杏吧原创

EVEN at 15 years old, Phoebe still likes playing with dolls. Unlike other kids of her age who are busy eyeing up the opposite sex or fast-talking their way into the cool gang, Phoebe speaks in simple sentences, avoids eye contact and finds it difficult to make friends. Her problem is autism. Like many other children with autism, she has limited social skills. Things like taking turns and imitating people, simple skills that most children pick up naturally, developed very slowly during her childhood. But if two British researchers are right, Phoebe can begin to learn the social skills she needs from a smart, robotic doll.

Not that robots have suddenly become masters of social nuance. Rather, the researchers believe, the very qualities that make robots so mechanical and, well, robotic, could be precisely the qualities needed to engage children with autism. By playing with this doll, the hope is that the children can 鈥渄iscover鈥� social skills for themselves, rather than being taught them explicitly 鈥� a method that has met with only limited success.

Autism is believed to affect as many as 2 in 1000 children 鈥� mainly boys 鈥� in Europe, Australia and the US, and the number of children diagnosed with the problem is rising rapidly. The disorder itself is difficult to define, but according to a British autism expert Simon Baron-Cohen at the University of Cambridge, it is thought to be a problem that affects the way the brain develops. 鈥淚n particular the brain is not developing the circuits for interpreting other people鈥檚 natural behaviour,鈥� he says. This manifests itself in a number of ways. Some children with autism have severe learning difficulties while others have high IQs. Others, such as those with Asperger鈥檚 syndrome, show normal language and intelligence but find it hard to communicate.

While it鈥檚 difficult to make generalisations, the common thread that links most people with the disorder is the problem they have interacting socially. 鈥淧eople with autism tend to prefer things that are highly predictable and highly structured,鈥� explains Baron-Cohen. But in our highly social world people simply don鈥檛 behave in predictable, mechanical ways. Eye contact, for example, plays a vital role in non-verbal communication but there are no simple rules for the way it works, he says. Empathy can also be a problem. 鈥淲hen we see someone behaving in a particular way the first thing we do is try to imagine what they are thinking,鈥� says Baron-Cohen. For people with autism this is extremely difficult.

To try to get round these problems, teaching sessions for children with autism are highly structured, says Kerstin Dautenhahn, a biologist and roboticist at the University of Hertfordshire in Hatfield, near London. 鈥淭he children are explicitly encouraged to make eye contact, take turns in a social setting and associate facial expressions with emotions,鈥� she says. Yet they still have difficulty generalising what they learn in the classroom into everyday experience. 鈥淚f you teach them to take turns in a classroom setting, how do you teach them that it also applies to other situations like conversations?鈥� says Dautenhahn.

One solution might be to place the children in a more natural situation where they are in control, she says. Working with roboticist Iain Werry from the University of Hertfordshire, she is exploring the use of robots that behave reactively but not completely predictably. The idea is that by playing with robots that behave according to some fixed rules, the children feel they retain control of the situation so they are more likely to feel comfortable. And giving the robots an element of unpredictability helps encourage the children to look for common threads in the robot鈥檚 behaviour and make generalisations from this.

When Dautenhahn and Werry started their research in 1998 their main concern was to make a robot that didn鈥檛 scare the children. 鈥淧eople with autism are often very scared of other people,鈥� says Werry. So they deliberately avoided building a humanoid and instead designed a robot that resembled a large toy truck. The robot is equipped with heat sensors that can detect nearby children and bumper switches that make it reverse when it hits something. The researchers programmed it with a few basic commands that allow it to play games like 鈥渢ag鈥� and 鈥渇ollow my leader鈥�. These games are not explicit, but emerge during the play sessions. For example, the robot will head towards a heat source but won鈥檛 get too close. If a child keeps moving away from the truck, this produces 鈥渇ollowing鈥� behaviour, and if the child then starts to move towards the robot, it will retreat. This allows games of tag to develop, and helps develop patterns of imitation and turn-taking behaviour, both of which are necessary for high-level social interactions, says Dautenhahn.

鈥淭urn taking is probably crucial because it鈥檚 the basis of dialogue,鈥� says Baron-Cohen. 鈥淧eople with autism who speak often don鈥檛 do dialogue.鈥� Imitation also plays a fundamental role in acquiring language, since we start off by copying what we hear. But this, too, is something many people with autism find difficult to master.

Working with schoolboys at the Radlett Lodge School (for Children with Autism) in Hertfordshire, the researchers filmed some playing with the robot truck and others playing with a similar sized toy truck. They then tried to quantify different types of interaction by analysing video footage from 10-minute play sessions.

They found that the boys, all aged between 7 and 11, enjoyed interacting and playing with the robot but also seemed to focus more of their attention on it than the normal toy truck. In fact, far from being afraid of the robot they formed simple bonds with it, even becoming protective and possessive.

