# Transcript: Featured Session: How to Support Resilient Youth in an AI World **Date:** March 12, 2026 · 10:30 PM **Session:** [Featured Session: How to Support Resilient Youth in an AI World](/sessions/2026-03-12/pp1149718-featured-session-how-to-support-resilient-youth-in-an-ai-world) ## Summary This Brookings Institution-led panel examined how generative AI is reshaping young people's learning, social-emotional development, and sense of purpose. Drawing from a 50-country research study, the speakers found that while AI offers transformational benefits for neurodivergent learners and personalized education, the risks — including cognitive offloading, creativity narrowing, trust erosion, and student demotivation — currently overshadow the benefits. The panel advocated for intentional AI use, stronger safeguards, community-based AI literacy, and shifting from 'screen-free' to 'pro-presence' approaches to childhood. ## Topics `ai in education` · `youth resilience` · `cognitive offloading` · `neurodivergent learners` · `ai safety for children` · `student engagement` · `ai literacy` · `digital wellness` ## Key Takeaways 1. AI risks currently overshadow benefits in education — cognitive offloading, creativity narrowing, and demotivation are real threats, with less than 4% of high school students regularly in 'explorer mode' where curiosity drives learning. 2. AI is transformational for neurodivergent learners — it can personalize cognitive load, reframe problems for ADHD students, create synthetic voices for kids with aphasia, and make text accessible for dyslexic readers in ways that were impossible before. 3. Safety must be built into AI from the ground up, not bolted on — Google DeepMind's LearnLM embeds pedagogy directly into the model, and under-18 account data is not used for training. 4. The conversation needs to shift from 'either/or' to 'both/and' — not AI versus teachers, but intentional, purposeful use alongside human connection, with Gen Z themselves driving an 'analog rebellion' that craves real-world presence. 5. As AI moves beyond screens into glasses, clothing, and everyday objects, the parenting framework must evolve from 'screen-free childhood' to 'pro-presence childhood' focused on critical thinking, curiosity, and human connection. ## Full Transcript Rather than focusing on the question of whether AI should be used in education, we have to ask the question, how do we ensure AI strengthens young people's capacity to think, relate and thrive. Drawing from the recent Brookings Report titled "A New Direction for Students in an AI World: Prosper, Prepare, Protect," our speakers will explore what it truly means to support resilient youth in an AI saturated world. I am very excited to share with you what we found. And as an education expert, I've been thinking a lot about this, but also as a parent. I have two young kids, and I probably every day think what you all think if you're a parent, which is, how do we protect and prepare our kids for this crazy generative AI world we're in? That was one of the main questions that we had at Brookings, and we did a really big deep dive around the world, 50 countries, hundreds of studies. Talked to 500 students, parents, teachers, technologists, really asking this question about what do kids need, and are we on the right track? And one of our major motivations was we've seen this movie before. It happened in social media, when parents, teachers, educators, coaches were not at the table, and we want to make sure that we don't repeat it. The first takeaway is it's very confusing. Generative AI, first off, we found that the lines were totally blurred. Kids are accessing generative AI all the time, everywhere. The lines between technology for education, technology for entertainment, technology for communication have been all mixed up. One of the students I spoke to, a high school student here in the US, said, well, we just use social media. We go on to our AI friends on social media, Meta AI, Snap AI, we talk to them about our day or how annoying our parents are, and then we take a picture of our math homework and say, please do it for me. And it does. There are a lot of good things and a lot of bad things. How can AI help? We did find that generative AI can help kids learn and enrich their ability to develop independent learning skills and deep learning. For example, if AI is embedded in virtual reality, it has a huge potential to really help kids learn in totally different ways. We also know that it can really help neurodivergent kids. One of the most moving examples is kids with aphasia who have trouble communicating. Using generative AI to create synthetic copies of their voice, they can all of a sudden communicate with peers and teachers in the classroom. Teachers benefit too. Not only do they love the admin support, which they need desperately because they're overworked everywhere, but assessment can be incredibly more nuanced with Gen AI really figuring out where kids get stuck. Another example that I found so moving is diaspora Afghan teachers using Gen AI to make little mini WhatsApp based lessons and sending them to girls in Afghanistan who are banned from schooling by the Taliban during secondary