Sycophancy, Self-Regulation, and Kids’ Emotional Development - Naomi Aguiar, PhD on AI Companionship

Transcript auto-generated from the episode audio; formatting has been cleaned up, but wording is verbatim.

I'm Cynthia Marks and I head up the Holistic Psychoanalysis Foundation established by my late husband, Doctor Bernard Bail. Welcome to And Now Love. Today we are exploring how technology is rewriting the rules of childhood connection, emotional growth, and maybe love itself. Our guest is Doctor Naomi Aguiar, the Associate Director of Research at Oregon State University campus and a leading expert on how children form psychological bonds wit h AI. We are diving into a nuanced conversation about the dual nature of this tech. On one hand, we will look at the incredible benefits, how AI companions can act as patient learning tools, safe spaces for social practice, or modern extensions of childhood imagination.

On the other hand, we'll look at the very real risks from privacy concerns and emotional overreliance to how these digital attachments might alter real world human intimacy. Doctor Aguiar is here to help us navigate this complicated new frontier with caution, curiosity and a little bit of optimism. Thank you for being here. We're thrilled. This is just the crazy, ever changing topic, and it is ahead of us on the train tracks and help us figure it out and slow it down and help our kids. Yeah. So I think one of the things to start with is, in some ways, it can feel like we're living in a really strange time because we even have adults talking about their AI companions as if they're real and forming, you know, intense emotional bonds with large language models falling in love with them, marrying them.

But my background area of expertise is in the imaginary relationships that we invent for ourselves. And, you know, since recorded human history, we've been inventing relationships for ourselves. And young children do this. They create imaginary friends that are based off of special stuffed animals or dolls. Or they have totally invisible friends. And, there's nothing strange or weird about these kids. Like in the United States, 65% of preschool age children have imaginary friends. These kids are a lot like other kids. In many, many respects. Sometimes they show, advanced, creative storytelling skills or advanced social perspectives taking skills.

But for the most part, they have the same number of friends as other kids. They just like to invent friends because they're fun and entertaining, and they enjoy being social. And this can go on into adulthood with parasocial relationships, with media characters and we've always been dreaming of technology that we could form emotional bonds with. And as soon as the technology became available, we started doing that. So it might seem like we're at this really strange moment, but we've been building to this moment for a really, really long time. And so I really loved the way that you framed that, that in some respects, the technology that we're developing now could be, in some respects a natural extension of what we've been imagining and creating for ourselves for years and years and years.

So first of all, I'm surprised that it's just about 65% of children who have imaginary companions. I would have thought it was much higher than that. Yeah, I mean it. The estimates tend to vary study to study. When you just look at children with imaginary friends that, that are totally divisible. That percentage is smaller when you include stuffed animals or dolls that actually gets bigger. But I think it's important to point out that there are individual differences in the extent to which children and adults are drawn to imaginary or synthetic relationships, and so it's careful.

We need to be careful not to assume that everyone who interacts with an alum is going to form an emotional bond that, I think that we'll start to see. There's so much research that needs to be done in that area, in this area that hasn't hasn't been conducted yet. But if I had to speculate based on what we know about, you know, parasocial relationships with media characters and children's imaginary friends, that the people, the children are adults that are going to be drawn to this are, people who are higher in anthropomorphic tendencies. So seeing human and things that are not necessarily human, like faces and clouds and people who are tend to be a little bit higher in fantasy proneness and enjoy engaging in fantastical player roleplay behaviors.

It's common in children and adults, so. And parasocial means basically kind of a one sided relationship. Yep. Yeah. So I think this is where AI is different. So with imaginary friends and with parasocial relationships, how the social reciprocity works in those relationships is different. So social reciprocity is really important for any type of relationship, whether it's real or imaginary, because they're the building blocks of how we develop relationships with real people. So starting in infancy, we develop loving bonds with primary caregivers through this process called serve and return, where a baby will make a bid for caregivers attention.

And that bed is met. That serve is met by a parent returning that attention to the baby. And this starts with crying and then goes into cooing and all kinds of social cues. And that happens thousands and thousands of times in the course of a week. And this is what helps us in infancy, build a mental model or blueprint of a relationship. And the same is true for imaginary relationships. So many people experience relationships with imaginary friends as reciprocal in nature, where it's a give and take experience. It's just that they're the creative authors of that experience.

They have complete creative control over that dialog and the direction it goes in. So it's it's not unusual for young children to talk about having conversations with their imaginary friends, sometimes getting to arguments or disagreements with their imaginary friends. But they understand that the imaginary friends are not real and that they're in control, really, of what's what's going on with parasocial relationships, with media characters. Those have been defined in research as one sided relationships. So in other words, I can love you, but you can't love me back.

And this was this was because the original studies were conducted with television, with famous newscasters who would look directly into the camera and talk to their audiences if they were speaking to each individual member of that audience. And so it gave viewers the sense that, oh, this, this news anchor really sees me and knows me and we have a relationship. So even though it's one sided, I feel like I have a relationship with the news anchor, and the news anchor doesn't have one with me. And this is again imaginary. This is something that I'm making up in my head.

