Reflections from the 2nd sub-group of Courageous Minority in August.

At a recent meeting of the 2nd sub-group of the Courageous Minority, I began with a question that has been troubling me more each day: If artificial intelligence can already do much of what schools teach students to do, what should schools teach?

This is no longer a speculative question. AI is developing far faster than most schools, policies, and curricula can respond. In only a few months, systems such as ChatGPT have become significantly more capable. They can research, write, analyze, program, explain, revise, and guide people through unfamiliar tasks. I recently asked ChatGPT how to transcribe a video. It did not transcribe the video itself, but it taught me the programming necessary to accomplish the task. Within a couple of hours, I had done something I did not previously know how to do.

The educational challenge is therefore much larger than deciding whether students should use AI to write an essay. AI is forcing us to reconsider what knowledge is, how people acquire expertise, what assessment should measure, and what schools are for.

Where Does Knowledge Begin?

A common response is that students still need foundational knowledge. They must know something, people argue, before they can judge whether AI is right or wrong.

That is certainly reasonable. Without some understanding of a field, people can easily accept an inaccurate or shallow answer. AI can produce plausible nonsense, and users who lack expertise may not recognize it.

But this answer produces another question: How does anyone acquire foundational knowledge in the first place?

Every learner begins without expertise. When we enter a new situation, meet a new person, visit a new country, or encounter an unfamiliar problem, we initially know very little. We act, make mistakes, revise our understanding, and gradually learn.

Before AI, teachers, books, parents, and other knowledgeable people helped us move from ignorance toward understanding. Could AI now become part of that process? Must students acquire all the necessary foundations before they use AI, or can they develop those foundations through interaction with AI, teachers, peers, experiences, and real problems?

I do not pretend to have a final answer. But I am increasingly unconvinced by the simple sequence that schools have traditionally assumed: first transmit a prescribed foundation, then allow students to do something meaningful.

Sometimes the meaningful problem comes first. It creates the need to know. Students search, experiment, ask questions, receive feedback, and build knowledge because the knowledge has become necessary.

Agency, judgment, and expertise are not prerequisites that suddenly appear. They are capacities that must be practiced.

AI Is a Tool, but That Does Not Settle the Question

Many people worry that AI will make students intellectually lazy. It may. But the same concern has accompanied nearly every important tool.

GPS reduces the need to memorize roads. Calculators reduce the need to perform some computations by hand. Agricultural tools save people from digging with their fingers. Each tool replaces some human effort while making other activity possible.

No tool is entirely beneficial, and no tool is entirely safe. A farming implement can cultivate land or injure someone. A search engine can expand knowledge or spread misinformation. AI can deepen learning or allow students to avoid it.

The fact that a tool can be misused does not tell us whether it should be prohibited. Nor does it tell us what forms of human capability should be retained, strengthened, or newly developed.

We cannot rely completely on AI. But we cannot rely completely on teachers, textbooks, parents, governments, or our own memories either. Human beings have always learned by distributing thinking across people, tools, symbols, and environments.

The important question is not whether students should rely on AI. It is when, why, how, and for what purposes they should use it.

AI Is Exposing Meaningless Schoolwork

One participant described an important response from students in an online school. When students believed an assignment was genuinely helping them learn, they wanted feedback from the teacher. They wanted to know what they understood and how they could improve. They did not want the teacher to evaluate a product created largely by an AI assistant.

But when an assignment felt like a hoop to jump through, students were much more willing to let AI complete it.

This distinction deserves our attention.

Perhaps AI is not merely encouraging students to avoid learning. Perhaps it is revealing how much schoolwork students have never considered worth doing. If an assignment exists only because a teacher requires it, and if an AI can complete it instantly, students will understandably ask why they should spend hours doing it themselves.

Schools may respond by increasing surveillance, closing laptops, banning tools, and scanning student writing. But such measures leave the central problem untouched: Why was the task worth doing before AI arrived?

If we cannot answer that question convincingly, the problem may not be the technology.

Assessment Must Change

Instead of pretending students will not use AI, some educators in the group are designing assessments that require them to use it responsibly.

Students might ask AI to produce an argument, case, explanation, or solution. They would then examine the result, identify its errors and limitations, compare it with other evidence, revise it, and explain their decisions. They might also submit a declaration describing how they used AI, what it contributed, what they rejected, and what they learned through the process.

