Chris Colley 0:08

Welcome back everyone. We are here at another episode of Shift Ed Podcast coming to you. Um I have a very special guest today. I was just mentioning before we hopped on here that um I've been following uh this person for a long time, a very unique perspective. Um I have Very Atkin, uh Dr. Very Atkin, uh who's a senior ethics fellow at the Alan Turn Institute, and also the co-founder of our AI collective. And her work focuses on AI ethics, but children's rights and children's relationship with AI, which I often find that we don't talk enough about how the kids are interacting with this new technology. Um and we're gonna kind of dive into that relationship um and how it's formed. Um for jumping on here today and sharing some of your thoughts with us.

SPEAKER_00 1:03

Yeah, no, thank you. Really pleased to uh to join the join the conversation.

Chris Colley 1:07

So maybe we'll start off very with a little bit of how do what were some moments that happened in your career that brought you to, you know, advocating for kids and and and AI and getting their voices heard.

SPEAKER_00 1:22

Yeah, um I guess I mean throughout all my career, my work's been around kind of emerging technologies um and particularly kind of um controversies around emerging technologies. And all my work's been really focused on participatory and inclusive approaches, so aiming to give a voice to the communities who are most impacted, but often the communities who are most impacted are also the communities who are least represented in decision making. Um, my career's had a few twists and turns. So my PhD was around renewable energy developments and looking at the role of rural communities around controversies around renewable energy developments. I then worked for quite a number of years looking at health data and looking particularly at patient perspectives in relation to decisions around health data before looking then at AI and finance and banking and then more and more turning to be focusing on AI. So lots of twists and turns, but for all of that, always focusing on um trying to give a voice to those people who are least represented in decision making. And in terms of AI, um the group who will be most impacted by advances in AI and who are already significantly impacted today are children. Um and children are also the group who are least represented in decision making, almost entirely excluded from policy-making discussions around AI, even when those discussions are explicitly about safeguarding children. Children's voices are so rarely part of those conversations. Um, so that's why for me I felt that this was um you know absolutely vital to develop approaches to give children and young people a meaningful voice, um, both in relation to decisions around how AI systems are designed, developed, and deployed, but also around policy making and regulation relating to AI.

unknown 2:57

Amazing.

Chris Colley 2:58

Wow. Great uh background, and I'm glad that your twists and turns have brought you here because we need your expertise right now so desperately. What have children been telling you about their experiences with AI? Like, I know it's a big question, but can you give us kind of an overview of like what are kids saying about it?

SPEAKER_00 3:19

Yeah, I mean, I suppose the first thing to say is that that kids, I mean, that's the most diverse group you could possibly think of, right? We're talking about kids from all different backgrounds, all different cultures, all different experiences. Um so the range of experiences with AI or digital technologies are hugely diverse. So we can't think of this as a homogeneous group and have a wide range of ideas, perspectives, concerns, um, interests around around AI and digital technologies. So I think that's the first thing to be really clear on that we we have to recognize that there's a diversity of experiences and interests. Um it's been interesting because I so I started working in this area five and a half years ago, almost six years ago, doing um engagement work with children and young people. Um, and that was before ChatGPT. So that was kind of before um before all the focus around generative AI. Um and so at that point we were talking more about how AI is used in decision making that impacts children's lives. So AI systems being used in healthcare or um in uh education or in you know kind of wide range of policy areas, um, and really talking about AI as kind of invisible computer systems that process data to make uh decisions, which yeah, maybe not as exciting as some of the headlines around AI right now. Um and then over those years it's been interesting to see how it shifted. And and now when we we talk to children and young people, AI is a very familiar term, but it's normally understood in terms of of generative AI, which is just kind of one category of AI technologies. Um and I think yeah, we we get a wide range of different responses. Um there are lots of kids who find it quite exciting, find it quite fun, enjoy playing around with AI tools. Um, more and more kids are telling us that they use AI tools as a as a way of finding out information, so as a source of information, asking ChatGPT or asking Gemini, um, or maybe just through having AI summaries or AI overviews that come up on on search engines like Google. Um so kids are definitely interacting with AI more and more, and um and and and that's many of them having kind of fun with it. Um but we also find that there's a lot of concern. I every time I do any kind of engagement work with children and young people, there are major concerns that come out. And any excitement about AI is always balanced by concern, like that is very consistent. Um some of those concerns are around um bias in AI models, and that bias in AI models leads to unfair decisions, or that children feel unrepresented or misrepresented through AI models. And that's particularly children of colour, uh, minoritized children, um, children who who don't feel that the outputs of AI models represent them. Um and that is a significant set of concerns. Um, we also find major concerns about environmental impacts of AI, generative AI. Uh, and that's something that we hear really, really consistently. I was at a um Edinburgh Book Festival last month and our AI collective. We held like a kind of drop-in session for kids to come along and talk about AI. And there were a significant number of kids who would just walk past our stand and shout, I hate AI because it's destroying the planet. And that's like without any prompting, that that's what that's what comes to mind.

