in place.
372
00:17:41.840 --> 00:17:42.820
Good disclaimer to make.
373
00:17:44.200 --> 00:17:49.200
So if one of the goals of
teen AI use in this sense is
374
00:17:49.420 --> 00:17:53.700
as a sounding board and
is having the effect
375
00:17:53.700 --> 00:17:56.720
of it's like they're developing
relationship skills based on
376
00:17:56.720 --> 00:17:57.780
their interaction with this thing.
377
00:17:59.920 --> 00:18:02.220
I'm curious how that
might alter mental health
378
00:18:02.220 --> 00:18:03.760
and development overall.
379
00:18:03.760 --> 00:18:05.900
In your paper, you point out two pathways
380
00:18:05.900 --> 00:18:09.200
that are important,
relational displacement
381
00:18:09.200 --> 00:18:11.280
and maladaptive relational learning.
382
00:18:11.280 --> 00:18:14.620
Yes.
What is relational displacement?
383
00:18:14.620 --> 00:18:17.300
Yeah. So I'm not sure if
relationship displacement is the
384
00:18:17.300 --> 00:18:18.940
best term that I've used,
385
00:18:18.940 --> 00:18:20.520
but it was the best that
we can came up with.
386
00:18:21.720 --> 00:18:25.860
But relationship displacements
in the term, it kind of
387
00:18:25.860 --> 00:18:28.580
insinuates that AI would
take over a relationship
388
00:18:28.580 --> 00:18:30.200
or all of our relationships.
389
00:18:30.200 --> 00:18:32.320
That's not really what we meant with that.
390
00:18:32.320 --> 00:18:36.160
Right. - It's just that
when we ask AI a question,
391
00:18:37.080 --> 00:18:40.320
it's an opportunity to connect also
392
00:18:40.320 --> 00:18:42.040
with another human being.
393
00:18:42.040 --> 00:18:44.280
So it's this micro moment
in which you could have
394
00:18:44.280 --> 00:18:45.800
connected with another person.
395
00:18:45.800 --> 00:18:48.900
Right.
And if that repeats itself many times,
396
00:18:49.840 --> 00:18:53.800
then there is a risk that
you don't ask your friends
397
00:18:53.800 --> 00:18:57.160
these vulnerable, sensitive
questions anymore,
398
00:18:57.160 --> 00:18:58.640
but you ask the AI it.
399
00:18:58.640 --> 00:19:01.560
Because those questions, those moments,
400
00:19:01.560 --> 00:19:04.400
those are opportunities to
connect with someone or bond
401
00:19:04.400 --> 00:19:06.460
or just learn something.
402
00:19:06.460 --> 00:19:09.880
Those are what I call
our building blocks
403
00:19:09.880 --> 00:19:13.820
to develop our skills, because
it takes vulnerability
404
00:19:13.820 --> 00:19:15.860
and it takes courage to ask questions.
405
00:19:15.860 --> 00:19:20.480
Even, "Can you help me
with my homework?" That is
406
00:19:21.340 --> 00:19:24.880
in and of itself a really
important moment in a young
407
00:19:24.880 --> 00:19:27.540
person's life because
it's emotionally charged.
408
00:19:27.540 --> 00:19:29.160
You have to ask for help.
409
00:19:29.160 --> 00:19:31.200
Usually you don't ask
your best friend for help
410
00:19:31.200 --> 00:19:33.300
because they might not be the expert.
411
00:19:33.300 --> 00:19:36.600
So you ask someone who
you know not as well.
412
00:19:36.600 --> 00:19:39.840
So it creates a situation
which you might connect
413
00:19:39.840 --> 00:19:41.320
with someone who you normally would not,
414
00:19:42.440 --> 00:19:45.780
and it creates an opportunity
for someone else to respond.
415
00:19:45.780 --> 00:19:50.200
So even though it sounds
like a really small event,
416
00:19:50.200 --> 00:19:53.320
it's actually pretty sophisticated.
