How your brain learns to cope with uncertainty

Laura Harrison didn’t want to evacuate. The neuroscientist and mother of two had finally gotten her young children settled into bed after their nighttime ritual. But it was the evening of January 7, 2025, and the skies over her home in Sierra Madre, Calif., glowed a haunting orange-gray. Outside, the Santa Ana winds sent debris swirling with powerful gusts.

In nearby Eaton Canyon a spark had already become a roaring fire. Harrison could see the flames advancing down the mountain, but she hadn’t received evacuation orders. She hesitated. She knew which escape routes were still open, but fleeing at the wrong moment could be catastrophic. She had little information about what choices would give her the best odds of getting out safely. Uncertainty shrouded the city along with the smoke.

But the rapidly worsening conditions convinced her to leave. As Harrison drove her family out, she sent photographs to her colleague Galen Buckwalter, who lived nearby, urging him to flee at once. Buckwalter, a research psychologist, became quadriplegic after a teenage diving accident and uses a wheelchair. That night the van he relies on for mobility was in the shop, and he physically couldn’t get into his wife’s car. He wondered whether he should just shelter in place.

On supporting science journalism

If you're enjoying this article, consider supporting our award-winning journalism by subscribing. By purchasing a subscription you are helping to ensure the future of impactful stories about the discoveries and ideas shaping our world today.

Then fire trucks began wailing their way closer, broadcasting urgent calls to evacuate. Buckwalter told his wife to drive away without him—he would use his motorized wheelchair to escape, hoping to reach a street where an accessible Uber could pick him up. He strapped on a high-visibility vest and fled. “I took off in my chair in this howling wind with ashes all around [and] fire on the hills,” he says. For around 20 minutes he motored through choking smoke on almost unrecognizable streets, trying to stay calm. “It was really quite an apocalyptic feeling.”

The Eaton fire would ultimately kill at least 19 people and destroy more than 9,000 homes and workplaces. Thankfully Harrison, Buckwalter and their families made it out safely, and their homes survived the blaze. Both scientists work for the Valor Institute for Neuroscience, which is part of a private equity firm. When the Eaton fire ignited, they were, fittingly, in the middle of revising a paper about the connection between the way our brains handle uncertainty and resilience to stress.

Brains evolved to handle uncertainty. They use whatever information they can gather to compute the best course of action for reproductive success and self-preservation—finding food to fuel metabolism, for example, or locating a safe place to sleep. We can never know enough to definitively guarantee our safety, but we can use what information we have to make a best guess. Some neuroscientists now theorize that the brain is a “prediction engine” at its core and that processing uncertainty is its central task.

Evolution prepared us for a world that is uncertain but with hazards that are often foreseeable and, to some extent, calculable. When we know the odds, we can run them and take our best shot at beating them. But the more the odds change on us, the more stressful life becomes—and, research shows, the greater our risk for mental and physical illness.

For many people, the world today feels less predictable than at any other point in their lives. Climate change is driving severe weather and disasters such as the L.A. fires. American norms about politics, science, education and law are being overturned, and artificial intelligence threatens further disruption. Understanding how the brain uses predictions to manage uncertainty may now be more critical than ever—and it may also help explain the causes of mental health disorders that are on the rise in the U.S. and show us how we can foster resilience in ourselves and our children.

In physics, one way to measure uncertainty is entropy—a way of quantifying disorder and randomness. As the bleak second law of thermodynamics states, entropy always increases over time in a closed system. Systems fail; unanticipated difficulties occur; things break or rot. To cope with this reality, philosophical and religious traditions such as Buddhism and Stoicism have long focused on transience, unpredictability, and the inevitability of change and loss.

Being alive allows us to briefly evade the second law. Living creatures dodge entropy for as long as we can by being organized as open systems with permeable membranes. We take in energy from outside to maintain metabolism and to reproduce; we expel disorder through waste. We die when entropy wins by ending—through aging, injury or disease—our body’s ability to stay fueled and organized.

According to one popular theory, we defeat chaos for as long as we do because information can bring some temporary order. In the 1940s mathematician Claude Shannon quantified information as the reduction of uncertainty. This founding insight from information theory can be united with physics and biology through the “predictive brain” theory, first proposed in the early 2000s by Karl Friston, a professor of imaging neuroscience at University College London. This theory, also known as active inference or the free energy principle, has become one of the most important frameworks in neuroscience for understanding how the brain works as a whole. From a computational perspective, Friston argues, attempting to reduce uncertainty—basically, working to avoid unpleasant surprises—is the brain’s primary goal. “I think uncertainty is at the heart of how we make sense of the world,” he says, and reducing it “is a fundamental drive for nearly everything we do.”

