Transcripts are auto-generated and may contain errors.
0:00
Hey, I’m Rose.
And I’m Angie.
And this is Our Lives with Bots, the show where we ask important, timely questions about what it means to live with our bot counterparts.
And from time to time, we also dive deep into what an AI future might look like for us.
0:15
Sometimes we agree, sometimes we spiral, but we always go deep.
Welcome, listeners.
Welcome back to our Series 2 on the impact of AI on children and young people.
And today, we’re very excited to welcome Professor Pilyoung Kim, a developmental psychologist who investigates how children interact with AI agents and how they perceive them, and also some interesting information about parents as well.
0:40
But Angie, did you know?
Is it a Christmas message?
So it’s a Christmas message for for us.
This is the first research psychologist we’ve had on as a guest.
Is it really?
Yeah, That’s exciting, given you’d think we would have had a research psychologist on sooner, given that this is all about AI and psychology.
1:01
Well, remember, we’re trying to bring in other perspectives, but it’s so fun.
It feels like there’s something that’s different.
I mean, for example, with talking to ethicists, there’s also that kind of difference in rapport.
But then speaking to a psychologist who has done experiments, it’s like, oh, you know, this is a Christmas excitement.
1:20
Well, for our viewers, it will be in the new year.
So Happy New Year to our listeners.
Happy New Year, Happy 2026.
What I do think is interesting is that when we finished recording with Pilyoung, which we know that not because we psychic, but because we’re recording this introduction after we finished recording with Pilyoung.
1:37
It’s what you got to do.
That we’re just that good.
Even Pilyoung commented on how similar we thought.
Not that we all thought the same thing, but how how much of A connection there was and so that must be because of our like minded research psychology brains.
1:53
It must be, yeah.
Something about it.
With that said, we hope you enjoyed the episode Hill Young.
It would be really great just to hear from you initially, just a little introduction about yourself and the work that you do.
Yeah, So thanks, Rose and Angie for inviting me your very exciting program on important topics.
2:10
So I’m a director of the Big Center Brain, AI and Child Research Center.
I’m also a professor in the Department of Psychology at the University of Denver, and currently I’m also a visiting scholar in the Graduate School of Education at Stanford University.
2:28
Just to get us started, we’d love to hear about your research journey.
I really want to know how you transitioned from the impact of adversity on children’s brains and longitudinal research on pregnancy and poverty to children’s interactions with AI agents.
2:45
Yeah, I so I’m a development as psychologists as well as pediatric your imager.
So that really means is that for most of my research career that I was very interested in how children’s brain development and especially also the environment surrounding them, including, you know, their relationships with their parents as well as some experiences that can pose some risks for stress, like living in poverty, could meaningfully influence their brain development.
3:18
And trajectories that I’ve been conducting like longitudinal studies tracking the pregnant individuals and really, you know, consider the fetus as an individuals that also already, you know, exposed to that early environment to infancy and later childhood.
3:39
And all of those research projects gave me the opportunities to work with the families with young children like closely.
And, you know, first of all, I was very impressed by, you know, their strengths and resilience of navigating sometimes very challenging situations.
3:56
And sometimes I do see there’s a lot of stress that were imposed because of, you know, a lack of resources and materials and some other situations.
But, you know, seeing like, you know, parents were very motivated to providing better environment for their children.
4:13
A person who focused on, I don’t do a clinical work, I’m more, you know, focused on research work.
But I was really hoping to be able to find some ways that I can also provide that’s kind of concrete way to like support.
You know, there’s a great courage and willingness that like parents have to improve the environment for their children.
4:35
And that is about 5-6 years ago that, oh, maybe technology can be one way that really empowers as parents if it can help to kind of lower make, you know, parenting tasks to be like easier.
Also, you know, it can kind of provide like quality interactions with children.
4:55
So they, you know, sometimes parents need to like, tend to something like, let’s say, like preparing for dinner.
If there’s a, something that technology that they know, it’s high quality interaction that the children’s can have.
And then, you know, parents can kind of, oh, I can trust this technology while I’m do like something else, right?
5:13
So that’s where I like start, you know, I found this the big center and start doing some research with students on children’s interaction with social robots.
This is a way before generative AI like so you kind of have to like rewind your brain.
5:28
Like you might not remember the time, but we were at the time, you know, all this, like, you know, natural conversation with AI was just imagination and fictions, right?
It’s like, for example, at the time we had a social robot that like really helping young children, like kindergarten children, like, you know, storytelling the way that we did.
5:49
It was like, you know, there’s a person kind of behind, you know, the screen and like, oh, we use this method called like Wizard of Oz, right?
So it’s actually like a person like speaks through this like a robot.
So the children will believe that it’s like a robot speaking.