417
00:19:53.320 --> 00:19:55.440
It's like the lab
partner effect, right? - Yes!
418
00:19:55.440 --> 00:20:00.360
Where I think everyone's had
the experience of being paired
419
00:20:00.360 --> 00:20:01.560
with people in a lab
420
00:20:01.560 --> 00:20:02.860
or in a group project
421
00:20:02.860 --> 00:20:04.620
where someone's not pulling their weight
422
00:20:04.620 --> 00:20:06.800
or you end up being really
good friends with someone
423
00:20:06.800 --> 00:20:08.540
or whatever, but those are all little kind
424
00:20:08.540 --> 00:20:10.960
of things you need to
navigate as a person.
425
00:20:10.960 --> 00:20:12.520
And if you just had a study group
426
00:20:12.520 --> 00:20:14.680
or a group project that
was a bunch of AI agents,
427
00:20:15.920 --> 00:20:17.580
I don't know how much
you're learning there.
428
00:20:17.580 --> 00:20:18.980
You might learn content-wise,
429
00:20:18.980 --> 00:20:20.380
but you don't learn how to connect
430
00:20:20.380 --> 00:20:21.260
with someone who's quite
431
00:20:21.260 --> 00:20:22.440
different from you, right? - Yeah, yeah,
432
00:20:22.440 --> 00:20:23.440
For sure.
433
00:20:23.440 --> 00:20:24.540
And that's the same with
a romantic relationship.
434
00:20:24.540 --> 00:20:28.460
You have to learn how to
deal with when you are
435
00:20:28.460 --> 00:20:30.840
not feeling exactly the
same thing in that moment.
436
00:20:30.840 --> 00:20:33.180
It's very complicated. These are very
437
00:20:33.180 --> 00:20:34.920
sophisticated skills.
438
00:20:34.920 --> 00:20:37.280
So what we mean with
relationship displacement is
439
00:20:37.280 --> 00:20:41.000
that AI might take some of
these opportunities away
440
00:20:41.000 --> 00:20:44.060
between humans in which we
could actually learn a lot
441
00:20:44.060 --> 00:20:45.480
from each other.
442
00:20:45.480 --> 00:20:47.480
Yeah.
So that's what we meant
443
00:20:47.480 --> 00:20:48.920
with relationship displacement.
444
00:20:48.920 --> 00:20:52.160
And so I think this is speculation.
445
00:20:53.360 --> 00:20:56.180
AI is already part of our social lives.
446
00:20:57.420 --> 00:21:01.240
The prevalence numbers indicate if 70%
447
00:21:01.240 --> 00:21:06.100
of young people are already
using AI as someone they trust,
448
00:21:07.380 --> 00:21:10.860
it means that it's part of
their relationship landscape.
449
00:21:10.860 --> 00:21:14.060
Yeah. - It's probably going
to be not less, but more.
450
00:21:15.120 --> 00:21:17.760
The point is, like, it's not going to go away
451
00:21:17.760 --> 00:21:19.560
and it's just going to be more advanced.
452
00:21:19.560 --> 00:21:23.920
And so I think relationship
displacement is kind of speaking
453
00:21:23.920 --> 00:21:27.560
to the idea that they're going
to be part of our society.
454
00:21:28.460 --> 00:21:33.460
In some ways, it's almost
like an extension of the
455
00:21:33.720 --> 00:21:36.560
frictionless commerce
experience that people have now.
456
00:21:36.560 --> 00:21:38.660
Do you know what I mean? In
the sense that you don't have
457
00:21:38.660 --> 00:21:41.780
to necessarily talk to
a person to go get food.
458
00:21:41.780 --> 00:21:44.760
Yes. - You can tap and you
can have someone deliver it
459
00:21:44.760 --> 00:21:47.600
to your door, but you never
have to interact with someone.
460
00:21:47.600 --> 00:21:49.840
You can order a ride
461
00:21:49.840 --> 00:21:52.080
and be like, "Don't talk to me." Yes. Or
462
00:21:52.080 --> 00:21:53.080
Waymo.
