How Generative AI Fails Kids and How to Fix It

This week has seen my thinking around AI in the classroom shift considerably. Working with year 4 pupils at Wellington College on an AI-supported project has opened my eyes to the massive potential AI holds as cybernetic classmate. Pupils used three platforms - Study Hall AI for deep research, Gemini for feedback on their design work, and Classforge for overall project support (this is the platform I've been building over recent months). As a tool for supporting pupil-led research (amongst many other things), Study Hall is superb. It's such a feature-rich platform and I cannot recommend it highly enough. Hats off to Sofia and team for what they have created. Interestingly, Gemini was more limited, and limiting, for pupils. On a number of occasions it refused to complete the simplest and most innocent of requests, which led to frustration and ultimately pupils preferring the other platforms. In Classforge, I designed a set of agents for project support inside one chat. They include an orchestrator, researcher, subject matter expert, designer and coach. Calling them by name brings them into the chat, along with their name and avatar. This was the biggest unlock of all. By the end of the project even the teachers were saying "let's see how Alex can help improve your design" or "why don't you ask Dr Sarah to help you understand that better." We talk a lot about ensuring we don't anthropomorphise AI, but for this sort of support, naming the different agent functions has been a revelation for all and has made the AI seem much less intimidating. But more than that, it has brought to light what I have been saying for some time, but have now seen first-hand. Bringing AI into the classroom, to work alongside pupils as cybernetic classmate, has enormous potential, providing we frame it as just that - a team of experts that the pupil can draw on whenever they need, that are waiting in the wings to help. There are many considerations to be borne in mind, and much to do in terms of understanding this incipient relationship. But it's a start, and a positive one at that. Pupils have had an absolute ball, and I know cannot wait to continue on their AI adventure.

This is really interesting. More of these types of studies need to happen. As written at the end of the article, '...educational AI must be held to a higher standard than technological novelty.' Teachers of all experience levels are right to be excited about the access to tools such as these but should also exercise caution and should diligently critique output before using tools ahead of human-created options. It does also challenge the wider discussion point, is storytelling in fact a truly human activity? One that AI (at this point) cannot recreate. The lack of 'soul' and a reliance on regurgitating pastiche versions of human-written texts would suggest yes.

🚀 Can AI Really Detect If a Kid is Bored or Overwhelmed in Class? Classroom engagement is evolving, and AI is stepping in to help teachers understand students better. From tracking attention to spotting signs of overwhelm, AI can provide real-time insights—but it’s not about replacing teachers; it’s about empowering them. In my latest article, I explore: ✅ How AI monitors student engagement ✅ Where it excels and where it falls short compared to human observation ✅ Ethical considerations and privacy concerns 💡 Takeaway: AI + teachers = smarter classrooms, but implementation matters. Read more here 👉 https://lnkd.in/gQXFspK6 Curious to hear from fellow educators and tech enthusiasts—how do you see AI reshaping classroom learning? #AIInEducation #EdTech #ClassroomInnovation.

Last week, I asked my son's principal if the school has an AI working group (or some group tackling how AI can best be used in learning) Turns out there's nothing. Not just in the school, but in the entire district. I was shocked! This is one of the top-rated public school districts in the entire state, with excellent teachers, engaged parents and solid resources. Yet there's no formal work on how AI is being applied in learning. It's pretty much the Wild West being led by individual teachers (adding one more thing to their already full plates). That’s terrifying. Because you know who is talking about AI? The kids. Even in middle school, many kids are using ChatGPT to write essays and research projects. Teachers are doing their best to detect it and stop it. But that's missing the point. We’re teaching them to avoid AI, or to make it look like they're avoiding AI. Instead, we should be teaching them to understand it and use it responsibly. This isn't just a K-12 problem. Colleges have no idea what to do either. and employers have no idea what to do. So they're focused more on getting assignments done with AI, than with actually learning (though I can guarantee you they'll have a solid grasp of the use of 6 and 7). But if we don't take action, we’ll end up with a generation that’s brilliant at gaming detectors — and terrible at learning. What are you seeing in your kids' (or town's) schools?

I have always been a translator... of child development and children’s media research into tools and products children and families can actually use and enjoy. And I’ve been lucky enough to see those tools in the hands of many children and families, enjoying them. When we started KidsAI, there were already enough tech experts in the AI space. What was missing were the kids people. So once again, I had to translate all that knowledge — this time into AI. Here are 3 principles anyone creating child-facing AI can use: -Translate, don’t just transfer. Research on how children think, learn, and play must be reimagined into experiences kids actually enjoy. -Build ethical guardrails. AI for children must spark curiosity without exploiting it. -Involve kids’ experts early. Educators, child psychologists, and play designers make AI safer and more meaningful. At the KidsAI Innovation Hub, we collaborate with organizations to apply these principles — translating both research and values into ethical, child-centered AI. *Takeaway: Every AI product for kids teaches a hidden curriculum. Children don’t just learn from what AI says, but from how it behaves. Design with that lesson in mind. If you’re building in this space, let’s connect. Together we can ensure the next generation grows up with technology that truly supports them.

