K-12 teachers need to understand AI. So do those who train them.

January 24, 2024

CEHHS Associate Dean Stein Brunvand was part of an inaugural fellowship program to help education faculty integrate this evolving technology into their curricula.

A graphic featuring a headshot of Professor of Education Stein Brunvand surrounded by icongraphy representing educational technology, including children on laptops.
Graphic by Violet Dashi. Images by Drazen via Adobe Stock

Recently, College of Education, Health and Human Services Associate Dean Stein Brunvand was in a virtual meeting when his colleague, Assistant Professor of ESL Education Kyongson Park, asked the group for input on a rubric for her class. She had the categories nailed down, but still needed to come up with specific criteria for what would be deemed "exceptional" versus "approaching standards" versus "below standards."

Multitasking as folks do from time to time on Zoom, Brunvand opened up ChatGPT and instructed: “‘I need you to be a professor of an undergraduate course in English as a second language methods and I need a rubric for a lesson plan.’ And I specified what the lesson plan had to have and the categories and said ‘I need criteria for a four-point rubric.’”

Before the meeting ended — in less than five minutes, actually — Brunvand was able to send Park the results. Not being a subject matter expert, he couldn’t ensure they were helpful. But overcoming the blank page is often the biggest hurdle, and he figured the info might provide a useful starting point. Park agreed.

When ChatGPT debuted just over a year ago, it ignited fears across K-12 and higher ed about the potential of generative AI to propel cheating on exams and essays into hyper-drive. But many educators — both nationwide and here on the UM-Dearborn campus — are focusing on its potential to enhance both learning and instruction, even as they remain highly cognizant of its limits and the ways it can be misused.

Brunvand and his colleagues in the Department of Education have a somewhat different relationship to generative AI than other faculty: not only do they have to think about the implications of the technology for their own students, they have to prepare those students to use it thoughtfully in other educational settings, including the K-12 classrooms where many of them will eventually teach. And, of course, they have to do so while continually learning about this quick-evolving technology themselves.

To know more about what he doesn’t know, as Brunvand puts it, he applied for the International Society for Technology in Education’s first-ever ”AI Exploration for Educator Preparation Programs” fellowship. Brunvand was selected for the eight-month program, which began in March 2023, along with six other educators from across the country. The fellows attended workshops with leaders in the field, learned about new resources and research, and collaborated on a presentation, “Critical Strategies That Prepare Teachers to Teach With and About Artificial Intelligence,” which they delivered at ISTE’s annual conference in Philadelphia. Building on that presentation, they offered a two-hour online workshop for educators nationwide in November. It was so successful, they are offering a second workshop — expanded to three hours — on Feb. 29.

Brunvand will also be sharing what he’s learned directly with the UM-Dearborn — and broader U-M — community. In addition to developing a workshop or other learning opportunity for colleagues across campus, he’s a part of two proposed presentations — one with Professor of Reading and Language Arts Danielle DeFauw and one with UM-Dearborn’s generative AI task force, of which he’s a member — at a “Generative AI in Education” conference being held at UM-Flint in March. And he’s working with Professor of Technology Mesut Duran to incorporate the best practices he’s studied into revamped Educational Technology undergraduate and graduate programs.

Brunvand shared some of the key takeaways from his fellowship experience with Reporter. 

It’s not what you say, it’s how you say it

The more specific the prompts given generative AI, the better the chances that the content will prove useful. Brunvand’s query on behalf of his colleague is a good example: it provides a context, assigns a specific role, creates a task and defines a format. It’s important to remember that a query will not usually be a one-shot exercise. If the output is promising but doesn’t quite suit an educator’s needs, revise the prompts and try again. Aspiring teachers can learn how to do this on the job by doing it during their time at UM-Dearborn, Brunvand says.

"We can be up front with students and say, 'I actually want you to use ChatGPT and I want you to use this right,'" he says. "'Here's the assignment that I'm giving you. I want to see the prompts that you put in. I want to see how you revise those prompts to narrow down the content that you got and I also want to see how you interrogated that content. How did you do your fact checking? Is it accurate? Are there hallucinations in the responses that you got?' and go through that process, and so that students can start to see, 'Okay, not everything I get out of this is going to be perfect. I need to verify, just like I would verify if I'm doing a search in Google.'"

Be aware of bias

The universe of information that generative AI draws from — essentially everything that is accessible on the internet — is vast. But it is also inherently biased. Educators need to critically assess the answers they receive and ask themselves who or what may be left out. “An example that someone gave was they did a search for the top children's authors, and they noticed that the first five names that came up were all white men,” Brunvand recalls. "They thought, 'Is that really true?' I mean, if I were to ask humans, if I were to approach this a different way, would I still come up with just a bunch of white guys as the top children's authors?"

Know students’ needs

One of the most fundamental skills taught in the education program is how to create a lesson plan. Although new teachers will be expected to use a highly structured curriculum, no matter what district they work in, there will likely still be room to customize. At that point, generative AI can be a big help. 

"Creating a lesson plan from scratch might not be necessary, but we want our graduates  to be able to use something like ChatGPT or other similar type of generative tool to say, 'I'm a third grade teacher, and I'm teaching about the water cycle. And I need some examples for how to teach this to students for whom English is not their first language, or I need some ideas for how to differentiate my instruction, because I've got students at different learning abilities, or I need some more ideas for hands on activities,'" Brunvand explains. “And it's not to say that you're going to get wonderful ideas spit back out at you, but as a teacher. if you've got an understanding of your students and know what their needs are, you can turn to some of these tools and say, ‘Okay, help me out.’”


Article by Kristin Palm