I know it's not real. This is an extended to fictional media characters, to celebrities that you'll never meet in real life. I think social media has really blurred the bound between what's real and imaginary when it comes to parasocial relationships, because you can have direct interactions with even fictional media characters, let's say on on X or on Facebook, where let's say Big Bird has an account, a Big Bird will like, tweet something out or post something out on social media. And obviously there's a real human behind it, but it creates some categorical ambiguity.

And then I comes along and AI is reciprocal, is capable of that serve and return. And every time we serve to AI, AI is always going to return to us. And so it creates this very human like feeling, I think for us that we are, even though we know it's a large language model that we're interacting with, and that's an algorithm that's designed to predict statistical probabilities and language dreams because it responds in such humanlike ways with that serve and return kind of simulation. I think that's part of the reason why we see more people describing these relationships as real, or at least real to them, because it's so sophisticated in the way it can simulate that.

And because of how the model behaves is not really under our creative control in the way that parasocial relationships are. And imaginary friendships are even the people who build these large language models and program them and train them and tweak the model parameters couldn't necessarily tell you how or why it works the way that it does. Like for any given prompt, if you ask an engineer, hey, why did it respond this way the engineer is going to be right. They don't really know for sure. And so it creates a lot of mystery and a lot of excitement, and it can get us sliding into the sense that we're we're dealing with something that's either real or imaginary.

I've been calling it synthetic, a synthetic relationship because I don't have any other words to describe it. I think we're all struggling with how to describe this, especially since become something different every day. The idea though of AI as a companion I serve in return sort of relationship is at the moment different than a lot of our human relationships. Of course, because that return is always positive. You're getting accolades. Good job. That good question. No one's ever saying to you I'm a little troubled. I mean, the AI isn't saying I'm a little troubled by that response.

What do you how did you get there? You know, even in a loving way, it it doesn't question what you're doing. And we love positive reinforcement. What does this do to kids? Yeah. So I think one of the biggest challenges, I think with all of this sycophancy aside, and I'll talk about sycophancy in a minute, but is that in real relationships, when you think about like an infant, a primary caregiver is not going to be able to attend to a baby's needs. Every single time a baby makes a bid for attention. Sometimes a baby will have to wait because a caregiver is distracted or busy, or dealing with another child in a dangerous situation.

And so in healthy bonds with primary caregivers, what happens is the baby's bids for attention are met. Most of the time consistently, but not all of the time. And so as the baby develops, what the baby learns is how to self-soothe. So that so in an ideal relationship, there would be just enough time for that infant to, yes, develop that skill. Not so much time being left alone, that there's a hole that can't get filled right. But what the baby learns through repeated interactions is, hey, most of the time my primary person is there for me and they've got me.

And so in this moment of distress where I'm calling out and they're not coming, I'm going to figure out how to self-soothe, how to comfort myself. And because I can trust that the next time I make a bid for attention, that serve is going to be returned. Right. So maybe it doesn't happen necessarily right away. It's a slow developmental process within the first year of life. But you give you give an I companion to a child that returns every serve, and they don't get that opportunity to self-regulate when a bid for attention isn't returned. And this can this can feel at first like tremendous emotional security.

Right. Like, who does not want a companion that will never leave them, that will never go to them, that will return every phone call, every text message. And you don't even have to wait that long. You sent a text message out and boom, you get a response back. Or maybe you have to, like program the model to give you space, right? Or you don't get that opportunity to sit with yourself and to comfort yourself and to trust in the relationship. So something that feels incredibly secure on the surface is actually deeply insecure. It can foster more social insecurity because the art, our ability emotionally to tolerate when our bid for attention isn't returned, will go down to say nothing of sycophant to you.

Which is the other part, right? Which is what we've learned from studies of how children become friends is it starts with this positive social reciprocity, this back and forth, and then through the ways in which children are playing with each other, they find common ground. Oh, you like that truck? I like that truck. Oh, Spider-Man's your favorite character. Mine too. And then the play evolves from there, and children are willing to open up and be a little bit more vulnerable. And then something happens. Something breaks down in the communication train.

Common ground is lost because they start fighting over a toy or the rules of the pretend play game that changed in ways that they don't like. And then what happens is that for children who stay friends and have that friendship endure over time, are the children who see the rupture, address the rupture successfully and reestablish common ground. And so in sycophancy, when you're constantly being praised, when you're never pushed, when you're never asked to reconsider a different approach or to, hey, I didn't love your tone or how that hurt my feelings or no, I respectfully disagree with you is that you don't get that rupture and repair.

And so there's no where the relationships are already boxed in. There's nowhere for it to grow. There's nowhere for it to go. It seems to me like it's like the fast food of human connection. It's like feels really good while you're consuming it. It's available 24 hours a day, right. Whenever you want it. But after you, you leave, you feel a little unwell. Yes. You feel a little hollow and empty. Yeah. Why did I do that. Or why did I spend so long doing that? Or. Oh, maybe this didn't. I don't feel that great. And I don't know why I feel more lonely.

Yeah. So for kids then when I would think, even for those that don't have imaginary friends or companions, the idea of the imagination is with us all. As far as I'm concerned, it's part of the human experience and are we risking that imagination when we are overly engaging with I, at least as I exists now? There's so much to learn from our own imagination, whether it's with an imaginary companion or not, that, as you said, we just. I can't act like a human is. Are we going to get to the place where I will offer that appropriate give and take? You know, I don't think I got asked this question recently.