“Uncertainty is at the heart of how we make sense of the world. Reducing it is a fundamental drive for nearly all we do.” —Karl Friston, University College London

Friston’s free energy principle mathematically describes how the brain uses the information it obtains from the senses to predict the state of itself, the body and the surrounding environment. According to the theory, we experience the world not through sensory information itself but through these predictions. If, for example, you see a gray blotch on the floor at night and think it’s your cat, you might conjure an image of your curled kitty before you realize it’s actually a discarded towel.

Our brain makes these predictions both consciously and unconsciously, relying on past sensory and physical experience to run the odds. This may explain why early life experience is literally foundational: to best learn the complex rules of being a functional human, children need a stable and generally predictable environment. We update our projections continuously when sensory information suggests they are inaccurate. Neuroscientists call these inaccuracies “prediction errors,” and they are associated with distinct signals in the brain. The erroneous cat changes seamlessly into the towel it actually is as prediction errors are overridden by new information.

Networks across the brain communicate with one another to create this perception, with information flowing between lower-level sensory regions and higher-level processing networks to accurately represent reality. Friston and his colleagues have found in studies across species that important brain circuits throughout the cortex seem to be wired to match information from the senses with data from memory to make predictions.

What’s more, dopamine pathways that were traditionally thought to govern pleasure and punishment also seem to be highly involved in prediction. Multiple studies now show that dopamine cells in these circuits signal prediction errors related to reward. For example, once a rat has been trained to push a lever to get a treat, its dopamine levels will rise at the sight of the lever before the treat is even available. But if it later pushes the lever and gets nothing, those levels will fall below baseline, telling its brain that its prediction was wrong. Dopamine doesn’t necessarily signal pleasure here; it just represents the difference between the expected and actual reward, which can be considered the mathematical expression of prediction error. New research has also shown that reducing another signaling chemical, norepinephrine, in certain brain regions can decrease rats’ ability to learn from situations where expected rewards are less available.

All of this means a highly unpredictable world is the brain’s worst nightmare—and the uncertainty that results can quickly become toxic.

Large prediction errors can be stressful. They signal that something major is wrong, requiring immediate attention and correction. They can activate the hypothalamic-pituitary axis, releasing glucocorticoid hormones such as cortisol to prepare the body to respond.

Resolving the prediction error typically requires bringing in new information. For example, if I hear a sudden loud noise, I may turn my head to locate the source. Or if I am trying to work and my blood glucose falls lower than expected, my brain attempts to resolve this prediction error about my body’s internal state, starting by producing the sensation of hunger. This new information will lead me to seek and eat food, and my glucocorticoid level will increase to focus my attention until I resolve the problem by returning my blood glucose to normal.

When levels of stress chemicals rise, they put the brain into a mode that uses more energy. In a mildly stressful situation—where, say, you can’t eat because you forgot to go shopping and there’s no food in your home—this rise is moderate and brief, and it increases learning. You will be less likely to forget to go to the store in the future. But in situations of ongoing uncertainty—such as chronic food instability—the stress is far greater.

Hormones such as glucocorticoids are one way that chronic stress “gets under the skin,” according to the late Rockefeller University neuroscientist Bruce McEwen, who pioneered research in this area. When released under conditions of acute stress, glucocorticoids power fight or flight by shunting metabolic energy away from processes such as digestion and toward the heart, lungs, brain and muscles. These emergency reactions consume energy and cause wear and tear throughout the brain and body.

In the brain, for example, elevated stress hormones can impair learning and memory by flooding the hippocampus with excitatory chemical signals. High stress can also shift the brain’s energy resources in ways that make it harder to regulate emotions and control behavior. Such wide-reaching effects may be part of why chronic stress, particularly in early childhood, increases risk for nearly every known mental illness. When people have little relief from stress in a chaotic environment, their brains have difficulty coping with the ongoing prediction errors, Friston, McEwen and neuroscientist Achim Peters argued in a 2017 article.

Chronic stress also increases risk for physical illnesses such as type 2 diabetes, heart attack and stroke in part because of the increased metabolic demands it imposes. A state of “metabolic emergency” can preserve life in the short term, but it can hurt the body in the long term. Metabolism involves constant trade-offs, and when the brain doesn’t have the information it needs to stay in balance, things break down—and entropy wins.

If the brain’s primary goal is to reduce prediction error, this way of being contains a trap. Wouldn’t it be most efficient to withdraw into the most predictable setting possible—an empty, lightless room, perhaps—and stay there? This is called the “dark room” problem.

