Play Live Radio
Next Up:
0:00 0:00
Available On Air Stations

CMU's RoboTutor To Teach Where Resources Are Sparse

Jennifer Szweda Jordan
90.5 WESA
Carnegie Mellon University researcher Jack Mostow observes a child using a program his RoboTutor team has developed that teaches addition and subtraction.

In many developing countries, teachers and classrooms are scarce.

In the poor rural village of Mugeta in Tanzania, Joash Gambarage grew up surrounded by children without access to education. While he’s moved on to graduate studies at the University of British Columbia, the picture in Mugeta hasn’t changed much. That’s why Gambarage founded a school that now has about 50 students.

“The kids were just staying home with their parents or with their grandparents selling eggs and stuff on the streets,” Gambarage said of his students.

It’s a school with few resources. The classroom is so small that one group of students has to wait outside while others learn.

Hoping to fill that need, two dozen Carnegie Mellon University researchers and coders and other helpers are working on another teaching option: RoboTutor, an Android tablet with software to teach literacy and arithmetic skills.

Team leader and professor Jack Mostow worked 25 years on a literacy teaching project that will be folded into RoboTutor’s work, which the group plans to send to Gambarage’s school through tablets outfitted with educational games. The longer term goal is to test and improve the software so it can be used across east Africa. 

“It’s to fill a tragic gap in terms of education in developing countries,” Mostow said. “Only half the kids have teachers, let alone quality instruction. The premise of the competition is that affordable technology can plug that gap.”

The team is employing what’s called intelligent tutor software that can adapt to individual students’ learning and change up questions or subjects when a child learns a skill.

“We can also aggregate that data and try to extract some lessons from it about how to refine the software,” Mostow said. “Both of those are things that teachers either can’t do or do in different ways. Good teachers learn, but no teacher will ever learn by teaching billions of kids, because no teacher will live long enough to teach billions of kids.”

To develop a software that can teach and interact with students without a teacher first requires a lot of education, psychology and coding experts carefully keeping an eye on how students interact with the software, Mostow said.

Before the group tests the software in Tanzania, it’s running trials at a preschool on the CMU campus. One activity has two lines of shapes to add or subtract. At one trial, some of the kids got it right away, but a few got confused.

“I didn’t even hear what he said,” said one child during RoboTutor's “kid testing.”

“Oh, he said let’s subtract some numbers,” Mostow told her.

“I don’t even know what that means,” she said.

“Well, this’ll explain,” Mostow said. “You know what? Let’s start with addition instead.”

Besides being closely observed by Mostow and colleagues, the session was also being recorded on the computer. Other members of the team will be able to follow along later and tweak the software.

The team has a number of hurdles to overcome. They have to teach kids not just in English, but also in Swahili.

Maria Robles, CMU graduate student in human-computer interaction, is working on finger writing on the tablet. It’s not easy, she said.

“We’ve come into challenges as to how to deal with punctuation,” Robles said. “For example, in Swahili, there’s things we don’t have in English. In Swahili, there’s apostrophes we don’t have in English.”

And Mostow said he’s been unable to find open-source Swahili literature to incorporate in the software that does not use archaic spelling and language — an equivalent of teaching medieval English to young children.

Mostow's team is also on a deadline. Next year, the group plans to enter RoboTutor into the Global Learning Xprize competition, which offers a purse of $10 million.

Despite the language barrier, Mostow said he's confident CMU's team will stand out among the other 200 teams competing. The university has been working on artificial tutoring for decades, he said.

“So we’ve got less than a year,” he said. “But this is just tailor made for Carnegie Mellon.”

In this week's Tech Report calendar: