Artificial teachers on the way
On the blackboard, somebody has drawn up tables and a few textbooks are on the teacher’s desk. It looks like any other classroom. Except that a robot is standing behind the teacher’s desk.
This is not a school, however, but the Social Robotics Lab at Uppsala University, and the classroom scene is temporarily rigged. Here, various kinds of robots are programmed and studied to be able to participate in teaching-related experiments.
One of them, Pepper, follows me attentively with his gaze. The large round eyes do not leave me so doctoral student Alex Yuan Gao ultimately takes hold of the robot’s head to break the sensor ray and thereby the motion setting. I feel relieved. Creaks arising in the machinery is not surprising, however. Research on social robots is still in its infancy. But there is great interest in artificial learning as seen from the number of projects at the robotics lab.
“In general, we work on teaching robots to interact with people in a socially intelligent manner,” says Ginevra Castellano, Senior Lecturer and head of the laboratory at the Department of Information Technology. “The goal is to develop robots that can be a support for people in everything from education and healthcare to entertainment. Our current focus is on creating scenarios for learning in classrooms and offices, for example, where robots can be instructors for children and adults.”
Her research team recently received SEK 5 million in the EU project Animatas, or Advancing intuitive human-machine interaction with human-like social capabilities for education in schools. The four-year project begins at year-end and includes participants from eight countries with expertise in robotics and artificial intelligence, as well as education and psychology.
“We are a good mix of people with different competencies in the project, as well as broad areas of expertise so that we can bring together different sides in the area,” says Ginevra Castellano.
Within Animatas, her team is responsible for developing and implementing the algorithms that the robots are programmed with. Then, the robots are to head out into reality for experiments with people. There are many challenges, mainly equipping the robot with the ability to read the student it is interacting with and provide responses appropriate to the situation.
“It’s always hard to build skills and abilities for robots that can serve in different scenarios and for different users,” says Ginevra Castellano. “Social interactions are very complex. But the assignment or the subject determines to a large degree how successful the robot is at teaching.”
In the laboratory’s latest experiment, the robots helped adults solve logical puzzle games and build Legos. One of the participants is Furhat, or Jona as he is also known. Jona is just a head, where facial features like eyes and mouth are projected from inside the head. Using various sensors and cameras, animations are created for the face. The robot can also be programmed to follow people with its gaze and read emotional expressions and behaviours.
On the table in front of Jona is a kind of board game.
“The puzzle is pretty challenging and there, the robot should serve as a supporting partner,” says doctoral student Maike Paetzel. “If you get stuck, the robot should be able to pick up if you get really frustrated and help you with hints about how to solve the puzzle. So it’s not a game against the computer, but a game where the robot and the human cooperate.”
The lab’s other two robots, NAO and Pepper, are made by a French-Japanese company, Softbank Robotics. At 50 centimetres, the full-body robot NAO is reminiscent of a toy doll, while Pepper’s 130 centimetre tall body stands on wheels. This is also of major significance to how people interact with them and how the robots can be programmed, says Ginevra Castellano.
“Since Pepper is quick and mobile, you can imagine that he helps by moving about the classroom. A robot face can instead be used for face to face tasks or as virtual support on a screen. Everything is about going back to the users and first investigating their needs before we program something.”
According to Ginevra Castellano, careful planning is necessary around which school students, teachers and subjects are to participate in the experiments, and what it is that is to be measured. One of the challenges is building abilities in the robots so that they can interact with people over long periods of time, like weeks and months. But it is difficult to test for several reasons.
“It’s a matter for example of me needing to work with ten children for ten weeks in a school, and that all of these children are really available every week. It’s one of the practical difficulties in our experiments.”
Ginevra Castellano’s research team is, however, already in contact with a school that they will cooperate with in the Animatas project: Katarinaskolan in Uppsala. More schools are waiting in line. “Our long-term ambition is to be able to integrate robots into the school’s curriculum where the teachers decide what roles the robots can play, not as a replacement for the teachers, but as a new way of thinking about teaching,” says Ginevra Castellano. She adds:
“It may involve subjects that are a little difficult to approach for some students and where special robots might be able to help by making the students more enthusiastic about certain things.”
Some of Ginevra Castellano’s projects:
- COIN – Co-adaptive human-robot interactive systems (2016-2021), financed by the Swedish Foundation for Strategic Research in cooperation with the KTH Royal Institute of Technology.
- Adaptive learning for personalized instructional robots (2016-2020), financed by the Swedish Research Council.
- EMOTE (EMbOdied perceptive Tutors for Empathy-based learning) (2012-2016), EU funded.
Find out more