"When Katrina struck New Orleans on Monday, August 29, notices began to appear online—on craigslist, on Yahoo, on the Red Cross Web site. But that information was too scattered to be useful. The queries were everywhere—on dozens of sites. So it was almost impossible to report that someone had been located and be guaranteed that the information would reach the people who needed it. On Saturday, September 3, as the catastrophe worsened, a handful of tech-savvy volunteers led by David Geilhufe started gathering data from these sites by "screen scraping," an automated process that involves grabbing the relevant information for each person—name, location, age, and description—and depositing it in a single database. Geilhufe and his team concocted a standardized method of organizing the data, which they called PeopleFinder Interchange Format.
Still, there were thousands of missing-person notices online the next day that hadn't been converted into the PeopleFinder format because they were not machine readable. Typically these messages were quite simple: for example, "I'm looking for my uncle John who lived in the Ninth Ward—Sarah Bowen." That requires a human to parse the syntax. So that morning two well-regarded online figures—Jon Lebkowsky and Ethan Zuckerman—decided to team up to coordinate a volunteer effort, with Lebkowsky recruiting people to scan through all the online posts and Zuckerman dealing out chunks of data to be analyzed.
By the next morning, PeopleFinder had attracted the attention of a few widely read bloggers. They spread the word to their readers about the need for volunteers. By the end of the day, thousands were volunteering. The group was briefly hamstrung by the database's being overloaded with activity. But by Tuesday night 50,000 entries had been processed, and the number continued to rise dramatically in the days that followed. Meanwhile, people looking for relatives or friends could visit www. katrinalist.net, where the PeopleFinder team provided a search tool that made it possible to enter a name, a zip code, or an address and get a list of names matching the query within seconds.
PeopleFinder was the kind of data-management effort that could have taken a year to execute at great expense if a corporation or a government agency had been in charge of it. The PeopleFinder group managed to pull it off in four days for zero dollars.
"The goal was not to overengineer our tools for the data-entry effort," Zuckerman says, "but to build something very quickly that would let people lend a hand. The solution we came up with was adequate to let 3,000 people participate. And 3,000 people, lightly coordinated, can do impressive things." Zuckerman is reluctant to compare the speed of PeopleFinder with the slow government response to Katrina. "We weren't pulling people from toxic waters," he says. But he does think that the decentralized nature of PeopleFinder has advantages: "One of the reasons PeopleFinder was deployed so quickly is that we had no one to answer to. When no one is approving your work, it's an invitation to solve problems in whatever way you want. I suspect there are many folks involved with the government response who wished they'd had that much flexibility and freedom."
There were other decentralized responses to Katrina. Barely 48 hours after the hurricane hit, two Web designers in Utah launched a site called www. katrinahousing.org to connect evacuees with people all across the country who had a spare bedroom or a guest cottage or even a foldout couch. Two weeks later, 5,000 people had found temporary homes through the site."