 |
| © MIT Media Lab The NAO robot tracking parent and child human facial expressions. |
By David Grossman, Popular Mechanics
Kids with autistic conditions sometimes work with robots to help them better distinguish emotions. Now, scientists at MIT are using artificial intelligence to make sure these robots understand the children they're working with.
When therapists use robots for kids with Autism Spectrum Disorders (ASD), they do so to demonstrate how to better understand emotions and socialize in a more general population. A study in 2012 showed several "potential advantages to using interactive robots" with ASD children. Robots like Milo or NAO can walk, talk and even mimic human facial expressions.
While these robots are built for human interaction, if the robot does not have a proper understanding of the context a conversation can only go so far. So researchers at the MIT Media Lab built a type of personalized machine learning that helps robots estimate the interest and engagement of every individual child during these interactions. Each data set accumulated would be unique to the child the robot was monitoring.
“The long-term goal is not to create robots that will replace human therapists, but to augment them with key information that the therapists can use to personalize the therapy content and also make more engaging and naturalistic interactions between the robots and children with autism,” says Oggi Rudovic, a postdoc at the Media Lab and first author of the study in a press statement.
The question of what children are really thinking is an eternal question for even the most seasoned child healthcare professional. When human observers monitor a child's emotional interaction with a robot, they're usually right around half or 55 percent of the time. Using the personalized algorithm, the robot was able to correctly interpret a child's reaction 60 percent of the time.
 |
| © Provided by Hearst Communications, Inc |
Using NAO, a robot almost 2 feet tall that can control the motion of its limbs and tone of voice, researchers led 35 kids, ages 3 to 13, in 35-minute sessions. They would show children flash cards representing certain emotions, and then NAO would interact with the children using those emotions. Sometimes the kids were energetic, sometimes they were bored and tired.
However, on the whole the children treated NAO "respectfully as it if was a real person," Rudovic says. Sometimes, children were scared of NAO at first but as their sessions developed they became more open to its presence and affectionate. The children on the whole loved hearing a story about NAO taking them out for ice cream and ended up laughing by the end of therapy.
NAO was using an algorithm based on of video of each child’s facial expressions, head and body movements, poses and gestures, audio recordings and data on heart rate, body temperature, and skin sweat response from a monitor on the child’s wrist. These reactions were then shown to five children's health experts, who then coded the children’s video and audio recordings.
Placed into an algorithm alongside other personal information like birthplace, gender, and specific abilities, NAO was given a personalized framework for each of these children.
One reason for the new algorithm's relative success is its consistency. "Humans change their expressions in many different ways, but the robots always do it in the same way, and this is less frustrating for the child because the child learns in a very structured way how the expressions will be shown,” Rudovic says.
The researchers found that plugging each child's algorithm into NAO gave the children a greater sense of interaction than a one-sized-fits-all approach. Robots have a lot of potential in a variety of health-related fields. When they're not working with children to better understand emotions, they can be performing brain surgery.
However, on the whole the children treated NAO "respectfully as it if was a real person," Rudovic says. Sometimes, children were scared of NAO at first but as their sessions developed they became more open to its presence and affectionate. The children on the whole loved hearing a story about NAO taking them out for ice cream and ended up laughing by the end of therapy.
NAO was using an algorithm based on of video of each child’s facial expressions, head and body movements, poses and gestures, audio recordings and data on heart rate, body temperature, and skin sweat response from a monitor on the child’s wrist. These reactions were then shown to five children's health experts, who then coded the children’s video and audio recordings.
Placed into an algorithm alongside other personal information like birthplace, gender, and specific abilities, NAO was given a personalized framework for each of these children.
One reason for the new algorithm's relative success is its consistency. "Humans change their expressions in many different ways, but the robots always do it in the same way, and this is less frustrating for the child because the child learns in a very structured way how the expressions will be shown,” Rudovic says.
The researchers found that plugging each child's algorithm into NAO gave the children a greater sense of interaction than a one-sized-fits-all approach. Robots have a lot of potential in a variety of health-related fields. When they're not working with children to better understand emotions, they can be performing brain surgery.