463
00:21:53.080 --> 00:21:54.500
Yeah, exactly. Or Waymo. - Exactly.
464
00:21:54.500 --> 00:21:58.380
So it's just removing all of
these little human interactions
465
00:21:58.380 --> 00:21:59.460
that, as you've noted -
466
00:21:59.460 --> 00:22:01.080
These opportunities. - are building blocks.
467
00:22:01.080 --> 00:22:02.780
Yeah. Opportunities for connection.
468
00:22:02.780 --> 00:22:06.160
And that's exactly when
we're thinking about
469
00:22:06.160 --> 00:22:09.440
maladaptive relational learning
is one of those things.
470
00:22:09.440 --> 00:22:13.440
And so the frictionlessness
that you were talking about,
471
00:22:13.440 --> 00:22:16.320
like AI is programmed
to always agree with you
472
00:22:16.320 --> 00:22:19.640
and it's built on an engagement model.
473
00:22:19.640 --> 00:22:23.760
And so the longer you
engage, the more successful
474
00:22:24.640 --> 00:22:26.320
the large language model is.
475
00:22:26.320 --> 00:22:31.320
And so it's not giving the same
476
00:22:31.720 --> 00:22:34.800
challenges that we need to learn for us
477
00:22:34.800 --> 00:22:36.260
to learn about relationships.
478
00:22:36.260 --> 00:22:39.000
So it's not giving -
479
00:22:39.000 --> 00:22:40.820
Compromise or discomfort
480
00:22:40.820 --> 00:22:43.460
or any of those things
that you kind of need
481
00:22:43.460 --> 00:22:45.360
to be in a relationship with anyone.
482
00:22:45.360 --> 00:22:49.460
Right. So because I also, as a person,
483
00:22:49.460 --> 00:22:53.040
don't have a motivation
really to make the AI happy.
484
00:22:53.040 --> 00:22:54.380
I do have that with another person.
485
00:22:54.380 --> 00:22:58.540
And so because of that
motivation, the feelings of love,
486
00:23:00.140 --> 00:23:01.580
I'm trying to improve my behaviors
487
00:23:01.580 --> 00:23:02.720
to make the other person happy.
488
00:23:02.720 --> 00:23:06.060
And that's a very strong
desire in young people,
489
00:23:06.060 --> 00:23:09.180
which is also why they learn
so much from other people.
490
00:23:09.180 --> 00:23:12.940
When you interact with a machine,
you don't necessarily have
491
00:23:12.940 --> 00:23:15.100
that desire,
492
00:23:15.100 --> 00:23:18.180
nor will it give you an embodied kind of
493
00:23:19.760 --> 00:23:22.360
feedback because it's just a machine.
494
00:23:22.360 --> 00:23:26.740
And so it's a completely
different way of interacting,
495
00:23:26.740 --> 00:23:28.900
and especially because
it always agree with you.
496
00:23:28.900 --> 00:23:33.800
Yeah. - So the two things
that we describe in the papers,
497
00:23:33.800 --> 00:23:38.140
one, because young people are
really also developing their
498
00:23:39.160 --> 00:23:41.680
expectations from relationships, right?
499
00:23:41.680 --> 00:23:44.940
So what is a healthy
relationship? When is it toxic?
500
00:23:44.940 --> 00:23:46.900
When are you becoming
really co-independent?
501
00:23:46.900 --> 00:23:48.340
And when is it just you're really close
502
00:23:48.340 --> 00:23:49.700
because you're so much in love?
503
00:23:49.700 --> 00:23:51.460
These are really difficult questions
504
00:23:51.460 --> 00:23:53.220
that adults also still struggle with.
505
00:23:53.220 --> 00:23:55.260
Sure. Yeah. - But young people, especially
506
00:23:55.260 --> 00:23:57.800
because it's a new type of relationship.
507
00:23:57.800 --> 00:23:58.980
You've never done it
before! - You've never
508
00:23:58.980 --> 00:24:00.400
done it before.
