While motivation and interest have always mattered to success, ChatGPT amplifies their importance. Because while ChatGPT can explain almost anything at any level, you have to want to learn for ChatGPT to work as a learning tool. As soon as you stop asking questions, the learning ends.

Motivation is something we build into every level of design at Rock Creek–whether it’s designing our core classes to spark interest, giving students the opportunity to uncover and pursue their interests in our Badging program, or teaching students what we know about human motivation in our Practical Psychology sequence so they can apply these ideas for themselves. We would have done all of this without the rise of AI, but the rise of AI gives this work more urgency.

The vector of motivation that I think goes most under-appreciated by schools is social. In a typical high school class that is mostly lecture, not only are students not doing the active, effortful work they need to in order to learn, their natural social motivation has been sidelined. When we make our classes more social, we are making them more motivating to students.

At Rock Creek, we make it a habit to discuss and workshop student work, often starting classes this way. This sort of collaborative critique makes learning social, which raises its perceived value and reduces the temptation to cut corners. Reviewing student work live also shows students you take their work seriously, crystallizes ideas about what makes work “good,” contributes to a culture of giving and receiving feedback, and fosters a more intrinsic orientation toward improvement. This is just one example of how classes can become more social–seminar discussions are another.

Making classes more social isn’t just tapping into a powerful force of motivation, it also shifts emphasis to social skills that will still be valuable in a world of AI, skills like having presence, being persuasive, responsive, and empathetic, and collaborating while negotiating differences.

LLMs like ChatGPT have advanced to the point where they can fully complete even high-quality school assignments. They can write an ethnography from notes, turn a few bullet points into a memo, and break down complex, open-ended math problems. Just as students benefit from building automaticity in basic math before they rely on calculators, they also need a strong foundation in writing before they can collaborate with an AI as a tool. Without a strong foundation, they’ll rely on AI to do all the work for them, ultimately resulting in adequate output but subpar learning.

If a learning task is crucial but might be easily farmed out to AI, we’ll do it in class. This approach allows teachers to observe each student’s process, help them get unstuck, and confirm that the work is authentic. By tackling these activities together, we help students build strong foundations from which to collaborate with AI.

If all you know how to do with AI is ask it to do your work for you, that’s probably how you are going to use it–as a crutch. If instead you learn to collaborate with AI–how to build and query notebooks, how to create and edit code with an AI copilot, how to set up retrieval practice to master a body of knowledge, etc.–you are more likely to use it as a lever instead. And the future belongs to those who can use AI as a lever.

At Rock Creek, we will explicitly teach students how to collaborate with AI. We will also set explicit guidelines for each assignment, something akin to this scale, which differentiates five different levels of AI use–from no use to full use with no need to cite. (HT Adam Browning.) On the middle levels where we allow some AI use, we will ask students to summarize their AI interactions as part of their assignments. By bringing the AI layer into the open, we are able to teach students to collaborate with AI more effectively and to deter misuse. We expect that these middle levels are where a lot of our assignments at Rock Creek will live.

With the help of AI, students are capable of doing more.
We are raising our standards and creating more complex assignments where students collaborate with AI to complete them.

Right now, a typical high school might look at the work of the Intergovernmental Panel on Climate Change by reading a generic curriculum resource about it or a headline news article with someone else’s analysis. In the past, when we taught the climate change unit of our Social Sciences for Social Problems class, we had students read the approximately forty page summary of the most recent IPCC report–and compare it to various articles written about it so that students learn the skill of doing their own analysis, not just relying on others’. This is a huge step up in rigor from what most schools do today, and we believe there is still value in learning how to read this type of report and compare it to coverage.

But now we’d add something like this to the assignment: we’d have students create a notebook of all of the IPCC reports ever released and query them--this is tens of thousands of pages. With the help of AI, anyone can come to understand these documents. AI can summarize each report, identify important changes across them, reveal the sources most relied upon, notice the people who are most involved and how this changes (or not) over time, or anything else that you ask it to. If the information it gives you back is too technical, you can ask it to explain it more simply until you understand it. If it makes you curious about something, you can ask it more questions. If you want to go to the source of a particular claim and read that piece of the report firsthand, you can.

Completing the task described would have taken hundreds of hours in the pre-AI world and would have required a good amount of background knowledge to do so–think of a team of New York Times investigative journalists working over a period of months. In our post-AI world, it could be completed by an advanced middle school student in under an hour. AI unlocks a whole new level of what students can learn and do across all our subjects.

We’re betting on a few skills that will become even more valuable in an AI-driven world: asking powerful questions, collaborating and building relationships, connecting ideas, persisting, persuading, and making decisions. What these skills have in common is that they all require judgment—navigating different paths, pursuing various lines of inquiry, and weighing complex choices.

These are the “gray area” skills, where there’s no single right answer. Today’s schools are often structured around clear-cut solutions and standardized tests, so at most schools, students do not directly learn these skills. Given that these “gray area” skills will be the most valuable in the Age of AI (and honestly, they always have been the important skills for life) we’re emphasizing these skills as we design courses at Rock Creek. We’re creating projects that develop these skills through real-world interviews, creative problem-solving, and analyses that demand nuanced, human perspectives.

AI is rapidly evolving, and learning how to collaborate with it is a valuable skill that is quickly becoming necessary for most ‘good’ jobs. At Rock Creek, we have a steady stream of learning experiments going on with ChatGPT so that we can determine how AI should shift what and how we learn.

As new capabilities are released, we’ll consider how we can apply them to the project of learning and working, and we’ll make our students collaborators in this endeavor. This habit will serve students for life as they graduate school and enter a workforce where human-AI collaboration is the norm.