The Extended Mind: The Power of Thinking Outside the Brain

by Annie Murphy Paul

Delving deeper, Pape discovered that distractions and interruptions of the pilot by other crew members accounted for a majority of airline “incidents.” Another correspondence came into view: interruptions of health care professionals, she knew, were also to blame for many medication mistakes.

Pape also became aware that aviation experts had devised a solution to the problem of pilot interruption: the “sterile cockpit rule.” Instituted by the Federal Aviation Administration in 1981, the rule forbids pilots from engaging in conversation unrelated to the immediate business of flying when the plane is below ten thousand feet.

In her 2002 dissertation, and then in a series of articles published in medical journals, Pape made a case for imitating this practice. “The key to preventing medication errors lies with adopting protocols from other safety-focused industries,” Pape wrote in the journal MEDSURG Nursing in 2003. “The airline industry, for example, has methods in place that improve pilots’ focus and provide a milieu of safety when human life is at stake.” Such methods could be adapted to the hospital setting, she argued, by creating a “no-interruptions zone” around medication preparation areas, and by having nurses who are administering medication wear special vests or sashes signaling that they are not to be disturbed.

At Kaiser Permanente South San Francisco Medical Center, for example, the introduction of no-interruption signaling in 2006 led to the “virtual elimination of nurse distractions for those wearing the vests,” according to the US government’s Agency for Healthcare Research and Quality. Over a six-month period, medication errors at the hospital fell by 47 percent. Nearly two decades after she initiated it, Pape’s lifesaving act of imitation has spread all over the country and the world.

Tess Pape figured out the correspondence problem on her own. But what if she had been taught how to imitate? Imitating well is a skill, one that Oded Shenkar believes should be deliberately cultivated. Shenkar, a professor of management and human resources at Ohio State University, studies how companies use imitation to gain a strategic edge in the marketplace. He maintains that we are living in a golden “age of imitation,” in which access to information about how other people are addressing problems similar to our own has made it more feasible than ever to copy effective solutions. Shenkar would like to see students in business schools and other graduate programs taking courses on effective imitation. He imagines companies opening “imitation departments,” devoted to identifying promising opportunities for copying. And he anticipates a day when successful imitators are celebrated and admired just as much as innovators are now.

In addition to the sterile cockpit concept Pape adapted, health care professionals have also borrowed from pilots the onboard “checklist”—a standardized rundown of tasks to be completed. In this case, too, imitation has worked wonders. In 2009, researchers from the Harvard School of Public Health and the World Health Organization reported that after the surgical teams in their study started using a nineteen-item checklist, the average patient death rate fell more than 40 percent, and the rate of complications decreased by about a third.

The medical field has also adopted the “peer-to-peer assessment technique,” a common practice in the nuclear power industry. A delegation from one hospital visits another hospital in order to conduct a “structured, confidential, and non-punitive review” of the host institution’s safety and quality efforts. Without the threat of sanctions carried by regulators, these peer reviews can surface problems and suggest fixes, making the technique itself a vehicle for constructive copying among organizations.

Developmental psychologists are increasingly convinced that infants’ and children’s facility for imitation is what allows them to absorb so much, so quickly. So efficient is imitation as a method of learning, in fact, that roboticists are studying babies in order to understand how they pull off the trick of observing an adult and then doing as the grown-up does. Imagine if a robot could watch a human perform an action—say, place a silicon chip on a circuit board, or make a repair on a space capsule—and then replicate that movement itself. Elon Musk, founder of Tesla and SpaceX, has invested in research on just such “one-shot imitation learning.” But, as University of California, Berkeley, psychologist Alison Gopnik notes, the most sophisticated forms of artificial intelligence “are still far from being able to solve problems that human four-year-olds accomplish with ease.”

For example, young humans’ copying is unique in that children are quite selective about whom they choose to imitate. Even preschoolers prefer to imitate people who have shown themselves to be knowledgeable and competent. Research shows that while toddlers will choose to copy their mothers rather than a person they’ve just met, as children grow older they become increasingly willing to copy a stranger if the stranger appears to have special expertise. By the time a child reaches age seven, Mom no longer knows best.

At the same time, children are strikingly unselective about what they imitate—another way in which our practice of imitation departs from that of animals. Humans are “high-fidelity” copiers: our young imitate adults to the letter, while other animals will make do with a slapdash approximation. This difference can make apes, monkeys, and even dogs look like the smarter species. Shown a procedure with an extra, unnecessary step—like touching a box with one’s forehead before prying it open and retrieving the treat inside—chimps and canines will skip the superfluous move to go right for the goods. Children, however, will faithfully imitate every step.

There is sense behind this seemingly irrational behavior. Humans’ tendency to “overimitate”—to reproduce even the gratuitous elements of another’s behavior—may operate on a copy now, understand later basis. After all, there might be good reasons for such steps that the novice does not yet grasp, especially since so many hu...

Because our own culture discourages imitation, American children aren’t granted similar opportunities to show how competent they can be; they also lack exposure to inspiring examples, or “models,” of the kind of work that kids their age are capable of producing. For decades, educator Ron Berger towed around a rolling suitcase filled with hundreds of such models: sketches, poems, and essays created by children, which he would pull out and share with teachers and students at schools around the country.

he has found that many teachers and parents object to the use of models, afraid that it will suppress students’ creativity and originality.

In this context, the emulation of model texts is valued for its capacity to reduce cognitive load—especially important when students are juggling new concepts and vocabulary while also trying to construct a coherent argument in writing. Following the contours of a prototype provided by the instructor permits students to process more deeply the material they are expected to learn.

Students were being asked to employ a whole new vocabulary and a whole new suite of concepts, even as they were attempting to write in an unaccustomed style and an unaccustomed form. It was too much, and they had too few mental resources left over to actually learn.

His solution: at the start of the course, Smith provides students with several sample legal memorandums like those written by working lawyers. Guided by a set of instructions and targeted questions, students are expected to detail their responses to various aspects of the memos—thereby relieving them of the burden of having to produce their own memos even as they are laboring to learn what a memo looks and sounds like. Only after several such structured encounters with legal documents are students asked to author their own memorandums.

As Smith notes, the emulation of model texts was once a standard feature of instruction in legal writing; it fell out of favor because of concern that the practice would fail to foster a capacity for independent thinking. The careful observation of how students actually learn, informed by res...

Kenneth Koedinger, a professor at Carnegie Mellon University and the director of its Pittsburgh Science of Learning Center, estimates that experts are able to articulate only about 30 percent of what they know. His conclusion is based on research like the following: A study that asked expert trauma surgeons to describe how they insert a shunt into the femoral artery (the large blood vessel in the upper leg) reported that the surgeons neglected to cite nearly 70 percent of the actions they performed during the procedure. A study of expert experimental psychologists found that they omitted or inaccurately characterized an average of 75 percent of the steps they took when designing experiments and analyzing data. And a study of expert computer programmers revealed that they enumerated fewer than half of the tasks they actually carried out when debugging a computer program.

This can be accomplished through what philosopher Karsten Stueber calls “re-enactive empathy”: an appreciation of the challenges confronting the novice that is produced by reenacting what it was like to have once been a beginner oneself.

Math education expert John Mighton has a suggestion: break it down into steps, then break it down again—into micro-steps, if necessary.

Experts have another edge over novices: they know what to attend to and what to ignore. Presented with a professionally relevant scenario, experts will immediately home in on its most salient aspects, while beginners waste their time focusing on unimportant features. But research shows that the expertise of experienced practitioners can be made more accessible by deliberately exaggerating it, even distorting it, such that the pertinent elements “pop out” for the novice as they do for the expert.

A number of years ago, the US Air Force sought the advice of psychologist Itiel Dror, now a senior researcher at University College London. Endeavoring to prevent friendly fire aimed at their own aircraft, air force leaders were looking for ways to improve the ability of pilots in training to instantly recognize the shapes of various planes. Dror observed that the trainees were becoming overwhelmed with details about the many airplanes they were expected to identify. He took a new tack, digitally morphing the outlines of the aircraft diagrams the pilots were given to study. Planes with a wide wingspan became even wider; sharp-angled planes were made pointier; snub-nosed planes appeared more rounded. The differences among the aircraft, once subtle and hard to notice, now leaped out at the pilots—and they continued to recognize these distinguishing features even when they viewed the planes at normal scale.

A third difference between experts and novices lies in the way they categorize what they see: novices sort the entities they encounter according to their superficial features, while experts classify them according to their deep function.

The undergraduates sorted the problems according to their surface features: whether the problems involved springs, or pulleys, or inclined planes. The graduate students, meanwhile, categorized the problems on the basis of the underlying principles of physics they represented: conservation of energy, the work-energy theorem, conservation of momentum.

That notion was behind a new kind of wine store created by Joshua Wesson, an expert sommelier and entrepreneur who founded the chain Best Cellars. “I heard the same questions all the time, and they all reduced to, ‘How can I make sense of the world of wine without having to master all the details? How can I deal with all these choices when all I want is a wine that goes with pizza?’ ” says Wesson. In most stores selling wine, he observes, bottles are arranged by grape (Chardonnay, Cabernet Sauvignon) or by region (California, France). Such classifications communicate little to the uneducated wine consumer.

Wine experts, meanwhile, know about surface-level characteristics like grapes and regions—but they think about wine in terms of function: wines that are luscious and fruity, good for pairing with spicy food; wines that are big and bold and can stand up to a hearty meal; wines that are fizzy and festive, fit for a celebration. “Luscious,” “Big,” and “Fizzy” are, in fact, three of the eight categories Wesson devised for his stores (the others are “Soft,” “Fresh,” “Juicy,” “Smooth,” and “Sweet”). Foregrounding these features is like giving customers a shortcut to thinking the way a sommelier does. Following Wesson’s lead, experts more generally can make themselves into accessible models for imitation by communicating the categories they use to organize information, categories that are themselves packed with meaning about how experts’ thinking operates.

These strategies—breaking down agglomerated steps, exaggerating salient features, supplying categories based on function—help pry open the black box of...

Research has shown that, across disciplines, experts look in ways different from novices: they take in the big picture more rapidly and completely, while focusing on the most important aspects of the scene; they’re less distracted by visual “noise,” and they shift more easily among visual fields, avoiding getting stuck. Within any occupation—among surgeons, pilots, programmers, architects, even high school teachers—experts’ gaze patterns are highly similar, while beginners’ are widely divergent and idiosyncratic.

In a sense, these innovations represent a technologically enhanced take on the teaching that has unfolded within apprenticeships for centuries—the twenty-first-century version of a master craftsman’s pointing finger or guiding hand.

In an age of knowledge work, the “mystery” of expertise is even more enshrouded, hidden by the scrim of automatization. Pulling this curtain aside requires experts to forgo the familiar conventions of brainbound instruction—to think outside the brain, where their cognition can be seen.

A major factor in the grad students’ transformation, he concluded, was their experience of intense social engagement around a body of knowledge—the hours they spent advising, debating with, and recounting anecdotes to one another. A 2019 study published in the Proceedings of the National Academy of Sciences supports Wieman’s hunch. Tracking the intellectual advancement of several hundred graduate students in the sciences over the course of four years, its authors found that the development of crucial skills such as generating hypotheses, designing experiments, and analyzing data was closely related to the students’ engagement with their peers in the lab, and not to the guidance they received from their faculty mentors.

Wieman is working to achieve wider recognition of an often overlooked truth: the development of intelligent thinking is fundamentally a social process. We can engage in thinking on our own, of course, and at times solitary cognition is what’s called for by a particular problem or project. Even then, solo thinking is rooted in our lifelong experience of social interaction; linguists and cognitive scientists theorize that the constant patter we carry on in our heads is a kind of internalized conversation. Our brains evolved to think with people: to teach them, to argue with them, to exchange stories with them. Human thought is exquisitely sensitive to context, and one of the most powerful contexts of all is the presence of other people. As a consequence, when we think socially, we think differently—and often better—than when we think non-socially.

the brain stores social information differently than it stores information that is non-social. Social memories are encoded in a distinct region of the brain. What’s more, we remember social information more accurately, a phenomenon that psychologists call the “social encoding advantage.” If findings like this feel unexpected, that’s because our culture largely excludes social interaction from the realm of the intellect. Social exchanges with others might be enjoyable or entertaining, this attitude holds, but they’re no more than a diversion, what we do around the edges of school or work. Serious thinking, real thinking, is done on one’s own, sequestered from others.

