Leap and learn: AI‑PLAY to turn movement into AI lessons

What if elementary school students could learn the fundamentals of artificial intelligence and large language models just by moving their bodies?  

As artificial intelligence is increasingly shaping how people learn, work, communicate and make decisions, University of Florida researchers are asking how best to help young learners develop a foundational understanding of how AI works and how it affects society.  

One of the answers may be what Kristy Boyer, Ph.D., calls “embodied movement.”  

The professor in UF’s Department of Computer & Information Science & Engineering, or CISE, is part of a recently funded National Science Foundation project called AI-PLAY. The project is an outgrowth of UF’s AI-Powered Athletics Initiative, which was originally funded through a presidential strategic initiative.  

AI-Play aims to connect AI concepts to learners' own movements and experiences, making abstract ideas more concrete while broadening participation in AI education at an early age.   

Boyer is joined on the project by Jennifer Nichols, Ph.D., associate professor in UF’s J. Crayton Pruitt Family Department of Biomedical Engineering, and Anthony Botelho, Ph.D., assistant professor in UF’s College of Education. The project represents a deep partnership between HWCOE and the University Athletics Association, with Spencer Thomas of the UAA serving as senior advisor.  

The AI-PLAY team will design, iteratively refine and study an after-school program, based in Alachua County public schools, that teaches fourth- and fifth-graders three foundational AI  concepts: how computers perceive the world through sensors, how AI systems learn from data and how AI can impact society.  

But how does movement teach kids about the inner workings of AI?  

First, the data. 

“The idea of helping kids create data through movement is that we use low-cost sensors and have the kids move around,” Boyer said. “It generates thousands of data points, and they know where it came from because they just watched it get generated by their own body.” 

Data creation and manipulation become personal, and even fun, in this context. Students observe how the data changes, depending on whether participants run, walk, jump or squat down. As different kinds of sensors are deployed, students see for themselves how the data change. 

For a wider variety of data, the activities call for additional sensors, including grip sensors, jumping pads or full-body cameras. 

Once students are comfortable with the creation of datasets, some of the higher-level concepts foundational to AI are introduced — in a fun way. 

“We get to grapple with all kinds of super interesting-but-hard-to-bring-home-for-kids things,” Boyer said. “Like, what if only the taller kids in the class fed their numbers into the AI. Is that going to be a very good model for the shorter kids in the class? And now we’ve introduced the idea of bias in training sets and systematic error. It’s hard enough to get the idea across to undergrads, but we’re hoping the fourth and fifth graders will get their minds around it by using these personally relatable data sets.” 

After the data is collected, the team plans to introduce students to the concept of self-supervised learning, widely used in today’s state-of-the-art AI. When students understand this core concept, researchers believe, the young learners will be well positioned to be AI-informed citizens. 

The members of the multidisciplinary team are focusing their expertise on how to best educate upper elementary students in the inner workings of AI.  

Boyer has completed numerous NSF-funded projects on upper elementary and middle school AI education and is also a recognized scholar in human-computer interaction and sports.  

Nichols’ expertise lies in biomechanics.  

“As a biomechanist,” she said, “my key contribution will be taking big problems — like how to analyze data from elite athletes while preserving their privacy or how measurements of human movement can be used to discriminate healthy versus injured states — and making them accessible for elementary school learners. Kids have a tremendous capacity to ask big questions and think big thoughts. Guiding them through how to think about movement and AI is going to be fun.”  

Botelho is an expert in the integration and instruction of artificial intelligence in K-12 settings. He leads research initiatives to improve instructional practices by leveraging integrated quantitative methodologies through the development of educational technologies.  

Researchers hope the project’s association with the UAA will motivate participants, apart from all the fun and movement. Some of the same sensors and testing methodologies used by Thomas's UAA staff will form the backbone of AI-PLAY.  

“We really believe that when the kids know that they are doing the same kind of data creation and tracking as collegiate athletes, and even professional athletes, there’s another layer of excitement, engagement and real-world relevance,” Boyer noted. 

For now, the team is looking forward to the project kickoff meeting, set for the end of September. Enthusiastic teacher partners from Alachua County have been identified. In the coming weeks, project staff will design the combination of hardware and software that will move the sensor data into the actual coding environment that students will use. The after-school program itself begins in January.