Plush Neurons Makes AI Friendly and Simplifies Neural Networks for Middle School Students

A team at Carnegie Mellon University is using soft, fluffy, LED-lit neural networks to help kids understand artificial intelligence.

“Just as it’s important to know the basics of how electricity works and what molecules are, everyone, even middle school students, needs to know a little about the building blocks of artificial intelligence,” said Dave Touretsky, research professor in the Department of Computer Science. “But these young learners haven’t even studied algebra yet. So how can we help them understand the computational power of these complex networks?”

Mr. Touretsky has been researching K-12 AI education for many years, investigating effective ways to teach young learners about AI technology at a level they can understand. Now, thanks to the facilities and expertise included in the Integrative Design, Arts and Technology (IDeATe) network, he and a team of staff, faculty, and students have brought Plush Neuron to life. These 3-foot brightly colored interactive computing devices allow middle school students to directly experience and influence how AI makes decisions.

Idea generation

Plush Neuron was designed to help students visualize the fundamental elements of neural network technology, the machine learning models that underpin much of modern AI. It is a physical counterpart to Neuron Sandbox, a browser-based tool created by Touretzky that allows students to work through a series of increasingly difficult decision-making problems and visualize how a computer arrives at a solution.

Touretzky programmed the neuron’s software, and IDeATe technical specialist Cody Soska developed its electronic hardware and served as the project’s design engineer as part of his graduate studies.

“This project is a true example of stars coming together to create something truly amazing to help AI education. It’s interdisciplinary at every level,” Soska said. Between his electronics and Touretsky’s code, the Plush Neuron framework quickly emerged, ready for the next production step.

Creation of neurons

Zarmond Goodman, a senior technical advisor at IDeATe and a fourth-year student in the Dietrich College of Humanities and Social Sciences, joined the team after a working prototype was completed. As a film and visual media specialist majoring in physical computing and art, Goodman’s role was to replicate Soska’s design and assemble the pieces into 10 individual neurons.

“Being able to work on projects like this is the epitome of what makes IDeATe cool,” they said. “Since getting involved with IDeATe during my first year at CMU, I have been able to combine my interests in different ways while working with people I have admired.”

Goodman followed Soska through the design and took over soldering the components. They also helped in other ways, such as filming and editing a demo video introducing Neuron.

“Hundreds of hours were spent building these neurons, and Zarmond was essential to completing the project,” Soska said. “Thanks to their work, everything is possible.”

Next, Soska hired Professor Olivia Robinson, who leads IDeATe’s soft technology minor, and Natalia Pinchuk, a soft technology lecturer. The team needed expertise in textiles to create Neuron’s final form: bright, tactile, and appealing to children.

The plush body of neurons had to accomplish three goals. First, it needed to be plush to pad the wires and other electronic components and feel soft when handled. At the same time, surface-embedded knobs and displays needed to maintain their position within the neuron to keep the electronic components stable. Finally, it had to resist the tendency to sag and maintain its shape when lifted into the air.

“One of our biggest challenges was designing a streamlined form that could hold the equivalent of a large shoebox of electronics,” Robinson recalls. “Underneath its uniform, bright exterior, the plush body is made with layers of foam and custom ‘shapewear’ that we create to keep everything in place and give shape to the outer material.” I still find satisfaction in putting it together and reflecting on all the inventive choices that came up along the way. ”

Pinchuk agreed.

“The vision was to provide children with learning tools that are tactile, interactive and fun. Textiles are uniquely suited to achieve that, providing sensory experiences that encourage contact and engagement,” Pinchuk added. “The stuffed animal form taps into the sense of play, comfort, safety, and the quiet magic of childhood, and has properties that make learning experiences more approachable, memorable, and emotionally resonant.”

active neurons

The finished Plush Neuron is a simplified version of how real neural networks work, designed for little hands and big questions.

This neuron has three inputs called “dendrites” at one end, which can be pressed to send a weighted input signal (adjusted by a rotary switch). The Dendrite’s LED display shows the current weight value ranging from -4 (lit red LED) to +4 (lit green LED).

The inputs are then sent to the body, or “soma,” where they are summed and compared to a threshold displayed on another LED array. If the input is greater than a threshold, a long axon at the other end flashes and makes a sound, letting the user know that the neuron is “firing.”

Weights and thresholds can be adjusted to pose a variety of simple decision problems. “Neuron explains ethics in AI by showing the importance of weighting decisions,” Soska explained. “It allows you to create and solve full-fledged logic problems using the same computational logic that AI uses: how it makes decisions and how biases are introduced.”

From symposiums to middle school classrooms, Plush Neuron has already piqued the curiosity of educators across the country. The project debuted this March at EAAI-25, the 15th Symposium on Educational Advancements in Artificial Intelligence. In June, the team brought the machine to Atlanta for the AI4MiddleSchools Cohort Curriculum Workshop, where it was used to train teachers for the first time. And in August, Soska presented at the International Symposium on Academic Makerspaces (ISAM) at the University of California, Berkeley.

This fall, the team sent the completed neurons to Christina Gardner McCune, a University of Florida professor who leads the AI4MiddleSchools project. Amber Jones, computer science education consultant and former middle school teacher in Atlanta. Will Hanna is a computer science teacher at Thomas County Middle School in Georgia. These early partners, who have previously championed AI education with Touretsky as part of the Artificial Intelligence of Georgia (AI4GA) project, will be the first to test Plush Neuron in the field.