Graham Clay isn’t just excited about the potential AI has for education in the future: He’s thrilled with the way it has helped save him time as a professor and enhanced the student experience in his classes.
Clay, a philosopher and PhD, has taught at The University of North Carolina at Chapel Hill and Fort Lewis College in Colorado. He also writes the AutomatED newsletter, which provides tips on how professors can incorporate AI into their teaching.
Clay regularly coaches fellow instructors in better utilizing AI in their classrooms. Here are some of his favorite ways to utilize AI to help save time, which range from simple-to-implement ideas to more complex strategies that require some extra work on the front end but pay dividends on the back end.
1. Using AI To Teach: Quizzes, Tests, and Other Content
AI can be really good at generating classroom content such as quizzes and slide shows based on educator lecture notes and recordings. Clay recommends Gemini 1.5 Pro, which anyone can access for free from Google AI Studio, because it allows for prompts that are very long.
“Maybe you record yourself using just the microphone at the front of the class so it can only hear you,” he says, which avoids any student privacy concerns. Then you pop that recording in Gemini 1.5 pro and ask it to create a 10-question quiz based on it.
“I start each subsequent day with a quiz about what happened the day before,” Clay says. “Normally, that’s a pretty heavy lift, because you have to remember everything you said, because maybe you deviated from your lecture notes because students ask questions, and so on.” Using AI in this way eliminates most of that work.
2. Writing Letters of Recommendation
Most college professors are asked to write regular letters of recommendation and have a template they already use, Clay says, whether that’s putting positive information first, then your qualifications and your relationship to the student.
To have AI help draft these letters, Clay recommends creating a prompt that describes that framework rather than adding an example of a past letter with any identity removed. “That shows the large language model exactly what you meant by your description,” he says.
The final step is adding a section of the prompt in which you use your favorite voice-to-text software to describe the individual student you are writing about in a stream-of-consciousness manner. Clay has done this while walking around his neighborhood with his dog. Then you put in the individualized prompt and edit the final letter of recommendation as needed.
3. Grading
Grading with AI is potentially a huge time-saver but also still has challenges when it comes to grading papers, etc.
Clay has a workaround for this. He still completes all the grading himself but then uses AI to save by creating a stream-of-consciousness response to each student’s work and then having an AI punch it up. For these, he prefers to write it out rather than use speech-to-text, but either approach can work.
“So it’s basically like having a TA sit there who has some understanding of the assignment and what the rubric looks like,” Clay says. “And here’s what you have to say about a given paper, and their job is to convert that content into something that a beginner student would understand and benefit from, and is professional and grammatical. That can be 10 to 15 minutes saved per paper. Half the time you know exactly what the problem is, and you just have to package it for this particular student. The AI can help you with that packaging.”
4. A ChatBot Trained For Your Class
For the more advanced AI teacher, Clay recommends creating a chatbot trained on data that is specific to your class and can help students when you’re not around.
Clay uses CustomGPT, which allows users to create their own tailored ChatGPT-style chatbots. Users can provide specific training data as well as set their own guardrails—for instance, Clay trains his not to give students the answer but instead act as a tutor. This doesn’t take any technical or coding skills; however, getting the chatbot right requires some fine tuning. Clay has written detailed instructions for doing this.
Despite the upfront work that is required, he says this can really help students once it is deployed. Students can ask it questions day or night, and it can be helpful during group work sessions during class. “Sometimes I have 40 students who are all pretty needy,” he says. “They are all beginners. They all ask a lot of questions. They’re not sure of how to proceed, and the CustomGPT can keep them going in class when I’m trying to circulate from group to group.”
He adds that as with many use cases for AI, a custom chatbot helps an individual instructor increase the impact they have on students. “In higher ed these days, there’s not that many of us who have TAs or a lot of support from assistants and other instructors,” Clay says. “So being able to kind of multiply yourself in some way through a CustomGPT is very appealing.”