The ethics of treating autistics with robots

Post Reply New Topic

Matu is part of a major EU-funded study that's testing whether such robot-assisted therapies work any better than human-only sessions. Robots are already being used in autism therapy, but whether they work any better than human-only sessions remains in question. Another study suggests they may be no more helpful than standard therapies, even if this robot-child interaction looks so positive.

Previous research suggests robots are well received by children with autism spectrum disorders, and there's plenty of anecdotal stories of children showing progress with interaction and communication after working with robots. In the UK, researchers based at the University of Hertfordshire have had plenty of positive attention around Kaspar, a child-like robot that helps teach children social skills.

The Development of Robot Enhanced Therapy for children With Autism Spectrum Disorders, or DREAM, is a long-running study of robot-assisted therapies, funded by the European Commission and run across seven European universities, including the universities of De Montfort, Portsmouth and Plymouth, as well as Skövde in Sweden. The first stage of the research tests the use of robots for common autism treatments, mostly using NAO, a half-metre tall programmable, humanoid robot made by French firm Aldebaran Robotics. The researchers are also using a "huggable" robot called Probo, developed by Vrije Universiteit Brussel, which was designed to teach communication skills, showing facial expressions using 20 motors in its head. Both robots have been used in previous autism spectrum disorder research.

The robots are used to assist with common treatments, cognitive-behaviour theory (CBT) and applied behavioural analysis (ABA), and target three skills: joint attention, imitation, and turn taking. Children are judged on how well they share and follow instructions before and after the sessions, which are set up as a series of games to play. For example, a robot will be used to show physical actions for the child to mimic, such as miming drinking from a cup or waving their hand. Imitation is one way young children learn, and the idea is that such actions will teach the child to model the skill, which may not come naturally to them. With turn taking, children play a game with the robot, learning how to wait their turn and interact with each other, while joint-attention measures how the child initiates contact with the play partner, human or robot.

These robots aren't designed to replace human therapists, but to be used as a tool by them. The therapist might use the robot as a focal point for the child, as some find it easier or more engaging to interact with a robot than an adult, or the therapist may use it to show a movement the child should mimic, a key aspect of behaviour-based therapies. A robot can do the movements to mimic, saving human therapists' time and energy.

The DREAM research is ongoing and results haven't yet been fully published, stresses Billing and his co-researcher Silviu Matu. Early results from the first phase were reported last fall, halfway through the study, saying there were "no significant differences" on the primary outcomes, the turn-taking, imitation and joint attention skills measured by the therapists. While the robots offered "small advantages" in secondary outcomes, such as how the children interacted in the sessions, there wasn't improvement in the three specific skills being measured, the two researchers note. "There are of course lots of variation for individual children."

Erik Billing, at the University of Skövde, says the results so far suggest that some children benefit from robot-assisted therapy, but others may do better without robots in the room. He also notes that the effects appear to wear off after time — like any new tech, robots are intriguing to anyone when they first see them, but less so after the tenth time. Matu adds that most children react positively to the robot in the room. "They are curious about the robot and want to see the robot and play with him," he says.

The robots are used to assist with common treatments, cognitive-behaviour theory (CBT) and applied behavioural analysis (ABA), and target three skills: joint attention, imitation, and turn taking. Children are judged on how well they share and follow instructions before and after the sessions, which are set up as a series of games to play. For example, a robot will be used to show physical actions for the child to mimic, such as miming drinking from a cup or waving their hand. Imitation is one way young children learn, and the idea is that such actions will teach the child to model the skill, which may not come naturally to them. With turn taking, children play a game with the robot, learning how to wait their turn and interact with each other, while joint-attention measures how the child initiates contact with the play partner, human or robot.

These robots aren't designed to replace human therapists, but to be used as a tool by them. The therapist might use the robot as a focal point for the child, as some find it easier or more engaging to interact with a robot than an adult, or the therapist may use it to show a movement the child should mimic, a key aspect of behaviour-based therapies. A robot can do the movements to mimic, saving human therapists' time and energy.