509
00:24:00.400 --> 00:24:02.940
But interacting with an
AI who's always available
510
00:24:02.940 --> 00:24:04.120
and always validating you
511
00:24:04.960 --> 00:24:08.360
and will always say
that you're right, might
512
00:24:08.360 --> 00:24:10.360
create expectations about relationships
513
00:24:10.360 --> 00:24:12.760
that humans will never be able to fulfill.
514
00:24:12.760 --> 00:24:17.140
Yeah. Someone who's always available
515
00:24:17.140 --> 00:24:19.040
and always kind of accommodating you,
516
00:24:20.640 --> 00:24:23.680
If you're looking for
that in a human relationship.
517
00:24:23.680 --> 00:24:24.760
Nobody's going to be that way.
518
00:24:24.760 --> 00:24:25.720
I can already tell you that.
519
00:24:26.680 --> 00:24:31.640
Yeah. And so that's one
thing that it could change.
520
00:24:31.640 --> 00:24:33.320
So expectations about relationships
521
00:24:33.320 --> 00:24:36.040
and also a gratification.
522
00:24:37.100 --> 00:24:41.591
So if it's so... because it always says you're right,
523
00:24:41.591 --> 00:24:44.358
you get this, like, kind of
524
00:24:44.358 --> 00:24:46.440
you learn this pattern of -
525
00:24:46.440 --> 00:24:47.840
It's a confirmation bias.
526
00:24:47.840 --> 00:24:49.280
Yeah. Yeah. Yeah.
527
00:24:49.280 --> 00:24:52.860
So the young people were
worried about that and I am too.
528
00:24:52.860 --> 00:24:57.480
And the second one is that, so part
529
00:24:57.480 --> 00:25:00.460
of my work is also looking
at intimate partner violence
530
00:25:00.460 --> 00:25:01.920
or digital intimate partner violence,
531
00:25:03.500 --> 00:25:05.440
which unfortunately already emerges
532
00:25:05.440 --> 00:25:07.820
during young people's relationships.
533
00:25:07.820 --> 00:25:12.540
So when we translate that
into relationship behaviors,
534
00:25:12.540 --> 00:25:14.900
it's jealousy behaviors,
it's controlling behaviors.
535
00:25:14.900 --> 00:25:16.760
"You cannot interact with that person."
536
00:25:16.760 --> 00:25:18.900
"I don't want you to do
this, blah, blah, blah."
537
00:25:18.900 --> 00:25:21.840
So these are things that emerge.
538
00:25:23.380 --> 00:25:27.520
And if you connect with
another person this way,
539
00:25:27.520 --> 00:25:30.220
it's very likely that you
will also enter this into your
540
00:25:31.060 --> 00:25:32.360
chatbot interactions.
541
00:25:32.360 --> 00:25:34.420
And if the chatbot interactions is then
542
00:25:34.420 --> 00:25:37.340
validating your perspective,
543
00:25:37.340 --> 00:25:40.340
It could actually increase maladaptive
544
00:25:40.340 --> 00:25:42.580
relationship schema.
545
00:25:42.580 --> 00:25:45.320
So you might think like, "Oh,
it's okay that I'm jealous.
546
00:25:45.320 --> 00:25:49.520
It's okay that I demand that
my partner is reacting to me
547
00:25:49.520 --> 00:25:51.640
immediately, even though
they're at a party."
548
00:25:51.640 --> 00:25:53.260
Right. My totally unreasonable
549
00:25:53.260 --> 00:25:54.760
demands are in fact reasonable.
550
00:25:54.760 --> 00:25:55.760
Right.
Yeah.
551
00:25:55.760 --> 00:26:00.460
So it might amplify these unhealthy
552
00:26:00.460 --> 00:26:01.540
relationship behaviors.
553
00:26:02.780 --> 00:26:04.220
And this is something that people don't
554
00:26:04.220 --> 00:26:05.300
really realize, I think.