In the long run, this strategy won’t work. You will eventually have to emerge from the small, safe place to which you’ve retreated. At that point the outside world will have become even more unpredictable because you haven’t been gathering new information about it. A more adaptive approach is to seek positive surprises when you can safely do so. Our very nature tends to encourage this behavior. “We are naturally born curious creatures,” Friston says. We play, solve puzzles and learn new skills even when they bring no obvious, immediate benefit for survival.

Sampling the world through our senses and our interactions with others lets us build rules and patterns for our predictions. But when these rules and patterns suddenly change, the models we built about the world can no longer resolve uncertainty. “We now have to change our world model,” Friston says. “What would that look like? It would effectively look like a grief reaction.”

In these situations, withdrawing into a literal or symbolic dark room—avoiding sources of stress and going to a safe place—can be adaptive in the short term. Friston notes that one theory on the origins of depression suggests that this is its evolutionary purpose: to give us time to consider behavioral changes when a survival or reproductive strategy is failing. If this withdrawal goes on for too long, however, it becomes harmful. Some core symptoms of anxiety and depression, such as avoidance and a lack of motivation, can be seen as the results of a brain stuck in the dark room, according to Friston.

Lisa Feldman Barrett, a professor of psychology at Northeastern University, argues that the brain is not necessarily trying to minimize cost—it’s simply trying to spend energy efficiently. States of uncertainty that cannot be resolved can be particularly inefficient. “In uncertainty, your brain has to assemble many plans and maintain those plans over a period of time,” she says. “And that’s just expensive because there’s more going on. There’s more signaling. There are more chemicals.” Symptoms such as a lack of motivation or pleasure and a feeling of being stuck can be “indicators of being strapped for energy,” Barrett says. When less energy is available to update predictions and you see the world as bleak, you may continue to have that view even when conditions are actually improving.

Alternatively, depression may involve placing too much confidence in outdated forecasts. Once you have decided, for example, that “no one will ever love me” or “everything always gets worse,” your brain may attend more to signals that support those predictions and less to those that don’t. Over time the brain can fall into a rut. An inability to update outdated predictions is one factor that may contribute to mental illness. Another way to describe this situation is as a lack of plasticity or changeability in the brain. Notably, most things known to have antidepressant effects increase brain plasticity—such as exercise, electroconvulsive therapy, selective serotonin-reuptake inhibitors prescribed as antidepressants, and psychedelic drugs.

Anxiety disorders may involve a different type of prediction problem that relates to reduced plasticity. Here you overgeneralize from one bad experience: if you get bitten by one dog, you might fear all dogs; if you fail one test, you may see yourself as a failure. “The more you avoid, the more you think your prediction was accurate,” says Philip Kendall, director of the Child and Adolescent Anxiety Disorders Clinic at Temple University. “‘I can’t go near that dog; it’s going to bite me.’ So [you] don’t go near the dog. [You] don’t get bitten. And you go, ‘See that—the dog would have bitten me, but I’m safe.’ That’s a false confirmation of a prediction, [and it] leads to more avoidance.”

Managing the stress of uncertainty is like building muscle. You start with small challenges.

Research suggests that having an especially hard time with uncertainty is associated with many anxiety and related disorders, including obsessive-compulsive disorder and generalized anxiety disorder. “We don’t know if it’s appropriate to call [intolerance of uncertainty] a cause or a symptom,” Kendall says. “The good news is that there are treatments that reduce intolerance of uncertainty,” and these treatments can be effective for anxiety disorders.

To treat mental disorders involving faulty predictions about the world, clinicians often use cognitive-behavioral therapy, or CBT, which addresses the intellectual content of the fear: exploring the real probability that the feared negative outcome will happen and instilling confidence in coping skills that can help. In practice, this means helping people to change the way they talk to themselves and to stop avoiding situations that make them nervous. Kendall teaches children and teenagers to break down anxiety-provoking situations, such as finding friends to sit with in the cafeteria or asking someone for a date, into smaller steps and planning what to do if something goes wrong. “When kids do things that they’re afraid of, they don’t fall apart,” he says. “They become proud peacocks.” This practice helps them override the predictions they made based on prior experience, which made avoidance seem like the safer choice. In a study published in 2019, Kendall and his colleagues showed that the more a child’s uncertainty intolerance was reduced by treatment, the more their anxiety fell and their ability to function improved.

Of course, sometimes events are genuinely unpredictable and threatening enough to jeopardize safety. Unpredictability is a hallmark of the circumstances that often produce post-traumatic disorder (PTSD), such as natural disasters, wars, terrorist attacks, vehicle accidents, and sexual or physical violence. These experiences can obliterate people’s ability to predict what will keep them safe. Events that make you feel betrayed or helpless in a life-threatening situation can contribute to the development of PTSD. The sense of continued threat that constitutes the experience of PTSD can be described as a prediction error.

