Meet Jennifer, she took her first TechChange course on Technology for Conflict Management and Peacebuilding in October and is now facilitating multiple TechChange courses.

Drawn by our teaching model, after completing her course, she wanted to become involved as a facilitator for our courses. She is currently co-facilitating TC111: Technology for Monitoring and Evaluation with Norman Shamas, and facilitating TC105: Mobiles for International Development. Jennifer will also be facilitating TC109: Technology for Conflict Management and Peacebuilding in the coming months, bringing her full-circle in her participant-to-facilitator involvement with TechChange.

Prior to joining TechChange, Jennifer participated in several research symposiums and conferences like the Institute for Qualitative and Multi-Methods Research, the Association for the study of the Middle East and Africa Annual Conference and more. She has also served as a guest speaker for the American Red Cross and has mentored several high school and undergraduate students regarding school-sponsored and independent international development and peacebuilding start-ups.

Jennifer is an emerging comparative politics scholar and methodologist focused on answering questions related to individual and community involvement in conflict, post-conflict, and peace processes. She holds a Bachelor’s in Political Science from the Colorado College, a Masters in Public Health from Indiana University, and is a Doctoral Candidate in Political Science with the University of New Mexico.

The Global Database of Events, Language and Tone (GDELT) has been gathering and databasing all the news events related to conflict and political protest dating back to 1979. GDELT continues to be fed new data through the various global news services, automatically updating every day. At the end of July GDELT released their Global Dashboard which visualizes all of their data collected from February 2014 to present on a map of the world. It’s a fantastic tool for conflict management and resolution professionals who are interested in big data, since it takes their information and puts it in a visually attractive, easily navigable format. This is an exciting development, so how does it work and what can peacebuilding practitioners get out of using GDELT’s event data?

The first thing to keep in mind is that the Dashboard is new. As it stands there are only two filters for event data (‘conflict’ or ‘protest’), but there are plans to expand these filters so that users can easily focus on the events that are of most interest. For now they’ve done a pretty good job of helping filter out conflict events, which are basically events involving kinetic violence, from protest events, which could end up being violent but are generally more along the lines of protests and social action. While basic, these are good starting points for an initial filter. The nice thing about the dashboard though is that if I have some expertise about the region or event I’m interested in gathering data on, I don’t need the filters because I can use geography and date to narrow my search. The Dashboard allows the user to take advantage of their contextual knowledge to filter the data, so while the built-in filters that come later will be helpful researchers can still use the database efficiently now.

Let’s say we’re interested in recent protest events in South Africa, but we want to know if there have been any in smaller cities, since we know that there’s likely to be a lot of political action in places like Cape Town and Johannesburg. I started with the Dashboard zoomed out to the maximum, so I could see the whole world, then went to the bottom left and set the date that I was interested in seeing news from. For this test I picked August 3, 2014. Below is what the screen looked like at this point:

GDELT Global Dashboard

We can see the whole world, and in South Africa there are big dots indicating aggregated data. Since I want to see what’s happened outside the main cities, I zoomed in until the dots started to disaggregate, then I selected the ‘protest’ filter to remove the ‘conflict’ events. Once I was zoomed in the filter was set, I found that there was a protest event in Port Elizabeth so I clicked on the dot and a box with the web addresses for news articles about a protest against money being spent on a museum appeared: GDELT Global Dashboard: South Africa

I clicked on the Google News link, which took me to the related articles that Google had collected about that protest and read one that had been reposted by a local news service from the Agence France-Presse:

"South African shantytowns residents force anti-apartheid museum to close," Agence France-Presse

I managed to do this in a few minutes using the Dashboard, work that would have taken longer if I was just doing searches for protest news out of South Africa. What makes the tool really useful is that I can search in a few different dimensions. If want to know if this is the first time there has been social action around the museum in Port Elizabeth, I can leave the map zoomed in to that location and scan through the dates going back to February. What we can do, relatively easily, is see events and narratives spatially and analyze how they change over time.

This is a big dataset, so I thought hard about what its value added is from a methodology perspective. As I dug through the data, I realized something important. I’m not sure this is a database that will be particularly useful for forecasting or predictive analysis. You might be able to identify some trends (and that’s certainly a valid task!), but since the data itself is news reports there’s going to be a lot of variation across tone and word choice, lag between event and publication, and a whole host of other things that will make predictive analysis difficult.

As a qualitative dataset though, the GDELT data has incredible value. A colleague of mine pointed out that the Dashboard can help us understand how the media conceptualizes and broadcasts violence at the local level. Understanding how news media, especially local media, report things like risk or political issues is valuable for conflict analysts and peacebuilding professionals. I would argue that this is actually more valuable than forecasting or predictive modeling; if we understand at a deeper level why people would turn to violence, and how the local media narrative distills or diffuses their perception of risk or grievance, then interventions such as negotiation, mediation and political settlements can be better tailored to the local context.

Big Data is a space that is both alluring and enigmatic for conflict resolution professionals. One of the key challenges has always been making the data available in a way that is intuitive for non-technical experts to use. GDELT’s Dashboard is a great start to this, and the possibilities for improving our understanding of conflict through the narratives we can observe in the media are going to grow rapidly in the next few years.

This post originally appeared in Insight for Conflict on September 19, 2014. 

