From the issue dated February 14,
2003
http://chronicle.com/weekly/v49/i23/23b00701.htm
|
Unraveling the Mysteries of the Connected AgeBy DUNCAN J.
WATTS
What is it about complex, connected
systems that makes them so hard to understand? How is it that assembling a large
collection of components into a system results in something altogether
different from a large collection of components? How do populations of fireflies
flashing, crickets chirping, or pacemaker cells beating all manage to
synchronize their rhythms without the aid of a central conductor? How do small
outbreaks of disease become epidemics, or new ideas become crazes? How do wild
speculative bubbles emerge out of the investment strategies of otherwise
sensible individuals, and when they burst, how does their damage spread
throughout the financial system? How vulnerable are large infrastructure
networks like the power grid or the Internet to random failures, or even
deliberate attack? And do norms and conventions evolve and sustain themselves
-- or alternatively get replaced -- in human societies?
As
different as all these questions appear, they are all versions of the same
question -- how does individual behavior aggregate to collective
behavior? As simply as it can be asked, this is one of the most fundamental
and pervasive questions in all of science. A human brain, for example, is in one
sense a trillion neurons connected in a big electrochemical lump. But to each of
us who has one, a brain is clearly much more, exhibiting properties like
consciousness, memory, and personality, whose nature cannot be explained simply
in terms of aggregations of neurons.
What makes the problem hard, and
what makes complex systems complex, is that the parts making up the whole don't
sum up in any simple fashion. Rather, they interact with each other, and in
interacting, even quite simple components can generate bewildering behavior. The
recent sequencing of the human genome revealed that the basic code of all human
life consists of only about 30,000 genes -- many fewer than anyone had
guessed. So whence comes all the complexity of human biology? Clearly it is not
from the complexity of the individual elements of the genome, which could
scarcely be any simpler; nor does it come from their number, which is barely any
greater than it is for the humblest of organisms. Rather it derives from the
simple fact that genetic traits are rarely expressed by single genes, but by
combinations.
What then of human systems? If the interactions of mere
genes can confound the best minds in biology, what hope do we have of
understanding combinations of far more complex components like people in a
society or companies in an economy? Surely the interactions of entities which
are themselves complex would produce complexity of a truly intractable kind.
Fortunately, as capricious, confusing, and unpredictable as individual humans
typically are, when you put many of them together, it is sometimes the case that
we can understand the basic organizing principles while ignoring many of the
complicating details.
Sometimes, therefore, the interactions of
individuals in a large system can generate greater complexity than the
individuals themselves display, and sometimes much less. Either way, the
particular manner in which they interact can have profound consequences for the
sorts of new phenomena that, from population genetics to global synchrony and
political revolutions, can emerge at the level of groups, systems, and
populations. It is one thing to say this, however, and quite another matter
altogether to understand it precisely. In particular, what is it about the
patterns of interactions between individuals in a large system that we should
pay attention to? No one has the answer yet, but in recent years a growing group
of researchers has been chasing a promising new lead. And out of this work,
which in itself builds upon decades of theory and experiment in every field from
physics to sociology, is coming a new science, the science of
networks.
In a way, nothing could be simpler than a network. Stripped to
its bare bones, a network is nothing more than a collection of objects connected
to each other in some fashion. On the other hand, the sheer generality of the
term network makes it slippery to pin down precisely, and this is one
reason why a science of networks is an important undertaking. We could be
talking about people in a network of friendships, or a large organization,
routers along the backbone of the Internet, or neurons firing in the brain. All
these systems are networks, but all are completely distinct in one sense or
another. By constructing a language for talking about networks that is precise
enough to say not only what a network is, but what kinds of different networks
there are in the world, the science of networks is lending the concept real
analytic power.
Understanding networks, however, is an extraordinarily
difficult task, not just because it is inherently complicated, but because it
requires different kinds of specialized knowledge that are usually segregated
according to academic specialty and even discipline. Physicists and
mathematicians have at their disposal mind-blowing analytical and computational
skills, but typically they don't spend a whole lot of time thinking about
individual behavior, institutional incentives, or cultural norms. Sociologists,
psychologists, and anthropologists, on the other hand, do. And in the past
half-century or so they have thought more deeply and carefully about the
relationship between networks and society than anyone else -- thinking
which is now turning out to be relevant to a surprising range of problems, from
biology to engineering. But lacking the glittering tools of their cousins in the
mathematical sciences, the social scientists have been more or less stalled on
their grand project for decades.
If it is to succeed, the new science of
networks must bring together from all the disciplines the relevant ideas, and
the people who understand them. The science of networks must become, in short, a
manifestation of its own subject matter, a network of scientists collectively
solving problems that cannot be solved by any single individual or even any
single discipline. It's a daunting task, made all the more awkward by the
longstanding barriers separating scientists themselves. Our languages are very
different, and we often have difficulty understanding one another. Our
approaches are different too, so each of us has to learn not only how the others
speak, but how they think. But it is happening, and the past few years
have seen an explosion in research and interest across the world in search of a
new paradigm with which to describe, explain, and ultimately understand the
networked world. We are not there yet, not by a long shot, but we are making
some exciting progress.
