Somalia Speaks: Lessons From Novel Journalism

Ushahidi
Feb 1, 2012

[This blog post was officially co-authored with Al-jazeera] The first 72 hours of the SomaliaSpeaks deployment were particularly intense. The purpose of this joint write-up with Al Jazeera and partners is to share some of our early lessons learned in this novel collaboration.  Every deployment teaches us a multitude of lessons, so our partners at Al Jazeera, Souktel and Crowdflower joined us in this effort to share these. We look forward to future collaborations with them as we share this story with you. This purpose of this pilot project was to let Somalis speak for themselves. For the first time ever, a prominent news organization, Al-jazeera, used crowdsourcing and SMS to let thousands of Somalis express for themselves how the crisis has been effecting their daily lives. More than 4,000 text messages were received within just a few days. Of these, over 1,000 were translated from Somali into English by about 80 translators. The resulting map of Somali voices received over 25,000 page views. Before reviewing our lessons learned, we first wanted to thank K’naan and Sol for the initial inspiration behind this project. They got in touch with Ushahidi last year because they wanted to use the platform to help amplify Somali voices and show how capable the Somali people are. The initial version of this project was a prototype that was not activated. But thanks to Al- jazeera, Souktel and Crowdflower, we were able to revive the project to help amplify Somali voices in the international media. Despite being a pilot, the project exemplified valid use cases in the application of pervasive technologies such as the web and mobile phone to news and information gathering. Somalia being a country run down by decades of neglect and war as well being rife with insecurity  provides one of the most challenging operational environments. Within a very short time we were able to curate information traversing through different parts of Somalia. This would have proved futile if not overly expensive or impossible had it been done using traditional news gathering techniques. Furthermore the information collected provides more insight on the realities of life in Somalia. Projects like this involve a lot of effort and goodwill from the community and a chain is only as strong as its weakest link. Therefore, building a strong community around the project goes a long way to solving problems and mitigating challenges. The community goes beyond the volunteer translators. We saw lots of support from the Somali blogging community, technology and media enthusiasts as well as innovators or innovation centric minds across the globe who all narrated or reported the Somalia Speaks project in their own ways. Having a strong community and inculcating a community ethos in  project operations and goals goes a long way influencing success. Without the community backing and promoting this project, it would have only been another temporary spotlight on Somalia. The project worked as follows. Al-jazeera editorial selected the following question for interview:

“Al Jazeera would like to know -- how has the Somalia Conflict affected your life? Please also include the name of your hometown in the response. Thanks."

Our colleagues at Souktel distributed the question via text message to 5,000 of their SMS subscribers across Somalia. The responses were then forwarded from Souktel’s SMS platform to a customized Crowdflower micro-tasking platform. There, Somali-speaking volunteers translated and geo-located the text messages which were then manually uploaded to Al-jazeera’s Ushahidi platform.

There are three points worth highlighting in terms of lessons learned: 1. Messaging While the question that was posed via SMS in no way asked for individuals to reply with their personal names, a small number of responders still added their names; some even added their full names. So these were deleted as quickly as possible. (Note that the numbers posted in the title of initial reports were not phone numbers but an assigned sequential number generated by the Crowdflower plugin). In hindsight, the SMS sent out with the question should have specifically asked that responders not include personal identifiers in their SMS replies. 2. Volunteer translation While we had recruited a small number of trusted volunteers to translate the incoming text messages using a Crowdflower plugin, a decision was subsequently made to make the call for volunteers public to cope with the 2,500+ SMS replies received. This means that anonymous volunteers could see the original text messages, some of which initially  included personal identifiers. So we immediately reached out to Crowdflower for guidance to take the plug-in offline. We then began to manually delete several dozen text messages inside the Crowdflower plug-in that contained personal identifiers. Our colleagues at Al-jazeera took over this process and set up their own micro-tasking platform, removing all personal identifiers from the text messages awaiting translation and geo-location. 3. Security One of Ushahidi’s community members tested the platform and identified a search security issue on Friday, December 9th. We quickly fixed this on the deployment. And, we issued a security patch to all deployers. (http://security.ushahidi.com) In the future, for this type of “The People Speak” project,  we recommend taking the following steps:

On large multi-partner deployments: Global organizations work in multiple timezones, so communications plans need to include 24/7 points of contacts for each organization.

Text potential interviewees to ask whether they agree to be interviewed and to have their responses made public before sending out the main question.

Text those individuals who have consented to being interviewed with the desired question and ask them to include the name of their town but not their personal names.

Recruit trusted translation volunteers well in advance and ensure that the micro-tasking translation platform has no personal identifiers.

Stagger the launch of the text messages and the live map. That is, start with the SMS broadcast and spend however many days/weeks doing the bulk of the translation with vetted volunteers. The system that holds the raw text messages should obviously be fully secure. When the majority of text messages are processed, launch the live map and gradually add the already translated text messages to grow the map steadily over a period of days/weeks.

At Ushahidi, we’ve also made some plans to help all deployers in our community:

We're building a program for privacy and security education for our users in 2012. (Blog posts, webinars, videos and meet-ups.)

Best practices for security and privacy will be included as essential documentation on our soon to be re-launched wiki.

Ushahidi is open source and the community is a large part of what makes it work.  We’ll build a security working group focused on our software, but it’ll take your participation to make it work.

Alongside our partners within the CrisisMappers community, we will participate in a security and privacy working group. This field is growing and collective lessons can only improve each map action.

Crisis mapping and journalism are both in the nascent stages of collaborating on real-time news connecting diaspora and citizens alike. Al-Jazeera is leading the fray in testing and implementing live maps into their fast-moving news cycle toolkit. We are thankful for all their efforts and look forward to further collaboration.