Chris Colley 6:29

So there's like an internal like issue with it, like like visceral, like yeah.

SPEAKER_00 6:36

I think it's really interesting because we often there's often this assumption that kids will be like will become over-reliant on AI systems, on digital technologies, that they're going to become uncritical, that they're just gonna be sucked in by all. And actually, I I find um that children and young people are often the most critical, like them have the most critical perspectives. And the concerns that they raise are informed and they're valuable, and they're concerns that we absolutely must take on board and address.

Chris Colley 7:04

Right. And and why is it that we tend to not listen to the kids? Like it seems like when frameworks are being put together and trainings and like all this stuff of that will directly affect them, that we just don't even bring them to the table. Like, what what's the deal with that? Like, where's the disconnect happening?

SPEAKER_00 7:23

Yeah, uh yeah, it's it's a huge, huge problem. Uh, you know, under the UNCRC, the United Nations Convention on the Rights of the Child, children have a right to a say in matters that affect their lives, and they absolutely should be part of the conversation and they should be shaping um you know the frameworks, the the safeguards that that are um you know ostensibly being developed for them, but without them. Uh and I think yeah, it's it's it's a huge problem. We have to have this you know these assumptions assumptions that like adults know best, like let's talk to the experts. Um but adults aren't the experts in children's experiences with AI. Like no adult today has experienced growing up in a world with generative AI technologies. No adult has ex has experienced that in childhood. Um so if we don't listen to children, we we miss those perspectives. Like we miss understandings about how kids are really experiencing those technologies, both positively and also what the risks are. And we miss like really important insights into like what would it mean to develop appropriate and effective safeguards regulations. If we don't understand how kids are actually experiencing the technologies, then we might miss really important things when we're developing the responses. Um so yeah, I think it's it's it's absolutely crucial. But all too often kids are kind of dismissed as um, you know, do they understand it well enough? Do they do they have the right information to make informed choices? And well, they do because what we're talking about is their experiences and and their needs.

Chris Colley 8:52

Yeah, I love that. But I mean, what changes when AI becomes a part of kids' kind of you know, social worlds? Um like are there is there research around that? Like once they start integrating a part of who they are socially um or how they connect with the world, like how does that change the the kid? How does that change the interactions? Like, have you done any kind of like thoughts on that or research around that idea?

SPEAKER_00 9:26

Yeah, I mean it's definitely a really um a really important dimension of this. Um and there is there's a growing evidence base, um, there's a growing evidence base very clearly of of some of the harms that have occurred, um, tragic, tragic harms that have occurred. Um, and there is a growing research field that is seeking to understand this. Um, but it's not enough yet. And I think one of the big problems that we have here is that um, you know, these AI models are kind of released into the world um before there's been the you know full research into what the potential impacts are, before the potential harms have been anticipated, and before the safeguards needed to address those harms have been developed and put into place. Um, that's like an incredibly dangerous, risky, irresponsible experiment when what we're experimenting on is is childhood and children's development. Um but yeah, when we see lots of uh, you know, everything from like smart toys that are marketed towards infants and preschool children that integrate um generative AI models, some of which, you know, there've been smart toys that have used ChatGPT, which is not designed for children. I mean, it's specifically, you know, there are there are minimum ages on it. And yet these LLMs are being deployed in um smart toys for very young children. Um, you know, and there'll be cases where that's led to these toys you know, talking about things like drug use, sexual fetishies, you know, like like just basic, basic things that there should be, you know, we should be able to stop this from happening.

Chris Colley 10:54

Yeah, some kind of you know, walls or something.

SPEAKER_00 10:57

It seems it seems so clear. Um and then there's the more kind of, I guess, subtle sides of it of that. Then these are systems that are they're designed to mimic human emotion, they're designed to mimic human communication. Um, and when we begin to integrate these into toys that that very young children interact with, we don't yet know fully what the impact might be on on children's social development, on the cognitive development, on their like their understanding of the world. Where you know, toys is about or play is about um learning and exploring and and developing understandings of relationships and and social dynamics. And when we have something that can very convincingly imitate human connection, we don't really know yet what what the impact of that might be.

