How to Use AI in the Classroom to Reignite Student Curiosity and Build Critical Thinkers

Henry Ford supposedly said, “Whether you think you can or you can’t, you’re right. The same logic applies to AI in your classroom. If you’ve already decided it’s cheating, it’s cheating. If you’ve already decided it kills critical thinking, it’s killing critical thinking. And if you’ve decided to ban it while your students use it anyway on their phones? Congratulations, you’ve made it worse.

AI is here. It is not going anywhere. The question isn’t whether your students are using it. They are. The question is whether they’re using it in ways that help them think and learn or in ways that let them stop thinking altogether.

“We have to create pathways that reward and celebrate critical thinking along with the use of whatever technology we have access to at the time. If we don’t critical thinking will diminish.”

Those different outcomes should feel like the uncomfortable truth that it is. Both of those outcomes are real, and they can coexist. Studies have found genuine cognitive decline when people outsource their thinking to an LLM. Other research, including two studies I ran, later confirmed through the California Community College System, that the use of AI course assistants increases student outcomes. When used thoughtfully within a course, a student’s self-efficacy and intrinsic motivation to learn can increase. Students also report feeling more supported and more genuinely interested in the material. This did not happen because AI did the work for them. It happened because it wouldn’t let them off the hook. Does that sound like a contradiction? It is only a contradiction if you ignore how the tool is implemented and used.

The Secret isn’t the AI: It’s How You Design the Conversation

A poorly designed AI assistant is a magic vending machine that dispenses answers. A student inserts a question or directions they receive, clicks a button, and thinks less. A well-designed AI assistant is Socratic. It answers a question with a question, pushes students to connect new material to what they already know, and refuses to dispense conclusions students haven’t earned yet. That creates friction. Friction is where learning happens. The Bjorks (psychologists) coined the term desirable difficulty to define the amount of difficulty that creates the greatest amount of learning that sticks (durable learning).

Most of us have probably been that student who waited until the night before to start writing an essay or some other work we were given in class. Picture this: You’re stuck on an essay prompt at 10 pm. You open the AI assistant and paste in the directions you were given, and voila, the vending machine spits out a well-written essay with references. A Socratic assistant doesn’t act that way. It will ask, “What do you think the central theme of the essay should be? What makes sense to you?” This forces thinking and a conversation that leads the student through a learning process that helps them think both critically and creatively. That’s the difference.

When students build on prior knowledge instead of bypassing it, they’re not just retaining information; they are developing the capacity to use it creatively to solve problems the AI can’t solve for them. I like to say we are cultivating critical and creative thinking to help people solve complex problems. How we design learning is the most important question. The way we designed and created learning that got us to this point will not propel us to where we need to go in the future.

Design Authentic Learning That’s AI Resilient

AI assistants open up active learning possibilities that did not exist five years ago: live role-play with historical figures, Socratic debate partners who argue positions students hate, escape rooms built around course content, and case-based scenarios with no clean answers. These are all simple and powerful ways students can engage more deeply with the nuance of topics and wrestle with gray concepts. These are not gimmicks. They connect directly to learning science, such as retrieval practice, elaborative interrogation, self-explanation, and so on. AI makes them scalable and available at 11 pm on a Sunday.

But AI Hallucinates Sometimes

Good. I mean that…good. LLMs get things wrong at times. The level of confidence they give the wrong answers with can be quite impressive. My thought on this isn’t to apologize or use it as a warning. Treat it as a pedagogical feature. Don’t allow students to take an AI’s output at face value. Have them verify it, question it, and push back on it. That process builds exactly the authentic critical thinking that passive consumption of correct content never does. Passive learning doesn’t work that well anyway. The embers of curiosity are not fanned by easy answers. Curiosity only turns into a bright flame by encountering something that doesn’t fully add up.

Curiosity is in Crisis & Has Been For a While

Decades of standardized teaching to the test, right answer culture, and content-delivery pedagogy have systematically trained the curiosity out of students. We’ve built school systems optimized for compliance and answer retrieval, but we are surprised when graduates struggle to ask original questions or sit with uncertainty. Curiosity requires practice. It requires the experience of not knowing something and making the choice to pursue it anyway. And pursue it not because it is part of an exam or something, but because the not knowing bothers you and fans those embers of curiosity.

AI, designed and implemented poorly, accelerates the curiosity crisis. Every time a student prompts an LLM to write their paragraph, they bypass the productive struggle. We want students to engage with that level of desirable difficulty. The level of difficulty that is desirable varies from student to student.

Either Extreme Will Cost You

Banning AI while students use it in the dark teaches them that AI is for cheating. Letting them use it without structure teaches them that AI is for thinking instead of them. Neither produces the critical and creative problem solvers the world actually needs. The world is complex. The problems we are solving are complex and don’t come with answer keys. We need people who can wrestle with ambiguity, synthesize across disciplines, and generate solutions that don’t yet exist. We are not currently building those people at scale.

The answer is not complicated: define how AI is used in your classroom. Deliberately build in friction. Create moments where students have to catch the AI being wrong, argue with it, or take its output and make it better.

No special platform is required to do any of this. Ask for a multiple choice quiz, explain a concept out loud and request feedback, or my favorite, ask the AI to explain something and intentionally include a content error, then make the student find and correct it. Low tech and high impact.

It is not about AI. It never was. It’s about rethinking how we teach, and being honest about the current model isn’t working as well as we’d like to believe.

“Curiosity is already there in every student. We just need to stop doing things that stop it.”