How is teaching going?

Thoughts after five weeks

I am teaching two courses as a visiting professor at UATX this term. One course is on Public Choice. The other course is on Political Psychology.

The term is only about half over. But many of my readers have asked for an update on how it is going, so here it is.

In Public Choice, my goals were to get across the ways in which Public Choice Theory differs from mainstream textbook economics. A simple way of putting it is to take the students from the mainstream theory that market failures create opportunities for wise and benevolent government intervention to “subsidize demand, restrict supply” as the most common form of intervention. Along the way, I wanted the students to understand the Coase Theorem, Hayek’s “Use of Knowledge” paper, Ostrom on Common Pool Resources, Buchanan and Tullock on the calculus of consent, and Olson on the logic of collective action.

Then I turn to some specific policy problems: social security and the effect of demographics; health care and the high rate of spending; housing and government intervention in housing finance. But these are mostly for later in the term.

I also wanted the students to practice “vibe coding,” by creating a virtual wax museum with some Public Choice economists and some left-leaning economists. Along the way, they will learn how these different economists answer questions that one can pose to the economists about public policy.

I wanted to conduct the class like a Swarthmore Honors seminar, with students writing papers and the class discussion centered around the papers. The written papers have been very good, but the discussions rarely work well. Part of the problem is that I have a lot of knowledge in these areas, and I end up having a lot to say as a result. So I tend to take over and lecture.

I think that the gap between what I know on the topic and what a student with only minimal economics background can pick up on short notice is too large to take the seminar-paper approach. Swarthmore seminars only begin in the junior year, so that students come in with a significant knowledge base to start with. I am teaching freshman and sophomores, so they are missing a lot concepts that would be helpful as background.

If I had it to do over again, I might instead give all of the students a list of concepts to study for each class. I could pose three or four short-answer questions for them to prepare in writing. The idea would be to establish a common base of knowledge in the first half hour of the class, and then we could proceed with lecture or discussion from there.

Once we were finished with the basic concepts of public choice, I collaborated with Claude to create a “practice exam.” I sent it to the students shortly before we had a snow day here in Austin (even a tiny snowfall jeopardizes transportation), so instead of trying to hold class on Zoom I had the students do the practice exam. I gave them a prompt to give to an AI of their choice to grade the exam, and I had them send me their answers and the AI grading. The AI grading was almost exactly the way I would have graded the exams. But note that the exam was practice, and as long as students turned in their answers with the AI grades I gave them credit.

In my view, the “practice exam” was a success. Students showed a good grasp of the concepts. The answers that they gave were evidently their own (not an AI’s), and most of them did well.

The wax museum project so far has not done as well. What I really hoped would that the students would report on what they learned by trying to do vibe coding. What was surprisingly easy, and what was surprisingly difficult? Few of the students have articulated their experiences.

In general, I do not feel that I have obtained as much of the students’ out-of-classroom effort as I would have liked. I think that the course has more to offer than what they have exploited. For example, I think that they could have learned something by going to a program that I attended at the University of Texas on energy policy.¹ I recommended it, but none of the students went. It would have added to the burden on their time. They also have not taken advantage of The Social Code, which has relevant sections, especially for political psychology.

This is partly the over-extended nature of young students. They find a lot of interesting things in college, both in and out of class, and they tend to take on too much. On top of that, it is quite common at UATX for students to have started businesses. And I’m not talking about selling cookies. One has a business to support service technicians. Another has a defense tech business, with funding from investors and from the defense department.

I might wish that I could capture more of my students’ mindshare, but it’s not so easy. Also, I am a very “soft” teacher. I do not cajole or threaten students to try to get them to do more. There is a joke about a Quaker with a balky horse. The Quaker, very frustrated, says to the horse, “You know I would never strike thee. . .But I could sell thee to a Methodist!” Sometimes, I feel like that Quaker.

The political psychology course uses material ranging from the 1950s (Theodore Adorno, Eric Hoffer, Richard Hofstader) to current substacks (Rob Henderson, Steve Stewart-Williams). I have advised the students to “vibe read” by asking for AI summaries of the readings.

The Swarthmore seminar approach works better in this course. I do not have so much advanced knowledge that I feel an urge to lecture. The only problem is that the class is at 8:30 AM, and sometimes the students are not fully able to engage that early.

I also gave a “practice exam” in that course, but it is not due for another few weeks. We will see how that goes.

The issue of how to use AI is very much in play at UATX, as it is at legacy universities. I am on the side of trying to make maximal use of it. I would like to see a freshman economics sequence that is mostly taught by an AI teaching assistant. I have a very perfectionist vision of what I would want this AI to do. I asked Claude if it is do-able, and Claude replied that it is a project that would take several months and up to a year. Claude thinks that my ideas are far more demanding than existing courses, such as Khan Academy, but that they are all within the realm of current capabilities of AI.

1. The program was put on by the Civitas Institute at the UT. The opening panel was anything but civil. Three of the authors of a controversial report issued by the Trump Administration were on a panel and a mainstream climate policy person was tasked with responding to each. The responder lost his composure. He argued from authority, used profanity, and scored relatively few substantive points. But I was not impressed with how the climate skeptics conducted themselves, either. After that, there were two more panels that were more satisfying.