Such assessments would not measure whether students could secretly reproduce something AI already does well. They would examine judgment, disciplinary understanding, reflection, transparency, and the ability to work intelligently with tools.

This approach requires expertise, but it may also help students develop expertise. Evaluating AI output can lead students to notice what they do not yet understand. Errors become invitations to investigate rather than merely reasons to distrust the technology.

Assessment should increasingly capture the process of learning: the questions students ask, the alternatives they consider, the feedback they use, the decisions they make, and the reasoning behind the final product.

Personalization Is More Than Choosing Courses

Another participant described the possibility of using AI with school counselors to help students create individualized graduation pathways. Instead of treating graduation as the accumulation of a standard number of credits, schools could help students assemble learning experiences connected to their interests, strengths, future possibilities, and desired contributions.

This is a much richer conception of personalized learning. It is not merely allowing students to move at different speeds through the same curriculum. It is helping them develop different pathways without prematurely closing future opportunities.

The participant offered a particularly powerful question: Are we preparing students only for the next stage of education and employment, or are we preparing them to become good elders?

To be a good elder is to think beyond immediate personal advancement. It means considering what kind of world one is creating for people who will live several generations from now.

AI may help students design more individualized futures. Education must also help them consider the consequences of those futures for others.

Curiosity Needs Space

The conversation also turned to young children.

At one small school in Tasmania, children between five and eight years old were redesigning the school’s kitchen garden. The garden had been treated mainly as a functional space: children planted, harvested, and supplied food for the cooking program. The new inquiry asked how the garden could become a place where children wanted to play, gather, explore, and spend time.

The children proposed an entrance covered with snow peas, colorful artwork, seating, and places where people could taste food directly from the garden. They measured spaces, built prototypes, presented ideas to older students, and sought feedback.

The teacher’s question was not whether the project was worthwhile. It clearly was. Her question was when AI should enter the process. Would introducing it too early interfere with the children’s natural creativity?

This is an important concern. AI does not need to direct the project. It might simply join the inquiry. Children could speak to it rather than type. They could ask about different garden traditions, plants that grow in particular conditions, ways to deter animals, or methods of designing welcoming spaces. They could compare its ideas with their own and decide what was useful.

There is no universal age at which AI suddenly becomes appropriate. The more important considerations are the form of use, the length of interaction, the presence of adults and peers, and whether the technology expands or narrows children’s activity.

Young children can play with AI without becoming dependent on it, just as they can consult a book without allowing the book to determine the entire project.

Do Not Explain Too Much, Too Soon

Another teacher proposed combining unstructured play with learning to ask better questions. Children might sit around an unfamiliar object and generate as many questions as possible. Over time, they could examine new objects and eventually more abstract ideas, noticing whether their questions became more open, curious, and complex.

Her idea points to a larger educational principle: curiosity depends on uncertainty.

When adults provide too many rules, children begin trying to guess what the teacher wants. When we demonstrate the correct way to use an object, children often imitate us rather than discovering other possibilities. When we provide the desired outcome too quickly, investigation ends.

If we say, “Here is a toy, and this is how you play with it,” children may follow our instructions briefly. If we say, “I found this object and have no idea what it does,” they are more likely to explore.

This principle should guide educational AI as well. An AI system should not always rush to supply an answer. Sometimes its role should be to introduce another possibility, pose a question, expose a contradiction, or invite the learner to look again.

Curiosity does not flourish when every uncertainty is immediately removed.

Students Can Help Teachers

Schools usually assume that teachers possess knowledge and students receive it. AI creates opportunities to reverse that relationship.

Years ago, I developed programs in which technologically capable students “adopted a teacher.” They helped teachers conduct online research, prepare presentations, and use unfamiliar technologies. Similar university programs paired undergraduate technology interns with professors.

Schools could revive this idea for the age of AI.

Students who are interested in AI could form teams to help teachers explore tools, create resources, address school problems, and support community organizations. They could receive academic credit for this work.

This would not simply provide free technical assistance. The process of helping others is itself a powerful form of learning. Students would have to understand a person’s needs, explain ideas, respond to constraints, revise solutions, and take responsibility for the consequences of their work.

