COMPUTING SCIENCE
Connecting the Dots
Can the tools of graph theory and social-network studies unravel the next big plot?
Brian Hayes
Plotting the Plotters
What do the principles of social networks and graph theory tell us
about the structure of terrorist cells? The very word
"cell" offers a clue: It suggests compartmentalization.
And indeed the lore of spy rings and resistance fighters speaks of
limiting communication so that if one person is captured others will
not be put in jeopardy. At the same time, however, the members of
the group have to keep in touch in order to make plans and carry
them out.
An illuminating case study comes from a rather different context:
price-fixing by manufacturers of electrical equipment in the 1950s.
The social network of the colluding managers and executives was
examined by Wayne E. Baker of the University of Chicago and Robert
R. Faulkner of the University of Massachusetts. They found that
"the structure of illegal networks is driven primarily by the
need to maximize concealment, rather than the need to maximize
efficiency." Nevertheless, the price-fixing and bid-rigging
simply could not be accomplished without communication among the
conspirators, especially in the case of the biggest machinery.
Despite the risks, executives had to meet face-to-face to coordinate
their plans.
Networks of terrorists apparently face the same conflicting
imperatives. Valdis E. Krebs, a consultant who usually applies
social-network analysis to business problems, has used the same
tools to map relations among the September 11 hijackers. In a paper
written just a few weeks after the attacks, he found the network
surprisingly sparse. Although every hijacker could be connected to
every other via some path through the network, many of the
paths were quite long, passing through three or four intermediaries.
This attenuated structure would make communication extremely inefficient.
Krebs later revised his analysis, as more information became
available. He has posted a new map at the Web site
http://orgnet.com/prevent.html. Here he reaches a different
conclusion. Starting with two men who were already under suspicion
in January of 2000, Krebs finds that known linkages lead to all 19
hijackers, and to other conspirators as well. Each node of the
network is tied to the two initial subjects either directly or
through a single intermediary.
José A. Rodríguez of the University of Barcelona has
created a similar network map for the bombing of commuter trains in
Madrid on March 11, 2004. Rodríguez recorded several kinds of
strong links among the conspirators. Some had ties of kinship or had
been childhood friends; others congregated at a shop owned by two of
the subjects; some were veterans of earlier wars or terrorist
actions. Looking at just the 13 men who actually placed and
detonated the explosives, Rodríguez found that the strong
ties produced a somewhat strange network. A core of six people
formed a clique: Each one was linked to all the others. But the
remaining members were only loosely associated or were completely
disconnected from the main group.
The outlook changed entirely when Rodríguez included some 70
persons associated with the plot in various ways and when he mapped
weak ties as well as strong ones. The weak ties denote pairs of
people connected by financial transactions, casual encounters and
the like. This larger and fuller network looks much like what
Granovetter's theory would predict. There are several dense
clusters, within which most nodes are strongly connected, but the
clusters communicate with one another only via comparatively loose
and unreliable couplings. For example, one cluster is made up of
Spanish citizens from whom the bombers obtained explosives; most
paths from this subgroup to the rest of the network pass through a
single node, a vulnerable choke-point.
» Post Comment