Since the days of ancient Greece, thinkers have grappled with the mystery of why humans and other animals play. Survival is hard, so why waste energy? Over the years, theories have ranged from Plato’s simple picture of play as skill rehearsal to ever more esoteric ideas about surplus energy, Freudian catharsis, and vestigial “primitive instincts.”
These days, developmental psychologists mostly see play as a way of gathering information about the world and our place in it. But there’s something missing from this picture, according to Marc Malmdorf Andersen, a cognitive scientist at Aarhus University in Denmark. “Yes, it’s about learning about the world,” he says. “Everyone agrees that play is really good for that. But no one really deals with the fun aspect. Why is it that it’s so rewarding?”
In 2023, Andersen and three colleagues published what they dubbed a “cognitive theory of play,” seeking to explain play within the framework of the newest, hottest theory of everything in brain science: predictive processing.
Play spurs children to learn about themselves and the world, but it also has important implications for adults.
According to this framework, people are wired to reduce uncertainty as we strive to perfect our predictions of the world around us — but when we play, paradoxically, we deliberately choose to increase uncertainty in order to experience the pleasure of reducing it. Andersen’s theory explains how play spurs children to learn about themselves and the world, but it also has important implications for adults: Our sense of play, it turns out, is crucial not only for having fun, but also for keeping our aging brains healthy and uncovering new ideas in ways that can’t be outsourced to an algorithm.
The new brain theory
The key idea in predictive processing is that, rather than trying to make sense of the firehose of information delivered by your senses, your brain makes its best guess about what’s happening around you and then uses all that sensory data to check its predictions. When you wander into the kitchen in the morning, you expect to see a fridge in the corner; your brain only takes notice of the incoming visual data if the fridge is missing.
This turns out to be a very efficient way for a brain to run. It also explains various quirks in our perceptions, since it means that vision is, as neuroscientist Anil Seth put it, “controlled hallucination.” In the 1860s, German polymath Hermann von Helmholtz invoked the idea of a predictive brain to explain certain types of optical illusions. More recently, this perspective explains why some people saw 2015’s viral photo of “The Dress” as gold and white while others saw it as blue and black: People who spend more time in artificial lighting have different baseline predictions about the photo’s background lighting than those who spend more time in natural light. What we see depends on what we expect to see.
In 2005, University College London neuroscientist Karl Friston began publishing a series of papers that proposed that the fundamental imperative of a predictive brain is to minimize prediction error. If your senses tell you that a prediction (“it’s warm enough to eat outside”) is wrong, you either update your prediction (“it’s not warm enough to eat outside”) or do something to make your prediction come true (put on a sweater). This seemingly simple principle turns out to be surprisingly powerful, and over the past two decades, it has given rise to a radically new mathematical conception of how the brain works. But it has a flaw.
The Dark Room problem
I first came across what has come to be known as the Dark Room problem when I was researching a book on the science of exploring. Various lines of behavioral and biological evidence suggest that humans are wired to seek out the unknown. Predictive processing, on the other hand, suggests that we want to minimize uncertainty to avoid prediction error. One way of achieving that goal would be to lock yourself in a closet and turn out the lights. Then it would be easy to predict what’s going to happen next: absolutely nothing. This chain of logic contradicts the idea of humans as instinctive explorers.
But survival isn’t just about predicting the present; it’s also about anticipating the future. The resolution of the Dark Room problem springs from the recognition that we want to minimize both present and future prediction error. To do that, we need to learn as much as possible about the broader world in order to anticipate what might happen next. In fact, Friston’s equations suggest that we need to preferentially seek out prediction errors in the near term in order to reduce prediction errors in the long term. We need to exit the closet, in other words, and explore.
British philosopher and cognitive scientist Andy Clark calls this “slope-chasing.” What we really seek isn’t the absence of prediction error, but the pleasure of surfing down a slope of decreasing prediction error. “You’re chasing these slopes because they improve your ability to predict the system,” says Mark Miller, a philosopher of cognition at the University of Toronto and a co-author of Andersen’s theory of play.
Building vs. chasing slopes
To Andersen, Miller, and their co-authors, play is a counterpart to exploring: slope-building rather than slope-chasing. Instead of looking for slopes of uncertainty to surf down, we choose the rules of our games to create the uncertainty that we can then resolve.
We’re looking for just the right amount of uncertainty. Too little, and we don’t get much pleasure from reducing it. Too much, and we’re unable to reduce it. This is why most casino games have odds that hover somewhere near 50%, and why the most satisfying sports contests are those whose outcome remains in doubt until the final seconds.
