Introduction: AI’s Growing Role in Learning
Concerns about artificial intelligence supplanting human thinking are rising amid the exponential growth of generative AI’s capabilities. One thing is clear: the technology is not going away. As AI tools become more ubiquitous in classrooms and homes, educators and parents face a pressing question — how do we harness AI’s educational benefits while protecting students’ ability to think for themselves?
In a recent episode of Harvard Thinking, host Samantha Laine Perfas sat down with three Harvard experts to explore this challenge: Michael Brenner, Catalyst Professor of Applied Mathematics; Tina Grotzer, cognitive scientist at the Graduate School of Education; and Ying Xu, assistant professor at the Graduate School of Education.
What Is at Stake for Student Development?
The Threat to Critical Thinking
Homework that once took hours of practice and comprehension can now be completed in minutes, potentially undercutting students’ development of basic skills. Grotzer emphasizes that learning is about far more than memorizing facts — it includes understanding how our minds work, what critical thinking looks like, and how to engage in creative reasoning.
“Kids don’t even realize that they need to learn to do it,” Grotzer noted. “They don’t reflect on the amazing learning that their minds are doing day after day.”
Student Self-Awareness of Over-Reliance
A survey of 7,000 high school students conducted by Xu’s team revealed a striking finding: nearly half of students reported feeling they rely on AI too much for their learning, and over 40 percent said they tried to limit their usage but failed. This points to a self-regulation crisis that educators and parents must address head-on.
How Educators Are Adapting to AI
Embracing AI to Raise the Bar
Rather than banning AI tools, Brenner argues that responsible educators must embrace them entirely. “I feel it would be irresponsible for me not to embrace it,” he said. “Anyone who doesn’t is going to lag in their careers and their ability to advance science. It’s just changed everything.”
Brenner redesigned his graduate-level applied mathematics course after discovering that leading AI models could solve his existing problem sets. Instead of traditional homework, he challenged students to invent problems that AI tools could not solve — and awarded extra credit to those who succeeded. By the end of the semester, the class had collectively generated 600 novel problems and co-authored a published academic paper.
Oral Exams as a Measure of True Understanding
To ensure genuine comprehension, Brenner replaced final exams with oral assessments. Students were required to solve their own invented problems at the blackboard and explain their reasoning. The result? A level of student understanding deeper than any previous cohort, driven by the necessity of pushing beyond what AI could accomplish.
The Human Element AI Cannot Replace
Relationship-Building in Learning
Grotzer and Xu both stress that learning is far more than an exchange of information. Studies by Xu found that while AI tutors and human tutors can produce similar knowledge outcomes, students consistently report higher enjoyment, greater engagement, and stronger confidence when working with human tutors.
In a compelling experiment, Xu gave students identical essay feedback — but told one group it came from an AI and another group it came from the instructor. Students who believed the feedback was from a human rated it significantly more useful. “What really matters is not only the kind of information students receive, but that they know the instructor really cares about them,” Xu explained.
Metacognition: Knowing Your Own Mind
Grotzer advocates for metacognition — understanding and reflecting on one’s own thinking — as a new core purpose of education. Her course, Becoming an Expert Learner, regularly asks students to compare human cognitive strengths against what AI can currently do. This Venn diagram exercise helps students recognize which tasks are best left to AI and which demand uniquely human capabilities.
Guidance for Parents and Educators
For Educators: Measure What Works
Brenner urges educators to experiment and track measurable outcomes. Can you prove students are learning more because of AI tools? That evidence-based mindset should guide every classroom decision.
For Parents: Consider the Whole Child
Xu encourages parents not to fixate on AI in isolation. Instead, consider it as one element within the larger ecosystem of a child’s life — alongside healthy relationships, time in nature, hobbies, and family connection. “AI probably doesn’t matter as much in isolation,” she said. “What does matter is how it fits into the larger ecosystem of a child’s life.”
Conclusion: AI as Part of a Larger Ecosystem
The challenge for educators is not whether to use AI, but how to use it in ways that enhance learning rather than replace it. Guardrails, when implemented, should serve students’ long-term cognitive development and sense of agency — not simply restrict access. As Grotzer put it, “Once you start to know what your mind can do that’s so much better than AI, it makes sense that some tasks are well-relegated to AI and other tasks are not.”
The future of education lies in empowering students to understand both the power of their own minds and the possibilities of AI — and to use each wisely.
