Deep Learning Equips Robots to Help Autistic Children With Therapy
Children who have autism often find it challenging to ascertain the emotional state of people surrounding them. For example, they have trouble differentiating between a scared and a happy face.
In order to resolve this concerning issue, some therapists have begun employing children-friendly robots who demonstrate these emotions and help them imitate these feelings so that they are then able to respond to them appropriately.
These robots are designed in a way that they engage autistic kids in a personalized way. However, this therapy can work only if a robot can accurately comprehend a child’s behavior and analyze his/her level of focus and excitement during the course of therapy.
This is where the researchers from MIT Media Lab come into the picture! They have designed a personalized machine learning network that assists these robots in estimating the interest and engagement of a child during a therapy session.
They use the data that is specific to each kid.
As a result of this deep learning network, it becomes a piece of cake for robots to perceive the response of each child in a personalized manner. This response is later verified with the human experts’ assessments.
The correlation score of this agreement is set to 60% as reported by the scientists in Science Robotics on June 27th. Being said that, human observers can have a hard time reaching a high level of agreement regarding a kid’s behavior and engagement.
That brings the correlation scores down to 50%-55%.
Oggi Rudovic, the first author of the study and a postdoc at the MIT Media Lab explains, “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.” He also mentions that these robots might be able to provide consistent estimates of the behavior of autistic kids someday.
A co-author of the study, Rosalind Picard uses a famous adage in respect to this study, “If you have met one person with autism, you have met one person with autism.” She says, “The challenge of creating machine learning and AI [artificial intelligence] that works in autism is particularly vexing because the usual AI methods require a lot of data that are similar for each category that is learned. In autism where heterogeneity reigns, the normal AI approaches fail.”
For the Robot-assisted therapy for Autism, the humanoid robot NAO by Softbank was used by the researchers. NAO is a unique robot who is almost like an armored superhero with a height of 2 feet.
It can express several types of emotions through the color of eyes, the tone of voice, and movements in the limbs.
The study consisted of 35 autistic kids, 17 of whom were from Japan, and the rest 18 were from Serbia. They all ranged between the ages of 3-13 years old.
Their reactions during the 35-minute therapy sessions to the robots were different, ranging from sleepy to bored and even jumping with excitement in some cases while touching or laughing with the robot and clapping their hands with delight. This study is definitely a step in the right direction and brings hope for autistic kids so that they can be helped with more-equipped ways in the future.
Via: MIT News