It also means that the sweet spot is a moving target, as what was once unfamiliar becomes familiar. That’s why our musical tastes change over time, Andersen points out, and why little kids enjoy going down slides at the playground — until they’ve been down enough times to resolve their curiosity about what it feels like, at which point they become more interested in the mystery of what it would be like to climb up the slide.
Exactly how this circuitry is implemented in the brain remains a topic of active research and debate. One key observation, from a series of experiments by German neuroscientist Wolfram Schultz in the 1990s, is that dopamine signaling in the brain functions in some contexts as a marker of prediction error. We get a dopamine hit not when something is good, but when it’s better than expected. “What the dopaminergic system seems to be doing is it tells the organism this strategy is better than the strategy you have been using so far,” Andersen says.
Of course, this picture of play might not capture all the nuances of what the word means to us. “We’re much more complex beings than uncertainty-based reward machines,” says Elizabeth Bonawitz, the head of Harvard University’s Computational Cognitive Development Lab. An infant playing peekaboo with its mother, for example, might be primarily motivated by social bonding rather than resolving the uncertainty of whose face is behind the hands.
Philosopher Bernard Suits defined playing games as “a voluntary attempt to overcome unnecessary obstacles.” To reach the finish line of a marathon, you can usually just walk a few blocks from the start line, or perhaps take a subway, but we choose to accept the arbitrary restrictions imposed by the course map. We impose these obstacles on ourselves to ensure that the outcome remains in doubt. Sometimes we fail, which means that when we succeed, the outcome is better than we expected. Andersen’s theory explains why this feels good — and it is a more significant insight than it first appears.
What is play good for?
The fact that play feels good is, you might think, a sufficient reason on its own to indulge in it. Suits famously argued that playing games is our highest calling since, in a hypothetical utopia where all our material needs are met, that’s all we would do.
But the pleasure of play also has utility: It tells us when we’re reducing uncertainty — that is, learning about the world — more efficiently than expected. In fact, it’s the most sensitive instrument we’ve got to tell us whether or not we’re choosing a fruitful path. “All else being equal, the organism is the only thing with access to its own knowledge,” Andersen says. “So fun should be the best signal for when an organism is either learning or dealing very efficiently with what the world is throwing at it.”
Even babies are able to tap into this instinct. A 2012 experiment led by Celeste Kidd, then at the University of Rochester, showed 7- and 8-month-old infants a sequence of shapes and used gaze-tracking to gauge their interest. When the sequence was too predictable, the infants looked away. When it was too random, they also looked away. But when there was an intermediate level of predictability, suggesting some sort of discernible pattern that they might figure out, they were hooked.
Many of the existing theories of play focus on children: the idea, for example, that a kitten pouncing on a ball of yarn trains the skills that a grown cat will later use to hunt. Andersen’s idea of play as slope-building fits with this picture of childhood as a period of uncertainty reduction writ large. Indeed, we spend less time playing as we grow up, and that’s not unexpected. “From the perspective of our model, that happens because adults find it harder to be surprised — because we know more,” Andersen says.
Play is for adults, too
We don’t stop building slopes entirely when we grow up — or at least, we shouldn’t. For one thing, there’s emerging evidence that staying playful helps keep your brain healthy as you age. The uncertainty inherent in play demands focus, flexibility, and adaptation, which in turn engages a region of the brain called the locus coeruleus, which is linked to cognitive health in older adults. One randomized trial found that taking up hobbies like quilting and photography was associated with slower cognitive decline.
The fact that it’s harder to surprise yourself when you already know a lot makes finding new surprises — new prediction errors — all the more important and rewarding. How else will you think new thoughts? The algorithms that control your social media feeds know how to titillate you, to give you the feeling that you’re surfing down a slope of descending prediction error. But they don’t have access to the internal cues that tell you when you’re successfully learning about the world. “We read books and act and do music and do arts and have hobbies and hike and explore the world,” Bonawitz says. “Those are the kinds of play that adults do, and I think that that’s a really important part of being happy and healthy and also just thinking.”
Of course, finding time to play more, amid the pressures and constraints of adulthood, isn’t easy. “Adults would play more if they could,” Andersen says. It also runs counter to our instincts. Ask people if they like uncertainty, and they’ll deny it. They want to minimize or eliminate it at all costs. But we’re not great judges of what we’ll find truly satisfying. If Andersen and his colleagues are right about the predictive processing theory of play, it’s not the absence of uncertainty that we love, but the process of reducing it: the journey, not the destination. And that involves seeking more, rather than less, uncertainty, and embracing unnecessary obstacles.
In other words, playing the game.
This article Why play brings us pleasure is featured on Big Think.