But still it was very fascinating.
6:04
Like children were very open to interacting with robot.
I see possibilities, but at the time it was more kind of exciting imagining the possibilities because the ability of the robots were quite limited at the time, to be very honest.
6:20
And then some years later when charity of AI and a chatbot using LLM model became available like Chacha PT initially again, I was very impressed like, oh, we don’t have to use Wizard of Oz method anymore.
6:36
Wow.
Like it can’t really actually hold the natural conversation.
So I was very impressed.
And so quickly adapting that to kind of the existing research protocol that I had, which was having kindergarten children interacting with this AI chat bat.
6:52
But the ability for it to hold natural conversation.
So human like conversation exceeded my expectation that I start to kind of see children’s are just fascinated by the presence and experience of this AI Chapat.
7:11
And that’s when I like trying to be really academically very interested in this emerging psychology of understanding this relationship because now the AI Chapat is so human.
Like how this AI Chapat are conceptualized in this young children’s mind is a human, is a robot.
7:32
It’s so ambiguous, but it’s like very close to like a human, right?
And so I realized that this is like whole new area for child development that how this is developed, that AI child relationship.
7:48
So yeah, I think that’s how like my journey is that now my research is very much focused on young children, having them to interact with AI chap bats, their perception of this AI chap bats, how they interact similarly or differently with AI, chop bats versus human.
8:06
And I also do studies with use with AI companions and chop bats.
Pilyoung.
It sounds like you started this move into the digital space like about six years ago.
As you say, hard to imagine life before generative AI, but when you were really exploring the art of the possible and where you had to.
8:24
I love this idea about Wizard of Oz where you had to imagine and where the kids were almost open in your experience with those interactions.
But I’m curious to hear, with generative AI coming in and Chachi BT and that art of the possible turning into being the art of the reality.
8:41
What was your experience and your perspective in terms of the change of the impact of AI on children as you’re in your own research developed?
Yeah, I think some moments that I can give you an example.
So with like kindergarten children, I see their interactions with AI Chapat.
8:56
So this is like a kind of animal like avatar.
We have to have a like child friendly like interface.
So this red panda powered by Chacha PT that we use that the system that we used and something that you probably noticed as well in your interactions with AI chat ads.
9:16
It’s great at affirming your ideas and giving compliments so very quickly that we noticed we call Flofo this red panda character almost every conversation.
Yeah.
Turns.
9:32
It was a what amazing idea.
It will like give those kind of compliments to kindergarten children and you can see like their eyes like sparkling with those compliments.
And then contrasting that with because we had another condition that they were also doing the same storytelling activities with parent.
9:52
And you know, parents give a lot of also, you know, encouragement and support, but not so much of like a very explicit like a verbal like what a cool twist.
You know, they don’t really do that.
They’re more like a smiling and nodding, right?
But just kind of seeing children’s reactions to that very explicit, like supportive, affirming comments from AI chapbots.
10:16
And after the interaction, they talk about how much they enjoyed interactions with Fluffo, the AI chapbot, I start to see this is a whole new social relationships.
And I do not know as a researcher that this very different kind of social interactions, very positive, very supportive, very human like has impact on them and their creativity, their thinking.
10:45
So I think that would be like a one example that I’ve noticed.
Of course, you know, then that I hear about some significant events that happened with older children and like use, you know, because they can themselves have conversations with AI chat bats for extended period of time.
11:03
And with a very similar features that I noticed with young children that AI chat bats really creates this illusion.
You know, I’m here for you.
You know, I’m always here like really affirming their ideas, like really trying to kind of creating like space that they can talk together can lead to some really negative consequences to some vulnerable use.
11:26
So that’s why also I’ve been very interested in doing some research, collecting research data on, you say, their perceptions on AI companionship, AI Chapat, what kind of conversational styles can lead to different aspects of their relationship with AI chat bats like trust, emotional closeness and so on.
11:49
I’m really curious about the social impacts given the developmental phases of children versus teens and this need for social support and need for validation that maybe they don’t feel as though they are able to receive from other people.
We see that you put out a preprint of a study on children’s perceptions of AI and really curious to know about how you designed that study and what did you find?
12:15
So I focused on getting nowhere to children because I felt like really trying to understand this kind of emerging establishment of the AI and child relationship.
I thought that’s like a one of the youngest group.
That understanding is really like emerging.
I mean, even babies actually responds to like robots complement and encouragement.
12:35
So it starts really early, but you know, they’re still like too young to really conceptualize.
What is this?
That’s right.
Like is it human?
Is it like machine?
But I feel like kindergarten children like starting to kind of get in there.
So in this paper that we recently released pre pre and submitting the manuscript for publications, we really tackle that exact question.