Chris Colley 11:41

It's like we don't really understand AI all that much yet, right? Like even the CEO of Anthropic was saying, like, we know like maybe about 3% of what happens inside that box, you know, inside that language model. Like it still seems like we are struggling, adults, with understanding, but I do think that we need to get more voices at the table. What do you think the adults are missing when they're making decisions about kids' interactions with AI and and futures with AI? Like what what is actually you know, like child-centered AI actually mean if we're if we're looking at, you know, are we making these tools or creating these environments that are gonna be safe um and regulated? Like, are adults missing something here in this in their decision-making process? Because it kind of feels like they're it's all over the place, you know, like somebody coined it as like the wild west, you know, where everybody's just kind of doing whatever they want, let it rip. Like there's like decisions that are being made that are going to affect our kids. So what does child-centered AI actually mean?

SPEAKER_00 12:54

Yeah, well, I think really importantly, child-centered AI is kind of addressing just exactly that situation that you're you're you're illustrating there, in that it's child center's child-centered AI is about really putting children's rights and children's well-being at the heart of development. So it's not just about identifying risks and harms and making sure that there's safeguards in place to to address the risks, but it's about, you know, how could we develop technologies better, how could we develop systems better so that they're actually positive for children and young people? Um, and that means that children's well-being is the starting point, not an afterthought. And I think at the moment that's the problem that we have. Many of these AI tools, they're developed, um, well, the generative AI tools, right? They're developed to be general purpose. They're developed to be kind of the biggest, shiniest AI model that can do everything. Um, and then it's deployed into lots of different contexts. It was not designed for. It was not um, it's it's not designed to um be safely used in relation to children's mental health or children's education. It's just it's not designed for that purpose. And yet we're deploying these systems in contexts where it's directly impacting on children's well-being, children's mental health, children's education. Um, and that is a that is such a dangerous, like dangerous approach. And I also say, like, I mean, what you were saying about like that we don't know um, you know, about AI, or that even that the kind of the heads of AI companies are suggesting that they don't know, I think that that is um that's not quite true, right? We we we do know, we should know, and those developing these systems absolutely should know what their models um can and can't do, and they should be anticipating what the harms are. Um, in terms of like impacts on um relationships or social dynamics or the ways that children might develop emotional attachments to AI models, like there's been 60 years of research into human relationships or chatbots, you know, from the early 1960s, where it was the first chatbot Eliza, um, where it was demonstrated that even where people had an understanding that this was a computer program, and it was a very rudimentary computer program, still people would develop a sense of attachment and a sense that it was having that it that it had emotions or it had intelligence or it had an a possibility to have empathy towards them. And over those 60 years, you know, like what we've developed now is is so convincingly replicating human emotions. And that's not an accident, it's not by chance, it's through um significant, like these AI companies employ significant numbers of um behavioral scientists, of psychologists to develop these systems that are, you know, they're they're explicitly designed to be uh to maximize engagement through mimicking human connection and human emotion. So some of these things about like, you know, this shouldn't this is not surprising. It should not be surprising. And I certainly don't think it's surprising to the those who are developing these systems because that's explicitly how they've been designed and and and developed.

Chris Colley 15:44

Um so yeah, I think that's it's similar to to social media in a way, how it was just thrown out there. They knew it was going to be addictive to kids, like they don't have the filters to okay, 10 minutes a day, like and just went head, you know, and now we're seeing the consequences, and now finally some regulation is starting, right? Like are we destined to go through this same process with AI, do you think?

SPEAKER_00 16:13

Uh uh destined, no, but I guess it's it's it's about you know it's about it's about business choices and business models. And sadly, what we have seen is that although there has been so much learnt about what's going wrong with social media and the negative impacts of social media, they're kind of repeating the same mistakes again with AI companions and um and you know AI um chatbots and systems that young people are interacting with. It's the same pattern, you know, the same thing where systems are being designed to maximize engagement, but with with incentives or persuasive design features that keep people going back. Um, because the you know, like we might use an AI tool or an AI system for like one particular purpose, but really the business model behind it is more about the the data that's extracted through that um interaction or getting people to subscribe on subscription models. It's about keeping people engaged. Um and they're designed explicitly for that purpose. Whatever it is that whatever it is that we're being offered, really what's happening is um creating a an engaging, addictive um interaction that we keep going back to.

