Robots help kids tell stories—with a little help from stuffed animals
Banner image: Layne Hubbard's collection of stuffed animals for helping kids tell stories. (Credit: Layne Hubbard)
Layne Hubbard knows the power of a stuffed animal.
In 2016, the artist and aspiring neuroscientist paid an impromptu visit to a children’s thrift store. On one of the shelves, she spotted a stuffed zebra with pink feet and a pocket with a zipper on its back.
[video:https://youtu.be/ZYGy2ulj1Ms]
Hubbard thought it was perfect. At the time, she was a doctoral student at CU Boulder pursuing a triple PhD in computer science, cognitive science and neuroscience. She had just launched a research and design studio that seeks to turn ordinary toys like this one into robots that can help kids tell stories—with the help of human-computer interaction technologies. Now, she's recently joined forces with the Digital Learning Lab and PBSKids on an effort to develop artificial intelligence for the TV show “Elinor Wonders Why.” Soon, the inquisitive bunny at the center of the show will be able to ask young viewers questions and listen for their responses.
“I think of storytelling as an ancient technology,” said Hubbard, who finished her doctoral degree in 2021 and is now a computing innovation fellow at the University of California Irvine through funding by the National Science Foundation.
It’s an ancient technology, but maybe one that new advancements can take in fresh directions.
The MindScribe process begins when children, usually around 4 or 5 years old, make a piece of art, such as a drawing or a Lego tower. They then plop their favorite stuffed animal on top of a smartphone (or tuck the phone into a convenient pocket). An app asks the kids open-ended questions about their creation: “Tell me a story about what you made,” or “then what happened?” And the all-important “why?”
Through a series of recent studies, Hubbard and her colleagues at CU Boulder have put the technology to the test, revealing the promise and limitations of storytelling technology. Her team has found that . In one experiment, for example, a 4-year-old spent 24 minutes telling a stuffed tiger a story called “The Space Story From Fly Guy.” Hubbard believes that such activities may help kids to develop important cognitive skills at a critical age for learning—and maybe even work through difficult life experiences.
She said the results show what little kids are capable of when they can take charge of their own stories.
“I remember what it was like being a young child and having a lot of strong ideas about the world,” Hubbard said. “Kids may be small, but their ideas are mighty, and they deserve a place alongside adults’ ideas.”
Learning to tell stories
Hubbard has seen a lot of those mighty ideas firsthand.
Before she earned a bachelor’s degree in computer science from CU Boulder in 2015, Hubbard taught preschool in Boulder. There, she learned how to engage in a storytelling exercise with her young pupils. She’d sit down next to them and admire their latest masterpiece—a crayon drawing, say, of a hungry, hungry caterpillar.
“So I’d say, ‘Tell me a story about that caterpillar,’” Hubbard said. “They’d say, ‘Well, it was looking for a leaf.’ And I’d write that down to show them that their words matter.”
The exercise taps into the fundamental human need to tell stories. Stories, the scientist said, allow kids to stretch their imaginations and reflect on their unique ideas about the world. But those skills also need to be learned, said Eliana Colunga, an associate professor in the Department of Psychology and Neuroscience at CU Boulder.
“One of the things that make human beings so special in comparison to other animals is our ability to think about things that aren’t real and to make plans for the future,” said Colunga, who was one of Hubbard’s doctoral advisors. “That’s not something that a 2-year-old can do naturally.”
Hubbard, however, noted that too many kids in the U.S. don’t always have supportive adults in their lives to listen to their stories. When she was between 5- and 6-years-old, she and her siblings lived in foster homes in Michigan.
“My imagination as a child was one of the main ways that I protected myself and created space for myself when I didn’t always have external support,” Hubbard said. “That time period helps me remember that young children have big ideas of their own. But young children also have big challenges of their own.”
Artificial intelligence may be able to help, she added—under the right circumstances.
From emotions to alien planets
The technology underlying MindScribe is simple by design. It listens to kids as they tell their stories, waits for them to pause, then prompts them with those open-ended questions. When the children say “I’m done,” the robot asks them to wrap up their story by giving it a name. Hubbard’s team has also programmed their robot to speak 14 different languages.
The idea, she said, is to put kids and their stories first.
To test how it works, she and her colleagues ran a virtual experiment in 2020 at the height of the pandemic. The team recruited 33 young kids (and their favorite stuffed animals) from across the U.S. to make a piece of art, then talk about it.
Hubbard was amazed by the response. The children, who were 4- or 5-years-old, told their stuffies a diverse range of often sophisticated stories. Some described how they made their art (“Then I made my papa. And then I made my mommy.”). Others told imaginative, even interplanetary stories (“Then he jumped to Jupiter. And then Saturn.”). Some processed events from their lives (“I 'goed' to school, and my friends were not listening.”).
“There was a lot of conflict with teachers and things that happened at birthday parties,” Hubbard said. “Even experiences like funerals and deaths in the family. For educators and parents, it’s really helpful to know that those things are on their minds.”
Hubbard and her colleagues presented their first set of results virtually this summer at the ACM Conferences on and .
But the exercise also showed where AI technology might fall short. Most artificial voices on the market, Hubbard said, sound like adults. In other words, they’re boring. Many of her or even silly characters. One young child was confused about why his Pikachu stuffed animal talked like a parent and not in the squeaky voice of that yellow Pokémon.
A menagerie of robots
Throughout the project, Hubbard and her colleagues have also strived to keep the children who use their technology safe. Her storytelling app can’t connect to the internet, doesn’t access your phone’s data and doesn’t record what kids say.
“It is one thing for technology to respect privacy and quite another for people to feel confident that privacy is being respected,” said Clayton Lewis, another of Hubbard’s advisors and a fellow in the Institute for Cognitive Science at CU Boulder. “This is a tough challenge, and I expect that Layne will be among those who contribute to resolving it.”
MindScribe isn’t available for download yet, Hubbard said.
But as she heads to California for her fellowship, she’s hoping to partner with toy developers and others to continue designing new educational products. Today, her stuffed zebra also has some company: Hubbard keeps a shelf full of animal robots, including a prairie dog, owl and raccoon.
“There’s no denying that our human-human interactions will always be the most important,” Hubbard said. “But toys let us do different things. They allow us to get messy with our ideas.”