AI is shaping our world—from voice assistants to personalized learning tools—and kids should grow up understanding how it works. But learning AI first, before heading into complex math or coding should be prioritized. Children should start with basic concepts. They need to learn how AI makes decisions. And why it’s important to use it responsibly.  Allow them to absorb AI tech slowly, though games or stories and even simple apps. This gives them the tools to understand and explore how machines "learn" and "think” instead of mindless usage. Maybe a child might want to use AI creatively as a “tool” like coding platforms or AI-based art and music apps. If they build, experiment, and ask questions, that encourages critical thinking to help them throughout their life. Help them understand that AI can make mistakes, even be biased and potentially affect our lives in big ways. Well, it already is, isn’t it? Most importantly, learning AI should spark curiosity.  We don’t need every child to become an AI expert, but they all deserve to understand the technology that is shaping their future, and well, shaping their “now”.

"I’ve interviewed nearly 100 teachers about AI in schools, and they unanimously say they need policy and professional development." https://lnkd.in/eG-V3K9U

Should I Let My Child Use AI for Homework? AI is now part of our children’s everyday lives; in classrooms, homes, and even their hands. From creative design platforms to large language models like ChatGPT and AI-powered math solvers, today’s learners have access to tools that can explain, summarise, design, and even write. And it’s not just about schoolwork anymore. Many employers today are already listing ‘knowledge of AI tools to enhance productivity’ as part of their hiring requirements. In other words, the future your child is preparing for is already being shaped by AI. So, the question isn’t can children use AI, it’s should they? And with that one question come many more. Here are a few and our take on them: Is AI harmful to my child’s learning Not really, but the wrong use of AI can be. When used without boundaries, it can weaken critical thinking, blur originality, and reduce the joy of learning. Should I just ignore this whole AI thing or stay neutral for now? Well, if you do that, you risk your child learning about AI from someone else; maybe a friend, social media, or the internet, and they may not learn how to use it responsibly. Maybe I should steer my child away from AI completely then? Hmm… we don’t think so. Because AI can be a powerful partner in learning if guided well. It can explain complex topics, encourage curiosity, and support revision. With the right guidance, AI can actually help your child learn better, think smarter, and create more confidently. What truly matters is helping them understand how to use it productively and ethically. So, to that question, should you let your child use AI for homework? ✅ Yes, but with guidance.AI shouldn’t replace thinking; it should expand it. So the real question is: What can you do as a parent to help your child stay digitally aware and safe in this new world of AI? Come back here tomorrow as we share 7 ways parents can guide healthy AI use at home. #JordanHill #BuildingConfidence #InspiringCreativity #AIandLearning #ResponsibleAI #FutureReadyKids #DigitalParenting #SmartLearning #RaisingThinkers #LifelongLearning

The Dangers of using AI to Grade Nobody Learns, Nobody Gains Marc Watkins Earlier this year, the AI detection company behind GPTZero made a pivot from trying to detect AI usage from students to offering educators an AI-powered grading assistant. You heard that right—three years since ChatGPT’s launch, one of the largest AI detection companies has decided to branch out into using generative AI to assess student work. AI as an assessment tool represents an existential threat to education because no matter how you try and establish guardrails or best practices around how it is employed, using the technology in place of an educator ultimately cedes human judgment to a machine-based process. It also devalues the entire enterprise of education and creates a situation where the only way universities can add value to education is by further eliminating costly human labor. https://lnkd.in/dN5k7cte

AI has entered the chat in Australian classrooms. Artificial Intelligence is rapidly reshaping how we teach and learn — and its potential in education is enormous. For many disadvantaged schools, where teachers are supporting students with diverse and complex learning needs, AI could be transformative. The ability to provide tailored explanations, generate engaging resources, and offer immediate feedback could help bridge long-standing gaps in opportunity and achievement. Yet, as with all emerging technologies, access and equity remain critical challenges. Too often, schools in under-resourced communities are the last to benefit from innovations that could make the biggest difference. Managing equitable access and ensuring every school, regardless of postcode or funding level, can use these tools safely and effectively is essential if AI is to support all learners, not just a privileged few. One promising example of how this can be done well is EdChat. A new generative AI chatbot developed by the South Australian Department for Education, in partnership with Microsoft, and custom-tuned for classroom use. Unlike general AI tools, EdChat is purpose-built for education. It operates safely within Australian servers, with strong safeguards and curriculum alignment, allowing teachers and students to explore AI confidently and responsibly. Why EdChat shows potential for equity-minded innovation: 🔵Personalised scaffolding: Students can ask questions at their own level. From simple clarifications to deeper conceptual prompts and receive guided explanations that build real understanding. 🔵Time back to teachers: Educators can use EdChat to draft lesson plans, suggest assessment prompts, or develop extension ideas, freeing them to focus on students who need extra support. 🔵Safe experimentation: Data stays within Australia, with filters and controls in place to protect student privacy and content quality. With over 10,000 students already involved in trials, EdChat is proving that AI, when thoughtfully implemented, can complement teaching rather than compete with it. The challenge ahead is to ensure that access to tools like EdChat doesn’t depend on a school’s resources, but becomes a shared foundation for learning across Australia. If we can get this balance right, AI could become one of the most powerful enablers of education equity in a generation — helping every student think deeper, learn faster, and explore with confidence. https://lnkd.in/gF73k4UB