At the AI Show last weekend, a three-day addition to the ASU+GSV Summit focused solely on artificial intelligence and its educational applications, she offered guidelines for when it’s best to use emerging technologies in the classroom — namely custom GPTs, AI podcasts and AI agents — and when it’s best not to use tech at all.
CUSTOM GPTS
For example, in her regenerative medicine course, Austin created one chatbot called the “regenerative medicine mentor” as a way for students to ask general questions around the clock. While Austin found that traditional AI chatbots often got science and math problems in this area wrong, her custom GPT gave students accurate answers to their questions.
Austin said custom GPTs are a good tool to use for giving students reliably correct assistance with low-stakes concepts. When students need help with high-order thinking, she said, skip the GPT.
“Use [custom GPTs] when students need specific advice exploring a complex lesson of your choice,” she said. “Skip it when you don’t have time to test and monitor the responses, and you need to personally make sure that they understood.”
Austin said she has also created custom GPTs to serve as coaches for students. To do this, she provides the GPT with all the same information she shares with students, and students can ask the chatbot questions to make sure they are on track with their assignments each week. She said this can be especially helpful for long-term projects that require pacing.
“Right now, I’m teaching four different classes in four different disciplines across three schools, so if I did not have AI, honestly it would be a whole challenge,” she said. “But it is important to keep an eye on everything.”
AI PODCASTS
Google’s NotebookLM, Austin’s tool of choice for AI podcasts, takes materials users upload and turns them into conversational podcasts. According to Austin, the podcast format can hook students on topics they otherwise aren’t very interested in and work with their mobile lifestyle.
Austin uploads lecture notes and uses the platform to create alternate ways of engaging with the material.
“Use it when you want to get creative, when you want to be accessible,” she said.
However, she recommended not using it when you don’t have time to review the finished product before sharing it, and not making the podcast too long. Seven to nine minutes is the sweet spot, she said.
“In my case, I had it talk about regenerative medicine. Halfway through, it said I was a cardiologist, it was mispronouncing some scientific names, it was going off the rails,” Austin said. “And there were some issues with the biases, where the female voice was like, ‘Oh, wow, I didn't know that’ and the male voice was very intelligent.”
Instructors can add parameters to the request and regenerate the podcast if they encounter similar problems, Austin said, and should keep in mind that some complex topics may be better served without it.
AI AGENTS
AI agents, which have the most decision-making power of these tools and can run on their own, are best for high-repetition tasks that instructors would like to automate, Austin said.
For example, Austin uses an AI agent to send students summaries of notes and discussions each week without needing additional prompting. It just takes granting access to the learning management system, in her case Canvas, once.
“I can actually turn off my laptop, go on vacation, and it will keep sending students summaries and notes every week,” she said.
Austin likes to think of these tools like chess pieces, differing in number and ability but working together. It might make sense to use one AI agent and two custom GPTs, for example, to best assist one instructor. Like chess, it takes a cohesive strategy to be successful.
