On November 28th, 2011 crowds assembled at 62,000 polling stations to elect DR Congo’s fifth president in the country’s first independently administered presidential elections since independence in 1960.

Polling Station in Bukavu, Ibanda Commune

Technology had improved considerably since 2006 with most people having cell phones and greater access to wireless. On the other hand, the remote areas most sensitive to violence and election fraud are the ones without cell service and electricity. Thus, many of the same challenges remained. Knowing all these challenges, we still decided that deployment would be useful since many organizations are looking to crisis mapping to assist in combating the ongoing insecurity in eastern Congo and massive sexual violence. We could all learn from the experience. The site went live one week before the elections, on November 18th, reporting in French and English.

Organizing Networks and Teams

Through the international partners, a U.S. based diaspora group, RDC2XTE, seeking an alternative vision for DRC through supportive actions to improve accountability and advancement, and our own contacts, we created a network of local independent observers and media sources based in Bukavu and Goma in the eastern Congo and Kinshasa to continuously report back what they were witnessing at various polling stations.

One of our main dedicated partners in the east, the Institute for War and Peace Reporting, trained and organized local women journalists to SMS or email direct reports. A second dedicated partner, Collectif D’Actions Pour La Defense Des Droits Humains (CADDHOM), a local NGO focusing on a wide range of programs concerning human rights, was an official partner of the CENI (the National Election Commission) deploying 200 monitors in South Kivu. CADDHOM sent us reports via email and SMS as they received phone calls and SMS from their official field observers. I also accompanied the executive director of CADDOM, Pasteur Joseph on election day to observe the elections.

Outside EDAP Polling Station in Bukavu with CADDHOM Coordinator, Pasteur Joseph (right)

A third dedicated partner was Catholic Agency for Overseas Development (CAFOD), who organized to station a volunteer in Kinshasa to receive reports from the 30,0000 network of Catholic Bishop monitors in the field. Although in contact with the EU monitoring team, there were no actual exchanges of information and the Carter center did not respond to emails.

We provided all partners with training material and produced post card sized printouts with SMS instructions. I only arrived on the ground on November 25th, but used my time there to meet with local partners none of whom, even though they had received all my materials before hand, had actually visited the site. Internet access is simply too complicated to spend time trying to view a site that takes hours to access. On the other hand, once I arrived and installed the local SIM cards, the Android phones with SMSSync worked perfectly. In fact, since I put the SIM cards on roaming, they work even here in Chicago.

Election Day, SMS Blocked and Lessons Learned

Even with the late start, once the elections got underway we began to receive reports via SMS and email. But then a major glitch came when the government shut down the whole SMS network on December 4th due to concerns of impending violence with the announcement of the election results scheduled for December 6th. Our independent deployment hit a major roadblock between the block on SMS and evacuation of internationals from Kinshasa. I was scheduled to fly to Kinshasa on December 5th, but could not get there since flights were cancelled. All work was suspended in Kinshasa and people ordered to stay home. As CENI continued to delay announcing election results and tension mounted, the CAFOD volunteer could not send any reports. CENI finally announced Kabila’s victory on December 9th, but with growing unrest from the opposition, SMS was not restored until December 14. As soon as it was, we immediately began receiving reports from the volunteer, but these were limited to the December 13th press release that the EU confirmed the Archbishop of Kinshasa’s declaration of irregularities in the elections and media reports.

Speaking with many organizations on the ground, it became apparent that Ushahidi was not widely known in the DRC and, although I physically sat with locals and showed them the platform there was interest, no one I met with could view the site on their computers given the slow Internet connection. A local radio station in Goma. Radio Mutaani, had also set up an SMS system and gmail chat feeding into their website, but were weary of sharing data with Ushahidi given security concerns. I visited their radio station and attended a program at the University in Goma and was impressed with the level of engagement surrounding the election. Until SMS was cut off, they had a lot of activity through their gmail chat.

Without SMS the deployment shifted to media monitoring coordinated by undergraduate Liz Casano of the NorthWestern University Student team and Bharathi Ram and Leesa Astredo of the Stand-by Task Force. The SBTF was instrumental in sorting through and reporting on more than 20,000 tweets. So far, a total of 320 online reports have been uploaded from over 70 locations, comprising 55 types of incidents. There are over 150 reports still waiting to be input into the system, most likely with more to come as the team reviews additional media sources.

One of our goals going in to the deployment was to be able to provide a broad platform to link up the work of local and international monitors and amass all the reports in “real time” for use by all. We learned that this was not in the best interest of the official monitors. Given the strategic role of the National Episcopal Conference of the Congo, the country’s Catholic bishops’ council as the largest monitoring group, it was in their best interest to kept their reports secret until ready to make an official announcement which they did on December 12th. In the end, the need of human rights and humanitarian organizations to report on and respond to crisis, did not coincide with the needs of official monitoring efforts to postpone release of information until they could better assess and negotiate the political implications.

