Sesame Workshop, IBM partner to use Watson for preschoolers
You’ve heard of Tickle-Me Elmo. How about Super-Intelligent Auto-Tutor Elmo?
A new, three-year research partnership between Sesame Workshop and IBM’s Watson division holds the tantalizing possibility of such jaw-dropping educational tools, as early-childhood educators get their hands on what’s arguably one of the world’s most advanced supercomputers.
The partnership, announced Wednesday, focuses on how the Jeopardy-winning computer can use Big Data to help parents and teachers give preschoolers the skills they need to succeed in school.
The effort seeks to develop advanced digital tutors that can interact fully with children, assess their skills — often through a brief conversation — and provide both spoken and written-word responses that help improve their skills.
Previous research has shown that one-on-one tutoring is the most effective way to teach, but most children — especially low-income children — rarely get this opportunity. So researchers will spend the next several years trying to duplicate one-on-one tutoring digitally.
“Imagine a really plush Elmo that engages directly with a child, listens and responds,” said Harriet Green, who manages IBM’s Watson commerce and education efforts, among others. The toy uses the information, in real time, to create activities for the child such as practicing ABCs or counting to a favorite song.
As it interacts, she said, “the toy begins to develop, it begins to adapt to the developmental skills of the child." Once the toy figures out that the child has mastered these skills, it moves on to more advanced lessons.
Jeffrey Dunn, CEO of Sesame Workshop, said Sesame Street, which debuted in 1969, used the power of advanced technology — nearly 50 years ago it was television — to customize learning for young children. Now it’s super-computing. IBM, he said, brings “technological capability that we don’t have.” Sesame brings an engaging curriculum and educational expertise that IBM doesn’t have.
Though it's arguably the most ambitious, it's by no means the first effort to use technology for personalized learning.
For nearly a century, educators have been trying to figure out how to create so-called “teaching machines” to make classrooms more productive. In some cases, the goal was to replace teachers altogether.
The first models, developed as early as the 1920s, operated mechanically and used punch card templates and multiple-choice lessons. One device, originally developed in 1925, actually dispensed Life Savers when a student punched in a correct answer. By the early 1960s, when Ohio State University psychology professor Sidney Pressey began championing it, he said the candies had grown too large for the machine. It dispensed Tums instead.
L. Todd Rose, an informal adviser to the project who teaches at the Harvard Graduate School of Education, said personalized learning has historically had a bigger problem: it relied on an “aggregation approach” to students. So, for instance, after researchers looked at the behavior of large groups of students, educators embraced “learning styles” that weren’t always helpful.
“And then we’re really surprised when it doesn’t predict individual kids’ behavior,” he said.
But a computer like Watson can use more data and more finely tune it, allowing developers to build lessons that more accurately predict how individual learners will react. It can then “adapt in real-time, rather than based on preconceived assumptions.”
He added, “I think it fills a role that doesn’t exist right now, and I think it’ll be really, really valuable in that sense.”
Watson is also powerful enough to help build tools that allow students to be creators themselves, not just passive media recipients.
“It’s never going to replace a human being, but I think the ability to meet a kid where they’re at and keep building their literacy through conversation, rather than just interaction with a screen, I think that’s incredibly powerful.”
Follow Greg Toppo on Twitter: @gtoppo