In the aftermath of a major disaster, it’s hard for aid workers to know what’s happening on the ground, and to direct resources where they are needed most. That’s when text messaging and social media can help. By analyzing tweets and other snippets, it’s possible to see trends–say, where people are trapped, or where there are water shortages–and do something about them.
The issue is the analysis part, says Lukas Biewald, CEO of CrowdFlower, a San Francisco company that finds people online willing to do “micro tasks” (normally for commercial purposes). One, you’ve got a huge amount of data to sift through, and not a lot of time. And two, all the text might be in a language–or filled with local references–that you don’t understand. You need some way of crunching it quickly, using people who aren’t put off by colloquial or foreign terms.
Patrick Meier, director of social innovation at the Qatar Foundation’s Computing Research Institute, and a member of a group called the Digital Humanitarian Network, says crowdsourcing can help. Following last December’s Typhoon Pablo, in the Philippines, DHN identified 20,000 relevant tweets, and then called on CrowdFlower to find volunteers to make the first assessment. The groups identified, one, messages with links to photos and video, and, two, messages that referred to damage that could be geo-tagged. From about 100 tweets, the UN Office for the Coordination of Humanitarian Affairs (OCHA) could then build a map plotting damaged houses and bridges, flooding, and so on.
“The entire project was carried out in less than 20 hours after OCHA’s request. This would not have been possible without the use of CrowdFlower as the first major filter of the 20,000-plus tweets,” says Meier, who writes about the response here.
CrowdFlower was also involved after the Pakistan and Haiti earthquakes. Following the latter, in 2010, volunteers set up a mobile number, 4636, that anyone could text with information. The result was thousands of messages in Creole, a language most aid workers didn’t understand. CrowdFlower organized for the text to be translated via its network–and at a fraction of the cost, and time, of using professionals.
“People would log into us, tag it and then send it directly to the right aid worker, depending on whether it was a medical problem, or a water shortage, or more of an infrastructure issue,” says Biewald.
For example, one message read: “NOU AP CHACHE, [NAME] KI AN FONDRE NAN YNIVESITE WAYAL.” Which, translated, means: “WE ARE LOOKING [NAME] WHO GOT BURIED UNDER ROYAL UNIVERSITY.”
By using Haitians to translate the messages, CrowdFlower could also funnel some cash back to people who needed it. Biewald and Meier say the aim now is to formalize their systems, so that when the next emergency happen they can act as quickly as possible, analyzing the data, and finding groups of online assistants. Biewald also thinks paying people a little–rather than relying on volunteers–leads to better results.
“In these disaster situations, information is at a huge premium–it’s really hard to get information from the ground,” says Biewald. “The text message system allowed us to map where all the problems were, so you could see patterns of crime and water. They could cluster it and deploy not only people, but see in real time where they needed to make systemic improvements.”