If you believe that the best way to learn is through simple trial and error, you’ll be happy to know that – despite numerous impressive advances in technology that have seen the introduction of not only computers, but also tablets, into the classroom – the educational establishment still agrees with you. It’s just that the failures that teach children belong to robots now, apparently.

A story in New Scientist explains that Japanese children are learning to speak English faster than before thanks to a new robot that tries to teach them, but gets things wrong. The discovery comes after Shizuko Matsuzoe and Fumihide Tanaka, two scientists at the University of Tsukuba, studied nineteen children aged between four and eight years old interacted with a humanoid robot known as Nao as it tried to teach them simple English words for shapes such as “circle,” “square” or “heart.”

The key discovery, according to the report, was finding out that the kids became more engaged with the robot as it made mistakes that they could recognize and try to correct. “When the robot got a shape wrong, the child could teach the robot how to draw it correctly by guiding its hand. The robot then either ‘learned’ the English word for that shape or continued to make mistakes,” the report explains, adding that “Matsuzoe and Tanaka found that the children did best when the robot appeared to learn from them.”

This was possible because the robot wasn’t actually a robot, as such. Instead, it was a remote controlled machine, with Matsuzoe and Tanaka operating it from the room next door so that they could adapt their approach to whatever their child subject did, allowing them to either play dumb in order to see the child’s reaction, act as a teacher to correct one of the child’s mistakes, or even carry out tasks set by the children appropriately, in order to create a control reaction.

The basic idea behind the robot theory is a simple one, and not an uncommon one in educational theory; by allowing the child to feel more in control of the situation and responsible for teaching the robot, it’s empowering the child and reinforcing the value of the lesson that they are now passing on, but in such a way that it doesn’t feel like a learning experience. That’s something that Andrea Thomaz, the director of the Socially Intelligent Machines lab at the Georgia Institute of Technology in Atlanta, agrees with: “Anything that gets a person more actively engaged and motivated is going to be beneficial to the learning process,” she explains. “So needing to teach the robot is a great way of doing that.”

Matsuzoe and Tanaka plan to present their findings at the Ro-Man symposium on robot and human interactive communication next month.