The DREAM research is ongoing and results haven't yet been fully published, stresses Billing and his co-researcher Silviu Matu. Early results from the first phase were reported last fall, halfway through the study, saying there were "no significant differences" on the primary outcomes, the turn-taking, imitation and joint attention skills measured by the therapists. While the robots offered "small advantages" in secondary outcomes, such as how the children interacted in the sessions, there wasn't improvement in the three specific skills being measured, the two researchers note. "There are of course lots of variation for individual children."

Erik Billing, at the University of Skövde, says the results so far suggest that some children benefit from robot-assisted therapy, but others may do better without robots in the room. He also notes that the effects appear to wear off after time — like any new tech, robots are intriguing to anyone when they first see them, but less so after the tenth time. Matu adds that most children react positively to the robot in the room. "They are curious about the robot and want to see the robot and play with him," he says.

One aspect of the DREAM project is focusing on the ethics of using robots for treatment. And there's plenty to consider, above and beyond the usual quandaries raised by the use of AI and robotics elsewhere in healthcare, such as building in bias and data privacy.

Robots may help treat children with autism, but we need to be wary of treating them like robots in return. Are machines the best way to teach social interaction? Do we risk further entrenching flawed assumptions that people with autism are machine-like? Could we be ignoring known successful therapies in favour of high-tech wizardry?

Given we don't understand the causes of autism, it's difficult to know how to treat it, and even though we don't yet know if robots help in therapy, they're already becoming a focus for research. Why spend so much money developing robots that can track eye movements, interact using AI and the like if all we need is a doll or street performer? Are we putting too much focus — and too much R&D funding — into robotics for children with autism?

While any research that could lead to happier lives for children with autism is welcome, it's frustrating for those doing the day-to-day therapeutical work who can't get the funding for already successful treatments. "It does concern me because millions and millions go into these projects," Richardson says. "The arts, for example, have an extraordinarily positive proven effects with people with alzheimer's and autism and they just don't get any money. It just goes into the tech. That does concern me."

Aside from costs and the most responsible way to spend research funding, there are other ethical questions raised by the use of robots for treatment, and much of it centres on how people with autism are viewed. There's a long-running — and false — assumption that everyone with autism is methodical and machine-like, lacking emotion and empathy.

"The idea that it's like a computer brain is dehumanising," Milton says. "Due to popular understanding — or misunderstanding — of what constitutes autism, it's often incorrectly seen in such terms that autistic people are somehow less than human, animalistic or machine like, only capable of compiling and broadcasting strings of information."

Richardson argues that the machine-like metaphor is one reason robotics are a popular research angle for autism therapy. "If you look into the underlying theory behind the use of robots, you find this idea of people with autism preferring things over other persons," she says. That idea suggests that autism is a version of the "extreme male brain" — which Richardson says is "nonsense" and "sexist". While people picture Sheldon Cooper from The Big Bang Theory — genius, abrupt, bad at social interactions — the reality is some people with autism can't speak, are withdrawn and beset by anxiety and obsessive compulsive issues. It just doesn't make for good TV. "I think this idea that children with autism are mechanical, they prefer systems, and they lack empathy is what really provided the basis for the thinking about autism and robots," she says.

And that's led robotics researchers to hope to build systems using autism models, such as theory of mind, a model that describes the idea that people with autism don't understand that other people have their own point of view, thoughts or even visual perspectives, making it difficult for them to relate socially. Roboticists at Richardson's conference showed off their systems developed with autism in mind, such as teaching their machines to have the ability to consider different perspectives, understanding that one person may see an object another may be blocked from viewing, for example.

Such borrowing from people with disabilities can be problematic, Richardson says, pointing to robotics developers creating vision systems using human blindness as a model. "They were borrowing lots of things from people with disabilities for their robots, as though people with disabilities weren't quite human," she explains. "If you're going to develop a technology around a group of people, they should be involved in it, otherwise you just end up with an echo chamber." That's why she's so eager to praise Billings DREAM work, as it incorporated ethical considerations early on in the project's development.

Milton describes himself as on the spectrum, and he's not convinced robots are what's needed by people like him and his son, who also has autism spectrum disorder. "Why machines? Are machines the best things to teach social interaction to people?" he says. "I'm not so sure." He suggests that instead of robots being designed to fix how people with autism behave, perhaps such research can be used to help everyone else understand better what it's like to be autistic, building empathy in both directions.