Still, the majority of people faced with traumatic experiences don’t develop psychiatric problems such as PTSD or depression. George Bonanno, a psychologist at Columbia University, explains that many people who have intense symptoms initially—such as nightmares and flashbacks—won’t develop PTSD. These symptoms may reflect the brain’s attempts to integrate and learn from what happened. “Your brain is trying to make sense of all that information,” Bonanno says. “This is one of these things that the public doesn’t hear about—that intrusive memories [or] even nightmares after something really disturbing or life-threatening are not uncommon, and they’re actually adaptive. But then, for some reason, that turns into PTSD in some people.”

It is surprisingly difficult to predict who will experience PTSD, a phenomenon known to researchers as the resilience paradox. Some features put people at higher risk for developing PTSD or more severe forms of the condition: for example, having suffered multiple previous traumas or having little social support. Determining who will recover quickly is also a challenge. But people who thrive in uncertain and threatening situations tend to have some things in common: traits related to flexibility, Bonnano says. These involve maintaining some degree of optimism about recovery, having confidence in your ability to cope, and seeing the world with what’s known as a challenge-oriented mindset. This is the ability to reframe thoughts and physical signs that could be interpreted as anxiety in a less threatening way.

For instance, if you’re about to give a speech, you can interpret your elevated heart rate as you approach the podium as a sign of fear—or one of excitement about being able to rise to the challenge. The more you can view uncertainty as opportunity rather than threat, the more experience you get successfully managing mild and moderate stressors, and the better you are likely to be at dealing with extreme situations.

In the paper that Harrison, Buckwalter and Friston were editing when the L.A. fires ignited, they describe four potential pathways or “phenotypes” of managing uncertainty. These paths aren’t unalterable; people can adjust them with therapy or other learning experiences. The least resilient is the “fragile” type. Genetic predispositions to sensory overload, which is often experienced by people with autism, attention deficit hyperactivity disorder or obsessive-compulsive disorder, can put someone on this path early in life. Their existence feels overwhelming and unpredictable, so they create their own dark room by adhering to rigid, repetitive routines that may limit opportunities for learning, social connection and updating of inaccurate predictions. Another on-ramp to the fragile pathway is a childhood of abuse, neglect, or other recurring traumatic experiences. These circumstances, too, create a perception of the world as unpredictable and may lead people to seek escape via alcohol or other drugs.

People with the second phenotype, called durable, might have grown up in a predictable world—for example, a strict religious sect or just a family that experienced little change—and developed only certain coping strategies as a result. They are highly functional in familiar settings but may become fragile or inflexible in new ones. The third possible path is that of resilience. Resilient people experience life’s typical ups and downs, but their senses don’t overwhelm them. In most cases, they will be able to respond even to potentially traumatic experiences by seeking social support. The final phenotype, confusingly labeled “pro-entropic,” describes people who proactively seek information. This makes them even more resilient because regular experience with significant stress allows them to become better at handling ever more extreme challenges.

Although both Harrison and Buckwalter initially denied they needed to evacuate during the L.A. fires—a potentially deadly response—they ultimately got themselves to safety. They used what they’d learned about paths to resilience to guide their choices. “Making the best inferences you can really, really came to the fore for me,” Buckwalter says. He knew the light on his motorized chair would help him see in the smoky air, and he grabbed a reflective vest so firefighters rushing to the scene could see him. Keeping himself calm allowed him to navigate his chair on a familiar street made unfamiliar by burning debris and smoke.

Buckwalter says his awareness of the predictive brain allowed him to cope better and regulate his emotional responses so he could do what needed to be done. “This process has been invaluable to me in my life,” Buckwalter says.

After classes resumed at her oldest child’s school, Sierra Madre Elementary, following the fires, Harrison gave a presentation there on the relation between resilience and the predictive brain. Her talk focused on being flexible and creative when responding to stress. First she praised the kids for being “brave superheroes” who had had to deal with a really difficult experience. “You guys were super,” she recalls saying. “You’ve all gotten through this.”

Then Harrison explained how they could get even better at facing future challenges. Stress isn’t always bad, she emphasized, and choosing to change course when something isn’t working is critical to learning. She walked the children through examples of situations when they might not get what they expect, such as not being able to have their favorite breakfast. The idea that they could come up with alternatives on their own, rather than just relying on their parents to fix it, energized the students, especially the youngest ones.

Managing the stress of uncertainty is like building muscle, Harrison says. You start with small challenges, and over time you learn to face larger ones more effectively. It’s important to be curious—to not become too attached to your expectations. “If you expect to be surprised, you can come back more quickly,” she says. “Stress or change in the world is opportunity as well.”