AI and the Future of Learning with Isabelle Hau

 

---

Su Jin: Everyone seems to be talking about AI and its potential to impact nearly every sector, including higher education. But what is artificial intelligence, and what do we really know about its potential impacts?

Isabelle: At its simplest, artificial intelligence refers to computer systems that can perform tasks that have typically been associated with humans, such as recognizing patterns, generating language, solving problems, and making predictions.

Its potential impact is really interesting for those of us in education, because it has the potential to reshape how we learn, how we teach, and how we support our students and families in at least three ways.

First, these algorithms have the ability to make support far more differentiated by providing feedback, translating across multiple languages, or offering tutoring and coaching—especially when human support is not available.

Second, AI can support educators and institutions in transforming how we assess learning—also a very important area.

Third, and most importantly, AI gives us a chance to rethink what learning is for. We are clearly moving into a world where jobs are going to evolve. So what does it mean to educate children, young adults, and adult learners for a world where education can’t only be about producing answers? It also has to be about agency, judgment, curiosity, creativity, ethical reasoning, and collaboration.

There are a lot of opportunities with this incredible technology, though I hope we’ll also speak about the risks.

AI has the potential to reshape how we learn, how we teach, and how we support our students and families.

Su Jin: In your book, Love to Learn, you write about the importance of relational intelligence, particularly as AI advances. Can you talk more about your perspective and how you define relational intelligence?

Isabelle: Right now I’m most focused on how AI is impacting our ways of relating to one another. I see a need to accelerate thinking about the future of human intelligence, perhaps in a different way than what we have shaped in the past, which has been very cognitively based.

When most of us think about human intelligence, we think about IQ, about tests in schools—forms that have shaped how we understand intelligence as a species, with a heavy cognitive emphasis. But I think that a future with AI will require us to recognize a different form of intelligence, which I call relational intelligence. It is our human capacity to relate to one another, to make meaning together, to collaborate across differences, and to address complex societal issues collectively.

Su Jin: And how do you see that connecting with AI? For higher education leaders and policymakers, how does that shape or reshape higher education’s focus?

Isabelle: Ideally, we would start rethinking success itself beyond credit, credentials, and completion. Those things still matter, but I believe we also need to ask: are our students in higher education developing the capacity to work on complex, systemic issues across difference? To build trust, support one another, participate in communities, and lead with empathy and responsibility?

I’d like to bring two areas of research to bear on this.

One is on relational skills. David Deming at Harvard University conducted a seminal study on the importance of different types of skills in employment and wages. He showed that a person with both high math skills and high social skills does, of course, very well. But the surprising finding is that someone with high social skills and low math skills does much better than someone with low social skills and high math skills. And his research shows that this trend has been increasing over the past 20 to 30 years. Relational skills are going to continue rising in importance as AI becomes better and better at knowledge accumulation and knowledge replication.

The second piece of research is more recent, published just a few days ago by an economist at the University of Chicago who is projecting future jobs between 2025 and 2050. He shows a meaningful rise in what he calls the “relational sector.” By meaningful, I mean that by 2050, he projects the relational sector to represent almost 50% of all jobs.

A future with AI will require us to recognize our human capacity to relate to one another across differences and address complex societal issues collectively.

Su Jin: Wow. What are relational sector jobs?

Isabelle: Relational sector jobs include a lot in education and a large share in the care economy. Those are probably the two biggest categories, with some in health as well. Think of professions like physical therapy or other areas of physical and mental health.

Those jobs are expected to grow from a small percentage of our economy to nearly half of it, which suggests we may be entering a future defined by what I would call a relational economy, one in which our capacity to relate to one another becomes increasingly important.

Su Jin: Let’s talk specifically about AI and student parents—that is, college students who have dependent children. We know that student parents often face unique barriers to enrolling in and completing college, including the availability and cost of quality child care. You’ve done a lot of thinking about the ways our higher education and child care systems could better serve families and learners. What does it take to build systems that truly serve student parents, and how, if at all, can AI play a role?

Isabelle: Student parents are one of the clearest examples of why we need more human-centered systems. They are not only students, they are caregivers, family leaders and, often, an extraordinary model of persistence. Yet too often, higher education systems were not designed with them in mind.

Child care, course scheduling, financial aid advising, transportation, mental health, and even basic needs are often very fragmented. We are essentially asking student parents to navigate a complexity that the system itself created.

