Emergence

by Steven Johnson

  Negative feedback, then, is a way of reaching an equilibrium point despite unpredictable—and changing—external conditions. The “negativity” keeps the system in check, just as “positive feedback” propels other systems onward. A thermostat with no feedback simply pumps seventy-two-degree air into a room, regardless of the room’s temperature. An imaginary thermostat driven by positive feedback might evaluate the change in room temperature and follow that lead: if the thermostat noted that the room had grown warmer, it would start pumping hotter air, causing the room to grow even warmer, causing the device to pump hotter air. Next thing you know, the water in the goldfish bowl is boiling. Negative feedback, on the other hand, lets the system find the right balance, even in a changing environment. A cold front comes in, a window is opened, someone lights a fire—any of these things can happen, and yet the temperature remains constant. Instead of amplifying its own signal, the system regulates itself.

  *

  We’ve been wrestling with information as a medium for negative feedback ever since Norbert Wiener published Cybernetics in 1949, and Wiener himself had been thinking about the relationship between control and feedback since his war-related research of the early forties. After the Japanese bombed Pearl Harbor and the United States joined the war in earnest, Wiener was asked by the army to figure out a way to train mechanical guns to fire automatically at their targets. The question Wiener found himself answering was this: Given enough information about the target’s location and movement, could you translate that data into something a machine could use to shoot a V-2 out of the sky?

  The problem was uniquely suited for the adaptability of negative feedback: the targets were a mixture of noise and information, somewhat predictable but also subject to sudden changes. But as it happened, to solve the problem Wiener also needed something that didn’t really exist yet: a digital computer capable of crunching the flow of data in real time. With that need in mind, Wiener helped build one of the first modern computers ever created. When the story is told of Wiener’s war years, the roots of the modern PC are usually emphasized, for legitimate reasons. But the new understanding of negative feedback that emerged from the ENIAC effort had equally far-reaching consequences, extending far beyond the vacuum tubes and punch cards of early computing.

  For negative feedback is not solely a software issue, or a device for your home furnace. It is a way of indirectly pushing a fluid, changeable system toward a goal. It is, in other words, a way of transforming a complex system into a complex adaptive system. Negative feedback comes in many shapes and sizes. You can build it into ballistic missiles or circuit boards, neurons or blood vessels. It is, in today’s terms, “platform agonistic.” At its most schematic, negative feedback entails comparing the current state of a system to the desired state, and pushing the system in a direction that minimizes the difference between the two states. As Wiener puts it near the outset of Cybernetics: “When we desire a motion to follow a given pattern, the difference between this pattern and the actually performed motion is used as a new input to cause the part regulated to move in such a way as to bring its motion closer to that given by the pattern.” Wiener gave that knack for self-regulation a name: homeostasis.

  Your body is a massively complex homeostatic system, using an intricate network of feedback mechanisms to keep itself stable in the midst of dynamically changing situations. Many of those feedback mechanisms are maintained by the brain, which coordinates external stimuli received by sensory organs and responds by triggering appropriate bodily actions. Our sleep cycles, for instance, depend heavily on negative feedback. The body’s circadian rhythms—accumulated after millions of years of life on a planet with a twenty-four-hour day—flow out of the central nervous system, triggering regular changes in urine formation, body temperature, cardiac output, oxygen consumption, cell division, and the secretions of endocrine glands. But for some reason, our body clocks are set a little slow: the human circadian cycle is twenty-five hours, and so we rely on the external world to reset our clock every day, both by detecting patterns of light and darkness, and by detecting the more subtle change in the earth’s magnetic field, which shifts as the planet rotates. Without that negative feedback pulling our circadian rhythms back into sync, we’d find ourselves sleeping through the day for two weeks out of every month. In other words, without that feedback mechanism, it would be as though the entire human race were permanently trapped in sophomore year of college.

  Understanding the body and the mind as a feedback-regulated homeostatic system has naturally encouraged some people to experiment with new forms of artificial feedback. Since the seventies, biofeedback devices have reported changes in adrenaline levels and muscle tension in real time to individuals wired up to special machines. The idea is to allow patients to manage their anxiety or stress level by letting them explore different mental states and instantly see the physiological effects. With a little bit of practice, biofeedback patients can easily “drive” their adrenaline levels up or down just by imagining stressful events, or reaching a meditative state. Our bodies, of course, are constantly adjusting adrenaline levels anyway—the difference with biofeedback is that the conscious mind enters into that feedback process, giving patients more direct control over the levels of the hormone in their system. That can be a means of better managing your body’s internal state, but it can also be a process of self-discovery. The one time I tried conventional biofeedback, my adrenaline levels hovered serenely at the middle of the range for the first five minutes of the session; the doctor actually complimented me on having such a normal and well-regulated adrenal system. And then, in the course of our conversation, I made a joke—and instantly my adrenaline levels shot off the charts. At the end of my visit, the therapist handed me a printout of the thirty-minute session, with my changing adrenaline levels plotted as a line graph. It was, for all intents and purposes, a computer graph of my attempts at humor over the preceding half hour: a flat line interrupted by six or seven dramatic spikes, each corresponding to a witticism that I had tossed out to the therapist.