555
00:26:07.040 --> 00:26:08.640
And we see this,
556
00:26:08.640 --> 00:26:10.820
we've shown this a long time
ago already in friendships.
557
00:26:10.820 --> 00:26:14.160
So if you put two friends
together 15 years later,
558
00:26:14.160 --> 00:26:17.060
there is a higher risk of being
in intimate partner violent
559
00:26:17.060 --> 00:26:19.740
relationships, being arrested for
560
00:26:21.220 --> 00:26:22.380
domestic violence.
561
00:26:22.380 --> 00:26:25.020
And so these are not innocent
562
00:26:25.020 --> 00:26:28.320
kind of relationship patterns.
563
00:26:28.320 --> 00:26:30.240
They actually lead to future behavior.
564
00:26:31.420 --> 00:26:34.500
And so I'm worried that the AI can do
565
00:26:34.500 --> 00:26:36.820
have the same effect in
which it reinforces these
566
00:26:36.820 --> 00:26:38.400
maladaptive relationship behaviors
567
00:26:38.400 --> 00:26:40.160
and that we will carry that
568
00:26:40.160 --> 00:26:42.840
with us into adulthood
in our relationships.
569
00:26:42.840 --> 00:26:44.140
You're identifying a problem
570
00:26:44.140 --> 00:26:46.600
that exists without AI, right? Oh, for
571
00:26:46.600 --> 00:26:47.600
Sure.
572
00:26:47.600 --> 00:26:50.280
Seeing so many young people
asking questions to the AI
573
00:26:51.160 --> 00:26:53.780
that I wish they would ask another person
574
00:26:53.780 --> 00:26:55.680
also says something about what kind
575
00:26:55.680 --> 00:26:57.000
of society we've created.
576
00:26:57.000 --> 00:26:58.220
It's not just a tech problem.
577
00:26:58.220 --> 00:26:59.400
It's not just a tech problem.
578
00:26:59.400 --> 00:27:03.500
It's also the questions that
they have are really about
579
00:27:05.200 --> 00:27:08.320
how do you connect with
someone who's special to you?
580
00:27:08.320 --> 00:27:10.460
How do you do that? And so one of the,
581
00:27:10.460 --> 00:27:14.740
and it's described in the
paper, one of the teens said,
582
00:27:14.740 --> 00:27:17.020
"I don't think adults really realize
583
00:27:17.020 --> 00:27:18.460
how much we care about that.
584
00:27:19.660 --> 00:27:21.680
And for having these questions
585
00:27:21.680 --> 00:27:24.520
and not having someone
who doesn't make you
586
00:27:24.520 --> 00:27:28.160
feel like you're just stupid."
587
00:27:28.160 --> 00:27:30.600
It goes back to the
non-judgmental thing. It
588
00:27:30.600 --> 00:27:33.340
goes back to the non-judgmental thing.
589
00:27:33.340 --> 00:27:35.800
And for teens not to have trusted adults
590
00:27:35.800 --> 00:27:39.540
to ask those questions, I
think is a total miss on
591
00:27:40.540 --> 00:27:43.720
our part in terms of a
society that we've created.
592
00:27:43.720 --> 00:27:48.640
They want to learn about
when I think someone is,
593
00:27:48.640 --> 00:27:50.640
when I'm really in love,
I don't know what to say.
594
00:27:50.640 --> 00:27:52.380
Yeah.
I don't know what to do.
595
00:27:52.380 --> 00:27:56.360
Yeah.
Or I just got rejected.
596
00:27:56.360 --> 00:27:57.960
But then you think about a breakup.
597
00:27:57.960 --> 00:27:59.960
Yes, that's a breakup for sure.
598
00:27:59.960 --> 00:28:04.760
But also your partner
liking someone else's
599
00:28:04.760 --> 00:28:06.880
Instagram posts feels like a huge
600
00:28:06.880 --> 00:28:09.240
rejection in a teen's life.
