"School is a gym, not a job."

On living up to a popular analogy

“School is a gym, not a job.”

You’ve probably heard it on podcasts or in conference presentations. It’s probably come up in meetings or discussions about AI integration. I’ve heard a version of it at almost every school I’ve visited.

At a job, the point is to get the work done, to deliver a product on a certain timeline that meets a certain goal and a certain level of quality. The process by which you arrived at that product matters less than the fact that you successfully delivered it. AI has the potential to be helpful at jobs because it can speed up delivery or improve the product or create multiple ways to arrive at that product.

At a gym, in contrast, success depends on our own effort. Altering the process to be more efficient, to find shortcuts, or to delegate effort undermines the purpose of the activity and has a negative impact on the outcome. Without the process of exercising, there is no product of fitness. I loathe going to the gym, but I still do because I just can’t seem to find a way around it.

When it comes to AI in education, the argument goes, we should think of school as a “cognitive gym,” a place where we exert our brains in order to strengthen them. Using AI to do cognitive work for us deprives our brains of the opportunity to exercise.

It’s a good analogy. Schools should be places of practice, of trying, places where it is safe to do things that are hard in order to become better at those things.

It’s also a good standard by which to assess ourselves. People go to gyms even though they don’t have to. People push their limits at gyms because they know that’s the only way to improve. People design workouts (weights, cardio, stretching, etc.) to align with their goals. People engage personal trainers for motivation, education, and accountability.

School is a system designed by adults for the benefit of students. In the age of AI, how do we ensure we’re designing it as a gym and not as a job? How do we help students see school as a place for training and improving, not delivering products?

Does your school signal “gym” or “job”?

In his book The Score, philosopher C. Thi Nguyen explains how our desire to quantify outcomes in metrics can incentivize the wrong behaviors. Jobs are increasingly defined by metrics: quantifiable outputs like products shipped, hours billed, workshops delivered, etc.

But metrics are not suitable for everything. The more dynamic and variable something is, the more resistant it is to assessment by metrics. Learning is a good example, yet we try to quantify learning in any number of ways. The example Nguyen offers is grades, which have become our metrics for learning. Over time, however, grades have become a classic example of Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure.

When a metric becomes a target, it creates a system where there is a gap between the signal (grades) and the actual work of the system (learning). To understand this idea, try to think of a student you know whose grades do not reflect what they have learned. Most teachers I know would not have to think for a long time to come up with a name.

Reliance on metrics can incentivize gaming the system, or, as Nguyen writes, exploiting “the gap between the signals a system uses to confer rewards and the genuine goods the system was built to promote.” Many students have been gaming the system of school for decades, seeking the shortest path possible to the highest possible grade. The arrival of generative AI makes the “job” of school even easier to game.

The darkest outcome of this is what Nguyen calls “value capture,” which is becoming so committed to gaming the system that you replace your personal values with the system’s values. He offers a few educational examples: seeing getting good grades as the point of school, or choosing a college or graduate school because of how a publication has ranked it.

We may say to ourselves and to our students that school is a gym, not a job, but what are the signals we send that back up this rhetoric? Do those signals focus on deliverables (job) or effort (gym)? Do our metrics of success incentivize gaming of the system or investing in it?

The arrival of AI offers an opportunity for us to audit our schools through this lens. It also offers us an opportunity to model a “gym” approach to AI.

ChatGPT: Vintage style image of robots working out in a gym. 1950’s era gym.

What does a “gym” approach to AI look like?

Teach Students the Process and Payoff of Learning

We teach athletes how to work out to maximize fitness, how to understand the impact practice has on their bodies and performance. We teach artists how technique improves their work and unlocks creative potential. We do not, for the most part, teach or model for students the relationship between schoolwork, learning, and expertise.

Blake Harvard, a teacher who has become an expert on applying cognitive science in the classroom, has watched his students use AI both to support their learning and to bypass it. He argues that teaching students about how learning happens and about what happens in their brains when they learn is one way to influence how they use AI (or don’t). The importance of this learning literacy extends beyond concerns about AI; it is part of our larger project to help students become independent, autonomous thinkers.

Simply teaching students how learning works is not enough, though. They also need authentic chances to see their cognitive workouts pay off. Athletes get to be on the field of play. Visual artists get to be in galleries. Performing artists get to be on stage. Students get… grades.

Andreas Schleicher, a director at the OECD, said in this recent Brookings Institute webinar that “AI has divorced task performance from learning.” As I’ve written about before, if we want to continue pursuing the core goals of education and to motivate students to do things that are hard, the work of school needs to change (and is changing!).

How do we reward students for knowing things? When do they get to see the benefits of their cognitive workouts?

Treat AI as an Object of Study

A mistake we have often made in education is treating technology as a solution to be integrated instead of a question to be explored.

What students need from us (and repeatedly tell us they want from us) is to learn about AI, to engage in the bigger issues, and to figure out how this technology fits into their lives and how they fit in a world increasingly shaped by AI.

Mike Taubman and Sam Kern have received a lot of well-deserved press for the AI literacy course they’ve developed for their high school students (this recent episode of “The Daily” podcast offers a good introduction). The course is built on the idea that students need time, space, and support to consider how they will exert agency over AI now and in the future. The course helps them get their “AI Driver’s License.”

Source: “The Rubber Meets the Road” by Mike Taubman

Along the same lines, computer science teacher Douglas Kiang and English teacher Rachel Blumenthal teamed up to develop a course where students study fiction on AI and practice writing with and without AI in order to think through whether or not AI has a place in human creativity.

These kinds of learning experiences do far more than ask students to use AI to help them perform a task. They ask students to engage critically with AI, to use their brains to develop informed opinions, to advocate for a particular point of view, and to connect AI to their visions for their future selves.

Use AI to Create, Not Consume

When you ask students how they use AI in their personal lives, the answers revolve around consumption: searching for information, creating content for fun, or homework help. If schools ask students to use AI, I hope they position students as creators, not consumers.

The skills gap that schools could fill is not “how do I use AI?” It’s, “how do I use AI to support interesting, creative work, not simply generate content for me to consume?”

A group of student writing tutors at the Waterford School in Utah, USA, created and conducted their own study to determine what led their peers to use AI and what support they needed on writing tasks. They then developed their own custom GPT to try to incorporate that feedback into a useful, school-sanctioned AI tool.

Mike Caufield used AI to develop his own film classifier as a way to illustrate how we can use AI to take constructivist approaches in the classroom. Teaching students how to use AI to build their own classifiers is not just creative, it is rigorous academic work that builds skills students will need in their futures.

The difference here is agency: the use of AI is driven by a student’s interest in exploring a certain idea in a certain way. AI tutors and other teacher-designed tools have their place, but we should be mindful of the role we are asking students to play in any AI interaction.

Cognitive Gyms Require More Than Our Brains

Camille Farrington’s research on developing a student’s “academic mindset” suggests that the design of the learning environment is a critical factor in whether or not a student will persevere through struggle. The most successful learning environments, Farrington argues, are ones where students can state:

• “I belong in this academic community.”

• “I can succeed at this.”

• “My ability and competencies grow with my effort.”

• “This work has value for me.”

These are what Farrington calls “noncognitive factors in learning,” issues of emotions, belonging, identity, etc. that have an impact on whether or not a student is truly engaged in academic effort. In other words, we must be mindful of, and perhaps prioritize, noncognitive factors in ensuring our schools are cognitive gyms, especially factors that we know are related to engagement.

This is my favorite part of the “school as gym” analogy. We all go to the gym for different reasons, we all use it in different ways, and we all have our ups-and-downs in our relationship with it (at least, I do). The question is, why do we keep going? Why do we work out? Why do we set the expectation that we will extend ourselves?

It’s because, above all, going to the gym is meaningful.

---

Upcoming Ways to Connect With Me

Speaking, Facilitation, and Consultation

• If you want to learn more about my work with schools and nonprofits, take a look at my website and reach out for a conversation. I’d love to hear about what you’re working on.

Links!

• The University of Chicago has decided to provide all of its employees and students with Claude Enterprise accounts. You should read the president’s letter explaining why.

• I love Jasmine Sun’s advice for college graduates about how to succeed in an AI world.

• Phillipa Hardman with a terrific overview of how AI is going to transform assessment practices.

• Software engineering professor Christa Lopes tells a troubling story about one of her student’s overreliance on Claude for a project. She also models verification through conversation, something that is increasingly important in an AI world.

• Common Sense Media has a new report on young people’s use of and views on AI, with data from children as young as nine.

• Carl Hendrick with a useful answer to his own question, “Is there a science of writing?”