My UATX term winds up

what worked and what didn't

Here is a video that includes a quick panorama (seconds 12-13) of one of my sections of Political Psychology. I taught that course as well as Public Choice. Elsewhere in the 90-second video, a few of my students in other classes appear.

What worked

In the Public Choice course, I assigned a “wax museum project.” I tasked each student with using AI to create a virtual wax museum of public choice economists (plus Richard Musgrove and John Kenneth Galbraith).

My assignment specified that I wanted the students to send me regular status reports. I wanted students to focus on learning from the process, not so much on trying to produce the fanciest end-product. But most of the students failed to notice that I cared more about the process.

Regardless of what you hear about “vibe-coding,” several of the students experienced frustrations along the way. That is a good experience, in my opinion. Overall, I would say that the wax museum project was an experiment that worked. And I think it was useful for students to experience what it is currently like to code with AI.

But a year from now, AI coding probably will be much different. So the same project probably would not be appropriate.

Another experiment that worked was scheduling one-on-one meetings with students. I thought that this was better than passively holding office hours. It helped me get to know students better and have a sense of how the classes were going.

Also, I was happy with the results of “practice exams.” In both the Public Choice class and the Political Psychology class, I made up exams that tested students’ knowledge of the important concepts in the course. I gave them as take-home exams, told the students that they would get full credit for completion, and I gave them a prompt to give to an AI to grade the exams. They turned in by email their answers and the AI evaluations. I was very happy with the results. I could tell that the students wrote the answers themselves, and most of the AI evaluations of the answers were justifiably good. Students understood concepts like rent-seeking (in Public Choice) and moral dyad theory (in Political Psychology).

The only thing that would make me hesitate to use this procedure for a real exam, without full credit for completion, is that the students would be tempted to use AI to help them write the answers. So supervision would be required. But AI grading definitely works.

I was also happy with an assignment that I called “three take-aways.” In each course, I had students write one paragraph on each of three concepts or ideas that they would remember from the course. For example, a student might write about the Coase Theorem, or the Dunbar Number. Because class discussions were free-floating, a take-away did not necessarily have to be a concept that was on the syllabus.

What didn’t work as well

The class discussions were decent. The students had intelligent things to say.

But they did not work as well as I had hoped. Fifty-plus years ago, when I was a student at Swarthmore College, students had more time to focus on course work. My memory is that the required readings at Swarthmore were a floor. Everybody went over the required material before class met, and often we found additional reading available in the library.

There was also a student culture that took pride in their dedication to serious study. A few years after I graduated, somebody came up with a T-shirt that said “Swarthmore college: guilt without sex.”

For my courses at UATX, I tried to follow the Swarthmore model. For each class, I had one or two students write papers on the assigned reading. I emphasized class discussions

If students had treated required reading like a floor, there was plenty of other additional reading available in substacks and other sources. Instead, they treated required reading more like a ceiling. Only a few of my UATX students did all of it. By the latter half of the term, only the paper-writer was doing the required reading.

As individuals, UATX students are as intellectually oriented as my Swarthmore contemporaries were. But the opportunity cost of studying for a Swarthmore student in the Philadelphia suburbs of 1974 was much lower than that for a UATX student in Austin in 2026.

One student suggested to me that if you want everyone in class to do the required reading, you have to do something like give short graded quizzes at the start of every class. But I do not want to use coercion to get students’ full dedicated effort.

I cannot honestly say that the most important use of their time is to work on my course. UATX creates a much higher opportunity cost, with a deluge of guest speakers and internships. They have opportunities to start businesses, get involved in the Austin tech industry, meet interesting people, and live on their own (students are housed several blocks from the classrooms, in fancy apartments recently built for University of Texas students).

The trauma of grading

I think of the Swarthmore Honors Program as close to the ideal form of education. When I was there, juniors and seniors could take two seminars in a semester. Each seminar met for one three-hour session per week. Seminars had about 6 to 8 students. The professor provided the syllabus, but students were supposed to play a prominent role. One or two students would write papers about that week’s material, and these would be the foundation for the class discussion.

Perhaps the most interesting feature of this program was that professors did not grade their own students. Instead, Swarthmore used outside examiners. The professor would send the syllabus to the examiner, who would then make up an exam. The examiner would mail the exam to Swarthmore, which would administer it to the students and then send their blue books to the examiner. The examiner also would come to Swarthmore to interview students in person as a follow-up to the written exams.

But UATX uses the old-fashion system of grading your own courses. And it also has a very strict curve.

If you are going to do grading, you need to be as explicit as possible about your grading criteria. I failed to do that. I had fewer than 12 students in each course, so I figured that I could form an impression of how students were doing based on their papers and discussions. That worked to my satisfaction, but this was far from transparent to students.

The worst part about it was the curve. I understand that UATX wants to position itself against grade inflation and the Ivy League. But the strict curve violates the “independence from irrelevant alternatives” postulate. Student A’s grade depends on student B’s performance.

This became painfully clear when a few students handed in assignments after I had drafted grades in our courseware (I still want courseware to die) but before grading became final. When I now gave partial credit to student B for their late assignment, I had to lower student A’s grade in order to maintain the curve. And this was all visible to student A in real time. The moral of the story is that I should keep grades off of the courseware until the date where I will no longer accept late assignments.