The Island of Manhattan. Twenty-two miles long
and less than five miles wide, it is, on the grand scale of the world, a speck,
a jewel in the mouth of the Hudson River as it pours into the North Atlantic. Up
close, it is more like a vast, roaring playground. Home to nearly a million
people and host to millions more every day, it is, and has been for more than a
century, Gotham, the quintessential metropolis, the city that never
sleeps.
But from a scientific point of view, it is something of an
enigma. Even on a daily basis, millions of people, along with the private and
commercial activity they generate, consume an awful lot of stuff -- food,
water, electricity, gas, and a vast range of materials from plastic wrapping to
steel girders and Italian fashion. They also discharge an enormous quantity of
waste in the form of garbage, recyclables, sewage, and wastewater; collectively
they emit so much raw heat energy, they create their own microclimate. Yet
almost nothing that the city requires in order to sustain itself is actually
produced, or even stored, within its own precincts; nor can it satisfy any of
its own disposal needs.
Another way to understand Manhattan, therefore,
is as a nexus of flows, the swirling convergence of people, resources, money,
and power. And if those flows stop, even temporarily, the city starts to die,
starved for nourishment or choking on its own excrement. New Yorkers are
renowned for their brash confidence, projecting an air of capability even in the
most trying circumstances. But really they are captives of the very systems that
make life in the city so convenient.
What would happen if this
infrastructure, or even part of it, were to stop functioning? Can it stop
functioning? And who is in a position to ensure that it doesn't? Who, in
other words, is in charge? Like many simple questions to do with complex
systems, this one lacks a definitive answer, but the short version is no
one. In reality, there is not even such a thing as a single infrastructure
to be in charge of. Rather what exists is a Byzantine mishmash of overlapping
networks, organizations, systems, and governance structures, mixing private and
public, economics, politics, and society.
No single entity coordinates
this bewilderingly complicated system, and no one understands it. Frankly, it is
a miracle that it works at all. If this is not a nerve-wracking thought, it
really should be. Complex, connected systems can sometimes display tremendous
robustness in the face of adversity and sometimes display shocking fragility.
And when the system is as complicated as a large, densely populated, heavily
built-up city, as vital to the lives of millions of people, and as central to
the economy of a global superpower, contemplating its potential break points is
more than idle speculation. So how robust is New York?
On September 11,
2001, we began to find out. The events of that day illustrate many of the
paradoxes encountered in the study of networks: how it is that connected systems
can be at once robust and fragile; how apparently distant events can be closer
than we think; how, at the same time, we can be insulated even from what is
happening nearby; and how the routine can prepare us for the exceptional. The
attacks of September 11 exposed, in a way that only true disasters can, the
hidden connections in the complex architecture of modern life. And from that
perspective, we still have some lessons to learn.
One important lesson
emerged from the severe organizational crisis that was precipitated by
what was essentially a physical attack. The mayor's emergency command bunker was
destroyed when Number 7 World Trade Center collapsed soon after the twin towers,
and by 10 a.m. the nearby police command center had lost every single phone
line, along with its cellular-phone, e-mail, and pager service. Faced with a
completely unexpected and unprecedented catastrophe, with almost no reliable
information available, and with the threat of subsequent attacks looming large,
the city needed to coordinate two enormous operations -- one rescue, and
the other security -- simultaneously. And less than an hour after the
emergency began, the very infrastructure that had been designed to manage
emergencies had been thrown into disarray.
But somehow they did it. In
what was, under the circumstances, an incredibly orderly response, the mayor's
office, the police and fire departments, the Port Authority, the various state
and federal emergency agencies, dozens of hospitals, hundreds of businesses, and
thousands of volunteers, turned lower Manhattan from a war zone into a recovery
site in less than 24 hours. In the rest of the city, meanwhile, everything
continued to operate in a way that was so normal, it was eerie. The power was
still on, the trains still ran, and up at Columbia, you could still go and have
a nice lunch at one of the restaurants on Broadway. For all the lockdown
security on the island that day, nearly everybody outside the immediately
devastated area got home that night, and deliveries of supplies and collection
of garbage resumed almost as normal the next day.
A few months after
September 11, I heard a remarkable story told by a woman from Cantor Fitzgerald
-- the debt-trading firm that lost 700 of its 1,000 employees in the
collapse of the south tower. Despite (or perhaps because of) the unfathomable
trauma they had just suffered, the remaining employees decided by the next day
that they would try to keep the firm alive -- a decision made all the more
incredible by the daunting practical hurdles they needed to overcome. First,
unlike the equity markets, the debt markets were not based at the Stock Exchange
and had not closed. So if it was to survive, Cantor Fitzgerald needed to be up
and running within the next 48 hours. Second, while their carefully constructed
contingency plan had called for remote backups of all their computer and data
systems, there was one eventuality they had not anticipated: Every single
person who knew the passwords had been lost. And the reality is that if no
one knows the passwords, the data are as good as gone, at least on the time
scale of two days.
