Scientists are using new artificial intelligence to uncover the secrets of infant learning and development

AI's analysis of infant movements reveals important insights into early developmental stages, emphasizing the importance of foot movements in learning. Recent advances in computing and artificial intelligence, as well as new understandings of infant learning, suggest that machine and deep learning techniques can be used to study how babies transition from random exploratory actions to purposeful actions. To date, most studies have focused on spontaneous movements in infants and have made a distinction between irritable and non-irritable behaviors.

Using artificial intelligence, the researchers explored infant learning by analyzing movements in infant movement experiments and found that foot movements are key to understanding infants' interactions with their environment. AI models, especially 2D-CapsNet, effectively highlight the developmental stages of babies.

While early infantile movements may seem disorganized, they reveal meaningful patterns in infants' interactions with their environment. However, we still lack an understanding of how babies consciously interact with their surroundings and the guiding principles of their targeted actions.

The researchers explored how babies move purposefully by attaching a colourful mobile device to their feet and tracking their movements using a Vicon 3D motion capture system. Source: Florida Atlantic University

To explore how babies begin to act purposefully, researchers at Florida Atlantic University and their collaborators conducted an experiment with infant cell phones, a developmental research technique that has been in use since the late 60s of the 20th century. In this experiment, a colorful mobile phone is gently tethered to the baby's foot, and when the baby kicks the foot, the phone moves, linking the baby's behavior to what they see. This setup helps researchers understand how babies control their movements and discover their ability to influence their surroundings.

In the study, the researchers tested whether AI tools could capture complex changes in infant movement patterns. Infant movements tracked with the Vicon 3D motion capture system are divided into different types – from spontaneous movements to reactions when moving. By applying a variety of AI techniques, the researchers studied which method best captures the subtle behavior of babies in different situations and how movements evolve over time.

The researchers explored how babies move purposefully by attaching a colourful mobile device to their feet and tracking their movements using a Vicon 3D motion capture system. Source: Florida Atlantic University

The findings, published in Scientific Reports, highlight that AI is an important tool for understanding early infant development and interactions. Both machine learning and deep learning methods can accurately categorize a five-second 3D clip of baby motion into different stages of the experiment. Of these methods, the deep learning model 2D-CapsNet performs best. Importantly, the foot movement was the most accurate of all the test methods, meaning that the movement patterns of the feet changed most significantly at all stages of the experiment compared to other parts of the body.

"This discovery is significant because the AI system was not told anything about the experiment and did not know which part of the baby's body was connected to the phone. "This suggests that the feet—as end effectors—are most affected by interactions with cell phones," said Glenwood Creech and Martha Creech Distinguished Science Scholar Dr. Scott Kelso of the Center for Complex Systems and Brain Science at Florida Atlantic University, co-author of the study. In other words, the way babies connect with their environment has the greatest impact on their touchpoints with the world. Here, it's 'feet first'. "

The 2D-CapsNet model is 86% accurate in analyzing foot movements and captures detailed relationships between different body parts during movement. Foot movements consistently had the highest accuracy rate of all test methods, about 20% higher than hand, knee, or full-body exercises.

"We found that babies explored more after disconnecting from their phones than they did before they had the opportunity to control them. It seems that losing the ability to control their phones makes them more eager to interact with the world in order to find ways to reconnect," said co-author Aliza Sloan, Ph.D., a postdoctoral research scientist at Florida Atlantic University's Center for Complex Systems and Brain Science. "However, the movement patterns exhibited by some babies during the disconnection phase contained cues of their previous interactions with the phone. This suggests that only certain babies are able to understand their relationship with their phones well enough to maintain these movement patterns, expecting that they will still be able to react from their phones even after disconnecting. "

If the accuracy of the baby's movements during disconnection remains high, it could be a sign that the baby learned something from a previous interaction, the researchers said. Still, different types of movements can mean that babies discover different things.

Co-author Dr. Nancy Aaron Jones, a member of the Center for Brain Science, said: "It's important to note that studying babies is more challenging than studying adults because babies are unable to communicate verbally. Adults can follow instructions and explain their actions, while babies can't. This is where AI can help. AI can help researchers analyze subtle changes in a baby's movements, even their static state, allowing us to understand how they think and learn, even before they can speak. Their movements also help us understand the vast individual differences that occur as babies grow. "

Observing how each baby's AI classification accuracy changes provides researchers with a new way to understand when and how babies begin to engage with the world.

"Past AI approaches have focused primarily on classifying spontaneous movements related to clinical outcomes, and combining theory-based experiments with AI will help us better assess infant behavior in relation to infants' specific environments," Kelso said. "This can improve the way we identify risks, diagnose and treat diseases."

编译自/SciTechDaily