By BENEDICT CAREY
and JOHN MARKOFF
LOS ANGELES


IN A HANDFUL of laboratories around the world, computer scientists are developing highly programmed robots that can engage people and teach them simple skills.

Several countries have been testing teaching machines in classrooms. South Korea, known for its enthusiasm for technology, is “hiring” hundreds of robots as teacher aides and classroom playmates and is experimenting with robots that would teach English.

So far, the teaching has been very basic, and the robots are still works in progress. Yet the most advanced models are fully autonomous, guided by artificial intelligence software like motion tracking and speech recognition, which can make them just engaging enough to rival humans at some teaching tasks.

Researchers say the pace of innovation is such that these machines should begin to learn as they teach, becoming the sort of infinitely patient, highly informed instructors that would be effective in subjects like foreign language or in repetitive therapies used to treat developmental problems like autism.

Already, these advances have stirred dystopian visions, along with the sort of ethical debate usually confined to science fiction.

“I worry that if kids grow up being taught by robots and viewing technology as the instructor,” said Mitchel Resnick, head of the Lifelong Kindergarten group at the Media Laboratory at the Massachusetts Institute of Technology, “they will see it as the master.”

Most computer scientists say that they have neither the intention, nor the ability, to replace human teachers. The great hope for robots, said Patricia Kuhl, co-director of the Institute for Learning and Brain Sciences at the University of Washington, great hope for robots, said Patricia Kuhl, co-director of the Institute for Learning and Brain Sciences at the University of Washington, “is that with the right kind of technology at a critical period in a child’s development, they could supplement learning in the classroom.”


Lessons From RUBI

“Kenka,” says a childlike voice. “Ken-ka.” At the University of California, San Diego, in California, a robot named RUBI is teaching Finnish to a 3-year-old boy.

RUBI’s screen-torso is mounted on a pair of shoes, with a bandanna around its neck and a fixed happy-face smile.

It picks up a white sneaker and says kenka, the Finnish word for shoe, before returning it to the floor. “Feel it; I’m a kenka.”

In a video of this exchange, the boy picks up the sneaker, says, “Kenka, kenka” - and holds up the shoe .

The researchers are finding that RUBI enables preschool children to score significantly better on tests, compared with less interactive learning .

Preliminary results suggest that these students “do about as well as learning from a human teacher,” said Javier Movellan, director of the Machine Perception Laboratory at the university . “Social interaction is apparently a very important component of learning at this age.”

Like any new kid in class, RUBI took some time to find a niche. Children swarmed the robot when it first joined the classroom. But by the end of the day, a couple of boys had yanked off its arms. Engineers re-programmed RUBI to cry when its arms were pulled. Its young playmates quickly backed off at the sound.

After RUBI’s re-arming, researchers from the University of California, San Diego, the Massachusetts Institute of Technology and the University of Joensuu in Finland reported in a paper last year that the robot significantly improved the vocabulary of nine toddlers.

After testing the youngsters’ knowledge of 20 words and introducing them to the robot, the researchers left RUBI to operate on its own. The robot showed images on its screen and instructed children to associate them with words.

After 12 weeks, the children’s knowledge of the 10 words taught by RUBI increased significantly, while their knowledge of 10 control words did not. “The effect was relatively large, a reduction in errors of more than 25 percent,” the authors concluded.

Social robotics is a branch of computer science devoted to enhancing communication between humans and machines. At Honda Labs in Mountain View, California, researchers found a similar result with their robot, Asimo. In one 20-minute session the machine taught grade-school students how to set a table - improving their accuracy by about 25 percent .


Making the Connection

“Before they have language, infants pay attention to what I call informational hotspots,” where their mother or father is looking, said Andrew N. Meltzoff, a psychologist who is codirector of the University of Washington’s Institute for Learning and Brain Sciences.

This, he said, is how learning begins. This basic finding, to be published later this year, is one of dozens from a field called affective computing that is helping scientists discover exactly which features of a robot make it most convincingly “real.”

