Let professors play: the case for sandbox classes in HE

Most classes, even electives, aren’t blank canvases. They come with constraints: required topics, departmental oversight, legacy syllabi, curricular goals and the invisible weight of student expectations. You’re not just teaching a class; you’re fulfilling a role in a larger educational ecosystem: covering “core content”, preparing students for downstream courses or meeting general education distribution criteria.

When I first arrived at Carnegie Mellon University, I was handed a pedagogical gift: a course that was free from those constraints. My department had plenty of coverage for its core courses but was looking to expand its elective offerings, ideally in a way that didn’t duplicate content covered in other classes. They invited me to propose a new elective that didn’t need to satisfy a major or distribution requirement and wasn’t tied to a pre-specified textbook or testing regime. Neither did it have to exist to feed another class. The result was Human Intelligence and Human Stupidity (HIHS), a course which started as a playful idea, the paradox of how a species can be brilliant enough to reach the moon but dumb enough to fall for email scams. But it became something more: a proving ground for pedagogy, a generator of institutional change and a joy to teach.

Of all the innovations that came out of the experience, perhaps the most valuable was the notion of a sandbox course itself: a structurally protected space where instructors can test ideas that would be too risky, too weird or too time-consuming to attempt in a load-bearing course.

The sandbox effect

We designed HIHS from the ground up. The teaching team experimented with everything: grading structures, collaborative norms, use of AI, even the role of the lecture itself.

One of the most successful experiments was a policy called optional-mandatory attendance. At the start of the semester, students choose whether attendance counts toward their grade. If they opt in, attendance is mandatory and tracked. If not, attendance remains fully optional. This simple intervention preserves student autonomy while also fostering commitment.

The results are unequivocal: most students choose to make their attendance mandatory, are more likely to show up to class and participate more when they do. Some colleagues and I later formally tested a variant of the approach in other classes using a randomised controlled trial. It has changed how I teach in all of my classes, and has now been adopted by other instructors at my institution and beyond. But that publication wouldn’t exist if I hadn’t had a sandbox class where the idea could first be developed, refined and tested.

• Spotlight guide: How to motivate students and staff in higher education
• Spotlight guide: Bringing GenAI into the university classroom
• Fostering module consistency to aid learning

Best practices can block better practices

Pedagogy is a living, breathing frontier. We have plenty of evidence-based methods, but we also have new technologies, new student populations, new findings in cognitive and education science and a changing cultural landscape. We should absolutely teach with best practices in mind, but we also need spaces where those practices can be bent, or even broken.

That’s what sandbox classes allow. Because they sit outside of curricular pipelines, they can be exploratory. Because they don’t satisfy requirements, there’s room to fail and to iterate.

One might worry that a sandbox becomes code for “anything goes”. It shouldn’t. The freedom I’m describing isn’t freedom from rigour. It’s freedom from inherited format. Sandbox courses should still be held to high standards: clear learning goals, frequent feedback and evidence that students are learning. Innovation is not the opposite of accountability.

HIHS is demanding. It pushes students to reflect deeply, engage critically and navigate controversial topics with intellectual humility. But it uses experimental and nonstandard techniques to get there. For example, one assignment requires students to debate AI-generated counterarguments to their own essays, improving metacognitive awareness, argumentative resilience, and AI skills. In a recent, peer-reviewed study on this assignment, we found that students learned to anticipate counterarguments to their claims, became more effective at prompt engineering and gained confidence in their ability to work with AI.

The freedom of the sandbox lets me match assessment to pedagogy in deeply intentional ways. For example, my final exam is decidedly non-standard. Students work in teams, rotating from room to room and solving challenges that seem, at least on the surface, as though they are unrelated to course content and fully unpredictable.  

That design is deliberate. My goal is to test whether they have mastered and can apply principles of collective intelligence. An intelligent group isn’t defined by what it knows in advance. It’s defined by its ability to coordinate, adapt and solve problems when the world changes without warning. So rather than testing whether students memorised course content, I test whether they have learned to work as a team well enough to handle tasks that they can’t directly prepare for. You can’t study for what you’ll be doing, because you won’t know what you’ll be doing. You can only study for how you’ll do it together.

It’s the kind of assessment I could develop only because the course itself is so unconstrained. But as I’ve refined and improved the exam structure over the years, it too has begun to generate transferable insights. The practicum final is now regularly observed by university administrators – not to evaluate me, but to better understand how to assess collective performance and smart teamwork.

The class uses experimental grading structures, non-standard classroom activities, live feedback loops and analytics-driven exit tickets that provide lecture-by-lecture insights on student engagement and content mastery. The sandbox isn’t just a teaching curiosity, it has become a research lab in its own right.

Let professors play

So, here’s a simple proposal: every department should sponsor at least one sandbox course. These courses could be taught by faculty members with a track record of good pedagogy, who are given full latitude to experiment and evaluated not by content coverage but by learning outcomes and instructional innovation. These sandbox classes shouldn’t satisfy requirements; they should be electives by design, honest about their experimental nature, and structurally protected from the forces that encourage conformity. When they succeed, their lessons can be integrated into other courses. When they fail, they fail safely and informatively.

Teaching centres could collaborate with sandbox instructors to document lessons learned and disseminate them more broadly. Promotion and tenure committees could recognise sandbox innovation as a contribution to the scholarship of teaching and learning, especially when it leads to peer-reviewed publications, as it has in my case.  Over time, these courses could become incubators for innovation across the university.

If we want better classes, we need to create the conditions where better classes can emerge. That means giving faculty not just permission, but encouragement to try things that no one has tried before. 

So let’s build some sandboxes. You never know what you’ll find when you start to play.

Danny Oppenheimer is a professor of psychology at Carnegie Mellon University.

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.