So what they did was this: They sat around in a group
and recalled everything they knew about their colleagues, everything they had
done, everywhere they had been, and everything that had ever happened between
them. And they managed to guess the passwords. This story is a little
hard to believe, but it is true. And it illustrates, in a particularly dramatic
way, that recovery from a disaster is not something that can be planned for in
an event-specific manner; nor can it be centrally coordinated at the time of the
disaster itself. Just as with the mayor's office, in a true disaster, the center
is the first part of the system to be overwhelmed. The system's survival
therefore depends on a distributed network of pre-existing ties and ordinary
routines that binds an organization together across all its scales.
What
was really so remarkable about the robustness of downtown New York was that the
survival and recovery mechanisms used by people, companies, and agencies alike
were not remarkable at all. In the immediate aftermath, nobody knew what was
going on, and nobody knew how they were supposed to respond. So they did the
only thing they could do: They followed their routines, and adapted them as best
they could to allow for the dramatically altered circumstances. From an
organizational perspective, therefore, what we should learn from the recovery
effort is that the exceptional is really all about the routine.
What can
the science of networks tell us about the properties of complex systems, and
especially their strengths and weaknesses? The honest answer, unfortunately, is
not too much -- yet. It is important to recognize that, despite 50 years of
percolating in the background, the science of networks is only just getting off
the ground. If this were structural engineering, we would still be working out
the rules of mechanics -- the basic equations governing the bending,
stretching, and breaking of solids. The vast storehouse of applied knowledge to
which professional engineers have access -- the tables, handbooks,
computer-design packages, and heavily tested rules of thumb -- are at best
on the distant horizon. But what the science of networks can do is give us a new
way of thinking about familiar problems -- a way that has already yielded
some surprising insights.
First, the science of networks has taught us
that distance can be deceiving. The first evidence in support of this
observation came in the late 1960s in the form of a remarkable experiment
conducted by the social psychologist Stanley Milgram. Milgram devised an
innovative message-passing technique in which he gave a few hundred randomly
selected people from Boston and Omaha letters to be sent to a single target
person -- a stockbroker who worked in Boston. But the letters came with an
unusual stipulation: They could only be sent to a personal friend, preferably
one "closer" to the target than the current holder. Each subsequent recipient
received the same instructions, thereby forcing the letters to traverse a chain
of social acquaintances from initial sender to target. Milgram's question was,
how many people would be in a typical chain? The answer was six -- a
surprising result that led to the famous phrase (and John Guare's 1990 play)
"Six Degrees of Separation."
That someone on the other side of the world,
with little in common with you, can be reached through a short chain of network
ties -- through only six degrees -- is an aspect of the social world
that has fascinated generation after generation. Now the science of networks
gives us an explanation in terms of the multidimensional nature of social
identity -- we tend to associate with people like ourselves, but we have
multiple, independent ways of being alike. And because we know not only who our
friends are, but also what kind of people they are, even very large networks can
be navigated in only a few links.
The second major insight we can gain
from the science of networks is that, in connected systems, cause and effect are
related in a complicated and often quite misleading way. Sometimes small shocks
can have major implications. Just as a single skier can unleash an avalanche in
the mountains, so too can influences that are initially small trigger, in just
the right network, a cascade of events that can propagate essentially without
bound. Other times even major shocks can be absorbed with remarkably little
disruption. In 1997, for example, a fire destroyed a key plant of the Toyota
company, halting the production of more than 15,000 cars a day and affecting
more than 200 companies whose job it is to supply Toyota with everything from
electronic components to seat covers. Without question, this was a first-class
catastrophe. But what happened next was every bit as dramatic as the disaster
itself. In an astonishing coordinated response, and with very little direct
oversight by Toyota, those same companies managed to reproduce -- in
several completely different ways -- the lost components, and did so within
three days of the fire. A week after that, the volume of cars rolling off the
production line was back at its pre-disaster level. Because Toyota managed to
escape the crisis relatively unscathed, the whole incident was largely
forgotten. But it could easily have failed, as could the next company faced with
a similar crisis. By accounting for the networks of connections between
individual decisions or events, we can see that predicting the future based on
previous outcomes -- even in situations that appear indistinguishable from
those in the past -- is an unreliable business.
Finally, by helping
us to understand better the relationship between cause and effect that pertains
to complex, connected systems, the science of networks teaches us a third
lesson: that such systems, from power grids to businesses, and even entire
economies, are both more vulnerable and more robust than populations of
isolated entities. Networks share resources and distribute loads, but they also
spread disease and transmit failure -- they are both good and bad. But
unless we can understand exactly how connected systems are connected, we
cannot predict how they will behave. And unless we know what kind of behavior we
are trying to understand, we don't even know what it is about the network that
is supposed to matter. In this manner, the science of networks may not only
provide deep theoretical insight, but also yield practical solutions to
currently intractable problems.
Duncan J. Watts is an assistant
professor of sociology at Columbia University and an external faculty member of
the Santa Fe Institute. This essay is adapted from Six Degrees: The Science
of a Connected Age, to be published this month by W.W. Norton &
Co.
http://chronicle.com
Section: The Chronicle Review
Volume
49, Issue 23, Page B7
|
Copyright ©
2003 by The Chronicle of Higher Education