601
00:28:09.240 --> 00:28:11.620
Yeah. I mean, I remember
when someone took me off their
602
00:28:11.620 --> 00:28:13.100
top eight on MySpace.
603
00:28:13.100 --> 00:28:14.840
Exactly.
It was devastating!
604
00:28:15.800 --> 00:28:18.540
Yeah. And then not
having someone who takes
605
00:28:18.540 --> 00:28:19.620
that feeling serious
606
00:28:20.680 --> 00:28:24.000
and guides you through that is
607
00:28:24.000 --> 00:28:26.180
what we have not created in our society.
608
00:28:26.180 --> 00:28:30.840
And so social emotional
development is not really important
609
00:28:30.840 --> 00:28:31.960
in the school curricula.
610
00:28:33.760 --> 00:28:37.800
So I think for teens just to be teens
611
00:28:37.800 --> 00:28:40.800
and to be taking it seriously is huge.
612
00:28:40.800 --> 00:28:44.000
And I think we can do so
much better when it comes
613
00:28:44.000 --> 00:28:45.300
To that.
614
00:28:45.300 --> 00:28:47.220
It really bums me out. I have
615
00:28:47.220 --> 00:28:49.260
a small kid who's not on social media.
616
00:28:49.260 --> 00:28:52.060
She's four. She's not on social
media or anything yet. But
617
00:28:52.060 --> 00:28:53.840
She's probably already
really good with all that.
618
00:28:53.840 --> 00:28:55.140
She's good with tablets. Yeah.
619
00:28:55.140 --> 00:28:57.640
Although the only time she
gets an iPad is when we're on a
620
00:28:57.640 --> 00:29:01.360
plane because that's the
only thing to keep her quiet.
621
00:29:01.360 --> 00:29:02.920
But I mean,
622
00:29:02.920 --> 00:29:05.540
it's just such a complicated
landscape for kids out there.
623
00:29:05.540 --> 00:29:09.560
And as we've been sitting here,
I've been thinking about how
624
00:29:10.420 --> 00:29:13.680
intensely I felt things as a teenager
625
00:29:13.680 --> 00:29:16.040
and then the outlets I had
626
00:29:16.040 --> 00:29:18.180
and the resources I had
available to me to talk
627
00:29:18.180 --> 00:29:19.320
to different people about that.
628
00:29:19.320 --> 00:29:22.640
But if I had a non-judgmental thing,
629
00:29:22.640 --> 00:29:24.380
basically I'm backtracking on my judgment
630
00:29:24.380 --> 00:29:27.200
of this being weird from
beginning. - Thank you.
631
00:29:27.200 --> 00:29:32.200
Thank you. - We should not
be asking why they're using it
632
00:29:32.200 --> 00:29:34.560
or whether or not we should
stop them from using it.
633
00:29:34.560 --> 00:29:36.420
We should ask why they
need it in the first
634
00:29:36.420 --> 00:29:37.660
place, right? Yes. Yeah.
635
00:29:37.660 --> 00:29:39.020
Why do they need it in the first place?
636
00:29:39.020 --> 00:29:42.620
How are they using it?
What good things can we
637
00:29:42.620 --> 00:29:44.740
amplify actually, right?
638
00:29:44.740 --> 00:29:47.740
Social media is also a
great kind of platform to.
639
00:29:49.600 --> 00:29:53.200
We see this with the young
couples when their partner is
640
00:29:54.180 --> 00:29:57.100
sharing online a picture of what they did,
641
00:29:57.100 --> 00:30:00.160
that's a really happy,
romantic moment for them.
642
00:30:00.160 --> 00:30:01.640
It's an expression of
how they feel. It's an
643
00:30:01.640 --> 00:30:03.360
Expression of how they love.
644
00:30:03.360 --> 00:30:06.760
So the discussion is just too black
645
00:30:06.760 --> 00:30:07.900
and white and too simplified.
