Let's Teach Our Students to Wrestle With AI, Not Surrender to It

I do not envy high school and college professors right now. They’ve been living with a reality in which AI is rapidly reshaping both their practice and their students’ experiences, and from an outsider’s perspective, it feels almost impossible to get a handle on the pace of these changes. Articles like this or this are steady reminders of our oldest students’ fears and challenges. As I read them, my empathy for the adults supporting these students remains high. As a teacher and parent of younger students, I take them as a mandate: figure this out now, so the high school and college years can continue to be places of fruitful learning, engagement, and growth.

To understand why this feels so urgent, it’s worth considering the pace of AI growth and its effect on even the most experienced educators. With any new development, we must rethink our practice and focus on transfer skills that ensure long-term success. But the speed at which AI has entered classrooms and students’ lives has been so rapid that even the most dedicated educators must feel overwhelmed, if not defeated. That overwhelm is not a failure of teaching; rather, it’s a reflection of how quickly AI has changed the landscape.

As a middle school educator, I am eager to discuss building the skills our students need to harness AI without compromising their autonomy or development. Teaching metacognitive strategies that keep students in control and at the center of this work is crucial. So, where do we begin? One useful entry point is a human-centered approach to AI.

I first learned about this principle from Eric Hudson, one of the leading voices in AI implementation in schools. From the start, Eric has emphasized building a school culture that prioritizes keeping the human at the center of generative AI use. Building on that, I recently came across another interesting reframing.

In his Substack article From Thinking Partner to Sparring Partner: A Better Way to Use AI, Mike Kentz writes: “A good human thinking partner brings experience, wisdom, and judgment to the collaboration. Ideally, they’ve been where you’re going. They can spot patterns you can’t see and offer insights born from years of trial and error.” He offers a useful boxing metaphor, arguing that AI is better suited as a sparring partner, which, when done well, leverages human advantages and leads to deeper thinking and better outcomes.

I liked this metaphor so much that I immediately tried this approach, and when I did, I noticed an unexpected benefit of AI: its ability to model humility and vulnerability. Even typing that feels uncomfortable (scenes from the movie Her come to mind, though I promise I’m not going there). But here’s what I saw: when I asked AI to complete a task and it failed, I pushed back. It quickly responded, “You’re absolutely right—thanks for the correction.” Later, after another mistake, it said, “Thanks for your patience—I’ve now carefully reviewed…” Yes, AI overuses em-dashes, but em-dashes aside, how often do we hear statements like “You’re right, thanks for your patience” in our daily lives? AI is sycophantic and eager to please—which can be dangerous—but when we’re in control of the task and the learning, its mistakes and corrections can be productive.

We all need to be better at admitting mistakes. If AI can do it, why can’t we? Our students deserve to see this modeled in a world short on leaders who show grace and humility. To be clear, I don’t think we should use AI to imitate human qualities, but this small benefit has real power.

Think about a learning interaction: you go back and forth with a teacher or peer, building understanding. At some point, you disagree, and your partner responds, “You’re right” or “Thanks for your patience—let me try again.” You stand a little taller, dig a little deeper.

Now imagine access to that kind of interaction all the time. A human thought partner is always best, and a human sparring partner is better than an AI one, but they’re not always available. So why not spar with AI?

The question, then, is how to translate these insights into practice for our students. This is where teaching metacognitive strategies for approaching AI and even small parenting choices matter.

It takes time to teach these skills, and building a framework with benchmarks is crucial. For now, I’m leaving breadcrumbs—at work and at home. In the car, I engage with Siri in front of my kids, calling it “it” (not “her”), and cheering when it messes up: “Siri is wrong! Let’s try again!” Instead of asking AI to do a task outright, I practice prompts like: “I need to [insert goal]. Before providing a solution, please ask me questions about my requirements and constraints so you can give me the best advice.” The more I practice, the better I can teach students to approach AI in this way.

These experiments, whether with Siri in the car or AI in the classroom, are small steps toward a larger goal: prioritizing metacognitive skills now, before students reach high school and college.

I’m grateful to exist in a space where I have little ones at home and work with kids before high school and college. If we prioritize metacognitive skills around AI now, we can shape real-world use that elevates learning, agency, and vulnerability by leveraging human strengths and staying in control of this technology.