A new study from Yale scientists shows how uncertainty helps us learn

Your brain adapts to its environment, deciding when learning is necessary.
Your brain adapts to its environment, deciding when learning is necessary.
Image: EPA/Erik S. Lesser
We may earn a commission from links on this page.

We may crave stability in life, but the brain benefits from volatility, according to a Yale University study on cognition that shows how uncertainty boosts learning.

The paper, to be published in Neuron (paywall) on Aug. 8, demonstrates that there’s more activity in the brain’s prefrontal cortex when a situation is difficult to predict, and less activity when the outcome is more certain. ”Perhaps the most important insight from our study is that the function of the brain as well as the nature of learning is not ‘fixed’ but adapts according to the stability of the environment,” Yale medical school professor of neuroscience, psychiatry and psychology Daeyeol Lee tells Quartz.

Our brains aren’t learning all the time. Instead, Lee explains, the brain determines whether learning is necessary in a given situation, and if so, what kind of learning is most beneficial. When situations are predictable, the brain doesn’t need to do as much. When circumstances are changing, it works harder.

One implication of the findings, according to Lee, is that we should seek out new situations to stimulate brain activity and learn more. “When you enter a more novel and volatile environment, this might enhance the tendency for the brain to absorb more information,” he suggests.

Monkey mind

To figure out where and how brains take in new information, and whether that changes depending on context, the researchers performed two experiments with monkeys—one that mimicked volatile conditions, and one in which conditions were stable. During the experiments, researchers monitored the monkeys’ neuron activity in three sections of the prefrontal cortex.

In the “volatile” trial, the monkeys were offered a choice between hitting a red and a green target. Hitting the red target gave them a reward of juice 80% of the time, while the green target gave them juice 20% of the time. After 20 or 40 attempts, the scientists reversed the probabilities for the targets—20% chance of juice for red, and 80% chance of juice for green. That way, the monkeys had to pay attention to which color they’d chosen before and recall whether that particular choice was rewarded to figure out which target was most profitable at any given point.

By contrast, in the “stable” condition, scientists used orange and blue targets, also assigned 80% and 20% probability rates. In this test, however, the probabilities didn’t change. Once the animal figured out that orange target rewarded 80% of the trials and blue only rewarded 20% of the time, they could ignore information about previous choices because reward probabilities were stable and signals about prior choices weren’t as important.

To make the tests more complex, there was also a visual cue for each target, indicating the magnitude of reward to be received if the target yielded juice. This made it more likely that a monkey would sometimes choose a lower probability target that offered more juice, choosing to take a greater risk for more reward based on expectations established over several tries. Indeed, says Lee, sometimes the monkeys chose the 20% probability target knowing that a successful hit, though more rare, would yield more juice than a successful hit on the 80% probability target.

While the monkeys were tested, the researchers recorded the activity of individual neurons in three separate areas of the prefrontal cortex—the dorsolateral prefrontal cortex, orbitofrontal cortex, and anterior cingulate cortex. Lee explains that  each of these areas show different types of choice and reward signals. But a common finding in all three areas was that signals related to the animal’s previous choices and outcomes were enhanced in the volatile condition when learning was necessary, especially in the dorsolateral prefrontal cortex and orbitofrontal cortex.

Uncertainty or volatility significantly influenced how much information about prior experiences—both choices and outcomes—were maintained in these areas of the brain. “In other words, when the environment is uncertain and therefore when the animal has to learn, signals related to their previous choices and outcomes (ie, whether they chose correctly and got rewarded) had a more robust effect on how active neurons in the prefrontal cortex would be,” Lee tells Quartz. “We therefore conclude that this is how the brain modulates learning according to volatility and uncertainty.”

The human takeaway

What all this monkey business means for the human mind is that we learn more in situations where we’re not sure. The uncertainty activates our brain and makes us more intent on signals about previous information, and causes us to listen to the most relevant of this messaging in an effort to predict the best future outcome.

The brain is like a gambling addict. It’s always betting on outcomes, predicting what will happen if you make a particular choice. This applies to everything from mundane daily choices like what to wear or where to shop to much more significant decisions like what studies to undertake for your career or who should be your mate.

Whenever the outcome is different from what you expected, you incorporate that new information so that you can make a better prediction later. If it is hotter today than you expected when you left the house this morning, then tomorrow, you’ll choose cooler clothing keeping in mind today’s surprising experience. Similarly, if you’re a liberal married to a conservative, for example, and the union turns out to be untenable because you constantly argued about politics, chances are good that after your divorce you’ll seek ideological alignment in any future love stories.

But Lee warns that there are situations when we shouldn’t necessarily learn too quickly from new knowledge. “If an environment is otherwise stable, then minor random events should not cause you to change your otherwise well adapted behaviors,” says the neuroscientist.

For example, if you have a personal favorite store, the fact that you purchase one defective product there probably shouldn’t prevent you from returning to the same store to buy other things. Or, more significantly, if your spouse is unusually annoying one day but sweet most of the time, don’t file for divorce based on the new information about a particularly irritating occasion.