NSF Bets $20 Million on AI Storytelling That Teaches K-12 Students

Artificial intelligence is moving rapidly into classrooms, but one of the most ambitious efforts to make that transition genuinely useful to learning has just secured a major vote of confidence. The U.S. National Science Foundation has awarded North Carolina State University a five-year, $20 million grant to continue developing AI tools that generate interactive narratives capable of engaging K-12 students and improving how they learn. The award renews funding for the NSF AI Institute for Engaged Learning, known as EngageAI, which was established at NC State through an identical five-year, $20 million grant and has spent its first funding cycle building a suite of AI-enabled learning technologies tested directly in classrooms.

The renewal is notable both for its size and for what it signals about the direction of AI-in-education research. Rather than focusing on chatbots that answer questions or systems that grade homework, the EngageAI Institute centers its work on story. Its researchers are building AI that can generate interactive narratives, story-based learning experiences in which students participate, make choices, and collaborate, with the narrative machinery adapting in real time to what students do and say. The premise is grounded in decades of education research suggesting that narrative context can deepen engagement, give abstract concepts concrete meaning, and sustain attention in ways that worksheets and lectures often cannot.

The institute draws on expertise across a broad coalition. AI and education researchers at NC State work alongside colleagues at Vanderbilt University, the University of North Carolina at Chapel Hill, and Indiana University, as well as the nonprofit organization Digital Promise, which specializes in translating education research into practice for schools. That mix matters, because building an AI narrative system that actually helps students learn requires more than strong machine learning. It requires specialists in learning sciences who can design studies, teachers who can ground the work in real classrooms, and technologists who can turn those insights into systems that run reliably for students and educators.

During its first five years, the institute reported significant advances in AI-enabled learning technologies and developed a suite of tools intended to serve as the foundation for creating highly effective and deeply engaging learning experiences for K-12 students. Crucially, those tools were not developed in a vacuum. The institute conducted a multitude of classroom studies to investigate how AI-generated supports actually make a meaningful difference for students, with a particular emphasis on collaborative learning, in which groups of students work together and the AI must respond to a dynamic social environment rather than a single learner. Studying these systems under authentic classroom conditions is one of the harder problems in educational technology, because laboratory results frequently fail to survive contact with crowded classrooms, limited devices, and the realities of a school day.

James Lester, the principal investigator of the new award and the Goodnight Distinguished University Professor in Artificial Intelligence and Machine Learning at NC State, says the continued funding will allow the team to build on that foundation in three distinct ways. The first is a self-improving system, one that can get better and better at both engaging with students and helping them learn more effectively in real time. According to Lester, the system will actually improve over the course of a school year as it works with students. That idea, sometimes described in the field as continuous or lifelong learning, is technically demanding: the AI must learn from its interactions without degrading, adapt to new students and classrooms it has never seen, and improve its narrative and pedagogical decisions from feedback signals that are noisy and indirect, since genuine learning gains unfold over weeks and months rather than seconds.

The second major direction taps into emerging world model technologies for narrative generation. World models are AI systems that maintain an internal representation of an environment, its characters, its objects, and the rules that govern how states change over time. In the context of interactive storytelling, a world model gives the system a coherent understanding of the story world, so that as students act, the narrative can unfold consistently, characters can behave believably, and the plot can branch in ways that remain meaningful rather than collapsing into contradiction. Lester describes these state-of-the-art storytelling capabilities as holding enormous potential for student learning, and the technical logic is straightforward: the richer and more coherent the simulated story world, the more authentic the choices students can make within it, and the more naturally educational content can be woven into the experience.

The third direction expands authorship. The researchers plan to broaden the ability of teachers and students to author the narratives generated by the technology, which will help tailor story-based learning experiences to individual classrooms. This is a deliberate answer to one of the persistent criticisms of AI in education, namely that automated systems can sideline educators and impose generic content on diverse groups of learners. By making teachers co-creators rather than passive users, the institute aims to let educators align narratives with their own curricula, their students’ interests, and local context. Student authorship adds another layer, turning learners from consumers of a story into participants in shaping it.

Underlying all three directions is a design principle the institute calls agency-centered learning. Wookhee Min, a senior research scientist at NC State and managing director of the EngageAI Institute, says the team is keeping that principle at the core of the work, ensuring that students feel empowered. Students need to have a sense of ownership, Min explains, and that sense of ownership can have a meaningful impact on their level of engagement, their learning outcomes, and the learning process itself. The framing reflects a growing consensus in the learning sciences that engagement is not merely a pleasant side effect of good technology but a mechanism of learning in its own right. A narrative system that dazzles students while leaving them feeling passive may entertain, but it may not teach; one that gives students real agency over how a story unfolds can convert that agency into motivation, persistence, and deeper processing of the material.

The stakes of this research extend well beyond a single institute. Schools across the United States and around the world are being offered a flood of AI products, many of them built quickly and validated thinly, if at all. Large public investments in careful, classroom-tested AI learning research provide a counterweight: models of how such systems should be built, evaluated, and governed. The EngageAI approach, combining narrative AI, world models, continuous improvement from real classroom data, and explicit attention to student agency, offers a template for what responsible educational AI can look like, one in which the technology amplifies teachers and learners rather than replacing them.

Lester frames the ambition in expansive terms. The research under this award, he says, will produce transformative advances in understanding how people learn, and will lead to technologies that fundamentally improve both the learning process and learning outcomes for K-12 students. Whether AI-generated interactive narratives can deliver on that promise will depend on the coming five years of classroom evidence. But with a renewed $20 million commitment, a multi-university coalition, and a first funding cycle’s worth of tools and studies behind it, the EngageAI Institute now has the resources, the partnerships, and the runway to find out, and to shape how a generation of students experiences the intersection of story, technology, and learning.

Subject of Research: AI-generated interactive narrative technologies for K-12 learning

Article Title: NC State receives $20 million NSF grant to develop AI-enabled learning technologies

Article References: NC State receives $20 million NSF grant to develop AI-enabled learning technologies. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: artificial intelligence, education, K-12, interactive narratives, EngageAI Institute, National Science Foundation, NC State, world models, agency-centered learning, collaborative learning, learning technologies, classroom research