While concerns about misuse and cheating persist, University of Michigan School of Information researchers argue that Artificial Intelligence, when applied thoughtfully, can transform how students learn, practice and grow.

“It really depends on the students,” says UMSI research professor Stephanie Teasley. “Cheating didn’t start with AI, and some will inevitably use it to cheat. But the real promise lies in helping students when a task is just slightly beyond their current ability.” 

Teasley, an expert in learning analytics, has been a strong supporter of using technology to provide personalized learning experiences and help educators make data-driven decisions that improve student outcomes. Teasley’s student-facing dashboard tool, My Learning Analytics, offers students the opportunity to gain insights into individual learning trajectories through actionable visualizations. The MyLA tool is now being launched on several other campuses in the United States.

Teasley and her team are currently exploring an AI component to MyLA. 

“The goal is not to do the work for the students, but to help them go a little farther than they would on their own,” she says. “If AI tools are used to support this kind of insight, like offering a suggested outline when students are staring at a blank page, that can be helpful. And ideally, over time, students would internalize this support and recognize what a good outline looks like and be able to create one themselves.”

This promise of guided support is echoed in UMSI associate professor Kevyn Collins-Thompson and professor Perry Samson’s, LearningClues, an AI-powered learning assistant. Designed to help students study more efficiently, especially those balancing jobs and coursework, LearningClues uses course-specific material to help students understand content on their own terms. Its virtual coach and adaptive question bank are part of a wave of tools aiming not to replace learning, but to empower it. 

The learning tool has been described as “life changing” for nontraditional students looking for additional support, says Collins-Thompson. 

“It’s about building students' confidence,” says Collins-Thompson. “Many of these students are non-traditional learners, working multiple jobs and returning to school after a long time. The software provides a structured way to boost their progress, helping them stay motivated.”

Since its inception in 2021, the platform has developed from an on-campus project to a growing startup with funding from the Michigan University Innovation Capital Fund. 

LearningClues has been used in research pilots at U-M and is being piloted at multiple other universities including UC Berkeley’s, the University of Wisconsin, and Georgia State University.

 “What’s also changed in the last couple of years is students can now use LearningClues to do personalized self-assessments using adaptive learning, based on high-quality questions generated from lecture materials,” says Collins-Thompson. “We’ve also introduced a 'coach' mode that guides students instead of giving answers directly, and a multi-course feature that can synthesize question answers and study review across all of a student's current and past courses.” 

UMSI associate professor Barbara Ericson has also been a longtime champion of using evidence-based teaching methods and technology to reduce student struggles and promote active learning, especially in complex subjects like computer programming.

An expert in computing education, Ericson has published numerous interactive ebooks with “ low cognitive load practice problems to help students learn difficult material.” Ericson has been teaching students to write code for decades. 

“The old approach of lecturing and then asking students to write code from scratch is often overwhelming,” she says.” I'm trying to reduce that by using what we know about how people learn, which is through active and social learning.” 

Ericson is researching how to leverage generative AI to help students learn.  If students struggle while writing code from scratch, she says, we can ask a large language model to find the closest correct code to their incorrect code and serve that code to them as a set of mixed-up blocks the students have to put in order.

“The goal is to help them learn, not just complete assignments,” she says. “Students prefer solving a code puzzle to just receiving the correct code because they still have to think through the problem.”

But what about the critics who say AI tools are doing students a disservice?

“That criticism has come with every new technology,” says Teasley. “I remember when pocket calculators came out. People said, ‘Kids won’t learn math.’ That didn’t happen.’” 

But still, Teasley acknowledges the risk of lazy use by students and instructors alike. 

“If you’re assigning busywork, and students use AI to complete it, maybe that’s more a reflection on the assignment than on the student.”

For Ericson, the key challenge is preventing over-reliance. 

“We’re seeing that students who rely too heavily on AI often struggle during exams,” she says. “That’s when they realize they haven’t really learned the material.”

Both Teasley and Ericson are working to steer AI use toward real learning gains. Teasley serves on the board of Michigan Virtual, a nonprofit helping K–12 districts explore AI responsibly. 

“There’s a real hunger for guidance,” she says. “District leaders know AI isn’t going away. They want to get ahead of potential problems, not just react to them.”

Ericson is also focused on expanding AI’s role in online learning platforms like Coursera, where LLMs may soon simulate interactive peer instruction for students who miss a lecture. 

“It’s not quite the same as a classroom but it’s a step toward more inclusive and accessible engagement,” she says.

As AI becomes more embedded in the classroom, Teasley, Collins-Thompson and Ericson agree: The goal is not to go backward, but to guide students forward with support and structure. 

 “Students are going to use these tools,” says Ericson. “The question is how do we help them use AI in ways that support learning?” 

RELATED

Barbara Ericson is an associate professor of information at UMSI and an associate professor of electrical engineering and computer science at the College of Engineering. 

Stephanie Teasley is a research professor emerita at UMSI and a professor emeritus in service at the Inter-university consortium for Political and Social Research at the Institute for Social Research. My Learning Analytics is in collaboration with U-M’s ITS Teaching & Learning Group. 

Kevyn Collins-Thompson is an associate professor of information at UMSI and an associate professor of electrical engineering and computer science at the College of Engineering. Learn more about LearningClues.