Bandit is designed to be helpful with autistic children. (Bob Chamberlin / Los Angeles…)
Robots aren't known for their soft side. They build cars and defuse bombs; they don't, as a rule, make friends or deal with feelings. But a few groups of researchers around the world are working to build robots for an unusual purpose: Making emotional connections with autistic children who often struggle to interact with humans.
There's something about machines that really seems to resonate with many kids with autism, says Maja Mataric, co-director of the Robotics Research Lab at USC. These children often have trouble reading human emotions and social cues — complexities they don't have to worry about when they're around a mechanical being.
"Robots are simpler than people," Mataric says.
Still, robots may seem like unlikely candidates for a job usually filled by therapists. As Mataric points out, the general public usually thinks of robots as either cold and efficient workers (at their best) or outright evil beings bent on enslaving humanity (at their worst).
The researchers at USC have a different vision. "We're trying to create something that's endearing," Mataric says.
The result: Bandit, a metallic-colored, child-sized robot that can win the attention — and even empathy — of hard-to-reach kids.
Bandit has a pleasant, inviting face with a movable mouth, archable eyebrows and camera eyes that let him "watch" his playmates. He also has proximity sensors to gauge whether kids are backing away or moving in. If they get too close, he can wheel away.
With his motor-driven arms, Bandit can automatically mimic the motions of children and lead a game of Simon Says. He can make sad sighs or happy chips, and he blows bubbles with the push of a button. He can also talk in soothing tones, although USC researches are just beginning to use Bandit's speech in their work with children with autism.
Bandit, who has been around in various incarnations since 2007, is human-ish but still obviously a machine, which is exactly the look that Mataric and colleagues were aiming for. If he looked too much like a robot, kids wouldn't want to be his friend. And if he looked too human, he would likely make kids with autism feel intimidated and overwhelmed. "It was a balance that we had to find," she says.
So far, a few dozen kids with autism spectrum disorders have spent time with Bandit in various small studies. Mataric would like to have more kids visit, but she says it's hard to find children and families who are willing and able to complete a study. Still, she has seen some real signs of progress. With Bandit's encouragement, children have learned how to take turns and initiate play with others. Bandit has even inspired some children to smile socially for the first time, she says.
Bandit has an overseas soulmate of sorts in KASPAR, a robot who works with kids with autism in a lab at the University of Hertfordshire in England.
With his baseball cap, black hair and child-like face, KASPAR (the name is an acronym for Kinesics and Synchronisation in Personal Assistant Robotics) looks more like an oversized doll than a robot. But he's still a big hit among the autistic.
While not every child is interested in KASPAR, "we've had a lot of successes over the years," says senior research fellow Ben Robins, who has been working with the robot for five years.
Robins has heard from parents and teachers that kids who always seemed to be locked in their own worlds suddenly showed an interest in other people after spending time with KASPAR. "I can't say for sure that the robot is responsible," he adds.
Unlike Bandit, KASPAR doesn't run automatically; a nearby researcher guides his actions with a remote control. Robins acknowledges that the bot isn't as advanced as Bandit or many other robots out there. But that suits him just fine. "I'm working from the standpoint of the children, not the technology," he says.
In the years since he first helped design the robot in 2006, Robins says he has removed features to make the robot simpler and easier to play with. "Children need something basic that is both reliable and repetitive. Everything else is already so confusing to them."
Robins envisions eventually building 15 or so KASPARs that schools or hospitals could keep for long-term therapy. Likewise, Mataric can picture a time when families could buy a Bandit or similar bot of their own to use at home.
For now, both goals are hampered by a lack of funding. Mataric says take-home Bandits could be a reality within five years if a venture capitalist would step up, but so far she isn't exactly swamped with offers.
Another challenge is that Mataric, Robins and other researchers lack the resources necessary to run the sort of large-scale clinical trials that could answer some key questions: How long do the benefits of therapy last? How do the social skills learned in the lab translate to the real world? Which children are most likely to benefit? And how can researchers design robots to get the best results with the fewest setbacks?