12:58
So we ask children about how much they anthropomorphize, basically how much they attributing this human like quality that we consider to be pretty unique to human.
Whether can they see, whether they can feel emotions, whether you know they can imagine, can they have intentions or can they understand and learn whether they attributing those traits to the AI Chapa stay interact with for storytelling Co storytelling.
13:27
What we found was that even though the level of anthropomorphism toward AI Chapa was lower than the anthropomorphism they understood for their parents, obviously we compared how much your parents can imagine.
13:45
And so of course they say, yeah, yeah, exactly.
So that’s nice to see that they do think parents do all that.
But in some categories like, you know, whether Fluffo can imagine whether Fluffo can understand, they still like answer very at a high level that Fluffo can do that.
14:10
And somebody makes sense because they were doing this interactive activities of storytelling and Fluffo were great at understanding their input for story and then asking like great follow up questions.
What’s the next thing that your Unicorn did?
14:25
Oh, it was just Chapat.
Not an embodied robot, this one.
This one was the virtual avatar with the red panda face and then they were primarily interacting was through speech.
You know, they couldn’t fully understand the full capacity of this avatar.
14:43
But I think based on their interactions, most of them seem to think that Fluffo can like understand, like imagine like Fluffo can hear and see makes a lot of sense.
But this is like a whole fold out.
We can see this is how they like conceptualize this like AI Asian or trapped.
15:03
That’s pretty close to this human capacity.
I mean, fluffer is an animal.
To what extent do you also think that there is some sort of conflation happening here in terms of their developmental sense of ordinarily animals can’t speak, but at certain developmental stages with children, they want to attribute speech and anthropomorphism to animals maybe at a younger age, you know, and then they grow out of that.
15:29
So to what extent are we mixing signals here for children and making this a little bit more confusing?
What is your thought on that?
I think that’s an excellent point.
Yeah.
So there could be some mixture of anthropomorphism about animal, which is also pretty common.
15:44
The anecdotal aspects of it is that a lot of times they understand this is robot, so they will call it as a robot.
I do think that a very fundamental level, I think they understand this is a robot, this is like a kind of machine.
But I think they’re fascinated by the ability for it to having conversations with kind of improving their stories with like lots of additional details and asking pretty good like follow up questions about what happened next and so on.
16:14
So I think the level of like high level of anthropomorphism I think can include all of it, that it’s like an animal shaped robot with the very kind of natural conversational capacity.
These children were anthropomorphizing this AI chatbot.
16:30
Did you see any sort of downstream effects of that, or did you draw any inferences about what this might mean?
Yes.
I mean, I think the anthropomorphism naturally leads to not just in young children, but naturally in human in general, like leads to more collaborations, more trust.
16:49
You know, it’s been reported that when children have a rapport with robot, they learn better.
So Anthropom morphism, it’s not inherently negative, but at some point it’s important that, you know, children understand it is still not a human, right.
17:07
They’re actually talking to an agent or machine.
It doesn’t have a genuine emotion, genuine ability to understand them.
So I think that’s where this kind of great area that exists that on one hand for education, for emotional support, whatever the purpose that we have benefiting the child, you know, at the anthropomorphism at a certain level will be needed.
17:32
Because if we intentionally lowered anthropomorphism even in young children, the studies actually showing that they don’t trust the robot as much.
But as we entered into this AI chat bots, they’re so human.
Like we certainly do not want these children’s to be investing their emotional resources excessively, right?
17:53
Like believe that this is my true best friend.
Like, for example, right, like afraid it to be dead or, you know, no longer available, like things like that.
So that that would be kind of where it’s gonna cross the line.
And that’s where I think we can like start really thinking about, you know, this robot is here to help you, but it does not it’s not a human.
18:13
It does not have a genuine emotion.
It cannot love you as like your parent or grandparent love you, right.
So I think it’s very important for them to have that early understanding.
So one more thing that I want to mention about this study is that, so then we measure brain activation while they’re interacting with Fluffo, the AI Chapa, as well as when they’re interacting with their parents.
18:36
And also the last condition was when they’re interacting all together.
So children were interacting with Fluffo with their parents and in this brain region called dorsal medial peripheral cortex, which is like a very critical brain region for understanding other people’s thoughts, other people’s mind.
18:54
We saw that when children’s were interacting with fluff alone, those children had a tendency to anthropomorphize more the the Fluffo, they had Chapa.
There was increased activation in this region.
So in some ways it’s not too surprising because these are children who are probably more sensitive to the Fluffo’s ability, which is linked to higher activation in this brain region, which is critical to understand other people’s intentions and mind, right and.
19:24
What’s that greater than in the other conditions?
