***** To join INSNA, visit http://www.insna.org *****
A U Mass Amherst press release, courtesy of ACM TechNet
Like almost all other write-ups, they do not note that the great majority
of networks did Not reach the target.*
Indeed, when I replicated via email with my cousin Lloyd** in Calif a few
years ago, I gave up when I realized how few were reaching the target.
But then again I am not a computer scientist or a physicist!
*Thanks to Charles Kadushin for pointing this out.
** The dropouts missed a wonderful guy and a hunk who may be on The
Bachelor this year.
Barry Wellman Professor of Sociology NetLab Director
wellman at chass.utoronto.ca http://www.chass.utoronto.ca/~wellman
Centre for Urban & Community Studies University of Toronto
455 Spadina Avenue Toronto Canada M5S 2G8 fax:+1-416-978-7162
To network is to live; to live is to network
“Six Degrees of Separation” Theory Explained in New Algorithm by UMass
Sept. 6, 2005
Contact: Rachel Ehrenberg
AMHERST, Mass. – University of Massachusetts Amherst researchers have
invented a new algorithm that solves a network-searching conundrum that
has puzzled computer scientists and sociologists for years.
The scientists created an algorithm that helps explain the sociological
findings that led to the theory of “six degrees of separation,” and could
have broad implications for how networks are navigated, from improving
emergency response systems to preventing the spread of computer viruses.
Dubbed expected-value navigation, the algorithm describes an efficient way
of searching a particular class of networks and was presented by doctoral
student Ozgur S,ims,ek, and David Jensen, professor of computer science,
at the 19th International Joint Conference on Artificial Intelligence in
The algorithm is applicable to a number of networks say the researchers.
Ad-hoc wireless networks, peer-to-peer file sharing networks and the World
Wide Web are all systems that could benefit from more efficient
message-passing. The algorithm could work especially well with dynamic
systems such as ad-hoc wireless networks where the structure may change so
quickly that a centralized hub becomes obsolete.
The work was inspired by research pioneered in the late 1960s that focused
on navigating social networks, explains S,ims,ek. In a now famous study by
psychologists Milgram and Travers, individuals in Boston and Omaha, Neb.,
were asked to deliver a letter to a target person in Boston, but via an
unconventional route: the message had to be passed through a chain of
acquaintances. The people starting the chain had some basic information
about the target individual—including name, age and occupation—and were
asked to forward the letter to someone they knew on a first-name basis in
an effort to deliver it through as few intermediaries as possible. Of the
letters that reached the target, the median number of people in the
message-passing chain was a mere six.
“What came out of that study was that we are all connected,” says
S,ims,ek. But the findings also raised a number of questions about how we
are connected, she says. What are the properties of these networks and how
do people efficiently navigate them?
The social network exploited by Travers and Milgram isn’t a
straightforward, evenly patterned web. For one thing, network topology is
only known locally—individuals starting with the letter did not know the
target individual—and the network is decentralized—it didn’t use a formal
hub such as the post office. If navigating such a network is to
succeed—and tasks such as searching peer-to-peer file sharing systems or
the navigating the Web by jumping from link to link do just that—there
must be parts of the underlying structure that successfully guide the
search, argue Jensen and S,ims,ek.
Participants in the Travers and Milgram study who efficiently sent the
message probably acted intuitively by combining two human traits that
apply to computerized network-searching as well, say the researchers.
People tend to associate with people who are like themselves, and some
individuals are more gregarious than others. “Searching” using both of
these factors, one can efficiently get to a target even when little is
known about the network’s structure.
The tendency of like to associate with like, or homophily, means that
attributes of a node—an individual in the Travers and Milgram study—tend
to be correlated. Bostonians often know other Bostonians, and the same
holds true for qualities such as age or occupation. The second important
characteristic of these networks is that some people have many more
acquaintances than others. This “degree disparity” leads to some
individuals acting as hubs.
Taking these factors into account simultaneously results in a searching
algorithm that gets messages to the target by passing it to gregarious
individuals who are most like the target. Or in the language of
network-searching, it favors nodes that maximize the probability of
linking directly to the target, which is a function of both degree and
homophily, say the scientists.
Previous research had explored these aspects separately, but S,ims,ek and
Jensen are the first to step back and incorporate both these qualities
into one broadly applicable algorithm with a strong basis in probability
theory. And the combination yields a powerful punch. It is remarkably
efficient at finding the short paths between nodes without knowing the
central network’s structure, say the researchers
“In this case, one plus one is more than two,” says S,ims,ek.
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.