Such experiences embody personalization, problem finding and solving, and human interdependence. Students develop their distinctive strengths by creating value for others.

What Authentic Agency Looks Like

A pilot program in Kansas City offered another illustration. Students between approximately 13 and 16 years old were given only a few days to identify problems that mattered to them and develop responses with the support of AI, peers, and facilitators.

Some examined community safety and belonging. Others pursued sports-related interests. One student created a children’s story about communication and manners. Another, who feared public speaking, researched anxiety and proposed a speech club designed specifically for students who shared that fear. Some students developed plans on paper; others built working websites.

The projects varied widely because the students varied widely.

That was not a flaw. It was personalization.

The facilitators resisted giving extensive “just in case” instruction. They offered examples and support when students needed them, moved around the room, asked questions, and helped students continue. The role of the adult shifted from delivering identical content to helping different learners make progress.

The most important feedback was not about the technology. Students said they felt listened to. They believed the adults genuinely cared about their ideas.

Schools frequently claim to give students agency while determining the topic, process, outcome, criteria, and schedule. Students are permitted to choose only within boundaries adults have already decided.

Authentic agency begins when we treat students’ interests and concerns as worthy of serious attention.

Designing a Different Learning System

During the meeting, we also demonstrated an AI-supported platform we have been developing. Its purpose is not to make conventional schooling more efficient, but to support a different form of learning.

The system begins by asking about the learners and their context. It then helps educators identify a real problem or tension students might investigate. What can students genuinely decide? Whose lives might their work affect? What would make the learning meaningful to the student, the school, or the wider community?

From there, the system helps develop learning journeys, challenges, smaller tasks, prompts, and opportunities for individual and group work. Students can document their actions, upload evidence, receive feedback, revise their plans, and pursue different paths.

The resulting evidence can help educators understand students’ interests, curiosity, persistence, choices, and contributions. Assessment becomes dynamic because it emerges from what students actually do rather than from a single predetermined response.

Required subject knowledge can still be included. But curriculum becomes a resource for accomplishing meaningful work, not the sole destination of learning.

The Danger of Policy by Panic

Many educators in the group work under restrictive policies. Schools close laptops, prohibit tools, and respond to public anxiety by limiting experimentation.

Some caution is justified. AI can cause harm. It can produce false information, expose private data, reproduce bias, and support unethical behavior.

But restriction also carries risks.

Students in some schools are developing sophisticated AI literacy, building tools, supporting businesses, and learning how to evaluate outputs. Students elsewhere are forbidden to experiment. By the time they reach university or employment, the second group may be significantly disadvantaged.

A policy designed to protect students may create a new form of inequality.

Banning is attractive because it allows leaders to claim that they have acted. But a ban rarely eliminates the technology. It often drives use underground, where students receive less guidance and schools learn less about what is actually happening.

The difficult work is not prohibition. It is creating conditions in which experimentation, ethical judgment, transparency, and adult guidance can coexist.

The Work of the Courageous Minority

No one in the meeting claimed to know exactly where AI is taking us. The technology is changing too quickly, and its social consequences are too uncertain.

But uncertainty is not a reason to do nothing.

The work of the Courageous Minority is not to produce another universal answer. It is to create small spaces in which educators and students can try something different, observe what happens, share successes and failures, and revise their thinking.

This work requires courage because the existing system rewards compliance. It is much easier to follow policy, assign familiar work, and avoid difficult questions. Those who attempt something different will encounter confusion, criticism, and mistakes.

But they may also rediscover the joy of education.

AI is disrupting schools because it can perform many of the tasks schools have treated as learning. That disruption gives us an opportunity—not merely to integrate a new technology, but to ask what learning should become.

Schools should help students develop knowledge, but also help them know when knowledge is needed and how to acquire it. They should cultivate judgment without pretending judgment develops before experience. They should protect curiosity rather than eliminate uncertainty. They should allow students to use tools while helping them recognize the consequences of their use. And they should help young people discover how their distinctive strengths can create value for others.

The important question is not whether AI belongs in education.

It is whether education is willing to become worthy of the world AI is helping to create.

This essay is based on a Courageous Minority discussion about AI, knowledge, assessment, curiosity, student agency, educational policy, and an emerging platform for personalized, problem-oriented, and interdependent learning.