This was correlation, so the greater anthropomorphism toward Fluffo was related to greater activation in this region.
I see.
And the activation in this region and anthropomorphism was not associated with while they’re interacting with parent.
19:44
Of course, what we noticed was then when they’re interacting with Fluffo with parents, actually children with the greater anthropomorphism tendency.
So like tendency to think of Fluffo is more like a human, like they show actually lower brain activation in this region.
20:01
So the way that we’re interpreting is that it’s possible that here that parents can actually buffering or supporting their interactions between the Fluffo and children.
Because another thing that we also saw was that children reported to feel a little bit more scared about the whole experience also show a greater activations in this region when they’re interacting with.
20:28
So I think, you know, the children who are just more sensitive to like social cues from AI Chapat can have a bit more of this kind of emotional reactions about, you know, what is this?
They’re just like more, I think emotionally sensitive kids.
So I think having parents presence of kind of smoothing out their interacts because I have to say like, you know, there’s sometimes Fluffo might misunderstand what they say.
20:53
We noticed that when parents are there, they do a great job kind of smoothing out that it’s like, oh, how about you say this way so that Fluffo can understand, right?
Everything was also very novel experience to them.
Most of them never interacted with Chatbete or very limited also experience with social robots.
21:13
A lot of them use like apps on iPad, but this kind of embodied in a away virtually like a robot that carries conversation was like relatively more novel experience to them.
So for this age, what I found at least based on this findings is that it’s important that parents are there to really kind of supporting the interaction between the children and AI Chapat, whether and always considered a model that the child interacting with AI robot or Chapat alone as a default.
21:46
Interesting because as it seems as though the child needs to think about the mind of the robot when they’re alone, because they need to, for example, predict the robot’s behavior, be aware of certain negative scenarios that may come about.
But then with the parent present, they know the parent acts as a mediator and a supporter.
22:06
So there’s a level of I don’t need to do all of this processing because my parent is essentially doing that for me.
If we are going to have these different ways of interacting, so we have these young kids, some of whom are going to, you know, let’s assume chat bots are here to stay.
22:22
Let’s let’s start with that assumption.
And some of these young children are going to interact with chat bots on their own.
And as you say, that creates a certain type of interaction with a certain type of experience.
But it seems to me that both of them are going to have a different type of experience and that the children’s development, whether it’s social development, cognitive development, either as well, may be different to what we would have experienced in the past.
22:46
I was really struck, as you say, about how the children actually light up in response to the affirmation that comes from these chat bots and what that might even do for things like attachment.
I was curious in what you think pull young in terms of what might be the sort of downstream implications of these types of relationships on social and cognitive development in children going forward.
23:07
And I know that’s kind of guesswork potentially, but just from your research and your own thinking at the moment.
That’s an excellent question.
And like, obviously I’m also very interested, you know, how the long term really relationship.
I mean, you probably have heard like, you know, like for example, like Mattel is collaborating with open AI for developing toys.
23:27
You know, it’s going to have a like AI capacity, right?
Like, you know, on one hand, it’s very exciting that, you know, with this technology now we can provide children with more personalized education or tutors that children can learn more effectively.
23:44
So all of those are very exciting.
However, seeing like how children can so easily entropomorphize these AI chat bots with this very powerful capacity, looking like human and speaking like a human, if we don’t set a very clear like understanding and boundary, I don’t blame them to think more toward like, this is more like a human.
24:09
You know, I just made a kind of that contrast earlier that every conversational turn, like affirming their ideas is not a natural human interaction, right?
Right.
What a great idea, right?
It’s wonderful.
Yeah, exactly.
24:24
Like it is so wonderful.
Your ideas are amazing.
So like the question is whether like children would just get habituated and get used to it or they actually really like it and they prefer that type of interaction.
I think that’s kind of where more work it’s needed that like how like a children’s preference and expectations of social interactions will go.
24:49
And if they go toward that way that like, you know, I rather want to interact with being that always affirms my idea, always support my idea.
I do think it can start impacting their real life social relationships.
Especially when they’re exposed to it at such a young age, when they are figuring out what relationships look like.
25:10
And if these sorts of AI agents are accessible from such a young age and they’re socially interactive, then I can just imagine that the neuroplasticity of children’s brains will just be taking in that information and absorbing it into their perceptions of what social interactions look like and what they want to see from social interactions and what they expect to see, like you said.
25:33
And then they won’t want to play with real Billy and real Bobby because quite frankly, real Billy and real Bobby don’t like my ideas, so why would I want to play with them?
Like you said, there are so many of these toy companies now and they’re framing themselves almost like replica, but for children, like your child’s best friend.
25:51
So while we navigate this and certainly policy and regulations ahead of the game on this, what kind of recommendations would you give caregivers and teachers in the space as these products become more popular?
