Team Wins Key Insights - and Second Place - in DARPA Challenge
Posted December 11, 2009 | Atlanta, GA
A national competition aimed at quickly locating 10 red weather balloons tethered at locations across the United States has netted a second-place finish for a Georgia Tech team -- along with a set of new insights into the use of social networks for gathering information.
Sponsored by the U.S. Defense Advanced Research Projects Agency (DARPA), the DARPA Network Challenge attracted hundreds of teams to tackle the problem of how to locate the balloons, which were positioned Dec. 5 at locations ranging from San Francisco and Portland to Memphis and Miami.
A team from the Massachusetts Institute of Technology won the $40,000 prize for correctly locating all 10 balloons. A team led by researchers at the Georgia Tech Research Institute (GTRI) found nine of the 10 balloons during the nine-hour competition.
DARPAâ€™s interest in the competition was in assessing how social networks could be used to address massive information-gathering tasks. In addition to its research component, the challenge also marked the 40th anniversary of the ARPANET, the forerunner of todayâ€™s Internet.
GTRI researchers Erica Briscoe and Ethan Trewhitt began discussing the challenge in early November, and quickly organized a core team of seven co-workers. They established a Web site and began using Facebook and word-of-mouth communications to build a network that eventually included more than a thousand people pledged to help.
One of their initial decisions was that if they should win, the prize would be donated to the American Red Cross -- rather than being split among the team members and balloon spotters. Team members believe that was important to attracting altruistic volunteers.
â€œOne thing that surprised us was that many balloon reporters specifically chose our team because we had decided to donate the winnings,â€ said Betty Whitaker, a GTRI principal research scientist who helped coordinate the team. â€œWe pledged any winnings to charity to encourage recruitment and avoid complicated issues with money after the contest.â€
Another key was establishing the Web site â€œI Spy A Red Balloon,â€ which built a high ranking on Google thanks to references on established Web sites. That allowed the team to attract people who may have seen a red balloon on Dec. 5 and wondered what was going on.
â€œThough we focused on getting the word out to the public prior to launch day, our strong presence on that day made it possible for people who were unaware of the competition to find our team after running across a balloon,â€ explained Trewhitt, a GTRI research engineer.
The team also connected established networks and used the news media to get information out to potential balloon-spotters. Beyond those who pledged to help, thousands more people knew about the effort and would have made contact had they seen a balloon.
But as with popular social networking services, not everybody could be trusted.
â€œBecause teams were commonly infiltrated by members of competing teams, one of the toughest parts of this competition was not being able to trust any particular members of the group,â€ Trewhitt added. â€œThis led us to realize that trust in large groups is a tricky issue -- and a topic for future research.â€
On competition day, which began at 10 a.m. with balloons being raised in the 10 previously-undisclosed locations, team members searched Twitter and Facebook for news of balloon sightings. They called friends, family and local businesses to validate alleged sightings, and analyzed incoming photographs to spot fakes and confirm the location of authentic red balloons.
They also used a variety of tools, some of which they built, to help track sightings. Their Web site, for instance, used Google Maps to summarize reports.
Though the GTRI team didnâ€™t win the top prize, its leaders believe the effort established credibility and planted seeds for future research projects.
â€œWe would like to study issues of trust in large social networks, as well as how to extract and validate useful and correct information from un-moderated online media such as Twitter,â€ said Erica Briscoe, a GTRI research scientist. â€œTwitter is often the fastest medium for notification of real-time events because it is unfiltered and raw. It would be useful to research methods for determining the accuracy and authenticity of rumors in this type of environment.â€
The competition also showed how much could be done on a budget of just $200, which was what the â€œI Spy A Red Balloonâ€ team spent in total.
For its part, the agency also seemed pleased with what the teams had done.
â€œ[The DARPA Network] Challenge explores basic research issues such as mobilization, collaboration and trust in diverse social networking constructs, and could serve to fuel innovation across a wide spectrum of applications,â€ the agency said in a news release. â€œDARPA plans to meet with teams to review the approaches and strategies used to build networks, collect information and participate in the Challenge.â€
Beyond those already mentioned, the team also included Stephen Cuzzort, Jessica Pater, Rick Presley and Miles Thompson, all from the Georgia Tech Research Institute.
Research News & Publications Office
Georgia Institute of Technology
75 Fifth Street, N.W., Suite 314
Atlanta, Georgia 30308 USA
Writer: John Toon