In a recent morning at Sparkletots preschool in Singapore, Natalie, Bryan and Mikayle, all four years old, knelt on the floor around a machine called Kibo and “programmed” it with a set of instructions printed on wooden blocks. Each of the blocks was printed with a command — “forward”, “backward”, “shake” — written in English and as a barcode that the robot could understand.
The children started the sequence. Natalie clapped and the others shrieked with delight as the machine wriggled. It had done just what it was programmed to do; hearing the child’s clap triggered its movement. “Again, again,” Natalie and her classmates squealed.
In Singapore, admired globally for its education system, authorities are trialling the use of robotic aides to teachers in kindergartens. Two humanoid robots, Pepper and Nao, assisted teachers in a seven-month trial at two Singapore preschools last year, while technology-enabled toys such as Kibo were deployed at 160 nurseries, including Sparkletots.
The city-state’s policy makers regard the androids as an adjunct to teaching, while the goal of the tech toys is to encourage children to be more creative with technology. Foo Hui Hui, an official with a Singapore government agency given the task of exploring the possibilities of digital technology, said last year: “We imagine a future not too far off, where interactive robots with the ability to perform multiple human tasks and provide visualisation of complex ideas can help children to learn and collaborate better.”
After decades in which robots have become commonplace on the factory floor, they are entering our society as comforters, playmates and, perhaps, teachers. Paro, the robotic baby seal first rolled out in Japan in 2003, has been adopted in care homes around the world. Gasparzinho, a stubby robot with a T-shaped head, plays games of tag with sick children in the paediatric wing of a Lisbon cancer hospital.
But it is education, and, in particular, teaching very young children, where some researchers and policy makers are focusing their energy. The early years are a period of intense learning and rapid brain development. Skills that will matter throughout life are developed at this point — from the acquisition of confidence and self-esteem to social skills such as learning to play well, sharing, taking turns and dealing with conflict.
Advocates of the trials raise hopes that the robots will, ironically, make us more human. In a statement, the Singapore government agency behind the pilots said that these robots “encourage collaboration and social interaction among young children”, citing research from the US and Israel.
But in a world in which people increasingly fret about the amount of time that we, and our children, spend swiping at screens, the trials also raise questions about the changing dynamics of our society. Some fear that “pretend relationships” with automata may come to be an acceptable substitute for human interaction.
The pioneer of efforts to combine robotics with education was Seymour Papert, a mathematician and computer scientist who worked with Swiss philosopher and child psychologist Jean Piaget at Geneva University before moving to the Massachusetts Institute of Technology in 1963. There, he developed a programming language for children and helped set up the MIT Media Lab.
Papert drew a distinction between the use of computers in education to simply instruct a child, and their potential to be an empowering, creative tool. “In many schools today, the phrase ‘computer-aided instruction’ means making the computer teach the child,” he wrote in his influential 1980 book Mindstorms: Children, Computers and Powerful Ideas. “One might say the computer is being used to programme the child. In my vision, the child programmes the computer and, in doing so, both acquires a sense of mastery over a piece of the most modern and powerful technology and establishes an intimate contact with some of the deepest ideas from science, from mathematics, and from the art of intellectual model building.”
Papert adapted an existing robotic “turtle” to follow instructions from a computer. This floor-level machine was equipped with a built-in pen allowing it to draw lines on a sheet of paper, building patterns from a series of simple routines. The patterns let children visualise the power of a computer program to generate complexity from a few lines of code.
Marina Bers, who studied with Papert at MIT and is now a professor at Tufts University in Massachusetts, designed the Kibo robot being trialled in Singapore preschools. She argues that the Singapore approach illustrates a deeper engagement with technology than western school systems, which emphasise computer science teaching.
“If you think of the world, we are surrounded by smart objects. The elevator door closes and opens because it has a sensor,” says Bers. “What Singapore has the vision to do is engage children in understanding how these objects are made. In the US, there’s a big movement for computer science for all — but computer science is not the same as understanding how these objects are made.”
The children using Kibo are encouraged to see the machines as devices that they can control and develop, to think about what makes them tick. “These children will be able to be makers of a new generation of robots,” Bers adds.
When Singapore first introduced the androids Pepper and Nao to the classroom last year, teachers noted that the machines often had gratifying results with normally timid children. Shyer boys and girls were more willing to put their hands up or volunteer to come to the front of the class if it meant they were able to interact with Pepper.