To truly serve student parents, we need to design around their real lives. That means flexible pathways, affordable and accessible child care, integrated benefits navigation, and family-friendly supports.

Where does AI fit into this picture? I think it could play a helpful role. It could help student parents find the right resources faster and navigate benefits more easily. One specific area where AI could make a real difference is helping students plan course schedules around caregiving responsibilities. There are also well-established benefits from receiving more real-time reminders, nudges, and coaching, another area where AI can have an impact.

And then there’s connection. AI could facilitate connection with peers, resources, and work.

For example, there is an organization in Florida that helps single parents find housing and shared housing arrangements, a meaningful area of need where more support is needed. I would also love to see more dedicated AI systems that help student parents navigate pathways toward jobs and career advising.

To truly serve student parents, we need to design around their real lives. That means flexible pathways, affordable and accessible child care, integrated benefits navigation, and family-friendly supports.

Su Jin: So a lot of it seems like smoothing administrative friction in ways that bring people together around shared goals or purpose.

Isabelle: Exactly. AI may play a role in helping institutions see student parents more clearly, reduce friction—both for institutions and for students—and connect students to the human and material supports they need to succeed.

Su Jin: AI sounds great, but it’s not without risks. What risks are you most concerned about?

Isabelle: There are two areas of deep concern I have with AI: one is critical thinking and the other is creativity.

On critical thinking: this is complicated, because we as humans have a very hard time distinguishing between what is machine and what is human.

Let me give you a specific example. I recently watched a short video embedded in a research study that showed two robots fighting each other. Participants were asked whether they felt any pain while watching. Even when they were explicitly told, “These are robots, not humans,” participants reported feeling pain. It’s a beautiful example of how it’s very difficult for us to recognize patterns of machines versus humans, even when prompted to do so. And it’s even harder for children.

We are observing this over and over again with AI, especially when AI is anthropomorphic—whether through a character manifestation like an avatar, or a robot, in my example, or through language that feels very human, such as an AI system using “I” as a pronoun. These qualities increase the risk of confusion. The research makes clear that we are not good at distinguishing between the two.

This is why we need critical thinking more than ever, and we need to start developing it early. I’m a strong proponent of AI literacy—not so much to understand AI itself, but to understand what is AI and what is not, what are the benefits and risks, so that our young people can shape these systems from an informed position and develop the ability to detect and question information that doesn’t feel right.

Research is actually quite divided and almost polarizing on the impact of AI on creativity. I personally think we need to lean in more in higher education on agency and ensuring that our college students across settings—community colleges, public institutions, private institutions—are equipped to create with these tools.

Creation with AI is increasingly becoming easier. My concern is that there is a new divide emerging—an agency divide, a creativity divide—where those who know how to use these tools can create incredible programs, new sites, and new models, while those who do not are essentially left behind as users rather than creators.

There are two areas of deep concern I have with AI: one is critical thinking and the other is creativity.

Su Jin: You’ve written about the Turing Trap, the risk that AI simply creates outputs based on the past rather than reimagines what’s possible. When College Futures Foundation Entrepreneur-in-Residence Maria Anguiano was on our podcast, she emphasized that we need to ask deeper questions about what we’re designing the system for. Can you tell us about the concept behind the Turing Trap?

Isabelle: The Turing Trap is a concept coined by my colleague Erik Brynjolfsson, who works in the Digital Economy Lab in Stanford’s engineering department. His idea is that the Turing Trap is the risk that we design AI to imitate human performance rather than expand our very human possibilities.

For context, in the 1940s and 1950s, the original Turing test, invented by Professor Turing, asked whether a machine could behave like a human, and whether we could tell the difference between the two. That framing has shaped a great deal of how we think about artificial intelligence.

In education, for example, we ask questions like: Can AI write like a student? Can it tutor like a teacher? Can it advise like a counselor? Can it grade like a faculty member? Those questions are essentially asking whether AI can replicate our very human capabilities.

But that framing itself can trap us, because it leads us to automate the past. And the past in education is not necessarily what we want to replicate for the future; we have an education system that has arguably failed too many children and adult learners.

Ideally, we take advantage of this technology not to automate the past, but to make it much better. If we think about AI solely in terms of replicating efficiency, we run the risk of replicating past systems rather than imagining what learning could become if every student had more support, every educator had more time, and every institution had better ways to understand and respond to learners.

