Study turns CPUs into STEM learners
Man and machine may work together to improve the educational outcomes of both, if a new project takes off.
Researchers in the US are teaching computers to learn the way young students learn, so that they can be used for effective education modelling.
Classroom research is a vital part of assessing or creating new policy, but such studies can suffer from a lack of big research groups, plus they are disruptive to students and difficult to control.
Now, a computational modelling project will see a computer “learning” student behaviour and then “thinking” as students would.
In a recent test, computers were used to examine student responses to science tasks – such as comparing liquid volumes – and then mimic the way students thought about it.
The synthesised students’ mind was created using a complex ‘artificial neural network’ - an artificial intelligence system that can simulate some functions of the human brain.
Students were given tasks to complete in an electronic game. The tasks were scientific in nature and required students to make a choice, with researchers then using statistical techniques to track each move and label each result as a success or failure.
The A.I system ‘learned’ their approach to science, rather than just how to do a specific task, so it could later try to solve a different problem the same way a student might.
Lead author for the Washington University study, Dr Rich Lamb, said the computer-based research system has a number of advantages over the human version.
“The computer is learning to solve novel or new problems, which means we can test different educational interventions before ever getting to a classroom,” he said.
“Even with a large research team, it’s usually too difficult to test more than one intervention at a time.
“Now we can run multiple interventions, choose the one that looks like it will work the best and then just test that one.”
He says it will help on the all-important bottom line, too.
“For me to get 100,000 students, teachers to administer tests, professors doing research and all the rest, we could easily look at about $3.5 million,” Dr Lamb said.
“We can now get those 100,000 students for the cost of running software off a computer.”