“It turns out that making a robot more closely resemble a human doesn’t get you better social interactions,” said Terrence J. Sejnowski, a neuroscientist at the University of California, San Diego. The more humanlike machines look, the more creepy they can seem.

The machine’s behavior is what matters, Dr. Sejnowski said. And very subtle elements can make a big difference.

The timing of a robot’s responses is critical. In recent experiments at a day care center in Japan, researchers have shown that having a robot simply bob or shake at the same rhythm a child is rocking or moving can quickly engage even very fearful children with autism.

“The child begins to notice something in that synchronous behavior and open up,” said Marek Michalowski of Carnegie Mellon University in Pennsylvania, who collaborated on the studies. Once that happens, he said, “you can piggyback social behaviors onto the interaction, like eye contact, turn taking, things these kids have trouble with.”

In a continuing study financed by the National Institutes of Health in Bethesda, Maryland, scientists at the University of Connecticut are conducting therapy sessions for children with autism using a French robot called Nao, a 61-centimeter humanoid. The robot, remotely controlled by a therapist, demonstrates martial arts kicks and chops and urges the child to follow suit; then it encourages the child to lead.

“I just love robots, and I know this is therapy, but I don’t know - I think it’s just fun,” said Sam, an 8-year-old with Asperger’s syndrome.

This simple mimicry seems to build a kind of trust, and increase sociability, said Anjana Bhat, an assistant professor in the department of education who is directing the experiment. “Social interactions are so dependent on whether someone is in sync with you,” Dr. Bhat said. “You walk fast, they walk fast; you go slowly, they go slowly - and soon you are interacting, and maybe you are learning.”

Personality matters, too, on both sides. Researchers have found that when the robot teacher Asimo is “cooperative” (“I am going to put the water glass here; do you think you can help me?”), children 4 to 6 did much better than when Asimo lectured them.

If robots are to be truly effective guides, in short, they will have to do what any good teacher does: learn from students when a lesson is taking hold and when it is falling flat.

Learning From Humans

“Do you have any questions, Simon?”

On a recent Monday afternoon, Crystal Chao, a graduate student in robotics at the Georgia Institute of Technology, in Atlanta, was teaching a 1.5-meter-tall robot named Simon to put away toys. She had given some instructions, but the robot was stumped.

Dr. Chao repeated her query, perhaps the most fundamental in all of education: Do you have any questions?

“Let me see,” said Simon, in a childlike machine voice, reaching to pick up a toy. “Can you tell me where this goes?”

“In the green bin,” came the answer.

Simon nodded, dropping it in that bin.

“Makes sense,” the robot said.

Just as humans can learn from machines, machines can learn from humans, said Andrea Thomaz, an assistant professor of interactive computing at Georgia Tech who directs the project.

The ability to monitor and learn from experience is the next great frontier for social robotics - and it probably depends on unraveling the secrets of how the human brain accumulates information during infancy. In San Diego, researchers are trying to develop a human-looking robot with sensors that approximate the complexity of a year-old infant’s abilities to feel, see and hear.

Researchers are aiming for nothing less than capturing the foundation of human learning - or, at least, its artificial intelligence equivalent. If robots can learn to learn, they can in principle make the kind of teachers that are responsive to the needs of a class, even an individual child.

Parents and educators would certainly have questions about robots’ effectiveness as teachers, as well as ethical concerns about potential harm they might do. But if social robots take off in the way other computing technologies have, parents may have more pointed ones: Does this robot really “get” my child? Is its teaching style right for my son’s needs, my daughter’s talents?

That is, the very questions they would ask about any teacher.


THE NEW YORK TIMES

Social robotics researchers seek to capture the foundation of human learning. RUBI the robot with toddlers. / ALAN DECKER/THE NEW YORK TIMES

Scientists are trying to find what makes a machine most “real.” A robot follows commands in Los Angeles. / KEVIN SCANLON FOR THE NEW YORK TIMES