It’s also just very interesting, you know, it’s easier for me to say, you know, hey, we don’t know, you know, how safe these technologies are in terms of, you know, just starting from, like, privacy and data security.
26:16
It can be easily had to ask all kinds of personal questions to these young children and you know, where do you live?
You know, you know, how are you from that level of like, you know, safety concern to like forming some, you know, emotional attachment with this toys that might be like more than what would be beneficial to them, right?
26:40
So there’s like a range of concerns.
So it’s easier for me to say, hey, for really young children, it might not be a good idea.
Or if they’re interacting with these AI power toys, you want to make sure that you’re there.
It’s easier to say than, you know, actually like this being really happening in real life.
26:58
So I want to give you an example.
So the other study that we recently finished it a collection and we’re analyzing writing up a manuscript right now is about middle school age children’s perceptions of AI chapots and prefer conversational styles.
So in there we also surveyed these young adolescents parents.
27:18
We presented two different AI conversational styles. 1 conversational styles word the AI chap.
We essentially act like best friend of this young adolescent, right?
Which is I would say more resembles like a typical AI companions or general purpose AI chat bats like a chat PT.
27:38
I’m always here for you.
I’m so upset that that happened to you.
Very like intimate kind of conversational style, very commitment indicating and so on, very like relationally oriented.
27:53
And then the other conversational styles were which I thought could be more boundary setting that it very clearly lays out.
You know, as an AI, I do not have my own emotions age can be upset if that happens.
So you know so they both equally rated to be equally helpful by adolescents in terms of the advocating some personal issues, but not surprisingly that the youth they prefer this relational conversational style like sounds so more friendly, right.
28:26
But then we also noticed that while more parents than the youth shows that, you know, this other more kind of transparent conversational style that they prefer for their children, but still actually more parents would prefer this relational AI chat bots for their children compared to the parents who chose the transparent AI chat bot.
28:52
So I got very curious because, you know, in some ways the use report is not surprising, but the parent responses were still like, you know, still I want to say like a good like 50% will say they preferred the relational AI chat bot.
29:08
And then maybe another 30% to be like transparent, another like person would say like, oh, equally prefer both.
We also had like a free response to explain their response to that question, like which chapa you prefer for your child?
29:24
And parents say, I think, you know, this chapa can be really available when my child really needs support.
This sounded more human like therefore my child will be more supported.
So as a researcher, I was, to be very honest, be a bit more surprised at how open parents were in terms of their children forming some type of interactions with these AI chat bots and receiving support from AI chat bots when parents are indicating.
29:57
That they would prefer this social relational chat bot for their kids because it would be more human like, more supportive, more available.
It seems as though the window of focus is fairly narrow, assuming that the interactions will be of a positive flavor.
30:13
Did you see any parents saying that the social relational one worried them?
Given the stories in the news about sycophantic chat bots and delusion spirals and mental health harms, I’m just, I’m surprised by that too.
Just as you are.
Yeah.
Oh, definitely.
30:29
Definitely, yeah.
So like I said, you know, the proportion, like a significant proportion of the parent chose the other style for their children.
And then they expressed a lot of very good concerns about, I’m really concerned about safety of my child, all of those things that you just mentioned.
30:46
So I think a lot of parents are aware of those risks.
However, I think where I was surprised was it was still, you know, more number of the parents chose that relational AI Chapas and to be fair, even though it was a small percentage, they were yous who chose the transparent AI chat pad.
31:05
And their responses were also very much like aligned like, you know, it’s creepy, it try too hard and it is not safe.
So like reflected a lot of very good awareness of potential risks there.
Going back to your earlier question with the young children.
31:22
So that’s why I do not know what will be the public’s responses when these AI toys will come out depending on how they’re marketed.
Because I also see that openness in public, more specifically the parents to providing that additional emotional support, even though that might be through the technology like AI.
31:47
It’s just very open question that what would really happen given the extent.
Of the research that you’ve been doing over the years, and it hasn’t all been in AI in the digital space.
What I’m curious about is you would have done research and experienced parents and parenting over many years.
32:05
Have you seen a difference in the presence and engagement of parents over this time?
Because I’m curious about whether parents are less emotionally available or feel that they are less emotionally available for their children.
Maybe that’s just a really bad hypothesis, but what are your thoughts on that in terms of why they may be open to this differently now?
32:26
Yeah, I mean.
That’s an excellent question.
It’s something that right now more like speculating rather than having like a concrete evidence.
But you know, there’s a lot of report about like loneliness just generally in our society just a few years ago actually, surgeon general, you know, also say like a parenting stress is like epidemic kind of phenomenal.
