Mark Rush and I gave two talks at AAC&U’s Conference on AI and Higher Education in Atlanta on October 30. The slides for both are up.
- Being Users Is Not Good Enough is the twenty-minute session (the program called it “Building the Agent”).
- Embrace the Revolution; or, How Professors Learned to Stop Worrying and Love AI is the pecha kucha, twenty slides at twenty seconds each. Mark led that one.
The first talk is about how I’m changing my classes, and it comes down to two drawings.
In the old drawing I assign something, the student makes it, and I grade what they hand in. That worked for a long time because an essay was expensive to write, so a good one was proof the thinking happened. Then the artifact got cheap. Last winter one of my seniors built an entire platform in four days, and another couldn’t answer an alum who asked her about security. All she could do was recount the prompts.
So I moved the grade one box upstream. In Foundations this fall, students build their own agent in the first week. It has a rules file it reads every time it starts, skills they wrote, and fences on what it’s allowed to touch. The agent makes the work, the work goes to a person off campus who owns a tedious job and can say no, and every week the student writes down what they decided and why.
I grade the rules file (it’s a paragraph, and I ask why each line is there), the weekly log, and whether the thing ran on the owner’s own information and the owner signed off. In the log, an honest “I tried it and it failed” passes and “nothing this week” doesn’t. Everything comes back Satisfactory or Needs Revision against a checklist they saw on day one. There are no points, because points let you skip the log and make it up with a nice build.
By week four, eight of my twenty first-years had written some version of “argue with me” into their agent, because they’d noticed it praised everything they did. Nobody assigned that.
The deck ends with three things you can try on Monday, and none of them need a CS department.



