Sat, Dec 25, 2010 - Page 9 News List

Software highlights what Facebook friends miss

Information overload has led an IBM lab to try to build software to make it easier to decide what material is worth our full, undivided attention

By Noam Cohen  /  NY Times News Service, CAMBRIDGE, MASSACHUSETTS

ILLUSTRATION: JUNE HSU

IBM researchers in a laboratory nestled next to the Massachusetts Institute of Technology have lately been forced to confront a limit on computing power: The human brain just isn’t processing data any faster.

It’s a shame, too, because the brain has never had more material to make sense of: More friendships, which come with more frequent updates, more bosses to report to, more news, more entertainment choices. And who knows when the next million-page WikiLeaks release may come along?

Overwhelmed by all the noise, some have simply chosen to block it out — to opt out, say, of social networks and microblog platforms like Twitter. Alternatively, others have hewn close to these social networks, counting on them to sort through all the information coming at us.

However, to be informed in the distributed world we live in, opting out isn’t really an option. For better or worse, we are watching a C-SPAN version of our lives trying to fast-forward to the good parts.

That is the task being squarely addressed last week at IBM’s Center for Social Software — a roughly 30-member lab that addresses “the modern-day challenges of collaborating across distributed, global enterprises.” The lab tries to use increasingly sophisticated computers to act as information advisers.

“I do think of computers as augmenting people, not replacing them,” said Irene Greif, the director of the research center. “We need help with the limits of the brain, but there are some things that our brains can do that computers can’t do.”

The researchers essentially create programs that find patterns (and outliers) in the “fire hose” of information. Once patterns are revealed, it becomes easier to decide who or what is worth our undivided attention.

A literal example of fast--forwarding through the clutter of government is a public Web site created by IBM researchers, Many Bills. Relying on sites that collect and format federal legislation, the IBM project uses textual analysis to summarize and display congressional bills as they move through the legislative process. This often reveals material that would seem to be unrelated to the business at hand (and probably was inserted as part of the bargaining process), like a provision on guns in a financial regulatory bill.

The added material isn’t necessarily a secret — someone had to insert it and often the dealmaking is quite public — but a computer trained to see patterns and to highlight in bright colors can often spot it instantaneously.

The creators of the program stress that all of its features are intended to allow people to dig deeper — the entire text is there to be read. In an academic paper on the project, its creators, Yannick Assogba, Irene Ros and Joan DiMicco, described other “open government” visualization techniques that seem to allow “only a few number of interpretations.”

By contrast, they wrote: “Our belief is that for citizens to become meaningfully engaged with government data, they need to be able to draw their own conclusions about it.”

Then there is the problem of searching for expertise, not necessarily Web page results. One tool for IBM employees, SaNDVis, does this by showing a Web of relationships around a search term to reveal who within IBM has expertise on a topic. It uses writings, meetings attended, personal profile information and previous work experience to map these connections with lines showing who is closest to whom.

This story has been viewed 986 times.
TOP top