  I walked away from the session without having improved myself in any noticeable way, and certainly I hadn’t achieved more control over my adrenaline levels. But I’d learned something nonetheless: that without consciously realizing it, I’d already established a simple feedback circuit for myself years ago, when my body had learned that it could give itself a targeted adrenaline rush by making a passing joke in conversation. I thought of all those office meetings or ostensibly serious conversations with friends where I had found myself compulsively making jokes, despite the inappropriate context; I thought of how deeply ingrained that impulse is in my day-to-day personality—and suddenly it seemed closer to a drug addiction than a personality trait, my brain scrambling to put together a cheap laugh to secure another adrenaline fix. In a real sense, our personalities are partially the sum of all these invisible feedback mechanisms; but to begin to understand those mechanisms, you need additional levels of feedback—in this case, a simple line graph plotted by an ordinary PC.

  If analyzing indirect data such as adrenaline levels can reveal so much about the mind’s ulterior motives, imagine the possibilities of analyzing the brain’s activity directly. That’s the idea behind the technology of neurofeedback: rather than measure the results of the brain’s actions, neurofeedback measures brain waves themselves and translates them into computer-generated images and sounds. Certain brain-wave patterns appear in moments of intense concentration; others in states of meditative calm; others in states of distraction, or fear. Neurofeedback—like so many of the systems we’ve seen—is simply a pattern amplification and recognition device: a series of EEG sensors applied to your skull registers changes in the patterns of your brain waves and transforms them into a medium that you can perceive directly, often in the form of audio tones or colors on a computer screen. As your brain drifts from one state to another, the tone or the image changes, giving you realtime feedback about your brain’s EEG act
ivity. With some practice, neurofeedback practitioners can more readily drive their brains toward specific states—because the neurofeedback technology supplies the brain with new data about its own patterns of behavior. Once you’ve reached a meditative state using neurofeedback, devotees claim, the traditional modes of meditation seem like parallel parking without a rearview mirror—with enough practice, you can pull it off, but you’re missing a lot of crucial information.

  *

  Were he alive today, I suspect Wiener would be surprised to find that biofeedback and neurofeedback technology are not yet mainstream therapeutic practices. But Wiener also recognized that homeostatis was not exclusively the province of individual human minds and bodies. If systems of neurons could form elaborate feedback mechanisms, why couldn’t larger human collectivities? “In connection with the effective amount of communal information,” Wiener wrote, “one of the most surprising facts about the body politic is its extreme lack of efficient homeostatic processes.” He would have diagnosed the pathology of Gennifer Flowers in a heartbeat. The Flowers episode was an instance of pure positive feedback, unchecked by its opposing force. Each agent’s behavior encouraged more like-minded behavior from other agents. There was nothing homeostatic about the process, only the “ever-widening gyre” of positive feedback.

  But if positive feedback causes such a ruckus in the media world, how can the brain rely so heavily on the reverberating circuits of neurons? One answer is a familiar term from today’s media: fatigue. Every neuron in the brain suffers from a kind of regulated impotence: after firing for a stretch, the cell must go through a few milliseconds of inaction, the “absolute refractory period,” during which it is immune to outside stimulation. Along with many other ingenious inhibiting schemes that the brain relies on, fatigue is a way of shorting out the reverberating circuit, keeping the brain’s feeding frenzy in check.

  It is this short circuit that is lacking in the modern media’s vast interconnectedness. Stories generate more stories, which generate stories about the coverage of the stories, which generate coverage about the meta-coverage. (Here the brain science seems wonderfully poetic: What better diagnosis for the 24/7 vertigo of media feedback than “lack of fatigue”?) A brain that can’t stop reverberating is one way of describing what happens during an epileptic fit; the media version is something like Steven Brill’s epic critique of the Lewinsky coverage in the first issue of Content: a high-profile media critic launching a new magazine with a high-profile indictment of the media’s obsession with its own reporting. If the problem stemmed from errors of judgment made by individual reporters, then a media critique might make sense. But since the problem lies in the media’s own tendency for self-amplification, it only makes the problem worse to cover the coverage. It’s like firing a pistol in the air to stop a fusillade. Once again, the Flowers affair illustrates the principle: the story wasn’t “real news”—according to the network wise men—until other outlets started covering it. The newsworthiness of a given story can’t be judged by the play the story is getting on the other channels. Otherwise the gravitational pull of positive feedback becomes too strong, and the loop starts driving the process, more than the reporters or the event itself.

