Children's eyes may hold secrets about how they learn from videos, and new research suggests that artificial intelligence can unleash it. By combining gaze tracking techniques with advanced neural networks, researchers are beginning to identify the exact moments when younger audiences grasp new concepts, or when those ideas escape.
The findings point to a future where video lessons are tailored in real time to suit the needs of learners and change how science is taught in classrooms, at home and more.
Young eyes working tracking
Researchers at Ohio State University conducted a large-scale experiment involving 197 children between the ages of four and eight. Each child watched a four-minute video sewn in from YouTube shows “Scishow Kids” and “Learn Bright.” This lesson focuses on animal camouflage. This is a concept that combines visual cues with scientific explanations.
Before looking, children were asked to measure baseline knowledge. They then answered similar questions to test what they had learned. During the lesson, a highly accurate eye tracker recorded where and how long each child saw the screen. Lead author Jason Coronell, an associate professor of communication, explained why the steps are important. “Eyetracking allowed us to measure our attention to video in real time, which is important for learning,” he said.
The team then fed this detailed data to two different artificial neural networks. One followed a standard approach, while the other included a theory-driven design that explained how new information interacts with older materials over time. The theory-based model has proven to be more accurate, especially when predicting which children will answer questions about camouflage correctly.
Important learning moments have been revealed
AI analysis revealed that certain parts of the video had a major impact on whether children understood the lesson. Seven “significant moments” stood out, each linked to a prominent change in eye movement and related to understanding. One of the most powerful predictors came early when the host asked viewers to help them find a comics companion.
Children following the queue with focused attention were more likely to grasp more difficult concepts later in the video. As the research notes, “Our machine learning and gaze tracking data show that child's eye movements at this early moment are one of the most powerful predictors of overall understanding of video.”
Related Stories
Another powerful moment occurred when the narrator explicitly defined the camouflage and combined the description with the printed words. These transitions, known as “event boundaries,” mark a point when one meaningful segment ends and another begins. According to co-author Alex Bonus, who specializes in children's media, seven key points coincided almost perfectly with these natural boundaries.
Why timing is important for learning
Learning is not just about receiving information. It's about how the brain organizes it over time. Coronel and his colleagues framed the work of machine learning around this idea of temporal interdependence. They argue that a seamless connection between past and new information improves understanding.
This theoretical guided approach allows AI models to detect more than surface-level viewing patterns. Instead, they emphasize how changes in attention at key moments shape later understanding. The results suggest the possibility of children working and designing videos with well-placed boundaries ready for more complex ideas.
Still, Coronel emphasized that the work is preliminary. “However, this approach can help experts design their messages at the boundaries of events that enhance learning,” he said.
Get a glimpse into personalized video education
This study, published in the Journal of Communication, is because it is more accessible to both eye-opening and artificial intelligence. Once booked in high-end labs, eye-tracking hardware is now cheaper and easier to use in classrooms and at home.
At the same time, machine learning models are more refined when processing streams of data, such as gaze patterns. Together, these advancements could change the way children learn from videos. It can take teachers days or weeks to realize that students are misunderstanding the lesson now. By then, the class may have moved ahead and left a knowledge gap.
Coronel expects a different pass. “Our ultimate goal is to build an AI system that allows viewers to know in real time whether or not they understand what they are seeing in educational videos.
This can mean that the video automatically provides a second example, changes the pace, or switches to a new explanatory style at the moment the learner is shaking. This adaptability can be more personal, efficient and scalable in a variety of classrooms and learning environments.
The road ahead
Much remains unknown about the exact cognitive processes behind these important video moments. Researchers need to learn why attention is surged at certain times and how that attention leads to stronger memory or problem-solving skills. The Ohio team plans to expand this work and combine more advanced models with longer and more diverse lessons.
For now, this study offers a glimpse into the next stage of educational technology. At this stage, real-time signals from the body guide the way information is provided. With nearly 200 children showing a consistent pattern in the short video, the results suggest that the eyes of younger viewers may reveal whether learning is actually a place.
As Coronel said, “Imagine a future in which eye tracking can instantly convey when a person doesn't understand a concept in a video lesson.
