In mid-August, Google announced a big change for this back-to-school season: AI features are now available in Google Classroom for all K-12 students — not just students 18 or older. With such easy access to AI tools and little guidance on how to use them, many teachers may be overwhelmed by this intimidating technology. 

With broad expertise in educational technology, literacy and teacher education, Associate Professors Stephanie Smith Budhai and Rachel Karchmer-Klein in UD’s College of Education and Human Development (CEHD) are here to offer support. With teachers in mind, Budhai and Karchmer-Klein share insight from their recent books on using AI critically in K-12 education and strengthening literacy across disciplines. 

First, do no harm

AI has the potential to powerfully advance innovative research and technological development. But students — who are only at the beginning of their educational journeys — are turning to AI for core academic tasks like writing topic sentences, identifying sources for a paper or analyzing literature. 

Bypassing these tasks can erode students’ confidence in their own abilities while giving them a false sense that they can succeed without these fundamental skills. The result is not just “de-skilling” — losing the skills they once had learned — but what Budhai calls “never-skilling.”

“Students’ cognitive capacities are being impacted as they offload tasks to AI,” Budhai said. “Memory loss, brain atrophy and decrease of critical and creative thinking are of great concern as students increase their use of AI.”

This is not to say that Budhai or Karchmer-Klein oppose the use of AI in the classroom. Instead, their goal is to help teachers and students use AI with purpose so that they “first, do no harm.” 

Purposeful integration

Budhai and Karchmer-Klein remind teachers that they are the experts in learning: they already know how to plan a lesson, how to measure students’ growth and how to do those things well. 

“AI is a promising and exciting tool that can serve as an intellectual collaborator, but it should support, not supersede, teachers’ professional judgment,” said Karchmer-Klein.

Budhai also reminds teachers of the SAMR model, which identifies four purposeful levels of technology integration to enhance and transform learning: substitution, augmentation, modification and redefinition. While simpler forms of technology allow a user to substitute one method for another — like replacing a chalk-based blackboard with a dry-erase whiteboard — the most advanced forms of technology redefine an activity, allowing a user to do what wasn’t possible before — -like using video conferencing tools to co-teach with teachers and students in a different country. 

“Before a teacher starts to use an AI tool, I always ask, ‘What is the purpose of using this tool? Is it really serving the learning objective, or is it more of a novelty device, or even a shortcut?” Budhai said. “We could plan lessons or write individualized education plans before AI. But how can you use AI to redefine learning in a way that you couldn’t do before?”

Karchmer-Klein encourages teachers to adopt a “literacy first approach,” which teaches students to use reading, problem-solving and critical-thinking skills in each interaction with AI. 

"AI is a problem-solver, but it’s also a problem to solve,” said Karchmer-Klein. “If you have a question that you can’t figure out on your own, it can help. But, if you don’t prompt it correctly, it will create an even bigger problem, giving you misinformation, pointing you in the wrong direction or replicating societal biases.” 

“So teachers have a real opportunity to teach critical thinking as their students are using AI in a safe, classroom setting with guidance from an adult who has already developed those foundational literacy skills,” said Karchmer-Klein. “This is much better than having students explore AI at home, in front of a phone, where they’re less likely to use those skills.”

Start with AI’s limitations 

But where should teachers start? Karchmer-Klein recommends starting with AI’s limitations. Consider lesson plans that first teach students how and why an AI tool’s response depends on the human prompt that it receives. 

“Begin by teaching different levels of prompting,” said Karchmer-Klein. “If you put in a very basic question, what kind of information comes out? If you put in the same question but now with more detail or context, how does the response change?”

Next, teach students to investigate the accuracy of AI responses by verifying information through other sources and to look for societal biases. 

“If you ask AI to show you an image of doctors, what or who is missing from that image?,” said Karchmer-Klein. “Does the image only depict men? How does the image change if you ask it to show a group of nurses?”

Once students develop these skills, teachers can use AI tools for more complicated compare-and-contrast activities. For example, students could workshop a piece of writing with a peer and then ask an AI tool for feedback on the same writing. Students can then analyze the response they receive with a peer, noting the differences in the feedback.

“The key is positioning AI as a tool, rather than something or someone that is doing the work for you, which, in fact, applies to any form of educational technology,” said Karchmer-Klein. 

To learn more about CEHD research in STEM education and teacher preparation, visit its research webpage.