Messenger taps social nets

By Kimberly Patch, Technology Research News

If your friends and colleagues don't know the answer to a given question, they often know of a better person to ask. Several teams of researchers are looking to the fast, easy communication of the Internet in order to leverage these social networks.

Researchers from the University of Michigan have brought the possibilities a step forward with their Small-World Instant Messaging System (SWIM), which extends instant messaging systems by identifying expertise and routing queries accordingly.

The Small-World Instant Messaging System aims to efficiently tap this expanded network of friends-of-friends, said Jun Zhang, a researcher at the University of Michigan. In trying to use instant messaging and email to get answers from friends and colleagues, Zhang noticed that many times a friend or colleague would defer to one of their friends or colleagues to answer the question. "So I thought why not build a new system to automate this process," he said.

The system is designed to make it easier to get information that is complicated, too new to be part of an organizational knowledge base, or too valuable for its owners to make it public, said Zhang. The system consists of instant messaging software and a pair of advanced functions that support the social network-based search process, said Zhang.

First, the program maintains a much more complicated user profile than most instant messengers, said Zhang. "Besides letting a user input his expertise and interests manually, SWIM can automatically mine a user's homepage and browser bookmarks to construct a keyword vector to represent the user's information identity."

Second, the program contains a referral agent that automatically handles the information-querying process, said Zhang.

To search for information, a user sends a question to his own referral agent, which broadcasts the query to all of the user's buddies' agents, said Zhang. A referral agent in the buddy's messenger searches its information identity profile to see if that person is likely to be able to answer the question. If not, the agent either returns empty results or forwards the query to its buddies, depending on how the user has set the software.

When a likely match is found, that person sees the question and the path the query traveled, said Zhang. This friend-of-a-friend or friend-of-a-friend-of-a-friend of the questioner "can either start chatting immediately, or discuss the questions... later if the answering person prefers not to be disturbed at that time," he said.

There are four challenges to be met in developing this type of software, said Zhang: identifying people's expertise or knowledge from their electronic presence, matching questions to the right people, motivating people to help others, and protecting privacy and avoiding problems with spam.

The system addresses a very timely issue: quantifying the value of informal social knowledge, said Jon Kleinberg, an associate professor of computer science at Cornell University.

There are two sides to this, said Kleinberg. "One is the issue of how to find people who have the relevant information." The researchers address this by building navigational techniques into their system, he said. Combining searching with social networks is an interesting idea, he added.

The more difficult issue is how to get people to participate, Kleinberg said. Friends help friends because they have a relationship, but once the network gets three layers out -- friends of friends of friends -- the person asking for help is a complete stranger to the helper.

In 2003, Columbia University sociologist Duncan Watts performed in experiment using email to test large-scale small-word social networks and found that many people were not motivated enough to participate. "The attrition rate tends to be high, and if the attrition rate is high at every step, then essentially you have this exponential decay in your ability to find faraway information," said Kleinberg. "And so you have to start asking about incentive mechanisms."

The Michigan researchers are showing that it's possible to build small-world navigation techniques into a working system in an interesting way and they are aware of the incentive issues, said Kleinberg. To assess whether the system will work, however, requires a large-scale experiment, he said.

The researchers are still developing and testing the tool, according to Zhang. They have a stand-alone program and implementation that works as a plug-in for existing instant messaging clients.

The widespread use of instant messengers in daily life and work environments provides the critical mass of people to make the scheme viable, said Zhang. Ultimately, the software is meant to become a human equivalent of the Google Internet search engine for answers to non-Boolean types of questions, he said.

The system is also part of a larger research project investigating how information flows influence productivity, said Zhang. As part of that project, the researchers have discovered that social networks make individuals more efficient.

The researchers analyzed a large data set that included communications flows and surveys of people's perceptions of email and found strong statistical correlations between social network factors and individual output, said Zhang. "Searching information from humans directly is the most traditional way that people seek information," he said. "We should really re-examine the value of this method and try to use new technology to promote it."

Zhang's research colleague was Marshall Van Alstyne. The researchers presented the work at the Association of Computing Machinery (ACM) Computer-Human Interaction (CHI) 2004 in Vienna, Austria April 24 to 29. The research was funded by the National Science Foundation (NSF) and Intel Corporation.

Timeline:  > 2 years
Funding:   Corporate; Government
TRN Categories:  Databases and Information Retrieval; Internet
Story Type:   News
Related Elements:  Technical paper, "SWIM: Fostering Social Network Based Information Search," Computer-Human Interaction (CHI) 2004, April 24-29, Vienna, Austria




Advertisements:



July 14/21, 2004

Page One

Messenger taps social nets

Quantum crypto network debuts

Teleport lifts quantum computing

Probabilities ease genetic logic

Briefs:
Ultraviolet powers pixels
Nanorods gain gold tips
E-ink drawing pad closer to paper
Retinal display guides near-blind
Multi-projector system gets diverse
Laser tweezer traps nanotubes

News:

Research News Roundup
Research Watch blog

Features:
View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 



Ad links:
Buy an ad link

Advertisements:







Ad links: Clear History

Buy an ad link

 
Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN


© Copyright Technology Research News, LLC 2000-2006. All rights reserved.