Teachers are often stretched thin. As classroom sizes get larger and resources dwindle, it can be a significant challenge for even the most qualified teacher to provide individual attention to every single child, especially those with special challenges or learning difficulties. 

One possible solution? Robots.

As part of the National Science Foundation (NSF) Expeditions in Computing, researchers from Yale University are developing socially assistive robotics— a new field of robotics that focuses on assisting users through social rather than physical interaction. A core part of their research is to design these robots to work with children, including those with challenges such as autism, hearing impairment, or those whose first language is one other than English.

The goal is not to replace teachers, but to assist them, said Brian Scassellati, a professor of computer science, cognitive science, and mechanical engineering at Yale University and director of the NSF Expedition on Socially Assistive Robotics.

“You may have a public school where there are 25 to 30 kids in a classroom and the amount of time that the teacher can spend one-on-one with each child is relatively limited,” said Scassellati in an interview with R&D Magazine. “What these systems are doing is really leveraging the time that the teacher can provide, not replacing their position entirely. Anytime where you need that little bit of extra one-on-one time, that is where these systems are successful.”

The NSF Expedition has been focused on this goal for nearly five years, with the mission to develop the computational techniques that will enable the design, implementation, and evaluation of robots that encourage social, emotional, and cognitive growth in children. These robots—which typically look like fun, cute, or silly toys— must have the ability to not just teach, but to adjust their teaching style and approach depending on each individual child.

“One basic approach is to customize the difficulty of the questions, the curriculum you are presenting, so as someone becomes more adept, the robot presents harder material and more challenging problems. That is something that we have done in intelligent tutoring systems and in robotics for many years now,” explained Scassellati. “What is new is that, not only are we customizing the content, but we are also now seeing systems that customizes the way in which they deliver that content.”

The team is working to create robots that recognize and respond to the smallest variations in learning styles, motivation, and personality based on its experiences with a child over time. For example, a robot could recognize if a child is motivated by collaborative goals or competitive goals. For a child motivated by collaboration, the motivation to stay engaged with the system may be based on achieving a specific score as a team, said Scassellati. For a child motivated by competition, a learning game where the student works to get a higher score than the robot or other students might be more effective.  

 

Robots are also uniquely positioned to learn more about students than a teacher might be able to, said   Scassellati.

In one instance, socially assistive robotics were utilized at a school where many of the students spoke Spanish or Portuguese at home and were learning English as a second language. Several teachers disclosed that their biggest challenge was not having a good understanding of exactly how much English their students knew and what areas of the language they were having the biggest challenges.

“If the teachers asked a kid a question, they would deflect or refuse to answer because it was embarrassing to say the wrong thing in front of the class,” said Scassellati. “But when we took those same kids and we put them one-on-one with the robots, even more so than when we put them one-on-one with other adults, the kids were willing to make mistakes in front of the robot. There is no social stigma attached with making a mistake in front of the robot.”

The robot then “used” these mistakes to better understand the children’s skills and customize an English curriculum for them.

“This had a tremendous effect in terms of how much those kids were able to learn, even in a relatively small amount of time–three, four sessions–with the robot,” said Scassellati.

This same benefit was observed in kids with autism. Kids who were afraid to attempt certain social interactions with peers based on past experiences of messing up and being embarrassed, were not afraid to do so in front of the robots. The robot could then understand what social challenges the autistic kids faced and help them to overcome them. Scassellati’s team is currently working on a project where socially assistive robotics were placed in the home of an autistic kid and his or her family for 30 days each. Each day the child spends 30 minutes playing social skills games with the robot as well as with their parents and the robot together. Scassellati’s team will then analysis the impact of this on the child’s social skill improvement.

“We know the kids like it, we know the kids are really excited by it; the question is if they learn over the long-term,” said Scassellati. “We have great responses from them over a few sessions in the course of a week, but this will be our first really ‘in the wild’ test in the home for 30 days.”

Other applications

While Scassellati’s team has been focused primarily on the use of this technology with children, socially assistive robotics could be deployed in a number of adult applications as well. The robots have been investigated for use with the elderly, to monitor them for signs of dementia and keep them cognitively engaged and socially engaged with those around them.

They have also been investigated for use as lifestyle coaches, helping people manage diabetes or motivate them to lose weight or exercise.

“Basically anything that requires a personal coach or trainer to work with you, those are exactly the kinds of application domains where we are seeing robots come into play,” said Scassellati.

The field has been growing rapidly in the last decade. According to Scassellati, the number of academic articles on socially assistive robots in 2006 was relatively small. When his team last took count in 2014 there were between 150 and 200 articles on the topic. He suspects that this number has nearly doubled today.

“We are still at the point where we are dotting all our I’s and crossing all our T’s and there is a big difference between these robots working in the lab and working out in the real world, but we are starting to have projects that really make that jump and that is exciting for everybody,” said Scassellati. “We are just at the tip of the iceberg with this. We are nowhere close to the peak of this wave. I think this is a really exciting domain to be in right now.”