AI research to improve K-12 STEM learning gets funding boost

Carolina’s computer‑science department is gearing up for another round of big‑ticket research, thanks to a fresh infusion of money from the National Science Foundation. The NSF has just renewed its five‑year, $20 million grant for the EngageAI Institute, pushing the program’s horizon out to 2031 and tacking on an extra $20 million to keep the work rolling.

The institute, headquartered at NC State but powered by a coalition of universities—including UNC‑Chapel Hill, Vanderbilt, Indiana University and the nonprofit Digital Promise—was born to tackle a simple, stubborn problem: how to make science, technology, engineering and math (STEM) learning more engaging and effective for K‑12 students. Over the past five years the team has already rolled out a handful of narrative‑driven learning environments, run classroom pilots, and built tools that can sift through video, speech and even gestures captured when kids work together on problems.

At UNC‑Chapel Hill, the effort is led by Mohit Bansal, the John R. and Louise S. Parker Distinguished Professor of Computer Science, who also serves as co‑principal investigator for the institute. He’s joined by faculty such as Xiaoming Liu, a distinguished professor, and Roni Sengupta, an assistant professor, each bringing a slice of AI expertise to the table.

“The NSF’s continued support is a clear signal that our work aligns with national priorities,” said Magnus Egerstedt, UNC’s executive vice chancellor and provost. “Artificial intelligence is reshaping everything, and universities have a responsibility to connect that technology with human purpose—especially in education.”

What’s new in this next phase? The researchers are aiming for lessons that can actually learn about each student, remember their progress across weeks or months, and tweak the material on the fly. Think of an AI‑driven tutor that watches a classroom, picks up on a student’s hesitation, and then offers a subtle hint or a different visual explanation, all while giving teachers a simple way to sprinkle in 3‑D storytelling worlds.

One of the standout tools already in the toolbox is EngageVP, a framework the UNC team created to decode video, speech, and gesture data when groups tackle problems together. This kind of multimodal analysis helps the system understand not just what a student says, but how they act—an insight that’s proving valuable for tailoring feedback.

Looking ahead, Bansal is optimistic. “We’ve built the core AI tech that powers EngageAI’s environments, and now we’ll focus on making it truly adaptive for each learner,” he said. “Our goal is to give teachers more engaging, personalized resources without adding extra workload.”

Beyond the labs, the project underscores a broader vision: leveraging AI responsibly to lift educational outcomes across the state and, eventually, the nation. As the EngageAI Institute rolls forward, UNC‑Chapel Hill hopes its research will ripple out, giving teachers practical tools and students a more inviting path into the world of STEM.