Just in time for the lucrative holiday shopping season, Mattel has released the latest and most technologically sophisticated edition of its star character — Hello Barbie. Those who have long dismissed the doll as a ditzy airhead will need to re-evaluate. Press a button on her belt, ask her a question and hear what she has to say.
"Barbie, what’s your real name?"
The doll answers back. "Oh! I thought you knew. My full name is Barbara Millicent Roberts, but you can call me Barbie."
Who’s the airhead now?
Hello Barbie is the newest member of an increasingly advanced set of interactive toys that use speech recognition software and natural language processing more typically found in automated personal assistants like Apple’s Siri and Microsoft’s Cortana to mimic real conversations, play games, or teach life lessons and learning strategies. Hello Barbie isn’t billed as an educational toy, of course. But she — or the technology inside her, to be exact — could represent a turning point in the category.
For years, child psychologists have recommended limiting children’s exposure to digital media — apps; videos; and beeping, flashing ebooks often labeled "educational," despite piles of research showing they aren’t. For children to learn, they must be engaged by a live partner, whether a teacher, a parent or even another toddler. It’s the give-and-take, the nuance of human exchange that guides a child toward a firmer grasp of language, they say.
Of course, even back in 1985, Teddy Ruxpin could talk, or at least act like it (he was really just playing prerecorded cassette tapes). But Hello Barbie — and a rash of new talking dolls hitting the market — may be altogether different. With the technology in place to make dolls and other toys truly responsive, there is the potential for a revolution in educational toys. Call it the post-Baby Einstein era, a time when "educational" really does equal learning.
"In the early years, verbalization and oral language are of primary importance in preparing children to learn to read," said Lisa Guernsey, director of the Early Education Initiative at New America. "So toys that encourage talking, especially with challenging vocabulary and interesting narratives, are worth looking at carefully."
Hello Barbie’s vocabulary and storytelling skills are impressive. She can hold a lifelike conversation, and she’s got a lot — a lot — to say. Curious parents can read all 8,000 lines of dialogue written for Hello Barbie on Mattel’s website (the company declined to comment for this story). Altogether, that’s over 83,000 words, longer than many novels. It’s the sum total of everything that can possibly come out of the doll’s mouth.
Hold down a button on Hello Barbie’s belt buckle, and a microphone inside the doll begins recording ambient audio. Tell her something about yourself, and that data gets transmitted through a home Wi-Fi connection to remote servers. Speech-recognition software determines what was said and algorithms process the audio information. Computers search for an appropriate response from the massive list of prewritten and vetted options and send the selection back to the doll’s location so it can reply, all in the space of a second or two — no small feat. "It’s part a voice acting challenge, it's part a writing challenge, it's part a runtime execution of cloud services challenge," said Oren Jacob, formerly the chief technology officer at Pixar and now the CEO and founder of ToyTalk, the software company responsible for giving Barbie a voice and writing the algorithms that power her witty repartee.
But even Barbie’s vocabulary pales in comparison to CogniToys, a series of adorable plastic dinosaurs whose responses are powered by the IBM-developed supercomputer Watson (the one that plays Jeopardy! and wins). Like Hello Barbie, the CogniToys dolls listen, transmit the input to servers via a wifi connection, then—hopefully—spit back an appropriate response as determined by a computer.
Unlike Barbie, the CogniToys dolls leverage Watson’s vast resources to formulate an original reply. Barbie is selecting from a predetermined list of responses, but without careful programming, CogniToys, and a number of other toys on their way to market, could say nearly anything.
Naturally, this has exciting potential, and terrifying pitfalls, for the educational kids market.
The closer a toy can mimic real conversation, the better, according to Sarah R. Lytle, director of outreach and education at the University of Washington’s Institute for Learning & Brain Sciences. "You want that interactivity to approximate a live human-to-human interaction as much as possible," she said. Even very young children notice when something is off.
"They're very quick to recognize that if something responds but it has a very standard response and isn't adaptive and contingent to the child’s actions, they seem to know very quickly that, ‘Oh, it doesn't really matter what I do,’ because the toy or the media is going to respond in this pre-scripted way," she said. "Kids don't tend to learn quite as much in that scenario."
And learning is increasingly good business. One study estimated the size of the market at $1.5 billion in 2012, projecting growth to $2.3 billion by 2017. While those aren’t all children’s games, technology developed for toys will inevitably be utilized in games, simulations and applications for older children and adults.
But simply slapping an "educational" sticker on a product without the goods to back it up can end up being costly. In 2009, Disney offered refunds to parents who had bought Baby Einstein DVDs in the hopes of giving their children a head start in subjects like reading, writing, music and art, after studies showed baby videos didn’t help develop language skills, and could even be detrimental.
Jacob is careful to note that his company doesn’t make those claims. "ToyTalk is an entertainment company. That's where we plant ourselves," he said. That’s not to say the company is opposed to the idea of Hello Barbie (or at least her technology) as teacher. In 2014, ToyTalk began a two-year partnership with Sesame Workshop, the producer of "Sesame Street," to investigate ways to use their technology to teach reading skills to preschoolers.
"Speech recognition for children is really, really hard," Jacob said. "To be in a place where you could, for example, teach spelling, teach reading, teach spoken foreign language using our technology stack is close, but not here in 2015.
But before you can be educational, you have to be fun, he notes. "It's only by building products that are meant for fun that you can get enough practice at your craft, data and algorithmic sophistication to meet the demands of the educational market."
The maker of CogniToys, on the other hand, is explicit about its educational mission. In March of this year, Elemental Path, which is based in Manhattan, successfully raised $275,000 on Kickstarter to fund the toy’s development, billing it as "educational" and a "true companion" for kids.
"Built into the play experience are a number of custom modules that engage the child in educational play; these including rhyming, spelling, vocabulary, mathematics and much more. As the interaction increase [sic] so will the challenge of the educational content continuing to become more challenging as the child learns," boasts the CogniToys website.
Whether it, or any other toy company can deliver on that promise, remains to be seen. Interactive toys and screen-based learning "can play an important positive role when they are used not as a babysitter, but in joint engagement with adults as a jumping off point for further conversation," said Michael H. Levine, co-author with Guernsey of the book "Tap, Click, Read," which examines digital media literacy in young children. "Toys alone — even those that are talking — cannot fulfill this basic function — at least not yet."
Ultimately, any app or toy, no matter how advanced, is just a tool — it’s only as effective as the person using it. "Technology can assist in building a literate citizenry," Guernsey said, "but high-powered human interactions are still the main show."