  It’s not overstating things to say that the story that emerged from this loop was a milestone in American history. It’s entirely possible that the Flowers controversy would have subsided had Clinton’s answer to Jim Wooten been ignored; the Clintons would never have gone on 60 Minutes, and a whole series of tropes that appeared around the couple (Clinton’s philandering, Hillary’s anti–Tammy Wynette feminism) might never have found their way into the public mind. Without Gennifer Flowers in Clinton’s past, would the Monica Lewinsky affair have played out the same way? Probably not. And if that’s the case, then we must ask: What really brought this chain of events about? On the one hand, the answer is simple: individual life choices made by individual people—Clinton’s decision to have an affair, and to break it off, Flowers’s decision to go public, Clinton’s decision to answer the question—result in a chain of events that eventually stirs up an international news story. But there is another sine qua non here, which is the decision made several years before, somewhere in an office complex in Atlanta, to share the entire CNN news feed with local affiliates. That decision was not quite a “pseudo event,” in Boorstin’s famous phrase. It was a “system event”: a change in the way information flowed through the overall news system. But it was a material change nonetheless.

  If you think that Clinton’s remarks on Gennifer Flowers should never have been a story, then who are the culprits? Whom do we blame in such a setting? The traditional critiques don’t apply here: there’s no oak-paneled, cigar-smoke-filled back room where the puppeteers pull their invisible strings; it’s not that the television medium is particularly “hot” or “cold”; there was a profit motive behind CNN’s decision to share more footage, but we certainly can’t write off the Flowers episode as just another tribute to the greed of the network execs. Once again, we return to the fundamental laws of emergence: the behavior of individual agents is less important than the overall system. In earlier times, the channels that connected politicians, journalists, and ordinary citizens were one-way and hierarchical; they lacked the connections to generate true feedback; and too few agents were interacting to create any higher-level order. But the cable explosion of the eighties changed all that. For the first time, the system started to reverberate on its own. The sound was quiet during those initial years and may not have crossed into an audible range until Jim Wooten asked that question. And yet anyone who caught the nightly news on January 24, 1992, picked up its signal loud and clear.

  Still, the top-heavy structure of mass media may keep those loops relatively muted for the foreseeable future, at least where the tube is concerned. Feedback, after all, is usually not a television thing. You need the Web to hear it wail.

  *

  In June of 1962, a full year after the appearance of The Death and Life of Great American Cities, Lewis Mumford published a scathing critique of Jane Jacobs’s manifesto in his legendary New Yorker column, “The Sky Line.” In her prescriptions for a sidewalk-centric urban renewal, “Mother Jacobs”—as Mumford derisively called her—offered a “homemade poultice for the cure of cancer.” The New Yorker critic had been an early advocate of Jacobs’s work, encouraging her to translate her thoughts into a book while she was a junior editor at Architecture Forum in the midfifties. But the book she eventually wrote attacked Mumford’s much-beloved Ebenezer Howard and his “garden cities,” and so Mumford struck back at his onetime protégé with full fury.

  At over ten thousand words, Mumford’s critique was extensive and wide-ranging, but the central message came down to the potential of metropolitan centers to self-regulate. Jacobs had argued that large cities can achieve a kind of homeostasis through the interactions generated by lively sidewalks; urban planning that attempted to keep people off the streets was effectively destroying the lifeblood of the urban system. Without the open, feedback-heavy connections of street culture, cities quickly became dangerous and anarchic places. Building a city without sidewalks, Jacobs argued, was like building a brain without axons or dendrites. A city without connections was no city at all, at least in the traditional sense of organic city life. Better to build cities that encouraged the feedback loops of sidewalk traffic by shortening the length of blocks and supporting mixed-use zoning.

  Mumford was no fan of the housing projects of the postwar era, but he had lost faith in the self-regulatory powers of massive urban systems. Cities with populations in the millions simply put too much stress on the natural homeostatic tendencies of human collectives. In The City in History, published around the same period, Mumford had looked back at the Greek city-states, and their penchant for founding new units once the original community reached a certain size—the urban equivalent of reproducing by spores. His attachment to Ebenezer Howard also stemmed from the same lack of confidence in metro
politan self-regulation: the Garden City movement—not entirely unlike the New Urbanist movement of today—was an attempt to provide the energy and dynamism of city life in smaller doses. The Italian hill towns of the Renaissance had achieved an ideal mix of density and diversity while keeping their overall population within reasonable bounds (reined in largely by the walls that surrounded them). These were street-centric spaces with a vibrant public culture, but they remained knowable communities too: small enough to foster a real sense of civic belonging. That kind of organic balance, Mumford argued, was impossible in a city of 5 million people, where the noise and congestion—the sensory overload of it all—drained out the “vitality” from the city streets. “Jacobs forgets that in organisms there is no tissue growth quite as ‘vital’ or ‘dynamic’ as cancer growths… . The author has forgotten the most essential characteristic of all organic growth—to maintain diversity and balance, the organism must not exceed the norm of its species. Any ecological association eventually reaches the ‘climax stage,’ beyond which growth without deterioration is not possible.”