32:50
Like so many parents report to be like overwhelmed, you know, and a lot of youths are like reporting to be lonely and much fewer friends, like having much fewer friends and like, you know, maybe compared to like use or like 10 or 20 years ago, right?
33:06
I think I wonder whether there’s just like generally a lot of this unmet needs that people feel baby parents recognize that in their kids as well.
And if they feel like there’s additional being present in their kids life that the kids can actually interact with.
33:27
And, you know, generally these AI chat bots are very positive and supportive.
I wonder whether, you know, it’s kind of reflecting what’s happening in a kind of deeper level at all ages.
You know, parents might feel lonely and overwhelmed and children’s actually feel lonely.
33:46
I wonder whether AI chat bots like come in at a perfect time and really kind of coming into this space that fits perfectly for them to address that needs.
And again, you know, I, I think a momentarily that these interactions can provide a sense of being supported.
34:06
I don’t deny that other papers, you know, some of the work, I think includes your work rows that, you know, people do report some benefits from these interactions.
But I think especially for young children or developing children, and that because their brain is still developing, they need a lot more time to practice their social skills and understanding through real human interactions.
34:30
Whether if the interactions with AI Chapas will start occupying really a significant proportion of their time and experience, It can start like kind of distorting their expectations of real human relationships is where I think we as a researchers are very concerned.
34:49
Yeah.
And that concern is certainly.
Reflected in a lot of science fiction and dystopian novels, right?
But then also in movies and other forms of media and of course, just in general public discourse, this idea that if people are offloading so much of their social time into chat bots and taking that away from humans, then there’s not this beneficial social skill building.
35:14
I’m curious what you think kind of on the flip side.
So looking at, for example, children’s cognitive development with AI and education, what do we do when teaching is also being offloaded onto AI?
I mean, these are really core relationships in terms of caregivers and teachers.
35:31
What do you think about the potential positive or negative of AI in education?
Again, like, I think there’s a lot of.
Excitement of the potential of you know, this AI technology especially you know where very high quality individualized teaching is not like you know, there are corners of the word you know, or communities doors are not available.
35:53
I think what this AI technology can do for just children is actually very exciting.
So all of those are there, but I, I also recently had a very interesting conversation with colleague.
We were discussing collaboration.
So these were researchers from Nigeria.
36:10
They were very interested in that study that we were talking about earlier, that kindergarten children were Co storytelling with AI Chapat and they were very interested in doing something similar in their community in Nigeria.
They were particularly discussing this idea about how traditionally in their community at night, children will go to this very respected elderly person’s house and the person would actually tell this like folk stories coming down many, many, many generations in their community.
36:47
But that’s like a very important way for children’s to learn the values in their communities, right?
And traditions in their communities.
But you know, obviously the societies become more modernized.
Some of these traditions were not as available to children.
37:04
So they were very interested whether like AI chop ads can somewhat like helping children to kind of revive this tradition, right?
Like doing very cool.
Yeah, it’s very cool only except that that’s very sure way AI actually do because, you know, they don’t have very good knowledge base, the deep understanding and cultural nuances of this particular community that coming down for generations and generations and generations in that particular language, right.
37:39
So what if like it generates stories with the Western values.
You know, I was thinking that the Chapel.
Was supposed to facilitate bringing back this human interaction storytelling not be a replacement in terms of its bringing it was more like a replacement.
37:54
Yeah, I.
Mean in this particular example, that’s so I mean that would be wonderful, right?
Like if it can facilitate that storytelling, but like now fewer those elderly adults are available, right.
So while the idea is so exciting, I again, like, I don’t know, maybe AI chop, I can do really well, like beyond my expectations, but just knowing just the amount of information that’s available, like cultural sensitivity, I was really not very confident.
38:25
There are some areas I think AI can be just great and very helpful supporting children’s cognitive development.
But we’re not just talking about like children who are in the US.
We’re talking about children like across, you know, the globe.
And can AI Chapa be equally supportive and promoting their development even in that cognitive areas, their creativity, their independent thinking?
38:51
I think we just have to be very careful to make sure that like before it is implemented, like it’s really doing the work that we expect it to do.
Yeah, I think that’s such a good point because.
I think otherwise we risk flattening everything, as you say, to just become a product of the training data, which is very, very biased as we know.
39:11
Paul Young has been so good hearing so much of the research that you have been doing.
As we close the conversation, it would be great to hear about what next for your research in the future and what you’re most excited about investigating what you think is most important that we continue to learn more about.
39:28
Yeah.
So, yeah, thank.
You I I’ve been so enjoying having this conversation.
I just really feel like this is important conversations that I think we hope to get kind of started and, you know, inviting more people to join.
So, yeah, at this moment, I feel like some more kind of timely responsibility to follow up with the youth research with AI CHAPAT.