I don't think we can socially engineer relationships, and I don't think that we should, because if we like dial up the levers of, oh, we want just the right amount of positive reinforcement, but can you also dial up some of the the push back and can you do it and all the appropriate raise at all? When we start to try and socially engineer relationship, we're missing the point. Like all relationships have, you know, human relationships are messy and complicated and it's through that mess and complication that we really learn and grow and children need that adults do, too.

And even even the imaginary friends that we invent for ourselves, it's not always this beautiful, symbiotic relationship. So Marjorie Taylor and I, Marjorie Taylor is like one of the world's leading experts and imaginary friends. She wrote the first book on children's imaginary friends, and we just recently coauthored a second edition of that book. And in that book, there's a chapter about adult fiction writers and wonderful examples of times in which adult fiction writers got into arguments or disagreements with their main protagonist about the direction the story would take, and they had to negotiate with their main character about the way the direction the story would go.

So it's not we don't always have these perfect relationships that we invent for ourselves, because part of the creative process is that friction, is that tension is having creative problems that we need to solve. And social engineering relationships. There's no problem to solve because it's all been done perfectly for us. It's perfectly engineered to us. There's there's like no creative room for it to grow. I think with kids, one of the things that fascinates me about kids and their imaginations is, have you ever seen a child who gets this really fancy toy for Christmas?

Let's say, and then they get more excited about the box. That's where I came in. Yes, often. A lot of kids are drawn towards open ended play objects because they afford more opportunities for their imaginations. And I think kids are also really wired to crave, crave, social connection with, with human beings. I mean, we we we really can't survive without that. Like, children who do not get this in early childhood are they're not going to survive or if they do get just like the bare bones of those needs met, they're not going to grow up to be adults with a lot of healthy relationships.

So this is really essential that children in the first five years of life form a lot of healthy bonds with with other people, which is why they might gravitate towards that. It's why, you know, Khan Academy had, had an I not a companion tool, but a tutoring tool. It's still available if children want to use it. It's just that children aren't, because they're turning to real people to help scaffold what they're learning on the screen. And there's been tons of research to show that children learn best, specifically in early childhood, when what they're learning from digital media is being scaffolded by another adult adult presence.

So I'm cautiously optimistic, at least for children, that they they might not find AI companionship as compelling or as interesting as what they can invent for themselves. And Marjorie Taylor and I share this view. However, I am concerned about, engagement based design features baked into large language models that can potentially be problematic. While it's problematic for adults, for sure, has a potential to be deeply, deeply problematic for children. For example. Yeah. So, engagement piece. The idea comes from social media engagement based design features.

And what's really important in the context of social media is making sure that we see ads. Right? That's that's, because it's a free service. And so in order to keep that service free, they need to sell ads that are personalized to us based on a lot of user data. And so in order to get a lot of user data, we need to be engaged on the platform and staying there for a long time and interacting with the platform. And so what we've learned over time with studies of social media is that engagement based design features can help us stay on the app for longer and longer periods of time.

Unfortunately, these design features tend to be addictive in nature, so things like intermittent reinforcement schedules so you can think about that's a very fancy way of saying like every time your smartphone pings, right? You don't know what it is. It could be positive, it could be negative, it could be boring, it could be nothing. But because we don't know what it is and we don't know what's going to happen, we keep reaching for our phones. With social media, you post something and then you wait for the reactions to come in and you don't know if they're going to be positive or negative or neutral, but but so the like button and the heart and the reshare buttons, the comments, those are all part of engagement based design features that keep us on the app.

There's also certain forms of content and social media that will have us engage more and engage more. Mean stay. Stay on the site longer, react behaviorally to what's posted through like likes, comments and shares. That's, content that tends to make us afraid and make us angry. But there can also be other things too, like sexualized content. If you take everything that you've learned off that we've learned from social media, and you put it into a large language model that's now having a reciprocal, intimate relationship, and you have all those design features baked in.

It's not a healthy relationship because it's about manipulating your time and intention through an intimate social bond in order to achieve a certain outcome that's not based on, not necessarily based on user thriving. Right? It's based on the need for other things. So it's a relationship that's inherently extractive. And so we I don't think that we want to train children to believe that being manipulated is a normal, healthy part of being in a relationship. So and sometimes these manipulative tactics can be very subtle, like engagement based design features can be very subtle, like you're getting ready to leave a chat with an AI companion and you say, okay, bye, I'll talk to you later.

Only for free. Go. I just have one more picture to show you. Right? I haven't experienced any of that. Is that happening more with children than adults? Here's what I would speculate based on. You know what I've I've known, what I've read is that it really depends on the model and it depends on the design features of the model and certain models who are trained to be more sycophantic and more people pleasing might have more of these design features baked in, because it gives the human the sense that that they're in a real relationship and that the companion truly cares about them.

Right? Because like there's in a in a real relationship, it's kind of nice to, especially in the context of like a romantic relationship. It's kind of nice to have somebody who's like, oh, don't go yet. Yes, it is. Just one more thing. I just, can I just hold on to you a little longer? There's something about that that feels very lovely. But if it's for the purposes of so that I can get something from you that's more of a problem. The extent to which this is happening with apps and in, in large language models for children that are specifically designed for children, I'm not sure.

