New science of learning offers preview of tomorrow's classroom

Related images
(click to enlarge)

Courtesy of Alan Decker and the Machine Perception Lab, UC San Diego.

Of all the qualities that distinguish humans from other species, how we learn is one of the most significant. In the July 17, 2009 issue of the journal Science, researchers who are at the forefront of neuroscience, psychology, education, and machine learning have synthesized a new science of learning that is already reshaping how we think about learning and creating opportunities to re-imagine the classroom for the 21st century. “To understand how children learn and improve our educational system, we need to understand what all of these fields can contribute,” explains Howard Hughes Medical Institute investigator Terrence J. Sejnowski, Ph.D., professor and head of the Computational Neurobiology Laboratory at the Salk Institute for Biological Studies and co-director of the Temporal Dynamics of Learning Center (TDLC) at the University of California, San Diego, which is sponsored by the National Science Foundation. “Our brains have evolved to learn and adapt to new environments; if we can create the right environment for a child, magic happens.”

The paper is the first major publication to emerge from a unique collaboration between the TDLC and the University of Washington’s Learning in Informal and Formal Environments (LIFE) Center. The TDLC focuses on the study of learning—from neurons to humans and robots—treating the element of time as a crucial component of the learning process. This work complements the psychological research on child development that is the principal focus of the LIFE Center. Both have been funded as part of the NSF’s Science of Learning initiative.

Among the key insights that the authors highlight are three principles to guide the study of human learning across a range of areas and ages: learning is computational— machine learning provides a unique framework to understand the computational skills that infants and young children possess that allow them to infer structured models of their environment; learning is social—a finding that is supported by studies showing that the extent to which children interact with and learn from a robot depends on how social and responsive its behavior is; and learning is supported by brain circuits linking perception and action— human learning is grounded in the incredibly complex brain machinery that supports perception and action and that requires continuous adaptation and plasticity.

As the only species to engage in organized learning such as schools and tutoring, homo sapiens also draw on three uniquely human social skills that are fundamental to how we learn and develop: imitation, which accelerates learning and multiplies learning opportunities; shared attention, which facilitates social learning; and empathy and social emotions, which are critical to understanding human intelligence and appear to be present even in prelinguistic children.

These and other advances in our understanding of learning are now contributing to the development of machines that are themselves capable of learning and, more significantly, of teaching. Already these “social robots,” which interface with humans through dialogue or other forms of communication and behave in ways that humans are comfortable with, are being used on an experimental basis as surrogate teachers, helping preschool-age children master basic skills such as the names of the colors, new vocabulary, and singing simple songs (see image).

“Social interaction is key to everything,” Sejnowski says. “The technology to merge the social with the instructional is out there, but it hasn’t been brought to bear on the classroom to create a personalized, individualized environment for each student.” He foresees a time when these social robots may offer personalized pedagogy tailored to the needs of each child and help track the student’s mastery of curriculum. “By developing a very sophisticated computational model of a child’s mind we can help improve that child’s performance.”

“For this new science to have an impact it is critical that researchers and engineers embed themselves in educational environments for sustained periods of time,” says coauthor Javier Movellan, Ph.D., co-PI of TDLC’s Social Interaction Network and director of the Machine Perception Laboratory at UC San Diego. “The old approach of scientists doing laboratory experiments and telling teachers what to do will simply not work. Scientists and engineers have a great deal to learn from educators and from daily life in the classroom.” Movellan is collaborating with teachers at the UC San Diego Early Childhood Education Center to develop social robots that assist teachers and create new learning opportunities for children.

What makes social interaction such a powerful catalyst for learning, how to embody key elements in technology to improve learning, and how to capitalize on social factors to teach children better and foster their innate curiosity remain central questions in the new science of learning.

“Our hope is that applying this new knowledge to learning will enhance educators’ ability to provide a much richer and more interesting intellectual and cultural life for everyone,” Sejnowski says.

Source: Salk Institute