Recent trials, which involved pairs of children playing with the robot at the same time, were even more encouraging, with the children exhibiting a number of social behaviours 鈥� both good and bad. For example, the children actually played with each other, although in some cases their behaviour became antisocial as they competed for the robot, says Dautenhahn. But that鈥檚 still a major step forward. 鈥淓ven this is positive because at least it forced them to acknowledge each other,鈥� she explains. Some children also instructed each other and cooperated by trying to get the robot to follow them, with one child even attempting to get the other to imitate the way he was playing with the robot.

The robot truck was tailored to help the children in other ways too. The researchers decided that it was important to include some central point of focus on the toy, much like an eye or head, says Werry. People with autism are perfectly capable of making eye contact, he says, they just prefer not to. In fact their eye contact behaviour is very selective, says Cheryl Trepagnier, a research consultant for the Autism Society of America Foundation and researcher at the National Rehabilitation Hospital in Washington DC. Children with autism only seem to maintain eye contact when they themselves are talking and they want to check that you are paying attention.

But when someone else is speaking, eye contact doesn鈥檛 seem important to them. They simply don鈥檛 seem to treat the face as a source of communication, and research has even shown that they sometimes make no distinction between normal faces and faces that have been turned upside down. 鈥淭hey are missing out on an important communication channel, and the experiences of a social expert 鈥� the adult,鈥� says Trepagnier. 鈥淚n effect they are developing asocially and aculturally.鈥�

Dautenhahn and Werry designed their robot鈥檚 sensor unit to resemble a pair of eyes on a head and located it on top of the robot at the front end. They found that the children tended to show great interest in it. 鈥淭he children often walk around the robot or step aside to test if the robot can see them, and check if it follows,鈥� says Dautenhahn. What is important is that they quickly figure out that if they want to get the robot鈥檚 attention they have to address it from the front, she says.

The plan now is to gradually increase the repertoire of the robot鈥檚 behaviour. By doing this they hope to guide the children towards more realistic forms of human-to-human social interaction. The positive results also encouraged Dautenhahn to try a more human-like robot. Working with Aude Billard, a roboticist at the University of Southern California in Los Angeles, she has begun to examine how children like Phoebe get on playing with a robotic doll called Robota, that has been developed by Billard.

Children with autism are known to enjoy playing with dolls, probably because of the level of control they can exert over them. But Robota comes with a few extras. Using an off-the-shelf doll, Billard has added motors, sensors and a simple processor chip that allows the doll to move, to sense movement, and even recognise gestures and respond to them.

After a few basic instructions from a teacher or parent, children can play freely with Robota and try some games, which involve imitation, turn taking and eye contact. For example, Robota will try to copy what a child does with its arms. The child is shown that the robot will copy gestures 鈥� a raised arm or a turn of the head, for example 鈥� and that these games won鈥檛 work unless it is facing the robot, encouraging face-to-face communication.

Although it is too early to tell how useful these dolls will be, Phoebe鈥檚 mother Anne believes that her daughter鈥檚 social skills have improved and she is already beginning to reap the benefits. Before she played with Robota, Phoebe could empathise in a limited way, says Anne. If she were playing with a toy doll, for example, she would pretend it was crying to indicate that she was unhappy. 鈥淏ut this has become more sophisticated,鈥� says Anne. She has watched Phoebe鈥檚 ability to empathise improve significantly since playing with Robota. When acting out a tea party with her dolls Phoebe will now pretend that one doll is aware of how other dolls are feeling, for example, pretending that one doll is concerned that another is hungry. She is essentially ascribing emotions to the dolls.

Dautenhahn agrees that Phoebe鈥檚 response to Robota has been encouraging, but admits that it is hard to establish a direct link between these play sessions and improved behaviour. 鈥淲e have not yet shown a therapeutic effect,鈥� she says. 鈥淏ut I鈥檓 optimistic that we will.鈥� And if larger trials are successful, robotic dolls and other toys could soon become a crucial tool in helping children with autism navigate our perplexing social world.

It鈥檚 an imaginative approach, says Baron-Cohen. 鈥淚t鈥檚 taking advantage of the fact that they like the mechanical world,鈥� he says, because it is predictable and tirelessly repetitive. But it could all backfire, he warns.

鈥淚f the robots are highly rule governed it may be that children with autism may not generalise their experiences into real social interactions,鈥� he says. If the robots are too predictable they may end up reinforcing their tendency to want to exert control on people and their surroundings. 鈥淢y concern would be that this behaviour would be perpetuated,鈥� he says. You could end up turning these children into control freaks.

Dautenhahn acknowledges these risks, but insists that the robots have been designed in such a way as to prevent this. 鈥淚f the robots were completely predictable then he would have a point,鈥� she admits.

Researchers will have to wait for further studies to find out whether this approach actually works. However Anne, for one, is convinced. Robota, she says, has helped Phoebe communicate her needs and anxieties more clearly. 鈥淚t has enabled her to live with the people around her, and enabled them to interpret her needs,鈥� she says. 鈥淭hat has to be a good thing.鈥