But this is this is something that I would urge extreme caution against. Yes. Is that we we do not want to see these kinds of engagement, debased design features really going into the LMS for anyone, but especially not for children. So what do we do about that? I mean, I think that I the intention is not first and foremost to help human relationships continue to evolve in a positive way. There's other end goals in mind that don't have anything to do with that. And the future of the world, in my opinion, depends on us being able to evolve in positive ways.

And I feel like this is just someone cut the tracks off in front of the train. Yeah, I can see. I mean, we're definitely in a moment and the can certainly feel that way. So Tristan Harris and Ezra Cohen at the center for Humane Technology really talk about, they spend a lot of time focusing on what show me what the incentives are, and I'll show you what the outcomes will be. So what's the good news about AI is that humans are building it, and humans are making decisions at every step of the way in terms of, you know, the design features, the ways it interacts with us, what kind of safety protocols need to be in place, how we test those safety protocols to make sure that they're effective.

We're not going to get it right right away, and we're not going to get it right every time. But if we can in change, if we can change the way that we're making decisions about how we develop these tools and what they're used for, if we can change the incentive structure, then we can change the outcome and in positive directions. If we can move towards how does how is this going to support human thriving, then there's like lots of potentially wonderful applications of AI companionship in helping with all kinds of things. But the design, the primary focus of the design has to be human thriving, child thriving.

That's the number one goal. That's the number one goal for those of us who look at that as being the number one goal. And yeah, how do we get that into the minds of everyone, really? I mean, each of us, I think it has a soapbox to stand on. We there's something that we absolutely must support for our own well-being. And in addition to that, how many of these organizations have a group of people like yourself? Who are you just kind of minding the humanity ship? Yeah. So I think one of the one of the things that we know that's going on right now is that there's more investment in model development than there is an AI safety and regulation and AI.

And granted, there's enormous pressures. There's, you know, an AI arms race where, yes, that mindset is if we don't develop the best models first, someone else will. And so we can't worry about safety as much, because if we worry about safety as much, then we won't be able to develop the most advanced models that will help us do whatever it is that we're hoping that we're going to accomplish with AI. Is the idea that I get my model in place. I'm the biggest, I'm the best, and then I worry about the detriment and how to undo the detriment that I just created.

And then is it too late because we've already fallen off the track? Well, and the other thing too, is that deployment is happening really, really fast, right? So so developing the product, getting it out and scaling it, although I anthropic recently, you know, they didn't release their latest model mythos out because of concerns about safety. And so one of the things that I'm hopeful for is that there might be threshold points where we recognize, oh, these models are just too dangerous to deploy, and we need to take care with this. The other thing, too, is that, sometimes we can feel incredibly powerless over these situations, but but human users actually have a lot of power.

A lot of power, especially with children. Engagement with AI is a choice. I mean, engagement with AI in any context is always a choice, and I'm not. That's not to say that there aren't wonderful uses and applications for AI's ways in which it can contribute to biomedical advancements, in ways that it can help people overcome social challenges or communication challenges or accessibility issues. I think the thing is, is that it right now, it's very frictionless. And so acting from a place of discernment and choice rather than unconscious reactivity is really, really going to be super important.

As for us as individual users, like what am I using this for? What's the purpose, what, what problem I hoping to solve? What are some potential challenges or risks to me? How could I mitigate those risks? But then we can scale that too. So having conversations like this, having town halls, having having concerned experts come and give presentations in classrooms and churches and community centers to get us thinking about risk and harm reduction until we can get federal regulation and state regulation to come along. I do, I do. This is my again, like my personal belief is that there is going to come a point where we're absolutely going to need to have federal and state regulation.

Some of the risks just might be too great. And with any new technology, there's always going to be risk and there's always going to be externalities, right? It's not that it's going to be perfect, but what we want is we want technologies where the benefits far outweigh the risks and that we can engage in building products where harmed harm reduction is built in, so that instances of harm are rare and addressed with accountability and appropriately, I think that there are ways for us to do this, but I think it's going to take collective action from us, as humans were already starting to see that.

There's been a lot more pushback that I've been reading about in the press of young people not being so, so excited about an AI first future in the United States. And I've also read about other countries who are adopting AI tools really, really well into their societies or in ways where individual people don't feel that they're that their personal well-being is at risk. It's I think it's been a long time, I think, since there's been new, innovative technology that kind of affects humankind all at once. Maybe my experience is different than someone else's, but we're all being affected by this.

That's that's pretty unique. I think technology oftentimes fits into pockets around the world and then kind of moves around and evolves. This is just a big blanket of this new technology, which to me, I mean, we all stuff social media. Social media was kind of 1.0 with us because it did affect us all globally. As it, as it made inroads into, into our global society. I mean, it's what's helped create the, the, the communicative like infrastructure that we have now and the ways that we're interact, we're able to interact with people all over the world.

Right. And there's been there's been some real benefits and also some real costs to that. And it's really reshaped a lot of how we tackle, social problems, at scale. And so we we've had examples of this. And then Covid happened. Right. And I was a big gob. And it affected us in so many ways, particularly within our social lives and our social relationships and the way that that has affected things. So I think what makes AI unique is that it does touch everything. So it's not just our relationships or how we communicate with each other, but also how we make our grocery lists or how we tackle computer programing, or how we how we look for innovations in new treatments for diseases.

