My Big Question on AI and Education for 2025

As the new year begins, I’ve been reflecting on the tremendous potential of generative AI in education. The question is: in which direction will it take us?

Could AI transform education for the better bringing knowledge acquisition and application closer together in the learning process? Imagine learning math by using equations to try and solve a goal in an immersive online world. Students could apply the math procedures they learn to accomplish a task in one context one day and another context the next day. This would make learning more relevant to students’ lives, foster creative problem-solving, and address the long-standing challenge of transfer—students’ ability to apply what they learn in one setting, like a classroom, to different situations. AI has the potential to radically personalize learning, including for children with learning differences, helping each child master skills they would otherwise get stuck on. It holds the potential to reinvent assessment and democratize the creation of quality teaching and learning materials. It could facilitate mother-tongue instruction in the early grades which is so important for inclusive education systems and so rarely done.

Or could AI steer education down a troubling path? Might it create more isolating, tech-driven learning experiences that reduce interpersonal interaction and hinder the development of critical thinking and collaboration skills? Could it foster a mindset where learning is effortless, undermining students' growth and resilience? Could it lead to more disengaged students because large-language models provide learning short-cuts especially for middle and high school students? Could it undermine children’s development of core knowledge needed to successful navigate the world, including assessing the reasonableness of generative AI responses? ?Will it proliferate fake news and deep fake videos that undermine learning and promote harassment? Will it fuel a backlash that discards not just harmful uses of technology, but its benefits too?

These are just some of the questions we’ve been grappling with in the Brookings Global Task Force on AI and Education. But there is one, deeper question I keep returning to.

What are the essential learning experiences children need to have on their own—free of generative AI?

I have spent the last several years researching student motivation and engagement. In our new book, The Disengaged Teen, my coauthor Jenny Anderson and I found that the majority of middle and high school students’ learning experiences inspire them to operate in what we call Passenger mode, coasting along and doing the bare minimum to get by. Will generative AI help them become more engaged or only provide a more powerful tool for short-cutting learning and fuel further disengagement? As neuroscientists routinely say, children’s brains develop based the way they are used. In schooling, what, if anything, do children need to learn to do on their own, at least initially before they collaborate with large-language models like ChatGPT?

There is currently no clear answer to this question. I’ve been discussing this with child development experts, including colleagues like Kathy Hirsh-Pasek and Mary Helen Immordino-Yang, EdD . It is a question that I will be paying particular attention to in 2025.

The "Miraculous Baby Walking Shoes" Analogy

In a recent discussion, Mary Helen and I came up with an analogy: Imagine a company inventing miraculous baby shoes that let infants walk without learning to crawl. Would you use them? Development experts would say no. Crawling isn’t just a step toward walking; it’s essential for building essential capabilities such as gross and fine motor skills, spatial awareness, independence, and problem-solving abilities.

What is the equivalent of crawling for children's learning in school in a generative AI world?

Balancing Cognitive Offloading

Humans offload cognitive tasks to tools all the time. As a child, I knew all my friends’ phone numbers, but now I don’t know a single one. Given my cell phone is usually within arm’s reach at any time, this hasn’t hindered me in any meaningful way. Except once, when I visited New York for meetings. Distracted on a call, I used Google Maps to navigate to my hotel in mid-town without paying attention to landmarks or directions. I dropped my luggage and carried on to my meetings. Later that night, my phone died, and I realized I didn’t remember the hotel’s name or address—just a vague memory of a green awning. I couldn’t call any family members for help because, as we established, I no longer know any phone numbers. Luckily, I grew up before the age of google maps and had basic spatial reasoning skills and methodically searched the neighborhood until I found it.

I use generative AI regularly—and I love it. ChatGPT is my always-available research assistant, and it creates visuals in a heartbeat (please admire my collaboration with it for the Calvin and Hobbes style graphic above). But I didn’t grow up with it. Instead, I had learning experiences that taught me how to fail and recover, think critically, and connect with others.

For children, when is cognitive offloading merely reflective of the natural course of social development? Afterall, I haven’t just forgotten phone numbers which has happened over the course of my lifetime. I can’t distinguish poisonous berries from edible ones which undoubtedly my ancient ancestors could.

But when is it not? Chanea Bond , a high school English teacher in Texas argues that generative AI hinders her students ability to develop their foundational thinking skills, which is something they need to do on their own first before engaging with AI.

This may be one way in which generative AI could limit students’ potential.

Learning as a Social Process

Another critical issue is the social nature of learning. Education is far more than transferring knowledge from teacher to student. It’s a deeply social process, shaped by emotions, relationships, and interaction. The COVID-19 pandemic revealed how much children need these connections to thrive.

The best way for children to learn to interact with other people is through face-to-face interactions, not by interacting with AIs through screens or voice. ?Will generative AI enhance or hinder the relational aspects of learning? ?How can we ensure it supports, rather than undermines, the emotional and social bedrocks of education?

The Role of Knowledge in Skill Development

Finally, I’ve been thinking more about the role of knowledge in developing skills. For years, I’ve argued that acquiring knowledge is necessary but not sufficient for thriving in today’s world—students also need skills like collaboration, communication, and creative problem-solving.

But knowledge can’t be skipped. Critical thinking and problem-solving don’t develop in a vacuum; they emerge when students immerse themselves in topics, issues, or problems. If children rely on AI for all their answers, how will they build the foundational knowledge needed to not only understand the world but think critically and collaborate effectively?

Looking Ahead

We don’t have all the answers yet. But when faced with uncertainty, asking the right questions is a good place to start.

As we enter 2025, I’ll continue exploring how to balance the benefits of generative AI with the foundational experiences and skills that children need to develop independently. Education must prepare young people not just to use AI, but to thrive in a world shaped by it.

#SDG4 #TransformingEducation #DisengagedTeen #LearnBetterFeelBetter #AIandEd

Pat Yongpradit Rachel Glennerster Jenny Anderson David Risher Michael Trucano Amit Sevak Neil Allison Punya Mishra Olli-Pekka Heinonen

?