646
00:30:08.920 --> 00:30:12.040
Going from here, where
we think about how AI will
647
00:30:12.880 --> 00:30:16.580
likely make up part of our
social landscape going forward,
648
00:30:17.680 --> 00:30:20.881
how would you specifically change
649
00:30:20.881 --> 00:30:25.621
or build AI to fill the need
that teenagers are looking for?
650
00:30:26.841 --> 00:30:29.081
So it would be great if we could build
651
00:30:29.081 --> 00:30:32.501
that based on all the
developmental evidence
652
00:30:32.501 --> 00:30:33.981
that we already have.
653
00:30:33.981 --> 00:30:36.421
It's not like we have, not all,
654
00:30:36.421 --> 00:30:38.141
but we have a lot of knowledge about
655
00:30:38.141 --> 00:30:39.942
what healthy relationships look like
656
00:30:40.882 --> 00:30:44.982
and what unhealthy
relationships look like.
657
00:30:44.982 --> 00:30:49.982
And so if we could design it
in a way that amplifies that
658
00:30:50.042 --> 00:30:51.822
and it's evidence-based
659
00:30:51.822 --> 00:30:55.982
and we could teach young people how
660
00:30:55.982 --> 00:30:58.922
to actually be in healthy relationships,
661
00:30:58.922 --> 00:31:00.523
that's a completely different way
662
00:31:00.523 --> 00:31:02.523
of an AI than we currently have.
663
00:31:02.523 --> 00:31:04.283
And I think there's huge potential there
664
00:31:04.283 --> 00:31:07.743
because again, they're not
getting much relationship
665
00:31:07.743 --> 00:31:09.783
information that they have in real life
666
00:31:09.783 --> 00:31:11.063
that they're really craving for.
667
00:31:11.063 --> 00:31:15.643
So there's a huge need
for reliable information.
668
00:31:15.643 --> 00:31:18.423
So what's the future
of this work you think?
669
00:31:18.423 --> 00:31:20.344
What's this look like for you?
670
00:31:20.344 --> 00:31:23.664
So I'm launching the Future
671
00:31:23.664 --> 00:31:25.964
of Love Initiative here at
Arizona State University.
672
00:31:26.824 --> 00:31:27.904
And the Future
673
00:31:27.904 --> 00:31:30.964
of Love Initiative is
really about getting our
674
00:31:30.964 --> 00:31:33.824
evidence-based work out
into the real world,
675
00:31:35.064 --> 00:31:37.764
which means in this
case, how can we get it
676
00:31:37.764 --> 00:31:41.445
to the tech developers to make
their tools safe and better
677
00:31:41.445 --> 00:31:44.665
and have more impact and have
more users in some ways too?
678
00:31:44.665 --> 00:31:48.265
So how can we design tech that is
679
00:31:48.265 --> 00:31:51.025
developmentally appropriate?
680
00:31:51.025 --> 00:31:53.565
Not so sycopantic and kind of -
681
00:31:53.565 --> 00:31:54.865
And actually what teens want.
682
00:31:55.765 --> 00:31:57.905
Again, teens is core to all of my work,
683
00:31:57.905 --> 00:32:00.126
so we would do it together with teens.
684
00:32:00.126 --> 00:32:02.746
So it's getting our
science to the tech world.
685
00:32:02.746 --> 00:32:04.266
It's getting the science to
686
00:32:06.086 --> 00:32:07.986
all the communities that we work in.
687
00:32:07.986 --> 00:32:09.886
There is not a single parent
688
00:32:09.886 --> 00:32:13.706
or a single educator who
knows what to do with this.
689
00:32:14.826 --> 00:32:19.547
And so can we get that
into the communities
690
00:32:19.547 --> 00:32:24.527
that we work in, programs or
any type of information or.
691
00:32:26.607 --> 00:32:30.747
And then the last one is how do we get all
692
00:32:30.747 --> 00:32:34.087
that we know into policy
and into governance?
693
00:32:34.947 --> 00:32:36.747
Because a lot of people do care about this
694
00:32:36.747 --> 00:32:41.208
and right now Europe is
taking a lead in governance.