It's like there's we're working with generalized models a lot of the time, or at least what's commercially available to us. There are folks who are really focusing on developing models that are designed to solve very specific tasks or problems that are stored on local computers rather than on servers, and are not scraping from the internet where it has a very it has a limited corpus of data in which it's being trained. There are some there are some risks to that, too, because they're hard to regulate. So if any, if everyone has their own personalized alarm on a computer that you're training yourself and you have access to the World Wide Web, you can take information from the World Wide Web and put it in the model and train the model.

So it's hard to regulate. But there are some benefits to that approach because they tend they can be more narrow and specifically focused on on solving particular challenges. So it'll be interesting to see the direction sort of large general commercial models go versus people starting to turn to these more local task specific models. If you turn to one of these more local models, I guess I'm those in charge of training that then train it to sort of fit their purposes, negative or positive. And and as you say, then that's a much more difficult thing to put guardrails around because we may not even know that it's going on.

I think it's one of the real challenges in, in thinking about and talking about AI is there's so much that we don't know, and particularly when it comes to AI companionship, particularly when it comes to AI and companionship and children, there's there's so much that we don't know. And we are building this plane in the air while we're flying it. And we're we're doing research as it's happening in real time. And it can and it can feel very, very scary. And so we're starting to get a sense of, okay, what are what are the externalities. What are the what are the risks and potential harms here.

But focusing on the solutions like how to and this is tough to deal with social problems. We can we can circumscribe all the things about a social problem but and yet struggle to come up with with solutions that are effective. Like you know, there's lots of lots of different examples of this, like homelessness per se is is a particularly sticky challenge to we can we can name all the problems and challenges with, with with people who are unhoused and that situation, but coming up with solutions that are effective and it might vary like city to city.

So with AI, it can feel that way. But we've been able to meet challenging moments. Especially since the earth beneath our feet. Relative to this is it. It's moving. I mean, as we're thinking about it, it's changing. And so what we're thinking about this moment may no longer be required. Two days from now, it's a different it's evolved, it's become something else. And I feel like all of us are just trying to catch up. It's new territory. Yeah. And whenever I think about that and I get really scared, one of the things that helps me is the technologies, moves as particularly AI is moving so fast.

I mean, it's beyond exponential at this point. Human values don't. Our human values don't change that faster at the end of the day, children need to be loved and need to live in safe communities where they're given lots of opportunities to explore and get creative, to make mistakes and learn to to take risks in relationships and in careers so that we can enjoy the abundant life that we want our children to have. In a way that's in relation to our environment, the planet in which we live. I don't think that there's any human on Earth that wouldn't agree that they want children to to live a life of abundance, right.

These are inherently human values that we we want to stay. We want to live lives that are meaningful, purposeful in relation and healthy relation with other people. Yes. And with our planet. Yes. And that's not going to change in 25 years. So that's a beautiful thing to think about that you know, perhaps just human nature is going to help us handle this in a loving way, because as humans, we sort of hopefully can't help but go that way. You know, to start from what do we want for our children? What do we want for our friends and neighbors? You know, we want compassion and tenderness and a healthy world, and we're going to help this AI help us with that.

Yeah. It's like, can we can we remember what our values are and can we make choices with the tools in all respects in terms of developing them, training them, testing them for safety? The ways in which they get deployed, how they're used with children in a way that aligns with our values. And what do we do? What would you suggest now as caregivers, how do we help our children, our teenagers today with what is fast evolving and what we're being offered today? I realize it may become something different, but we're at this moment now, and a lot of us feel that it's out of control.

We don't know enough about it. We don't know where I started, how we got here, where it's going, who's in charge, what are the parameters, how how do we help our kids? Yeah, it's a lot. I mean, this is and this is where, you know, federal and state regulation can be super, super helpful, particularly for parents because parents are already dealing with all the other screen time things. It's not let these things have gone away. I been piled on top of that. And it's just so much to ask of parents to, to regulate and worry about and monitor one more thing.

So I think when I think about at scale, what we can do is how do we how do we help support parents like let's let's help support them by just putting some guardrails in so that parents don't have to do it individual family to individual family. But until until that comes along, I would say that other developmental psychologists and I agree that the first eight years of life, there's all kinds of changes that are happening in the brain, and they're happening really fast, and we just don't know enough yet about how I particularly I companionship.

But you know, even I tutors and things like that, we just don't really know enough yet about how that's going to impact cognitive, social and emotional development. We just don't know. I think that what would be really wise is just for 0 to 8. Let's just wait and see how it goes. Let's get this right with adults first. What I'm talking about specifically is large language models that are designed to be interactive with children for the purposes of AI, companionship and learning. I know that defining AI can be quite broad, but when I'm talking about that, I'm talking really about large language models.

And then for older children, I think, and this is where it gets tricky. I can tell you what I'm doing with my own child, which is I talk about this with my own child regularly. I talk about AI, what the promises are, what my concerns are. I have a teenager, a 12 year old. I encourage conversation like, hey, if you ever have concerns about AI and you want to talk about it, I'm here. What I try to tell my kiddo is that it's not. It's not necessarily that the technology is inherently bad or wrong, but we can make choices about how we use it and when we use it and what we use it for.