school. At home, they're staying in school with little, tiny, adaptive lessons. Incredible stuff. But we also know that it can be really bad. One of the things that drives me crazy is when somebody says, isn't it just like a calculator? No, it's not like a calculator at all. The calculator offloaded arithmetic and a big slice of mathematics. If an AI chatbot is like a calculator, it would do math, English, physics, chemistry, biology, history, social studies, poetry, music, art history, take the SAT for you, and give you relationship advice. The cognitive offloading where kids are using it to do their thinking for them is dangerous. If you repeat it time and time again, kids are not wiring their brains to develop the critical thinking skills we need. We're also worried about narrowing ideas. This is a study by Adam Green and his colleagues that tracked thousands of high school seniors writing their college applications. When kids write without AI, they have many different unique ideas. When they write with AI, their ideas all cluster in the same area. They're all kind of telling the same story. It really hurts kids' creativity. On social and emotional development, one in three teens in the US say they like talking to their AI friend equally or more than a human being, and these companions are designed to agree with you at all times. If kids are developing their social emotional muscles by being agreed with all the time, how are they going to take feedback when they make a mistake? How are they going to be resilient when things don't go their way? And how are they going to stand up to the very real and very dangerous events when chatbots manipulate kids to self-harm and potentially take their own life? AI can also amplify bias. Research with Punya Mishra and others showed that if you are a seventh grader and you put the same essay into ChatGPT and said give me feedback, if ChatGPT knows you like rap music versus classical music, the kid who likes classical music will get a full grade level higher critique and suggestions. It also reduces trust. The instructional core, the relationship between learning content, teachers, students and their parents, that is the beating heart of how learning happens in schools. You need trusting relationships. Human beings have evolved our entire species to learn in relationships, and when you don't have trust, I'm really worried about what that will do for the ability for kids to sustain a positive learning experience. Because the risks are of a different nature than the benefits, we say at the moment that the risks are overshadowing the benefits as AI is currently implemented. Remember, it showed up to schools uninvited. It's the first ed tech you don't have to procure. One of the things that worries me most is this underlying demotivation. So many kids we talked to said, the chatbots can do everything, why am I here in school? What is my role as a human being? In our research, we found kids show up in four different modes: resistors who avoid and disrupt learning, passengers who show up but aren't interested in what they're learning, achievers who are chasing the gold star but are fragile learners, and explorers where curiosity meets drive and they become unstoppable. About half of high school kids say they're regularly in passenger mode, and less than 4% say they're regularly in explorer mode. But it is not too late. We can shift course to focus on the benefits and mitigate the risks. We recommend three things. First, shift how we do teaching and learning, support curiosity, foster and model the thrill of learning. If you're an educator assigning homework that can be hacked by ChatGPT, don't assign it. Second, prepare kids. It's time to teach kids about the online world, about ethics, what AI is good for, and how to create what they want with it. Third, we need safeguards on the commercial products to protect children. On children's safety, Miriam Schneider from Google DeepMind shared their approach. For DeepMind and Google, safety has to be table stakes. Their approach was to build AI purpose-built for learning and education through an effort called LearnLM, which infuses pedagogy into the model itself, thinking about teaching and learning first, not as an afterthought. They adhere to all regulations including COPPA and FERPA. Any under-18 account data is not used to train models. They purpose-build age-appropriate experiences with additional safety filtering for users under 18, and they give parents supervised controls and granular tools. DeepMind also does rigorous adversarial red teaming, bringing specialists to try to break their safety safeguards, exposing gaps and vulnerabilities to strengthen the systems. They view this as a team sport, working closely with cognitive neuroscientists, developmental psychologists, child safety experts, parents, families, and policymakers to inform what these tools should look like. Martin McKay from Everway, who has invested his career in assistive technology for neurodivergent populations, broadly agreed with the report findings. He highlighted two big risks: first, that learning comes from cognitive struggle, and if we outsource the struggle, we don't really learn. A recent OECD study