And the ubiquitous depictions generated by fMRI in turn perpetuate that very orientation: the scans offer a vivid visual affirmation of the assumption that everything worth observing happens within the bounds of a single skull. Scientists who might have wished to investigate the role of social interaction on cognition have until recently been hampered by technical constraints; for many years following the introduction of fMRI, researchers were all but required to examine the individual in seclusion, shut inside the solitary bore of the MRI machine. Thus the neuroscientific study of how people think has been, for decades, the study of people thinking alone.

Technologies such as electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) are allowing scientists to scan multiple people’s brains as they interact in naturalistic settings—making deals, playing games, or simply talking to one another. Using these tools, researchers have found persuasive evidence for what is known as the “interactive brain hypothesis”: the premise that when people interact socially, their brains engage different neural and cognitive processes than when those same people are thinking or acting on their own.

In experiments that track brain activity while subjects are not reading or passively listening but actually talking to other people, a third and heretofore unknown language-related neural circuit has been identified. Studies using fNIRS—a brain-scanning technology that works via a flexible band encircling the head—demonstrate that this newly recognized network, called the subcentral area, is specialized for predicting and responding to language as it is used moment by moment in conversation. This discovery adds to accumulating evidence showing that engaging in real-time conversation involves much more nimble and nuanced cognitive processes than does the simple recognition of discrete words. It requires us to anticipate the language our conversational partner will use in speaking to us, and to improvise the language we ourselves will muster in response.

our society remains mired in a brainbound approach to cognition; our activities at school and at work still treat thinking as the manipulation of abstract symbols inside individual heads. We are asked to produce facts (on tests, in reports) without the presence of a person to edify. We make arguments (write essays, author memos) without the presence of a person to debate. We are asked to set information out (log entries in a knowledge management system), or take information in (read manuals and instructions), without the presence of a person with whom to trade stories.

toddlers under the age of two and a half readily learn new words and actions from a responsive adult but pick up almost nothing from prerecorded instruction delivered on a screen—a phenomenon that researchers call the “video deficit.” Humans learn best from other (live) humans. Perhaps more surprising, people learn from teaching other people—often more than the pupils themselves absorb. Consider this finding: firstborn children have an IQ that is on average 2.3 points higher than that of their younger brothers and sisters. After disconfirming several potential explanations, such as better nutrition or differential parental treatment, researchers concluded that firstborn children’s higher IQs stem from a simple fact of family life: older siblings engage in teaching younger ones.

students who learn information in preparation for teaching someone else review the material more intensively and organize it more thoroughly in their own minds than do students who are learning the same information in order to take a test. For social creatures like us, the prospect of engaging in an interpersonal interaction—with all of its potential for feeling admired or embarrassed—is far more motivating than the relatively anonymous activity of supplying written answers on an exam.

When fielding the pupil’s questions, and posing questions of her own, the tutor is obliged to adopt a “metacognitive” stance toward the material, consciously monitoring what her pupil knows and what she herself knows.

Hoogerheide has found, the act of teaching on video enhances the teacher’s own learning, improves her test performance, and enhances her ability to “transfer” the learned information to new situations. Writing out an explanation of the same material for an imagined tutee does not generate the same gains. Hoogerheide theorizes that teaching on camera generates persuasive feelings of “social presence”—the sense that there is someone watching and listening. Explaining oneself while being recorded, he notes, measurably increases the explainers’ physiological arousal—a state that is associated with enhanced memory, attention, and alertness.