Su Jin: We haven’t talked about a couple of the other major risks that are in the headlines.

Isabelle: There are two areas that we haven’t discussed: the environment and ethics. And those two are super important.

I love that the activism around AI’s environmental impact feels like it is being led by young people. I see more and more clubs in high school and on college campuses focused on the risks this technology presents for the environment. And I love that this is coming from young voices leading our future.

The environmental impact of this technology is a very serious consideration. The power required to operate AI systems is astronomical and will have a significant impact on this planet if we do not accelerate improvements in compute performance or find other ways to address those energy needs.

The other big question that is also emerging from many of our students is ethics. There is a rising question not only about how young people should use AI ethically, but also about faculty. When is it appropriate for faculty to use AI? Is it ethical for faculty to grade with AI? Is it ethical for faculty to use AI tools to create a lesson plan? When are appropriate disclosures required? How are we modeling for our young people and future leaders what it means to use these systems ethically?

These are big questions I hear young people asking more and more.

How are we modeling for our young people and future leaders what it means to use AI systems ethically?

Su Jin: We named this season Crystal Clear because we’re trying to cut through complexity to reveal what’s really needed to strengthen California’s future. When it comes to AI and redesigning higher education, what’s the biggest misconception you’d like to clear up?

Isabelle: I think the biggest misconception is that AI is primarily a technology issue. We tend to frame it that way, but it isn’t. At the end of the day, it’s almost a design issue—a learning issue, an equity issue, a human development issue.

Very often, we start the conversation with the tool: which platform should we use? What tech should we invent? But I think the real questions are: what kind of learning do we want? What human capacities matter most? Which students are being left behind in the current system, or would benefit most from greater support? Once we answer those questions, we can ask how technology can help us build institutions that are more supportive, more relational, and more equitable.

The biggest misconception is that AI is primarily a technology issue. It isn’t. It’s a design issue, a learning issue, an equity issue, and a human development issue.

Su Jin: I appreciate the focus on learning.

Isabelle: Yes.  Research and learning science should be at the center of all of this.

This is my profound belief, that we know a lot about how humans learn from centuries of research. How can we connect that scientific knowledge to practice so that we accelerate human potential?

Su Jin: Let’s talk about concrete steps. What are higher education institutions doing—or what can they do—to enable greater AI literacy, creativity, and deeper understanding of the risks among students, faculty, and staff?

Isabelle: There are two things we are doing at Stanford.

The first is something we call the AI Tinkery, and it has surprised me how much people love it. The AI Tinkery is very similar to a makerspace. It is a physical space where people come to learn together, learn from each other, and learn by doing by using and tinkering with AI tools.

It has created a space on campus for people to learn about AI at a moment when many of us feel overwhelmed by the pace of change. Having a space where you can reflect and learn from others about how they are using different tools, and what is helpful or not, is a really interesting model that I would love to see replicated at other universities and community colleges. It is not expensive. We have also welcomed educators and groups of students from non-Stanford institutions. Anyone is welcome.

For example, for me personally, I recently used the Tinkery to learn more about OpenClaw, a tool that can connect different systems. I was intrigued but also concerned about data privacy and connecting different data sets. I attended a session alongside colleagues from the School of Medicine and several students; we discussed how people are using OpenClaw and some of the questions people had about this tool. That’s just one example of using this incredible resource that I would love to see made available to many more people.

The second initiative is a program we are launching in partnership with the Center for Teaching and Learning to provide more support to faculty in two areas: one is thinking through where pedagogy can be transformed—or not—by this technology; two is conducting research on what is working so we have more evidence going forward, not only for us at Stanford, but for higher education at large.

Su Jin: As we look ahead, what gives you the most optimism for the future of learning in California?

Isabelle: I have a lot of optimism about California, because I feel the state has all the assets needed to truly shape the future. We have extraordinary public institutions, community colleges and universities. We have large language model companies headquartered here, or with meaningful presence. We have philanthropies, community organizations, and this incredible, beautiful diversity as a state.

If we can design learning systems that work for California’s learners, including student parents, multilingual learners, and those who have been historically underserved, I think we can create models for the nation, then maybe for the world. I’m very optimistic that we have everything it takes to build systems where every learner can thrive.

If we can design learning systems that work for California’s learners, we can create models for the nation, then the world. We have everything it takes to build systems where every learner can thrive.