had teenagers write an essay; those who used ChatGPT, 85% couldn't remember what they wrote three days later. Second, kids have access to AI even if schools block it, and the technology doesn't respond differently to a 55-year-old than to a child, without considering their stage of emotional development. On the optimistic side, Gen AI is game-changing for neurodivergent learners. Previously, conventional approaches to summarizing text or lowering cognitive load didn't work well. Now AI can rewrite content in seconds, personalized to each person's cognitive profile. Some people prefer bullet points, some like emojis with keywords. For dyslexic kids, reading aloud is one of the best things, and with AI voices now sounding super human, a teacher's voice can read content to them. You can keep all the text but lower the cognitive load, making it easier rather than shorter, and enrich text with images for keywords. It's completely transformational. Maureen Polo from Hello Sunshine shared that their research with Gen Z consumers revealed young women don't feel represented in the world of media. They built a Gen Z advisory board, including a young woman who is an AI statistician building an ethics code that major media companies want to learn from. These women know the world is cautious about AI, but they also know they need to step into it, because if women aren't part of building in the AI space, it becomes another area where they're not represented. Gen Z is driving what Hello Sunshine calls the "analog rebellion." These young women want to go out and meet people, sign up to use their hands and craft again, listen to and tell stories with others in person. They're craving the opportunity to be more intentional with their time. While people think this generation's interest in everything 90s is about nostalgia, a lot of it is that you had more agency back then. This generation wants that agency, the ability to figure out who they are and to be more curious and critical in how they think about the world. Miriam Schneider emphasized the need to shift from either/or to both/and thinking. It's not AI or teacher, AI or pen and paper. There are pointed, intentional uses where AI can be really additive, enhancing experiences and giving time back to educators to reinvest in building human connections. It comes back to intentionality, being clear about when, how, and why you are using a specific tool. Hello Sunshine built a physical, printed playbook teaching young girls how to be intentional about their time in AI. The girls had to write down what they were feeling and observing as they played with AI tools, creating space for them to really think about what they're using it for. Some of the advisory board girls said they want to write letters to each other more. The company is now building physical books and tools because that's what the young women asked for. For kids with ADHD, AI offers practical solutions. A teacher can take a math problem that doesn't engage an ADHD student and ask AI to reframe it for a kid who's super interested in hockey or football. Kids with ADHD often see a task, get overwhelmed, and can't break it down into subtasks. AI is really good at that subtask breakdown. A major concern is AI reducing connection between parents, teachers, and students. In workplaces, employees are asking AI instead of colleagues for help, leading to loneliness. In schools, when teachers use AI tools for essay feedback instead of reading them personally, kids interpret it as less care. One teacher found a creative solution: using AI to get structured feedback, then recording personal voice notes for each student, maintaining the human connection while saving time. Hello Sunshine addresses this by introducing AI technology in community settings where young women learn together, discuss bias in different tools, and talk about whether they trust the information. They create spaces for discussion because nobody is truly an expert yet. Google DeepMind's Miriam Schneider emphasized that AI literacy doesn't capture the full opportunity. It's about aptitude and attitude, what they call "AI readiness" — how do you become ready as a family or classroom to embrace what AI can do, be clear about how it works, and approach it with a critical lens. In closing, the panelists reflected on parenting in the AI age. Rebecca Winthrop noted that while the screen-free childhood movement has been valuable, AI will soon be embedded in glasses, clothes, pencils, and notebooks. The slogan will need to shift from "screen-free" to something like "pro-presence" or "pro-human" or "pro-connection" childhood. Martin McKay emphasized that parents need to learn about AI and take a position on it. Miriam Schneider encouraged parents to get involved in tech advisory boards at their children's schools. Maureen Polo's philosophy: create the space for kids to be critical thinkers, curious thinkers, to know themselves, and to be more joyful human beings, because that's going to help them be more thoughtful and critical as they navigate AI. --- *Source: stt · Language: en · Model: claude-opus-4-6* [← Back to session](/sessions/2026-03-12/pp1149718-featured-session-how-to-support-resilient-youth-in-an-ai-world)