Research suggests that each of us should aim to be wrong a certain percent of the time if we want to absorb new information
Math has never been my strong suit. When I was applying to graduate school a few years ago, thinking about the math section of the GRE left my stomach in anxious knots. (Yes, this is a story about a standardized test, but bear with me.) In a move that was part prudence, part masochism, I signed myself up for a refresher course. To my own shock, found myself actually enjoying the interactive practice tests, which start out easy, but get harder and harder as your accuracy improves.
I learned math more quickly during my GRE prep than I ever did in a traditional classroom. I suspect this had to do with the fact that the practice tests were constantly adapting to make sure that they were just hard enough, but not too hard — a balance that one recent study calls the “sweet spot” for learning.
For the study, published earlier this month in the journal Nature Communications, researchers taught computer neural networks (artificial intelligence systems that loosely mimic animal brains) to do simple tasks like sorting patterns and numbers, paying attention to how the difficulty affected the computers’ learning speed. The optimal difficulty for learning, they found, was an accuracy rate of about 85%. In other words, the study suggests, it’s possible that the fastest way to learn is to be wrong about 15% of the time.
The idea that mistakes are essential to learning hearkens back to one of the most famous theories in educational psychology: the “zone of proximal development,” a concept first proposed in the 1930s by the Russian psychologist Lev Vygotsky. Vygotsky believed that children learn best while doing tasks that require some guidance and support — not when they’re left to do things they can easily accomplish by themselves, or, on the other end of the spectrum, when they’re expected to try things they definitely can’t do on their own.
Other researchers have since built on that idea. For example, the psychologists Elizabeth and Robert Bjork, who run the Bjork Learning and Forgetting Lab at UCLA, have found that forcing students to struggle a little helps them learn faster, a phenomenon they call “desirable difficulty.”
But while most scientists today agree that there is a sweet spot for learning, there’s no real consensus about exactly what that sweet spot is. The team behind this latest study hasn’t yet tested whether the 85% rule applies for people, although they’re planning to, says Robert Wilson, the study’s lead author and a psychologist and cognitive scientist at the University of Arizona.
Wilson predicts that the formula will hold up for tasks similar to the ones in the study, where people have to choose between two options, and get more accurate with experience. One real-world example might be a dermatologist learning to determine if certain moles are cancerous or not, he suggests. Future studies could explore whether the principle applies in people attempting to learn a wide variety of skills, like learning to play the violin or perform brain surgery.
For those of us trying to cram for the GRE or learn anything else, the lesson is that each of us will probably learn best when we find our own sweet spot, rather than trying to match the pace of a group. You’re probably getting close to your optimal level of difficulty when you’ve achieved a state of “flow,” and are feeling engaged, but not discouraged, Wilson says.
And no sweet spot, he adds, is going to be universal: “What’s 85% for one network is going to be 55% for another.” So try performing your own experiments to see what level of challenge works best.