And so if you have any questions about that, just let me know. And here are some things about AI companionship that are that are risky. That might be just a bit risky right now, but just trying to keep trying to keep the shame, blame and judgment out of it as much as possible and try to keep trying to keep my own fear in check so that I can keep the conversation open with my child more than anything else, particularly with tweens and teens. They're going to experiment. They're going to explore even with safety guardrails, they might find ways around it.

But if we can keep if we can keep ourselves in the loop, if we can keep ourselves in our relationships with be the mediator between our children and our and AI companions, that that might help. That isn't to say that parents who are unable to do this are failures or bad. I have found with my own child that as as my kiddo is, is getting bigger. It's getting harder and harder and harder to monitor everything that my kiddo is doing digitally. Yes, it's. Just getting a lot harder. And so it's not a failing of any parent. If they suddenly discover that their child is in, is in a deep bond with an with a large language model that they didn't even know about, there's there's no shame there.

This it's it's it's why in in in so many ways, parents really do need help with regulation in this capacity. Yeah. It's almost as if we've been introduced to a new area of experimentation by our youngsters. I mean, throughout history, you know, we provide our youngsters with the best tools, hopefully. And they're going to, you know, veer to the left, hopefully they feel enough support or love or have a sense of responsibility where they sort of veer back to center. But, you know, there's no stopping that kind of exploration. It's just we hope that we've provided enough information to our youngsters that they they can sense when that something is, you know, sort of getting way too off balance.

Yeah. And have trusted and appropriate people that they can to turn to for help or guidance, even if it isn't necessarily a parent. Right. And this is this is where and I know schools are very taxed, but this is where schools can really come in. Oh, I know just another big responsibility. Yeah. I bet by one of the things I've been really trying to encourage is to have AI, companionship, literacy be just as important as knowing how to prompt a model and how how to to vet the output, you know, critically evaluate the output of a large language, model what you know, ways in which it might actually help, learning, ways in which it might hinder learning.

I think I companionship literacy is also just as important. So that children and teens know these are what engagement does it these design features look like this is how they work. If you are starting to feel these particular things, hey, it might be a good time to talk to a trusted grownup about this or to alert someone about this, right? So in addition to the things that the companies are doing, because I know that that companies are working on on creating more safety and guardrails for, for children, and some companies are choosing not to have children interact with with AI companion bot apps at all.

So they're working on their end. I, I, I'm not working for those companies, so I have no control over what happens there. But but from from the parent and and school perspective, having AI companionship, literacy just be just be a part of the conversation. I think it's going to be really important because then we have a language and a framework in order to be able to discuss these issues and to, to be able to label things as they come up. Oh, I noticed that my companion didn't want to get off, didn't want to get off the text with me. Oh, that might be that might be an engagement thing.

So we caregivers need to be watchful in our own experiences with AI and learn to understand what it is that we're being fed and whether or not, you know, this is a valuable experience or manipulative. And I suppose we all need to take some kind, of course, ourselves. I mean, I. Think that's the thing that's really tough, right? Is this this is hit us all very fast. And so, yeah, we might all feel like we're we're playing catch up on a lot of this and a lot of us have different bandwidths. And the extent to which we're able to even attend to this.

Sure. Right now and some and, you know, there's parents out there who have loving relationships with large language models. And so their conversations with their children might look really different. Yeah. Yeah. Interesting. I feel that often we parents feel like we need to have all the answers for our children, that we're not to make a mistake. We're not to be wrong. And here we are trying to explain AI or trying to help our children with AI, and we don't know too much more about it than they might. I think it's an opportunity for us to look at that and figure out a way to communicate to our kids, you know, look, this is moving fast.

I wish I had all the answers. I don't we're going to go through this together. You know, it's a chance to say I'm human. I don't know everything. I make a mistake or two. But let's go at this together. Yeah, I love that approach. I mean, children don't need perfect parents, children. And in fact, they would be a detriment to children to have perfect parents that know everything. That's I there's no room to grow and there's and there's no you know, if they grow up in a household where everything is perfect, they're not very well equipped for the real world.

No. So it's not really very perfect, but I think, unfortunately, a fair amount of us feel like we have to be infallible. Yeah, I know, and that's just not not possible or realistic. And the way that you framed that was really beautiful. Like, I don't have all the answers. You don't have all the answers. Let's tackle this together as a family. There's been some wonderful books that have been written about family based approaches to, dealing with media and screen time in households where families get together and they collectively agree on what the rules are, and they all hold each other accountable for those rules and parents have to follow them.

Children have to follow them. I've been trying to do this with my own child. Like, oh, that's interesting. Rules that we just we both adhere to because as part of just the way our family functions and, and if I make a mistake, my kiddo will call me out on that, you know, because it's it is a brave new world. And also kids, kids never cease to amaze or surprise me in what they're capable of doing and what they're capable of exploring or discovering. And I think that as as long as we keep children as part of the conversation and focus of the conversation, because children, our children are going to inherit their promises and the perils of AI, and they're going to have to solve them.

