Dirk Helbing is Professor of
Computational Social Science at the Swiss Federal Institute of
Technology (ETH) in Zurich and heads the FuturICT and Nervousnet
initiatives.
CAMBRIDGE – In game
theory, the “price of anarchy” describes how individuals acting in their
own self-interest within a larger system tend to reduce that larger
system’s efficiency. It is a ubiquitous phenomenon, one that almost all
of us confront, in some form, on a regular basis.
For example, if you
are a city planner in charge of traffic management, there are two ways
you can address traffic flows in your city. Generally, a centralized,
top-down approach – one that comprehends the entire system, identifies
choke points, and makes changes to eliminate them – will be more
efficient than simply letting individual drivers make their own choices
on the road, with the assumption that these choices, in aggregate, will
lead to an acceptable outcome. The first approach reduces the cost of
anarchy and makes better use of all available information.
The world today is awash in data. In 2015, mankind
produced
as much information as was created in all previous years of human
civilization. Every time we send a message, make a call, or complete a
transaction, we leave digital traces. We are quickly approaching what
Italian writer Italo Calvino presciently called the “memory of the
world”: a full digital copy of our physical universe.
As the Internet
expands into new realms of physical space through the Internet of
Things, the price of anarchy will become a crucial metric in our
society, and the temptation to eliminate it with the power of big data
analytics will grow stronger.
Examples of this
abound. Consider the familiar act of buying a book online through
Amazon. Amazon has a mountain of information about all of its users –
from their profiles to their search histories to the sentences they
highlight in e-books – which it uses to predict what they might want to
buy next. As in all forms of centralized artificial intelligence, past
patterns are used to forecast future ones. Amazon can look at the last
ten books you purchased and, with increasing accuracy, suggest what you
might want to read next.
But here we should
consider what is lost when we reduce the level of anarchy. The most
meaningful book you should read after those previous ten is not one that
fits neatly into an established pattern, but rather one that surprises
or challenges you to look at the world in a different way.
Contrary to the
traffic-flow scenario described above, optimized suggestions – which
often amount to a self-fulfilling prophecy of your next purchase – might
not be the best paradigm for online book browsing. Big data can
multiply our options while filtering out things we don’t want to see,
but there is something to be said for discovering that 11th book through
pure serendipity.
What is true of book
buying is also true for many other systems that are being digitized,
such as our cities and societies. Centralized municipal systems now use
algorithms to monitor urban infrastructure, from traffic lights and
subway use, to waste disposal and energy delivery. Many mayors worldwide
are fascinated by the idea of a central control room, such as Rio de
Janeiro’s IBM-designed operations center, where city managers can
respond to new information in real time.