Although more time to conduct networking on the ground would have produced more buy in from humanitarian, civil and human rights organizations, it seems that in the end we might have had the same problems given the SMS and need of the largest observers to withhold their reports until the best strategic moment.

The project team is now shifting its focus to analysis. In the next three months our goal is to issue two reports on the elections. The first will examine the question of transparency that was central to the rejection of the election results by the opposition and critiques of the official monitors. The second will layer in additional data to seek to answer questions broader questions such as did NGO civics work result in reduced problems? Were their areas that had violence in 2006 but not this time? And what kinds of responses were most successful and why?

Voix des Kivus

A crowd-seeding system in Eastern Congo that uses cell phones to obtain high-quality, verifiable, and real-time information about events that take place in hard-to-reach areas. This pilot project is led by Peter van der Windt and Macartan Humphreys from the Center for the Study of Development Strategies at Columbia University.

The pilot

Atrocities in hard-to-reach areas – for example many areas in Eastern Congo – often go unnoticed because of the lack of accessibility, both due to poor infrastructure and to the simple fact that fighting makes it too dangerous to get close. The inability of international organizations and humanitarian NGOs to collect information under these conditions hampers the provision of assistance in a timely and effective manner.

There is fast growing recognition of the role that technology can play in addressing these problems. But a real challenge faced by many approaches is the difficulty of getting data that is not just real time, but representative. Columbia University (with support from USAID) began the Voix des Kivus pilot project in summer 2009 to assess the technical feasibility of a decentralized, representative, SMS-based information system in the region and to assess the utility of the program to participating communities and potential users. Presently (beginning 2011) the program is operating in a random sample of 18 villages from four territories of the war-torn province of South Kivu in the Democratic Republic of Congo.

Phoneholders and the goal

It works like this. In each village participating in Voix des Kivus there are three cell phone holders: one representing the traditional leadership, one representing women’s groups, and one elected by the community. Holders are trained extensively on how to send messages to the system. They are provided with a phone, monthly credit, and a codesheet that lists possible events that can take place in the village. Sending messages to the system is free but it is also voluntary – while users do not have to pay for each message they do not get any financial rewards for sending content to the system.

For participating communities Voix des Kivus provides a system for creating histories, archiving testimonies, and communicating with the rest of the world about events that affect their daily lives. For researchers and practitioners working in the region the information gathered forms an important resource to learn more about the situation on the ground in hard-to-access areas.

Technology and the data

The technology for Voix des Kivus is cheap to set up and simple to use. Built on the freely available FrontlineSMS software, the system allows holders to send numeric or full text posts from almost any cell phone. On the receiving side a standard cell phone linked to a laptop linked to the internet comprise the necessary equipment. With other freely available software (R and LaTeX – our code is available upon request), messages received are automatically filtered, coded for content, cleaned to remove duplicates, and merged into a database. Graphs and tables are automatically generated which can then be automatically mounted into bulletins spanning any period of interest and with different levels of sensitivity. Translations of non-coded text messages (often from Swahili into French and English) are undertaken manually.

Over the last 18 months phone holders have sent thousands of pre-coded and text messages ranging from reports of attacks and abductions to reports of crop diseases and floodings. The constant flow of data from our phone holders is kept in a database and captured in weekly bulletins. Each Monday a bulletin is produced and disseminated that presents events that took place in the preceding week. These bulletins are shared with organizations that have received clearance from Voix des Kivus and its phone holders. The latter includes several development organizations based in Bukavu, DR Congo who can use the data to evaluate the situation on the ground throughout the region.

The figure gives the cumulative number of messages sent per phone holder since the summer of 2009. Phone holders show no sign of reporting fatigue (the slope of the lines do not decrease). As you can see, Voix des Kivus expanded from June 2010 onwards.

Crowseeding vs crowdsourcing

An important question for a system like this is whether the messages received can be trusted. Here we find the true value of crowdseeding. In most crowdsourcing approaches anyone can send information directly to the system. Crowdseeding works in a more restricted way with phone holders that are pre-selected, and only they can send in information. Crowdseeding has three main advantages for data quality: 1. The data is received from a representative set of areas; 2. All senders are known to the system and are in a  long term relationship with the Voix des Kivus program; 3. Because more than one holder is selected in each village “internal validation” is also possible. The system can also be used for sending information to holders and for engaging in more interactive forms of data collection. There are also disadvantages of this approach relative to crowdsourcing, the most obvious is that because of their relation with the program there may be concerns about the security of holders.

What we learned from the pilot

We have learned a lot from the pilot. The technical and social capacity is there right now. Interest in participating areas has been very great as witnessed by the steady stream of messaging. Technical barriers were also not as great as expected; solar technology can be used to power phones in the most remote areas and cell phone coverage is much greater than some maps suggest. Data quality appears good with fairly high levels of internal validation. Two questions though are still unanswered. First although we encountered no security concerns we do not know how safe the system would be for holders if it operated on a larger scale. Second, we don’t know whether this information will get seriously used. At the scale in which we have been operating many organizations expressed great interest in the concept and the data; but we do not know of any serious reactions from international actors to the messages coming in, including real time reports of attacks and abuses. Phone holders have continued to engage with the system despite the poverty of reactions, but we cannot expect that to continue forever.