Chris Colley 17:17

Uh and that is um how you make money, right? Like, I mean, they see the writing on the wall that if we want to make prop profit, we gotta get kids addicted to this so that they use it throughout their whole, you know. And I do see the traces of that with AI, as you as you've echoed as well, um, that there are issues that we're kind of like sweeping under the t, you know, we just not dealing with them. Which brings me to this question, Very, that uh where does like does it have to start at AI literacy? Like where we're not talking about the tools and but we're talking about the ethical, the safety, the be careful of like I feel again we're just kind of missing the boat. And can it start at AI literacy? Like, could we integrate AI literacy across all school? And like like, wouldn't that be a better idea than kind of like I hope they learn about these models and what they do? Good luck. Uh see you later. Like, yeah, where do you sit at at that? Like, where where's our starting point, I guess? Like, where can we grab onto listening to kids for sure? I love that idea, but we gotta start like putting action to the words that we're saying. Like, where do you see like AI literacy fitting into this whole kind of like system that we're creating?

SPEAKER_00 18:37

I mean, I think it's absolutely vital. Um, and it has to be critical AI literacy. So I think there's a lot of um yeah, you'll you'll find lots of approaches, lots of programs that are emphasizing the importance of teaching kids about AI. And what they're really teaching is about um maybe a little bit about coding, but they're teaching about prompting and then teaching them how to use these tools, and maybe there's a bit about how to use them responsibly or how to use them safely, but it's about using them. Um, and it's sort of suggesting that inevitably we're going to use them. Inevitably, this is um, you know, these tools are gonna be part of how we how we work, part of how we access information, a part of how education is delivered. Um, and that misses a really big fundamental part of this, which is do we want them in the first place? Like, should we be using them? For what purposes? Um, and I think that that is absolutely fundamental. So when we teach kids about AI, it first of all, we shouldn't approach it as a technical subject, right? There are technical dimensions, but you can have an opinion on AI and you can um feel confident to engage in discussions about AI, even if you don't understand the technicalities of how those models are developed or you know how they're how they function. Um, because these are systems that impact increasingly every aspect of our lives and the systems we interact with and policy areas. Um so I think that that's really, really important. And then when we talk to kids about AI, um, we need to talk about um, yeah, not just the technic how they technically work, but like the design choices that have gone into those systems, the business models behind those systems to help children to make informed choices so that they know that if you're interacting with an AI model, really what's happening is it's collecting that data. And that data might be used for marketing, it might be used to develop future models, it might be integrated into training model of future AI models. Um, all of that is really, really important to I guess to empower kids to make informed choices, to understand what's really happening. Um and then also to talk about how it impacts their rights and and what their rights are in relation to digital technologies. I think that that is you know, that's really fundamental so that kids learn from from as early an age as possible that where AI systems are being implemented or where they're being used in ways that impact their lives, they can say no. You know, they can ask questions, they can um demand explanations for how those how those technologies are being used and how they've informed decisions. Um, and that's really important because otherwise, it seems particularly in education where AI systems are being deployed and rolled out in lots of ways, where kids are never asked for their views on it or never really meaningfully asked whether they consent.

unknown 21:11

Right.

SPEAKER_00 21:12

And it kind of normalizes the use of digital systems in such a way that kids might grow up and when they leave school, when they go into the kind of future workplace or even to kind of future society and communities where AI tools are used to survey um communities, or used to monitor them, or used to sometimes make unfair or discriminatory decisions about them, they need to know that they have rights and they need to know that they can um you know hold decision makers accountable for the ways that these systems are used. Uh and to do that, to equip them with that confidence and that ability to do that, we need to start in in early education in showing them that where these tools are used, that they have a right to say no, and they have a right to demand explanations as to why and how those systems are being used.

unknown 21:56

Well, I love that.

Chris Colley 21:58

That's um an amazing way to end this. Um, very thank you so much. Um you've made me a hell of a lot smarter, I'll say that. Um but your input to this is so um needed right now. Um so I really appreciate your thoughts and your ideas. And I'd love to have a part two of this because I have a ton more things that I'd love to talk to you about. Um so I'll invite you back for a part two down the road, and uh we'll see where we're at too, because this stuff is changing so quickly. Like the evolution of it all. So thanks so much. I really appreciate your time.

SPEAKER_00 22:33

Yeah, no, thank you. Looking forward to part two.