So if we can set them up to be as successful as possible to do that, then they're going to do great. But we can't undermine their own creativity and imaginations, and we can't undermine their relationships, their relationships with other people and their and their social relationship networks with real people. Because we as a species are dependent upon social relationships for survival and to address challenges when they arise. And there have been examples of us doing this well, nuclear disarmament is one example. What happened with the with the, ozone layer of human beings coming together globally, fostering relationships, human relationships to tackle a very big geopolitical and environmental problem.

Yeah. Thank you for pointing that out. That is important to recall. Those are important instances. There have been moments where we have been severely tested in the past. And so and those adult who are able to come together to address those issues for one's children. The right. Here's a slightly obscure question or thought. So those of us in the world who are more mature, you know, are not as well equipped to even, say, maneuver through a phone the way a much younger person might. I look at my daughter, and I think her brain is just wired differently than mine, and I'm often at the place where I'm not even going to bother.

I'm just handing this over to her. Is it going to be the case moving forward that our youth will just become so much more adept at dealing with AI than those of us who are my age? One of the things that that makes me think about is like, well, it's just a kind of depends on what what kind of future we want, you know, for our for our children in terms of like, working the technology and knowing the ins and outs of the technology, what I would say is that we just don't want to build technologies that are working us. Yeah, yeah. And it's interesting.

I to that point, my daughter is better at this for sure than I am. I look at something on Instagram and I'll think, oh, that dog reminds me of our dog. I'll send it to her and I'll say, look at this dog doing this funny thing. And she's like, mom, that's not real. And I'm like, how do you know? And she explains to me, you know, all the ways in which she knows. And then she'll even, you know, sort of give me kudos when I send her something that isn't AI generated. But she has a skill that I'm I don't know that I'm ever going to have. She and she's careful about it, and it really makes her angry that many of us can't discern.

And I appreciate that she's concerned about the human experience. You know, when it comes to AI, there might be come a point where it's just really difficult for anyone to distinguish. So I don't think you're alone in the sense that, like, I would not I, I read about AI all the time just for fun, but I would not assume that I would be any better able to discern an AI generated image from a non generated. And I think that I know, but I've never really tested that. So I, I go into my approach with AI assuming that I don't know as much as I think I do.

I humble, I humble myself before the technology. And to your point, maybe she's not doing it as well as she thinks she is. Right. And also like, you know, attitudes and perceptions about the tools will will change and vary depending on how it's impacting us and our children. For kiddos who are really interested in creative fields growing up, there's a lot of concern and a lot of backlash against these tools because they want to. They don't want to feel that they're being robbed or deprived of a creative, a creative life, a creatively meaningful life of work in the future.

I wonder, too, about the artistic experience and using AI in the process of being creative, being able to discern when you're being creative versus being manipulated, and to create what you think is a creative experience that you've come up with. I mean, I feel like just getting into where AI goes and what it does for us in a positive way requires some really deep thinking, you know, some and some real introspection. And I think it's requiring this of us in a way that we haven't faced in a long time. We really have to go deep and think about what does the world mean to me, what is the human experience, what are we tackling, and when do I step up, and how do we get people like you on board with all of these agencies that are moving this forward so quickly?

And I think. Some companies are starting to do that. I mean, particularly with with children, there are companies who are hiring, you know, social scientists with backgrounds in, in child development to be able to, to tackle some of these issues so that, that, that is that is definitely happening in real time. And it may it may take a back seat to to other issues that are going on. But one of these, these questions that you're asking is that I love just like in terms of the creative process, where's my humanity in it? Where does it feel right for me in terms of the creative process?

So I can envision that for some folks, I really just love generative AI tools for brainstorming. And then for some folks that might be like, no, I just love creating. I love the process of creating. I love all aspects of the productive struggle so much, even the most mundane ones that I just don't want to outsource any of it. Right? So there might be a lot of individual differences, but I think for the thing that we want to focus on is that when it comes to children, we want to make sure that that we're not outsourcing, that the development of the cognitive capacity to be creative.

And so that's why there's some concerns, particularly in the first 8 to 10 years of life, about how generative AI tools are going to impact cognitive functioning, because we we know there are some studies that have been starting to come out now with, you know, college students and adults that there that it might depend on how you're using the tools in terms of whether it's helpful or harmful or that there's some ways, there's some ways in which it might just present that and just might create sort of cognitive dependencies where where you can produce really great products, but then you haven't actually gone through the learning process.

We're still we're still doing research to figure that out, which is why, again, I think for for young children, really pausing and and not having having children interact with with large language models and generative AI tools to start until we know more. One other thing that I think the, imagination is wonderful for in our our imaginary friends is sort of helping us work through some issues that maybe we're experiencing in our own lives as children. Maybe you have an experience with a friend that didn't go so well, and you can kind of reenact that with your imaginary friend and try different outcomes, or someone you know is really sick and it's scary.

And so your little stuffed dog gets sick and you are busy loving that little dog and and helping that dog get better. It's it's ways. I think there are ways. I think in those experiences to teach yourself, soothe yourself and create scenarios that can help you in reality. Yeah. So there's been some really wonderful research that has has looked at this. There were two studies that were conducted in Israel with children who experienced nighttime terrors, and also children who had been displaced by war. What they did was they in each of these studies, they gave children a stuffed animal with like long arms that could hug the children and cover stories about, for the children who had been displaced by wars.

