an ai policy designed with my students

posted on Jan 27, 2025

danny caballero

we do lots of modeling

I’m teaching Classical Mechanics this semester. In that class, students learn to make models of physical systems. We study the mechanics of physical systems; their motion, energies, stabilities, and so on. But, instead of framing the course as a series of exercises and problems, I’ve been trying to frame it as a way of modeling the world – through classical mechanics.

With that framing, I can open new avenues for students, and bring in more modern physics and approaches. I’m sure it’s that I’m partial to nonlinear dynamics and chaos theory from the group I joined at Georgia Tech. I started working in experimental fluid dynamics and nonlinear science before working in physics education. Taking those experiences and blending them with my professional work in physics education research has made me consider that physics courses for majors are great opportunities for students to explore and investigate systems.

we do lots of coding

While we start the class with the elementary motion of particles, like projectile motion, we quickly move to more complex systems. By focusing on models, we begin to build a general process and set of analytical and computational techniques that helps us investigate systems and their behaviors. Students began to explore the stability of systems, how they evolve in phase space, and how they respond to perturbations. By working to build their ideas and skills by focusing on modeling, they began building confidence in exploring systems that are new-to-them.

In this class, students learn to write Python code that can integrate equations of motion, plot their solutions in real and phase space, investigate the numerical error in their solutions, and perform Fourier analysis on data. For their final project, students develop deep analysis of a new system of interest to them using both analytical and computational techniques. They write up a report to present their work in a form of computational essay. My colleague, Tor Ole Odden, has worked a lot with these essays in physics. And he’s found that they can provide students with significant agency over their learning.

ai has enabled students to explore more

In my prior teaching of this course, I found that students were using AI tools to help them answer questions, write code, debug solutions, fix code, outline reports, edit reports, and write them completely. Their final projects were extensive and exploratory, but sometimes with mathematics and code that we had not discussed in class. I nominally had a policy that students could use any resources, but the rapidity with which students found use cases for AI tools was surprising. And the concern I had for their understanding of the material led me to reconsider my policy.

As I’ve written before, I wanted co-design an AI policy with my students this time. I wanted to ensure we had a discussion about the benefits and the problems with AI. In class, we talked about energy consumption, data privacy, and the potential for bias in AI tools. The students discussed a few questions in small groups and then reported their answers in a survey, which I turned into four proposals.

• Proposal 1: We adopt a policy that does not allow AI use at all.
  • Violation results in a failing grade on assignment.
  • Repeated violations result in failing the course.
• Proposal 2: We adopt a policy that allows AI use for brainstorming, help, and editing.
  • AI cannot be used for direct answers or completion of assignments.
  • We expect documentation of AI use, but it can be informal.
  • Violations are discussed with Danny; the first violation requires a redo of the assignment, and repeated violations result in a failing grade.
• Proposal 3: We adopt a policy that allows AI for use in nearly any way.
  • We require detailed documentation of use; this means screenshots, prompts, responses, and outcomes.
  • Violations are discussed with Danny; the first violation requires a redo of the assignment, and repeated violations result in a failing grade.
• Proposal 4: We adopt a policy that allows AI for use in any way with no documentation required.
  • Violations of the policy are limited to sharing answers or solutions with others.

My students used rank-choice voting to select the policy that they wanted to adopt. They results appear below.

Here’s a brief summary of the results (complete results are in the image above):

1. Proposal 2 (AI for helping; documentation needed): 61.5% first choice; 12.8% second choice
2. Proposal 3 (AI for anything; documentation needed): 26.5% first choice; 61.5% second choice
3. Proposal 4 (AI for any use; no documentation): 46.2% third choice; 23.1% last choice
4. Proposal 1 (No AI in this class): 17.9% third choice; 74.4% last choice

Students selected a justifiable, pragmatic, and transparent policy. They will use it to enhance and support their learning with brainstorming, seeking help, editing, and so on. But they will provide documentation of their use. Their documentation will help me understand how they are using AI and how it can be a productive tool in class. I can share those examples with the class, so more students can benefit from the productive uses of AI tools.

But what was more interesting to me is that the least popular options were a free license to use AI tools in any way without documentation, and no use at all. This vote is really interesting to me because it indicates that students have views about the use of AI tools that are nuanced and complex. I do want to know more about what they think about using these tools, especially over time.

ai policy going forward

The final aspect of the policy is that students can amend it. If they collect 1/3 of the class with a proposed amendment, we will discuss it, and students will vote on it. This is a way to ensure that the policy remains responsive to the needs of the students, but equally importantly, it matters to me that students see the course as a place where we can explore not only classical mechanics, but how we approach learning about it.