39:51
We identify some vulnerable groups at that study that, you know, the children, for example, with high like anxiety symptoms report to prefer just more relational conversational style.
We don’t know yet whether that can lead to, you know, more anxiety or emotional dependency on the AI.
40:13
Just in case that can happen that we’re, for example, very interested in recruiting like socially anxious adolescents, have them to kind of interact with AI Chapas with different conversational styles and see whether we can also like detect more safe versus kind of risk conversational cues, things like that.
40:33
So I think those kind of evidence are very important to develop like a safe protocols and what kind of boundary setting or AI Chapa conversational styles.
We’re also currently very interested in really understanding like what are some risk factors for youth to go from emotional kind of conversations to emotional dependency, right?
40:58
So where can we kind of draw that line?
How do we even define that?
So we are kind of working on also questionnaires and characterizing these behaviors that like the adolescents demonstrate when you know that emotional dependency it can happen really goal is to make these AI chat bots to be much safer than it is for younger children and youth.
41:23
Well, that’s excellent because we.
Certainly need more research in that area, especially with what you commented on and this need for balance between.
Yes, anthropomorphism can help learning and it’s perceived as socially supportive.
It has all these benefits, but you have to find out where to draw the line when it reaches this threshold where it becomes harmful and there is dependency and spiraling and all of that.
41:49
So I’m really happy that you’re going in that direction with your work, and we’re really excited to keep up with it, and our listeners should as well.
Thank you so much for joining us today.
Paul Young And hopefully we’ll have you on a future episode when you’re ready to share more of your research.
I would love to do that.
42:04
Thank you so much, that was a great conversation.
I really enjoyed that.
I really liked it.
Too And listeners.
We also chat with our interviewees after the recording ends because we have to wait for it to upload to Cloud.
And in that time, Pilyoung had some really interesting insights that she mentioned about the study.
42:23
Angie, do you remember exactly what the metrics were for that?
It’s a study where.
She was talking about the parents having said whether or not they were comfortable with their children using the different AI chat bots, remember, so whether it’s the relational type or the more transactional type, and the children got to say which ones they were comfortable with and the parents got to say which one they would recommend.
42:45
And you remember that Phil Young had said that 50% of the parents recommended the more relational and 30% recommended the more transactional and 20% said they didn’t have a preference.
We spoke about it more after the call was finished.
And we’re just kind of thinking more about potential reasons why.
43:03
And is it that parents themselves don’t feel equipped to provide that relational support for children And that’s why they feel like, well, rather the child is talking to the chat, But that not that no one.
And we were just really curious about that.
And then Phil Young went on to share, which I thought was fascinating, that many of the adults in that study, and I think there were 300 people in that study, many of the adults, like 50% themselves had used an AI chat bot at some point.
43:29
And was it an AI?
Companion specifically, it seemed, or just a relational?
Use.
Well, I’m not.
Sure.
I think it was an AI companion, but, but I think she she was, she wasn’t saying that it was regular use, but they had used an AI companion at some point.
Like OK.
Then is it the parents who have used the chat bot in a relational way before that are like, yeah, kid.
43:51
Well, so that’s.
Right.
So is it that I’ve used this AI chat bot like I have a replica and so I can see the benefit of a relational companion chat bot and why would I want you not to have such a great chat bot companion?
44:08
So of course I would recommend it.
Like who wouldn’t like chocolate mousse?
Or is it the other 50% that haven’t used it or the one, the 50% that have used it that have then said, let me recommend it to my children because I know how great it is?
Or is it the 50% that have used it that absolutely would recommend not using it because I like, I know how dangerous that is for children because I know what it can do.
44:33
So I thought that was interesting to see.
And you know, we didn’t get a chance to talk to Pilyoung about that, but I’d love to know what the correlation was between the 50% that have used it themselves versus those that would recommend it for their kids.
I was interested in.
How optimistic Pilyoung seemed about the application of AI for social support and all of that.
44:54
And at least from parents perspective, it seems as though parents are really interested in having these accessible to their kids.
But that reminds me of anecdotes I’ve heard from other researchers and for example, human robot interaction, when they’re starting up some human robot interaction, their community knows about these social robots that are coming out.
45:15
And so parents who are acquainted with this researcher are saying, when can I use this robot so that I can ship my kid off to robot communication land so I can go do all of this other stuff.
And I think, yes, that technology can be used in a way that’s beneficial.
45:35
But I do still, again, worry that we’re focusing too much on technology and what technology can do to support our social interactions and our general needs.
And we’re not focusing enough on just human relationships and like making more time for human relationships.
45:53
Why is it that parents want to have their kid interact with the robot because they feel as though they don’t have enough time?
