LK: With so many messages coming in through Crowdmap, especially in a time of crisis, how could a news organization verify that the information is correct? And how well can Crowdmap filter these messages, so readers don’t suffer from information overload?
BH: Well, those are two separate issues. The verification is up to each individual deployment. Depending on the situation, it might be incredibly important that everything is 100 percent verified and accurate, and there are different ways of doing that. When someone submits an SMS report, the phone number is stored with the message, so you could call them back if you need to do the follow-up, or you can actually respond to the SMS directly from the Ushahidi system, so you can just type in a response, like “Please give us more information,” or, “Are you really there?” Then it’s really up to the administrators’ journalism skills to verify events that are happening.
But in a lot of cases, it’s really important to just get information out there. In the case of Haiti, there were people sending messages that they were stuck in a collapsed building or something. You don’t really have the time to send out a team and make sure the building is really collapsed; in that case it’s better to just get the message out that, “Hey, there’s probably somebody stuck in there, so if you’re over there, you should check it out.” So it really is deployment-specific.
Another cool thing about Ushahidi is that [an administrator] can approve a report [to be published], but not verify it. Then on the list of reports, on the far right, it will say whether it’s been verified, so you can make your own determination about whether you’re going to believe that report.
And the second question was how to handle the massive flow of data. Ushahidi is also developing a platform called SwiftRiver, and one of the betas [was released Monday]. It’s a suite of API tools for developers to use to help you manage your data. If there’s way too much data coming in, there’s a tool, for instance, that will determine which Twitter messages are re-tweets or duplicates. It also has some ability to determine which messages are probably important to look at. It doesn’t take the human out of the equation, but it filters things so that you can make your determinations much easier and faster.
LK: How does a piece of software determine which messages are important?
BH: It’s a combination of machine algorithms, looking at the location of whoever is submitting the message, people voting things up and down, like Digg. A combination of factors goes into creating what they call a “veracity score,” a score between zero and a hundred—one hundred meaning, “this is probably valid,” zero being “not valid.” We want to stay away from saying something is true or not, because just because a lot of people say something is true, doesn’t mean it is.
LK: The Washington Post’s “Snowmaggedon” website seems to be the first major news organization to take advantage of Ushahidi on the local level, in their case, to connect snowed-in DC residents to volunteers who could help shovel them out during a blizzard earlier this year. Do you happen to know the results of that experiment? I see about 320 problem reports on the site, but I can’t tell how many were actually resolved.
BH: I have no idea, but I do remember seeing a report that somebody ran out of beer. So I hope they got that taken care of….[laughs]
LK: Do you have any advice for local news sites who might want to use this software? Are there things that news organizations should be doing that maybe they aren’t thinking of?