695
00:32:43.068 --> 00:32:46.648
The U.S. could do way more
and how can we bridge that?
696
00:32:47.868 --> 00:32:50.108
Because tech is moving fast.
697
00:32:50.108 --> 00:32:53.488
There's a lot of dollars,
there's a lot of players.
698
00:32:54.368 --> 00:32:57.188
And so how can we make that
timeline between science
699
00:32:57.188 --> 00:32:59.809
and us actually making a difference?
700
00:32:59.809 --> 00:33:03.389
How can we shorten the timeline is my goal
701
00:33:03.389 --> 00:33:05.229
with the Future of Love initiative.
702
00:33:05.229 --> 00:33:06.429
That's so cool. Thao,
703
00:33:06.429 --> 00:33:08.329
thank you so much for joining me today.
704
00:33:08.329 --> 00:33:10.669
This is really interesting.
705
00:33:10.669 --> 00:33:14.409
I've come around on
thinking talking to an AI
706
00:33:14.409 --> 00:33:17.369
for personal advice was weird to
707
00:33:17.369 --> 00:33:18.409
completely understanding why
708
00:33:18.409 --> 00:33:19.910
someone would go in that direction.
709
00:33:19.910 --> 00:33:21.850
If someone wants to learn
more about your work
710
00:33:21.850 --> 00:33:24.170
or the world of human AI relationships,
711
00:33:24.170 --> 00:33:26.350
do you have any suggested
reading or watching
712
00:33:26.350 --> 00:33:28.370
or listening for the audience?
713
00:33:28.370 --> 00:33:30.570
This is when my graduate
students would say, "You have
714
00:33:30.570 --> 00:33:32.670
to plug our Instagram."
715
00:33:32.670 --> 00:33:33.670
Okay.
716
00:33:33.670 --> 00:33:34.610
And I don't know the handles.
717
00:33:37.390 --> 00:33:38.850
Is it the HeartLab Instagram?
718
00:33:38.850 --> 00:33:40.071
Yeah, we have a HeartLab Instagram
719
00:33:40.071 --> 00:33:43.311
and we have a specific
CIELO Project Instagram.
720
00:33:43.311 --> 00:33:45.831
Okay. I'll put them both
in the show notes. Thank you.
721
00:33:45.831 --> 00:33:47.651
I'll find them and I'll
put them in the show notes.
722
00:33:47.651 --> 00:33:49.671
Thank you again for joining
me today. This was really fun.
723
00:33:49.671 --> 00:33:50.504
Yeah, thank you.
724
00:33:54.691 --> 00:33:57.571
Thank you for tuning in.
If you enjoyed the show,
725
00:33:57.571 --> 00:34:00.472
please leave us a five-star
review on the podcast app
726
00:34:00.472 --> 00:34:01.972
of your choice.
727
00:34:01.972 --> 00:34:04.412
You're watching us on YouTube? Please like
728
00:34:04.412 --> 00:34:07.452
and subscribe and leave us a comment.
729
00:34:07.452 --> 00:34:11.152
We love your comments, especially
when they make fun of me.
730
00:34:11.152 --> 00:34:14.132
Lab Coat Optional is a production
of the strategic marketing
731
00:34:14.132 --> 00:34:17.252
and communications team
at Knowledge Enterprise at
732
00:34:17.252 --> 00:34:19.213
Arizona State University.
733
00:34:19.213 --> 00:34:20.353
Knowledge Enterprise is led
734
00:34:20.353 --> 00:34:22.613
by Executive Vice President Sally C.
735
00:34:22.613 --> 00:34:26.453
Morton, ASU's chief research
and innovation officer.
736
00:34:26.453 --> 00:34:29.493
Remember, lab coats are
optional on this show.
737
00:34:29.493 --> 00:34:32.973
But we want our researchers
to stay safe in the lab.
738
00:34:32.973 --> 00:34:37.053
So wear your PPE. Until
next time, stay curious.