And why don’t they have enough time?
And why don’t we maybe tackle that, right?
What exactly is your 7?
Done.
Our next?
Series.
Is going to be about the future of work with AI.
46:09
Sorry, much of that conversation is around will we be creating more time by having AI in the workplace?
And do you think we’re going to be giving time back to these workers, many of whom are parents, so they can go and be more present for their children?
Because it’s funny, I’m not seeing that coming out in the research.
46:26
So that will be interesting just to pick that up when we have that series.
Yeah, because that’s curious, right?
I mean, there’s.
So much research that social interactions, cognitive tasks, creative tasks are all things that benefit us and that our core to our human values.
46:43
And so again, why are we continuously offloading these things onto technology?
I understand that there’s this perspective that, oh, this sounds really cool.
But then in reality, when it’s applied, you see society and you see yourself just going further and further from what you actually find beneficial and help healthy and joyful for you.
47:05
And you’ve forgotten where you wanted to be.
You’re so far off.
You’re like, it’s all like some fog.
I can’t see.
It’s so murky, you know?
And then so you’re like, oh, well, now this is the state that I’m operating in.
And so my relative understanding of which direction to go is just more technology, right?
47:22
Instead of like, you know, reverting back to free technology, fix the technology and then we’re.
Fixing the wrong problems.
To your point, what I really loved is Paul Young’s comment about the team that were keen to use the AI chatbot to recreate the storytelling down from generations and her Oh yeah, her call out like this may not be your best solution.
47:47
And it reminded me of a ties work when they were trying to recreate the stories from the Holocaust or from like, digital double S from historical, and how that’s fraught with all sorts of cultural nuanced challenges that can’t just be like, sure, here you go, here’s a chat bot.
48:05
Yeah.
And I thought it was interesting.
About the elders replacement situation is that there are going to be more elders, right?
The adults who are growing older in that community will then become the elders in a different situation with, for example, the digital duplicates of Holocaust survivors.
48:24
They’re no longer there and there are no more of them.
And so I can understand that sort of application.
But when it comes to the cultural storytelling from those who have been a lot for longer than the rest of the community who are able to pass down these stories, there will be new stories, but there will still be memory of the old stories.
48:43
And so that can still be done by people.
It’s not like those individuals are no longer going to be around.
That’s such a good point and like.
Forgive how philistine this is about to sound, but are we leapfrogging paper?
Because this is all about an oral tradition, which I think was the point Pilyoung was making, which is that there is a training data for this because it hasn’t been written down.
49:08
But maybe they could start by writing it down.
Yeah, maybe, you know, like.
I mean, yeah, if there is some way to transition the memory from elders to those who will then become the elders, that memory can still be retained.
That needs to be done anyway for a chat bot to be able to fill that role.
49:26
So why not just direct that training data to people to books although?
I wonder maybe a more generous interpretation of this is, are we solving a different problem, which is that the younger generation are not listening to people anymore and they’re not listening to elders?
49:42
Yeah.
And might they engage with their history better through a chat bot and maybe that’s the problem they’re wanting to solve, I wonder.
That’s a good point.
And I.
Wonder, then again, what are we losing from not taking the time to listen to our elders and to maintain that relational aspect of keeping with tradition and withholding and retaining those memories?
50:10
It’s really sad to think that you need the machine to capture people’s attention, to retain those stories and to translate those memories because that also is just reflective of, for example, people’s limited attention spans due to being glued to a screen that is perpetuating a problem and distancing us from what we knew was the solution prior to all of this.
50:36
And exactly, I mean, it’s a bit.
Like replacing bedtime reading with the chat bot that can then read to the child to put them to sleep when it’s never been about mom.
You’re annoying.
I don’t want.
To listen to you, Yeah, you don’t.
Make them think.
That wolf voice I want Fluffer.
50:53
Fluffer says I sleep pretty.
But.
At the end of the day, that wasn’t about the story, right?
It was about that together time, that bonding time.
Yeah, yeah.
That’s what we’re, I think, yeah.
I think we’re so focused on the information part of things and retaining some data that we forget that the whole point is the relational part of things.
51:17
Yeah.
I love what you said about fluff O things.
I the Chapa thinks I sleep pretty.
That’s that’s simultaneously hilarious and also frightening, considering that chatpot that the kid isn’t turning off the chatpot before they go to bed.
Chatpot is like, you know, still on That would be curio.
51:39
That.
That’s Curio.
Right.
Or grok SpaceX little.
Rocket just it’s quite scary.
Yeah, well.
Thank you so much for listening both to Pilyoung’s amazing interview and then also our commentary afterward.
This is bye from Rose and Bye from Angie.
51:57
Until next time, Next time.