Such outcomes may be due, in part, to the experience of what psychologists call “productive agency”: the sense that one’s own actions are affecting another person in a beneficial way. Actually seeing the fruits of one’s labor is especially gratifying; research finds that tutors learn more, and derive more motivation, from a tutoring session when they have the opportunity to watch their tutees answer questions about what they’ve learned.

Holly Chiu, an associate professor of business management at Brooklyn College, reported in a study published in 2018 that employees who engage in sharing job-related knowledge with their co-workers enlarge their own expertise in the bargain. By “systematically going through the knowledge, examining it, understanding it, integrating it and presenting it,” Chiu notes, these workers increased the depth and breadth of their knowledge, and subsequently turned in job performances that were rated more highly by their supervisors.

Cognitive scientists refer to stories as “psychologically privileged,” meaning they are granted special treatment by our brains. Compared to other informational formats, we attend to stories more closely. We understand them more readily. And we remember them more accurately. Research has found that we recall as much as 50 percent more information from stories than from expository passages.

One study found, for example, that a 1 percent reduction in efficiency, allowing time for “unstructured employee interaction,” produces a threefold increase in group performance over the long term. During such interaction, it may seem as though employees are simply exchanging gossip. “But what is gossip?” asks Sandy Pentland, a computational scientist and MIT professor who has conducted many studies demonstrating the benefits of workplace interaction. “Gossip is stories about what happened and what you did” in response. He adds: “If you think about what needs to happen for a healthy organization, people need to know the rules of the road. They need to know how things are done. Which means they have to hear the stories.”

Such nitty-gritty details constitute what psychologists call “tacit knowledge”: information about how things are done, when, and under which circumstances. It’s what gets left out of the depersonalized information employees encounter in more formal meetings and training sessions. It’s also where the “knowledge management systems” in which so many firms have invested go wrong: the information such systems make available is devoid of context, stripped of detail, and thereby rendered all but useless. “Much of the knowledge needed for employees to learn and thrive at work is not the kind of formal, codified information that is typically documented in online repositories or knowledge management systems,” Myers notes. “Instead, what is often critical for success is mastery of the tacit knowledge of the organization—the complex, often subtle interpretive knowledge that is difficult to capture or write down.”

At first, Hutchins observed, Richards reached for ways to spread the burden of the task across his own body and across the tools he had at hand. He “subvocally rehearsed” the numbers he was computing, repeating the digits under his breath—using his voice and his auditory sense to expand the capacity of his working memory. He traced the columns of numbers being added with his fingertip, using his hand to help keep track of the masses of information he was managing. With a pencil, he jotted down intermediate sums in the margin of the navigation chart, fixing in place a kind of “external memory,” in Hutchins’s phrase. And he pulled out a calculator, using it to relieve his brain of the burden of carrying out mathematical operations. Still, laboring on his own, Richards began to fall behind. He recruited yet one more resource: the mental ability of his teammate, Quartermaster Second Class Silver. The addition of another mind created a new challenge, however: how to figure out, on the fly, the best way to divide up the complex and fast-paced task.

Uncritical group thinking can lead to foolish and even disastrous decisions. But the limitations of excessive “cognitive individualism” are becoming increasingly clear as well. Individual cognition is simply not sufficient to meet the challenges of a world in which information is so abundant, expertise is so specialized, and issues are so complex.

Yet our schools and companies are increasingly doing just the opposite. Aided by technology, we are creating individual, asynchronous, atomized experiences for students and employees—from personalized “playlists” of academic lessons to go-at-your-own-pace online training modules. Then we wonder why our groups don’t cohere, why group work is often frustrating and disappointing, and why thinking with groups doesn’t extend our intelligence.

We then allocate more mental bandwidth to that material, processing it more deeply; in scientists’ terms, we award it “cognitive prioritization.” In a world of too much information, we use shared attention to help us figure out what to focus on, then direct our mental resources toward the object that the spotlight of shared attention has illuminated. As a result of these (mostly automatic) processes, we learn things better when we attend to them with other people. We remember things better when we attend to them with other people. And we’re more likely to act upon information that has been attended to along with other people.