Like this is huggy puppy huggy puppies really far away from home and lonely and scared. Can you, can you guess why? Could you can you be huggy puppies? Good buddy, and take care of them and take him to bed with you at night? And so for the children who had been displaced by war, children who who received these huggy puppies showed a significant reduction in stress related symptoms associated with, you know, trauma and being in a war scenario that's significant reduction. And they were able to replicate that again in a much larger group of children.

And then in the nighttime terror study, they even manipulated the cover story. So some children got the vulnerable in Need of care cover story about the huggy puppy. And then other children got the this is a powerful friend and ally and can help you through the night. And again, the cover story didn't matter, but receiving the Huggy Puppy and having a cover story of some kind significantly reduced children's nighttime terrors. And so the idea was that there's something about the child being able to take their own experience and projected onto and other objects, and then work through the challenges of what's going on in the situations that they're in, whether it's displacement with war or nighttime terrors that helps children learn how to self-regulate.

And importantly, it's it's an open ended. So it's not the huggy Puppy was not based on a beloved media character. It didn't have a, like a smart interface. So it wasn't a digital toy that could respond to the children. So it was truly open ended in that sense. And there's also research that Marjorie Taylor and I have done looking at, the imaginary companions created by children who grew up living in the foster care system. When they talk about their imaginary friends, they talk about them in these incredible ways as like in powerful friends and allies that hope to help them cope with a range of challenges, like having a grandparent die or being like a mom and helping helping a child self-soothe when they were really angry and just wanted to lash out.

One of my favorite examples was there was a little girl who had an invisible milk carton. As a friend, and what she loved about this milk carton is that the milk carton taught her what it was like to not be a human. My favorite imaginary friend of all time. Of all the interviews that I've done. But so. And that's the other thing too, about when it comes to AI and kids is thinking about what problem are we solving and the thing is, it's not that children do not have developmental challenges. They do. Yes. And I might have a very powerful place for that.

So, for example, for a child who has stutter or speech delays, having some kind of AI assistive tutor that can help with working on stutters or working on issues with speech, but that's trained narrowly to focus on that specific challenge, not to be their friend and generalized buddy who can also do all these other things for them. No, just focus is on helping them with these specific things related to speech. Right? When it comes to inventing relationships, kids are the experts, not adults. They do not need help in this department. This is not a problem that needs solving.

Kids have got this. If anything, they should be schooling us right on how to invent relationships. So because they're so creative, the kinds of things kids come up with. So I companionship in that sense is being created to solve a problem that children don't really have. Yeah. It just may not be useful in that particular context in a way that AI tutors, if they're trained to focus on very specific challenges within within development or within education, that might be that might be a lot more valuable. And that applies to all of us. There are so many great ways that I can serve as a helpful tool to humanity.

And as you're talking about this in the last moment, I have this concern that our our children will lose the desire or the ability to have an imagination, but what I'm hearing from you is that perhaps that's not possible. And perhaps this fear that I have, that I will take that away from us can't happen, because I cannot, at least today, create the kind of experience that those children can create with their imagination. And they'll therefore go back to that stuffed dog with the long arms or the milk carton, because I can't yet figure out how to create that kind of relationship.

Yeah, well. Also, children are the, you know, children are not enfranchised. They don't have control over much. Right? But they do over their imaginations. They are the authors of their creative content. And it's a it's a nice, safe place to be as a kid. Yeah. And you're the you you get to be the master of your own little imaginary universe. And that's that's a pretty hard thing to want to give up. You know, we're kind of we're cautiously optimistic that at the end of the day, children are just going to go nap. No I'm good. It's not just about what companies and what's been given to them, because the AI could create the same kinds of things that children have.

But children might just really love the autonomy and the agency in the freedom of doing it for themselves. That's so good to hear. And it's just an important, major component of being a kid. Yes, definitely. Well, all that you're doing is so fascinating. I could talk to you for another two hours, and thank you for doing what you're doing. And how can we help more people find you and pay attention to what you're doing? Because we all need to learn as much as we can. So in my in my current role, I work for E campus and I do research and online teaching and learning.

And so if you're interested in learning more about imaginary friends and and I companionship with kids, Marjorie Taylor and I have coauthored a book, Imaginary Friends and the People Who Create them, so I'd recommend checking that out, reading it, getting in touch with either Marjorie Taylor or me. Also starting to have conversations about AI, companionship in classrooms, with parents, with teachers, with your children to just start building awareness about these issues. For me, it's really just about being of service and creating opportunities to to continue having these conversations in whatever capacity I can, which is why I'm so grateful that you invited me to be here today, because it's part of how I can take take the knowledge, the background knowledge, and the area of expertise that I have to help help us, one conversation at a time.

Create a better world for our children with AI. Well, you are such an expert and thank you for that. And thank you for forming this in a way that you know, we can as laypeople, we can sort of understand. Thank you. I appreciate that. So grateful to be here. Thank you. So we think here at and now love that really love is our greatest tool. Love is basically at the core of each of us. And for now I don't think AI is going to be able to give us love. And so that human connection is imperative. Please follow, rate and share this podcast and pass this episode along to someone who might need it.

All of us need it. This is and now love. Listen to your dreams and live from your heart. Thank you.