Education researchers should throw out the traditional linear learning science process in favor of a more flexible process that allows them to work more closely with educators and students in the classroom, according to Javier Movellan, director of the Machine Perception Laboratory at the University of California, San Diego,
“There’s a tradition that you do the science and then you take it into the field,” Movellan said at the American Association for the Advancement of Science conference in Washington. “I really think that’s a bunch of baloney in some fields. Developing a mature science before you apply it is too rigid.”
Movellan leads the RUBI project, which is building a robot capable of interacting with preschool children and teaching them basic vocabulary and language skills. The researchers initially focused on building a “socially capable” robot, not on the interaction with children, and as Movellan said, when the team took the robot into an early childhood center, “Nothing we had worked, but the good news is the failures that we had transformed our science.”
For example, the RUBI team had developed award-winning technology to allow the robot to tell when a child is smiling, so that it can respond positively or continue what it is teaching. Unfortunately, the sensor had been tested only under controlled conditions. In the classroom, “children were not posing; they had spontaneous smiles.”
The researchers had to sit in the classrooms, observing and testing one sensor after another to find one that caught the glancing expressions that mark a normal smile in a constantly moving 2- or 3-year-old. The testing led not only to more information about how young children interact socially, but advanced expression-sensing technology that Sony now uses in its camera design.
Similarly, the children initially did not follow the robot’s cartoon “gaze” when it turned to look at something, a critical indicator that they were not interacting with it. During routine maintenance, the researchers sped up the motions of the robot’s head turning—and were surprised to find an “explosion” in the number of times children interacted with the robot, following its gaze, pointing, and talking to it. Over time, the team has been able to calculate the exact speed of movements during social interaction; again, developing “pure science” out of iterations of applied interactions.
The researchers have “mopped floors and changed diapers” in the center, and in the process, found out more about what the parents, teachers, and children in the center actually see, do, and need from research, Movellan said. It continues to change his research practice, as the engineer explores the importance of emotion and motivation in learning. He urged more researchers to become “embedded” in the schools and classrooms they study. “We need students and teachers to enrich our science,” he said.
Barbara Means, an educational psychologist at SRI International, agreed that starting from applied research and working backward can help scientists understand the real problems of practice they are trying to solve and avoid years of research on interventions that ultimately don’t work in the classroom.
“You’ve got to attach those new ideas to real important needs out in the world,” she said, “which you really can’t do unless you are out interacting with people in the world.”
For a Q&A and video on the RUBI project, see this.