March 9, 2000 Vol . 17, No. 5
By Marianne Meadahl
Tracking the social networks of HIV needle users in the Baltimore area recently presented a problem for Johns Hopkins University researchers. They wanted to analyse data on more than 4,000 individuals, but couldn't find a computer analysis program capable of handling that capacity.
So they turned to SFU communication professor Bill Richards (below) and graduate student Andrew Seary, both specialists in creating social network analysis software. The pair had designed a program that could analyse information on up to 2,000 individuals. Faced with the challenge, they successfully extended the program's capability to include 5,000 people.
That enabled the U.S. researchers to access quickly new and important analysis of data on drug user activity, including tracking 750,000 bar-coded needles used in their needle exchange program. "What used to take a week to do was soon being done in about an hour," says Seary, "and what was impossible last year is now easy."
Seary, a mathematician and computer programmer who has worked on various projects at SFU for two decades, is currently a graduate student in a special arrangements doctoral program spanning communication, math and physics. He has spent the past decade involved in network analysis research with Richards, who is well-known in the field and whose software has been used worldwide in a variety of applications.
Richards' initial software, developed in the late 1970s was known as NEGOPY. He and Seary created MultiNet in the early 1990s and are about to release a new version of the software. Richards estimates it is capable of increasing data size to include as many as 32,000 people. "There is currently no other software that can handle this much data in the way that MultiNet can," he says.
"We take advantage of the fact that, in most social networks, especially large ones, each member of the network is likely to have connections with only a small fraction (usually less than 5 per cent) of other members. The methods we're using enable researchers to deal with very large networks quickly and efficiently, and at various levels of detail," he adds. "MultiNet can handle many networks at the same time by keeping track of only what is essential in each one."
Richards plans to launch the new version of MultiNet at an international conference in Vancouver in April. The Sunbelt Social Network conference, being coordinated by Richards, will be held from April 13-16.
The Johns Hopkins researchers will be among more than 300 international experts from 24 countries who will present papers during the conference. Summaries can be found at the conference web site, www.sfu.ca/~insna
Richards says since 1970 the study of social networks has become
an international effort, with its own professional organizations,
text books, journals, research centres, training centres and computer
programs designed specifically to facilitate the analysis of structural
© Simon Fraser University, Media and Public Relations