In the seven-month pilot study, in which the robots were loaned for free by their Japanese maker SoftBank Robotics, Pepper questioned children after they were told stories. Following a reading of the Aesop fable The Tortoise and the Hare, the robot asked them: “How did the tortoise feel about winning? How do you think the hare is feeling?” The questions, which had multiple choice answers, were aimed at building comprehension but, curiously for an artificial teacher, were also supposed to help the children learn about emotion.
The robots assisted in maths lessons too — setting up a shop in class, scanning purchases and then checking whether the children had paid the right money. The children, already exposed to the idea of intelligent machines through television and films, were thrilled by meeting humanoid robots, according to Melissa Bea, a teacher at MY World Preschool. “None of the children were afraid of it,” she tells me by email. “In fact, they had many questions for Nao like ‘How old are you?’, ‘Why are you so short?’, ‘Will you stay at the same height when you reach a hundred years old?’”
Like any new educational gadget, the robot grabbed children’s attention, Bea says, but she questioned its ability to direct a child’s education with human skill. “The robot, not being a thinking individual, is unable to pose questions, elicit responses, prod for discovery learning and point a child towards engaging in a new train of thought or develop new ideas. It is also unable to meet the emotional needs of children.”
Singapore is not the only country to explore the potential of robots in the classroom. Nao has been deployed at a primary school in Birmingham in the UK, where it was used to help children with autism. The trial at Topcliffe Primary School in 2014 was aimed at helping children to pick up social cues. The idea was that, by simplifying interactions, robots could make social exchanges such as playing games easier for autistic children to follow. Researchers in Israel have also used Nao to tell stories to kindergarten children.
In autumn 2002, American researcher Javier Movellan brought Robovie, a lanky bipedal Japanese-made android, to his son’s childcare centre in Kyoto. Movellan had become aware of the powerful and positive feelings of emotional connection that social robots could arouse in humans, but later confessed that he was shocked at how intensely the children reacted — they were terrified of the machine.
Back at the University of California, San Diego, where he worked, Movellan launched a project to study the potential use of robots to assist teachers in early childhood education. Perhaps mindful of how fearful the children in the Kyoto centre had been, the machine that his team developed, called Rubi, was cute and a little klutzy. A boxy, bug-eyed contraption with a touchscreen on its front, Rubi performed songs and interactive games aimed at improving the children’s vocabulary.
While the children were watching the robot, it was also watching them, learning to recognise whether they were smiling or looking blank. Rubi could also display emotion. After an early episode in which the children pulled off a robotic arm, it was programmed to “cry” when it came under physical pressure — signalling that it was getting hurt.
In a paper delivered at a conference in Osaka more than a decade ago, Movellan argued that “the development of social robots is revealing and making explicit aspects of human nature that have been ignored for the past 40 years”.
A key assumption of scientists who study the human mind is that people are fundamentally “thinking creatures”, he said. But social robots, and their ability to “touch the human heart”, were showing us the importance of our feelings and responses to sensory inputs. In 2012, Movellan and three other UC San Diego scientists launched a start-up, Emotient, with the goal of using artificial intelligence to read emotions through facial recognition. It was acquired by Apple in January 2016, with the management team, including Movellan, joining the company.
Why are we working so hard to set up a relationship that can only be “as if”? The robot can never be in an authentic relationshipSherry Turkle, professor at MIT
The growing use of robots with small children is causing some disquiet. Sherry Turkle, professor of social psychology at MIT, who has studied robot and human interaction for decades, warns of the risk of creating false relationships between children and mechanical helpers. In an emailed response to questions, she wrote: “Why are we working so hard to set up a relationship that can only be ‘as if’? The robot can never be in an authentic relationship. Why should we normalise what is false and in the realm of [a] pretend relationship from the start?”
The damage is potentially great, Turkle warns. “Not to what children ‘learn’, but to who they are, their capacity for relationship.” In her 2011 book Alone Together, about the isolating effects of human reliance on technology, she warns of the risks of employing robots as companions for children. “It is from other people that we learn how to listen and bend to each other in conversation,” she writes. “The developmental implications of children taking robots as models are unknown, potentially disastrous.”
Singapore began its second trial of educational robots in January 2016, at a cost of S$1.5m (£840,000), placing Kibo and another technology-enabled toy, Bee-Bot, into 160 nurseries to give children insight into how to master and manipulate machines. The impact will be assessed to see whether every preschool in the country should have access to the machines.
The experiments form part of Singapore’s “Smart Nation” programme, launched by prime minister Lee Hsien Loong in 2014 with the goal of making effective use of technology to enhance daily life. Other initiatives include deploying sensors and cameras across the city-state to gather and compare data on an unprecedented level. “As Singapore becomes a Smart Nation, our children will need to be comfortable creating with technology,” said Steve Leonard, a Singapore government official, at the launch of the second pilot last November.