Continuation?

After operating for more than  a year and a half as a pilot in Eastern Congo, the Voix des Kivus experience suggests that obtaining verifiable, high-quality data in real-time from these hard-to-reach areas is not only possible, but needs much less expense and oversight than previously thought. Our pilot is now coming to an end and Columbia is bowing out from Voix des Kivus. The big question we face now is whether and how to continue the system after the pilot, whether this should be run by a domestic group or an international group, whether this should continue as an open resource or as a resource tied to the operations of organizations that can respond. Please post your thoughts here.

For more information see: http://cu-csds.org/wp-content/uploads/2009/10/Voix-des-Kivus-Leaflet.pdf and www.cu-csds.org

The most important part of the site to have translated first is the front-end (the part that everyone sees), the admin area is also important, but not nearly as urgent. We couldn’t just go in there and change the words out, what we needed was a programmatic way to do this so that any other future translations in any other language could be done fairly simply (and much faster).

The first step was to pull all the English text content from the site and put it in a flat file document to be translated into French. Kaushal and Ory worked particularly hard finding people to do that translation and then to create a process, using Google spreadsheet docs, so that the next time we have to do it the transition will go a lot smoother. Once those text fields were all finalized we simply funneled that back into the platform and made an option to access it by clicking on a French flag in the site’s header.

For the administrators

It’s important that the administrators are able to translate incoming English reports into French (or other languages in the future). When on the “reports” page of the administration side, you’ll see this new button:

Once clicked, that button takes you to the following page where you can select the language you would like to translate the report into, and begin translating each section:

Challenges

We’re about 80% of the way done. One area that has proved to be particularly challenging is categories. Those are set by the administrator. This means it’s dynamic, not static like the rest of the content on the front-end. We’re working on a way to for administrators to add their categories in multiple languages.

Some languages take more characters (or words) to say something than others. We’re trying to figure out how to make sure that type of change doesn’t break our formatting completely.

Lastly, we’re trying to decide where and when the right time is to use automated computer translation of content. Requiring administrators to translate and retype reports, news headlines and local help organizations is neither efficient or time-saving. We’re interested in figuring out a way to programmatically help administrators manage this.

Thoughts on ways we can do this better?

Over the last week, while the world watched the US elections (as were we), we were also watching the Democratic Republic of the Congo (DRC). Since last Wednesday we’ve been quietly scrambling towards the first deployment of the new Ushahidi Engine into a crisis zone. Today, we’re announcing this instance to the world.

The DRC deployment can be found at http://DRC.ushahidi.com, and the mobile number to send SMS reports to is +243992592111.

Note: This is the alpha software for Ushahidi, so there are likely quite a few bugs and kinks that need to be worked out. If you find any, please submit them to bugs.ushahidi.com.

Help get the word out

We need to get the message to the people on the ground in the Eastern Congo that this tool is now available for them to report incidents in on. If you have contacts there, or can help spread the word through some other means, it would be a great help if you did so – even if it’s just point them to the website or telling them about the SMS number.

Setup and lessons learned

We don’t believe in waiting for software to be perfect before a launch, so we’ve tidied it up as best we can, and are doing constant updates along the way. This instance we’re running on our own servers for speed and ease of updates.

SMS Reporting Through Ushahidi, originally uploaded by whiteafrican.

We run what we’re calling our “tech hub”, which features a mobile phone with a DRC SIM card, tethered to a local PC running FrontlineSMS. Whenever this computer connects to the internet, it auto syncs with Ushahidi.

David Kobia worked very hard through the weekend to get the SMS functionality right. We’re using a combination here of FrontlineSMS to receive local SMS from anyone in the DRC at +243992592111, with a completing feedback loop from the admin area to the incident submitter that goes through Clickatell. Ken Bank’s team at FrontlineSMS has been very quick on the ball to provide us updated code samples to get everything working in time (big thanks!).

At the same time, this isn’t Kenya, so none of the Ushahidi team is well positioned to manage any kind of approval or verification of reports that come into the system. We’ve found a couple local groups that are willing to help us with this, including HEAL Africa. Their members are helping to report incidents, as well as verify incoming reports.

The categories being used in the DRC are different. One of the new ones is “verified reports” which only come from users who are known quantities by the groups managing the Ushahidi instance. It’s a form of more “trusted” reports. While we’re optimistic about gathering reports from people all over the Eastern Congo, we know that there is a certain amount of disinformation that goes on. We are actively looking at ways to neutralize false reports and encourage factual ones.

We’re also working on the French translation, and we need to be proactive here in doing other major language translations before they’re needed. If you’re an expert in a major global language, we’d love to hear from you.