Crowdseeding in the Congo: Using Cell Phones to Collect Conflict Data

High-quality information about what happens on the ground during war is difficult to collect. In the attempt to solve this problem, NYUAD Assistant Professor of Political Science Peter van der Windt and Macartan Humphreys of Columbia University have developed a system that uses cell phones to collect real-time data from areas plagued by violence.

The system may provide researchers with information that is difficult or even impossible to gather with traditional methods. van der Windt undertakes much of his research in the east of the Democratic Republic of the Congo (DRC), an area that has witnessed widespread violence for decades and his findings were recently published in the Journal of Conflict Resolution.

The history of fighting in the region is complex. The 1994 genocide in Rwanda led to an influx of refugees; migrants strained local resources and altered the social and economic life of villages in the region. The genocide was followed by the First (1996-1997) and Second (1998-2003) Congo Wars.

Because of scarcity of land, ethnic tensions, the presence of natural resources, and grievances related to prior conflicts, these wars led to the death and displacement of millions. Although the war officially ended in 2003, thousands of civilians still die each month, especially in the South Kivu province, the location of van der Windt’s cell phone project.

To learn about conflict, researchers traditionally conduct surveys. In the DRC, as in many other places, this activity would be logistically challenging because a research team would have to travel for days to a remote village to ask villagers about an attack by a militia that happened months before. But even with great effort, "general conflict data is often bad because you get biased information," van der Windt said.

A house in the Democratic Republic of the Congo, the site of van der Windt's research study.

Selection biases affect data because only certain members of a community will be able to provide information about past events: "People move, they are displaced, you can't get to them, they die." Recall biases occur as people's memories change over time: humans are prone to "telescope" events, or run together several moments as being related when in fact they occurred at different times in the past.

Data is also influenced by reporting bias: people who have witnessed horrific events may be unwilling to talk about them because they are traumatized and fear for their own safety. Moreover, "people often reply to surveys strategically in the hope of receiving resources from NGOs," van der Windt noted.

People move, they are displaced, you can't get to them, they die.

Peter van der Windt, assistant professor of political science

To overcome these problems, van der Windt and Humphreys developed a cell phone-based data collection system called "Voix des Kivus" (Voice of the Kivus). The project employs an innovative process called "crowdseeding" to gather data. Crowdseeding builds on the popular "crowdsourcing" approach, in which anyone can send a SMS to a system about events they witness or hear about. Implementing a crowdsourcing project in Congo, however, would not have worked: few people would know about the project; few people have a phone and credit; moreover, people might send incorrect information to obtain access to NGO resources.

Crowdseeding combines crowdsourcing with standard social science techniques. Prepaid cell phones with credit are given to randomly chosen villages and villagers are asked to provide information. "Like crowdsourcing, we've got the benefit of real-time information," van der Windt said, "but the additional benefit with Voix des Kivus is that we have data that is representative," since the villages are chosen randomly. A further benefit is knowing who is at the other end of the line. "We build a relationship over time with the reporters, which is not the case with crowdsourcing or with traditional survey methods."

Voix des Kivus began in four villages in August 2009 and was expanded to 18 villages a year later. Three phones were given to each village: one went to the village chief, another to the women's representative, and a third to an elected member of the community, allowing the researchers to get information that spanned different social groups. van der Windt and his team extensively trained reporters so that they knew how to use the cell phone and accurately report events.

From the start of project to its end in July 2011, researchers received 4,783 messages, which reported on a total of 4,623 unique events. The most common were the presence of troops, looting, violence between villagers, kidnappings, and deaths related to conflict.

The participants were heavily invested in the project. A few months after the Voix des Kivus began, the researchers traveled to the villages and spoke with participants. They found that some would walk three hours to a tea plantation simply to charge their cell phone. van der Windt solved that problem by buying USD 25 solar chargers so that the reporters could charge their phones.

van der Windt started the Voix des Kivus simply as a "purely academic exercise," to see if high-quality conflict data could be gathered in real time. And it worked. The data they received contributed excellent information about conflict in the participating villages. But once he realized that the system allowed villagers to provide data about events that were influencing their lives, "We immediately moved away from an academic exercise to a more activist agenda," he said.

We immediately moved away from an academic exercise to a more activist agenda ... We wanted to see if those that received aid had lower or less conflict, and we found that that was the case.

Peter van der Windt

The researchers initiated a program to provide bulletins to NGOs in Eastern Congo that provided data received from Voix des Kivus about conflicts and other humanitarian events. These organizations were carefully vetted and the team only shared data that would not jeopardize the safety of participants of the study.

There are numerous applications of a project like Voix des Kivus. In their paper, van der Windt and Humphreys use the collected data to answer the question whether aid leads to more or less conflict. Surprisingly, perhaps, academics have difficulty answering this question. While there is an enormous amount of scholarship on the question, the findings are equivocal, van der Windt said. Some studies suggest that aid works, while others argue that aid increases conflict because it introduces more resources to fight over.

van der Windt and Humphreys built on a development project that took place in the South Kivu province before Voix des Kivus began. Moreover, because the development project worked only in randomly selected villages, it is possible to learn about the causal impact of development aid. "We wanted to see if those that received aid had lower or less conflict, and we found that that was the case," van der Windt said.