Kala Miniandi, principal of Sparkletots, tells me the children’s initial reactions to the toys varied. “Some were impatient,” she recalls. “The children who were always swiping at home — they were all ready. Some with low self-esteem were reluctant to try. These are the children who did not have computers at home.”
But she found that the toys proved effective as a means of extending children’s creativity in new directions — something that echoes Seymour Papert’s early work at MIT. “They can play around,” she says. “Their imagination just comes alive.”
Gathered around Kibo, a rectangular machine with sockets at the top and each flank, the children at Sparkletots are being taught the syntax of coding. The “repeat” command in particular is a subtle way to teach them about automation and the scale that can be achieved through getting a machine to sequence a single human command. As the command begins and ends with the words “repeat” and “end repeat”, the children are told to imagine making a sandwich. “It must have bread at both ends. What do we want Kibo to repeat in between?” asks one of the teachers.
There are commands that allow the children to explore greater complexity. One block is printed “if”, so the children can construct a sequence such as “if light sing”; the robot will then respond to the shining of a flashlight by singing.
In another classroom, three children are being taught Mandarin with the aid of the Bee-Bot, a dinky black-and-yellow-striped machine about the size of a toy car. Pressing buttons on its back, the children program the device to move across a plastic sheet on a floor marked with Mandarin characters. It’s harder than it looks. One of the children, in a blue T-shirt and red shorts, is struggling with the task. He counts out the right number of squares to reach the word “monkey” but the Bee-Bot heads in the wrong direction. It becomes obvious that the child is confused between left and right and is therefore orienting the machine the wrong way.
At a table in a narrow room connecting the two other classrooms, children are busy making circuits out of copper tape, using the tape to link a miniature battery with an LED. Another group is building small machines out of tiny electronic components — a motor that can rotate a shaft, a pressure sensor, an LED — that are snapped together like Lego blocks. One of the children has made an interactive Valentine’s day card; a heart scrawled in purple felt tip with an LED creating a point of light at its centre. “I love you to the bones,” the child has written on the front.
The robot is mainly a teaching assisting tool — not to replace the teacher. I don’t think this will automate the classroomChen I-Ming, Nanyang Technological University
These children are encountering robots as objects to be manipulated rather than mechanical teaching assistants. Yet, here too, the teachers say the machines often spark an emotional reaction. Louiline Manahan, a teacher at Sparkletots, notes that one child — more reserved and less fluent in English than some of the others — has opened up with the use of robots. “In normal lessons he will not talk so much [although] I know he knows the answer. He is actually doing good — opening up to share what he has done. He wants to share his achievement.”
The initial results of Singapore’s pilot with Pepper and Nao have been encouraging but inconclusive, according to a robotics specialist involved in evaluating the trial. Chen I-Ming, director of the Intelligent Systems Centre at Singapore’s Nanyang Technological University, says the measuring of the pilot was more qualitative than quantitative — rather than looking at whether robots might improve children’s math scores, for example, the focus was on social exchanges.
“We looked at how the robots were engaging the children’s creativity,” Chen says. He observed a gap between human expectations and the robots’ abilities. “People think the robot is very clever, but every movement needs a human programmer to code that,” he says, adding that the educational trial had to be customised “to the limitations of the hardware… The robots can do a simple dance but they cannot jump.”
Asked whether robots might one day replace human teachers in a classroom, Chen sounds deeply sceptical. He notes that the expense of robots is one constraint on expanding their use in education. Another is parental anxiety about screen time, and whether too much technology in the classroom may be inappropriate for children. “It is mainly a teaching assisting tool — not to replace the teacher,” he says. “I don’t think this will automate the classroom. It’s just … how are we going to introduce robots into education [so that] kids get a better understanding of this technology?”
For future generations, as robots grow more sophisticated, the distinction between a social machine and a human might become fuzzy. In the Sparkletots classroom, the children seem happy enough to accept robots as companions. For Chinese New Year, at the end of January, the Kibo robots at Sparkletots were given paper “heads” in the style of Chinese dancing lions, mounted on rotating turntables. On a video playing on a laptop in the teachers’ breakroom, the robots dance alongside the children, who faithfully mimic their moves.
Jeevan Vasagar is the FT’s Singapore and Malaysia correspondent
Photographs by Amrita Chandradas; Cynthia Solomon/MIT
Please post comments and questions at the end of this article, or email email@example.com