Project Learning in the age of AI

trying to think it through

Hollis Robbins writes,

1. Adopt a Project-Based Learning (PBL) Model: Structure the curriculum around CSU-faculty-approved “main course” projects, where the project is the primary vehicle for learning, not a final activity. This approach naturally fosters critical thinking, collaboration, and self-directed learning.

There is much more (too much more?) at the link.¹

I agree that the fit between AI and project-based learning is a good one.

When I taught AP economics in high school and “Economics for the Citizen” at George Mason, I liked giving students a project to outline a business. I told them to think in terms of a “strip-mall business,” to avoid anything too complex or online. An ice cream store. A barber shop. A video game arcade.

I wanted the students to calculate costs in terms of the main factors of production: labor, land, and capital. I wanted them to identify key ratios that would determine success or failure of the business. Sales per hour is a good example. I wanted them to identify key roles and responsibilities. I wanted them to think about issues related to marketing, including location. I wanted them to identify key unknowns, such as local demographics or potential competition. I wanted them to think about inventory management.

I assigned this as a group project. On the plus side, this allows students to brainstorm together. But otherwise, group projects are a terrible idea. The main reason I did it was to conserve on my grading time. At GMU, enrollment in the course was 100 students! I was paid $1200 to teach the course, minus the fee I had to pay for parking.²

Today, I would give the assignment to students individually. They would not need other students with whom to brainstorm, because they could use AI. I could handle thousands of students, because assessment would be done by AI. If necessary, we can assign a monitor to be present to make certain that the student is not cheating by using an AI during the assessment.

I would have the AI interview each student about that student’s business idea. The AI would assign a grade and offer feedback based on a rubric that I would provide.

Of course, students would get help from AI in doing the assignment. My guess is that the AI would enable them to avoid the mistakes that my students typically made. They had difficulty with the concept of breaking costs down to a common time unit. Often, a student would say that rent was the highest cost, because rent might be $1000 a month, whereas two workers only cost $20 an hour each. Another mistake to avoid was assigning zero to the cost of the student’s own labor.

The way I look at it, getting help from the AI is not a problem, as long as the student understands the thought process involved. In other words, as long as the student is learning from the AI, not just copy-pasting.

I figure that student learning can be assessed by my AI interviewer. It would be able to challenge the student’s understanding by asking questions about different scenarios and alternative choices.

substacks referenced above: @

Professors Behaving Badly

@

How to Deliver CSU's Gen Ed with AI

1. She is thinking in terms of using AI to lower the cost of providing general education courses. But maybe lowering the cost is not an objective. Note the following cynical view:

Gen eds generally don’t work. Faculty exploit students by forcing them to take useless and ineffective courses that students don’t want to. The purpose of many gen eds is to transfer money from students to professors.

That is Steve Stewart-Williams, writing about a takedown of higher ed from Jason Brennan and Phil Magness that came out in 2019.

2. Yes, I am very bitter about the experience. GMU did not deserve me, and I did not teach there very long. The high school did not pay me anything, but the students were better and the school did not charge me for parking, so I stuck with high school teaching much longer.