BLUE, a fluffy robot with cartoon eyes, is telling a little girl a story about a snowman. The tale plays out in pictures on a tablet that sits between the preschooler and Blue鈥檚 soft plastic feet. Blue鈥檚 eyes look down at the characters as it describes them, then up at the child to check that she鈥檚 paying attention. The 5-year-old is one of the first children to learn language skills from a robot, and she is captivated.
The girl is part of an eight-week experiment by MIT鈥檚 Media Lab now taking place in a preschool in the Boston area. The idea is to see how well children learn from robots 鈥� it is one of a handful of similar experiments being run by graduate student .
Kory and colleagues are testing children to see how well they remember new words learned from robots over time. As the weeks go by, Blue and other robots like it, dubbed , will adapt their stories as a child learns new words, keeping pace with individual development. Kory and the team leader, , will be logging everything and tracking how well they learn.
Advertisement
The team believe robots represent a powerful new way to enhance children鈥檚 education. Unlike educational TV shows, say, the robots are physically present and have some of the same social skills as humans. That gives them the potential to tap into a child鈥檚 appetite for one-to-one communication and help kids learn in many of the same ways a human teacher does. This is especially important when it comes to language skills.
Through experiments conducted earlier this year, Kory has already shown that the robots鈥� social abilities make a huge difference to how well children engage with what the robot says. In one experiment, children talked to two different robots about toy animals. One robot鈥檚 reactions and movements were programmed to be similar to a child鈥檚 own, while the other鈥檚 were more random. Children seemed far more willing to trust the robot that mimicked their body language, learning better from it.
The team鈥檚 work is still in the early stages. When the robot prompts a child to tell it a story, for example, the goal is for the bot to listen and gauge how well their vocabulary is developing. But speech recognition for preschoolers is still an emerging area, hampered by a lack of data on which to train machine learning software.
The DragonBots鈥� hardware will also need upgrading. Right now they only really work in a controlled research environment, but the next generation could find a place in homes and schools.
Kory and Breazeal see a future where young children鈥檚 learning can be supercharged by socially intelligent robots that fit smoothly into their lives. Children already bond strongly with their stuffed animals, for example, so what if those toys had the ability to talk back? Previous work at the Media Lab has even used the DragonBots to teach preschoolers little snippets of French, centred on role play where robot and child share a meal together.
鈥淚 love the idea of delightfully designed, childlike robots scooting around and befriending children. I think there鈥檚 a lot to say for this kind of effort,鈥� says Michael Levine, who directs research on educational technology at the in New York, the non-profit behind the popular children鈥檚 television show Sesame Street.
Levine says Breazeal鈥檚 robots overcome one of the fundamental problems with developing educational content for television and tablets 鈥� individualisation. Even when testing the most popular episode of a kids鈥� TV show, there will be a child in the room who doesn鈥檛 like it. Robots can pick up on social cues and offer a solution, letting a child learn at their own pace, he says.
聯Robots that can pick up and learn from children鈥檚 social cues allow a child to learn at their own pace聰
Breazeal has grand plans for the socially aware robots. 鈥淚n the US there鈥檚 a big issue with children entering kindergarten not ready to learn. And if you don鈥檛 enter kindergarten ready, you never really catch up,鈥� she says. 鈥淭here is a tremendous opportunity to develop new technology to support children鈥檚 education, especially early childhood learning.鈥�
This article appeared in print under the headline 鈥淜indergarten bots鈥�
