For all the promise of user-generated content and contributions, one of the biggest challenges for journalism organisations is that such projects can quickly become victims of their own success. As contributions increase, there comes a point when you simply can’t evaluate or verify them all.
One of the most interesting projects in 2008 in terms of crowdsourcing was Ushahidi. Meaning “testimony” in Swahili, the platform was first developed to help citizen journalists in Kenya gather reports of violence in the wake of the contested election of late 2007. Out of that first project, it’s now been used to crowdsource information, often during elections or crises, around the world.
Considering the challenge of gathering information during a chaotic event like the attacks in Mumbai in November 2008, members of the Ushahidi developer community discussed how to meet the challenge of what they called a “hot flash event“.
It was that crisis that started two members of the Ushahidi dev community (Chris Blow and Kaushal Jhalla) thinking about what needs to be done when you have massive amounts of information flying around. We’re at that point where the barriers for any ordinary person sharing valuable tactical and strategic information openly is at hand. How do you ferret the good data from the bad?
They focused on the first three hours of a crisis. Any working journalist knows that often during fast moving news events false information is often reported as fact before being challenged. How do you increase the volume of sources while maintaining accuracy and also sifting through all of that information to find the information that is the most relevant and important?
Enter Swift River. The project is an “attempt to use both machine algorithms and crowdsourcing to verify incoming streams of information”. Scanning the project description, the Swift River application appears to allow people to create a bundle of RSS feeds, whether those feeds are users or hashtags on Twitter, blogs or mainstream media sources. Whoever creates the RSS bundle is the administrator, allowing them to add or delete sources. Users, referred to as sweepers, can then tag information or choose the bits of information in those RSS feeds that they ‘believe’. (I might quibble with the language. Belief isn’t verification.) Analysis is done of the links, and “veracity of links is computed”.
It’s a fascinating idea and a project that I will be watching. While Ushahidi is designed to crowdsource information and reports from people, Swift River is designed to ‘crowdsource the filter’ for reports across the several networks on the internet. For those of you interested, the project code is made available under the open-source MIT Licence.
One of the things that I really like about this project is that it’s drawing on talent and ideas from around the world, including some dynamic people I’ve had the good fortunte to meet. Last year when I was back in the US for the elections, I met Dave Troy of Twittervision fame who helped develop the an application to crowdsource reports of voting problems during the US elections last year, Twitter Vote Report. The project gained a lot of support including MTV’s Rock the Vote and National Public Radio. He has released the code for the Twitter Vote Report application on GitHub.
To help organise the Swift River project for Ushahidi, they have enlisted African tech investor, Jon Gosier of Appfrica Labs in Uganda. They have based Appfrica Labs loosely on Paul Graham’s Y Combinator. I interviewed Jon Gosier at TEDGlobal in Oxford this summer about a mobile phone search service in Uganda. He’s a Senior TED Fellow.
There are a lot of very interesting elements in this project. First off, they have highlighted a major issue with crowdsourced reporting: Current filters and methods of verification struggle as the amount of information increases. The issue is especially problematic in the chaotic hours after an event like the attacks in Mumbai.
I’m curious to see if there is a reputation system built into it. As they say, this works based on the participation of experts and non-experts. How do you gauge the expertise of a sweeper? And I don’t mean to imply as a journalist that I think that journalists are ‘experts’ by default. For instance, I know a lot about US politics but consider myself a novice when it comes to British politics.
It’s great to see people tackling these thorny issues and testing them in real world situations. I wonder if this type of filtering can also be used to surface and filter information for ongoing news stories and not just crises and breaking news. Filters are increasingly important as the volume of information increases. Building better filters is a noble and much needed task.