This robot might look like a Furby. But, using artificial intelligence, it can learn to understand a child’s emotional state to better engage them during a lesson.
This Cute Robot Uses Sophisticated AI to Help Teach Kids
Published: June 6, 2019
Tega: Hi, My name is Tega. It’s nice to meet you
Onscreen: This robot is learning to read human emotions to help teach kids.
Tega: I’ll tell you two stories and you tell me your version of those two stories.
Tega: (reading aloud) “Was it you?” He said, spying Possum.
Sophia: I like how his eyes sort of change. I’d rather read with the robot. I like how he’s kind of acting it out.
Onscreen: Tega does more than read stories. It uses artificial intelligence (AI) to read non-verbal cues, like a smile and adapt its response over time
Rana El Kaliouby: It’s not enough for our AI for have IQ.
I’m a big believer that our AI systems need to have EQ as well.
Onscreen: E = EMOTIONAL
Tega: Possum emerged from his home…
Rana El Kaliouby: They need to understand humans.
Tega: What is the of the meaning of emerge? … Oh come on, don’t be shy. Tell me what you think.
Sophia: To come in? Or go out?
Tega: Uh huh .. emerge means come out, or come into view
Hae Won Park: Tega is using the camera on his forehead to assess the child’s facial expressions. Tega can detect if the child is looking at the tablet where the storybook is shown, or at the robot and from that it can detect the engagement of the child.
Hae Won Park: If the child was disengaged, it’s probably something the robot did wasn’t engaging enough for the child so the robot will learn to try something different.
Tega: Clifford can see the cat. Can you?
Tega: You can too!!
Rana El Kaliouby: These non-verbal signals can act as early predictors for learning outcomes. If a child is consistently not paying attention … and you can interject early on, then you are increasing the likelihood for future success
Onscreen: Tega is a product of a collaboration between MIT and emotion AI company Affectiva
Rana El Kaliouby: We build artificial emotional intelligence -- basically algorithms that understand all things human. We are the social and emotional brain of [Tega].
Rana El Kaliouby: We use machine learning and deep learning to train algorithms to understand if people are happy or surprised or confused or tired
Onscreen: Deep learning A.I. works by letting tega study lots of examples.
Rana El Kaliouby: We give the algorithm hundreds of thousands of examples of people smiling. Hundreds of thousands of examples of people smirking.
And then the algorithm determines the difference between these two states so the next time we show it an example of a face, the algorithm kicks in and says, “aha I see a smile!”
Or, “I see a smirk!”
Onscreen: Tega uses those distinctions to inform it’s decision about how to engage the child
Rana El Kaliouby: If it truly understands these different states, it can adapt in real-time to make for a much more engaging, and much more effective, learning experience.
Sophia: (reading aloud) He came out to see what all the racket was...
Onscreen: Developers say Tega isn’t meant to replace interaction with humans, but make time spent with learning tools more effective
Rana El Kaliouby: The relationship with these social robots, especially for kids is more like a peer-to-peer kind of relationship. It’s more like your friend, or your companion, less so as an adult figure.
Onscreen: Researchers surveyed children who interacted with Tega and asked where they’d place the robot among their existing relationships.
Hae Won Park: They will actually acknowledge a robot being very different from the relationships they currently have. That the robot is not a person, but a very different thing. It’s not a pet, it’s not a teacher
Onscreen: Tega is a prototype. Before it’s ready to market widely, developers plan to further test Tega in classrooms.
Digital Producer: Emily Zendt
Production Assistance: Greg Kestin and Rachel Swansburg
© WGBH Educational Foundation 2019