 

This is a guest post by Dhairya Dalal. If you are interested in using crisis mapping and using technology for humanitarian relief, conflict prevention, and election monitoring, consider taking our course Technology for Conflict Management and Peacebuilding.

Overview

Recently, I had the opportunity to run an election monitoring simulation for TechChange’s TC109: Conflict Management and Peacebuilding course. Led by Charles Martin-Shields, TC109 taught over 40 international participants how mapping, social media, and mobile telephones could effectively support the work of conflict prevention and management.  Robert Baker taught participants how the Uchaguzi team leveraged crowd-sourcing and Ushahidi, a web based crisis mapping platform, to monitor the 2013 Kenyan elections.

For the simulation activity, my goal was to create a dynamic hands-on activity. I wanted to demonstrate how crisis mapping technologies are being used to promote free and fair elections, reduce electoral violence, and empower citizens. To provide students a realistic context, we leveraged live social media data from the Kenyan elections. Participants walked through the process of collecting data, verifying it, and critically analyzing it to provide a set of actionable information that could have been used by local Kenyan stakeholders to investigate reports of poll fraud, violence, and voter intimidation.

Below I’ll provide a brief history of election monitoring in the context of Kenyan elections and provide a more detailed look at the simulation activity.

Brief History of Election Monitoring and Uchaguzi

uchaguziIn 1969, the Republic of Kenya became a one-party state whose electoral system was based on districts that aligned with tribal areas. This fragile partitioning often generated internal friction during the electoral cycle. The post-election violence of 2007-2008 was characterized by crimes of murder, rape, forcible transfer of the population and other inhumane acts. During the 30 days of violence more than 1,220 people were killed, 3,500 injured and 350,000 displaced, as well as hundreds of rapes and the destruction of over 100,000 properties. 2

Ushahidi was developed in the wake of the 2008 post-election violence. Ushahidi, is a website that was designed to map reports of violence in Kenya after the post-election fallout. However, Usahidi has since evolved into a platform used for crisis mapping, crowd-sourced data gathering, and many other things. Since then, the name Ushahidi has come to represent the people behind the Ushahidi platform. 2

Uchaguzi was an Ushahidi deployment, formed to monitor the 2013 Kenyan general elections held this past March. The Uchaguzi project aimed to contribute to stability efforts in Kenya, by increasing transparency and accountability through active civic participation in the electoral cycles. The project leveraged existing (traditional) activities around electoral observation, such as those carried out by the Elections Observer Group (ELOG) in Kenya.3

Election Monitoring with CrowdMaps

TC109 Simulation Figure 1: TC109 Simulation map (view official Uchaguzi map here: https://uchaguzi.co.ke/)

For the simulation activity, we used Ushahidi’s CrowdMap web application. CrowdMap is a cloud-based implementation of the Ushahidi platform that allows users to quickly generate a crisis map. Crowdmap has the ability to collect and aggregate data from various sources likes SMS text messages, Twitter, and online report submissions.

To provide the participants a more realistic context, our simulation collected real tweets from the Kenyan elections that had just occured the prior week. Our simulation aggregated tweets from Uchaguzi’s official hashtag, #Uchaguzi, as well several other hashtags like #KenyanElections and #KenyaDecides. In addition students were tasked with creating reports from Uchaguzi’s facebook page and local Kenyan news sites.

The aggregated information was then geo-tagged, classified and processed by the participants. The participants created reports, which described incidents licrowdmapke instances of voter intimidation, suspected poll fraud, and reports of violence. The CrowdMap platform plotted these reports on a map of Kenya based on coordinates the participants provided during the geo-tagging phase.  The resulting map showed aggregation patterns, which would have allowed local actors to see where certain types of incidents were taking place and respond accordingly.

Conclusion: Going beyond the Technology and Cultivating Information Ecosystems

workflow   Figure 2: Uchaguzi Workflow

While technological innovations have made it easier to collect vast amounts of data in real-time during a crisis or an live event, a lot of process and human capital is still required to ensure that the data can processed and acted upon. Prior to the Kenyan elections, the Uchaguzi team established a well-planned information workflow and local relationships to ensure that information was ultimately delivered to the local police, elections monitors, and other stakeholders who could take action on the reports received. This workflow also delineated volunteer workgroups (based on Standby TaskForce’s information processing workflow) which were responsible for different parts of information collection process from Media Monitoring and Translation to Verification and Analysis.

To provide the participants an understanding of the full picture, we had them assume the role of various workgroups. They were challenged to identify how the information would be gathered, verified, classified, and distributed to local stakeholders. Participants followed the official Uchaguzi workflow and learned more about the challenges faced by the various workgroups. For example how would you translate a report submitted in Swahili? How would you determine if a report is true or falsely submitted to instigate provocation? How would you escalate reports of violence or imminent danger like a bomb threat?

Overall, the participants were able to learn about both the technology that enables the crowd-sourcing of election monitoring and the strategic and deliberate structures put in place to ensure an information feedback loop. Participants were able to gain an understanding of the complexity involved in monitoring an election using real data from the Kenyan elections. They were also given an opportunity to recommend creative suggestions and innovations that were sent to the Ushahidi team for future deployments.


About the Author:
Dhairya Dalal is a business systems analyst at Harvard University, where he is also pursuing his master’s degree in Software Engineering. Dhairya serves a curriculum consultant for TechChange and is responsible for teaching hands-on technical workshops centered around crisis mapping and open gov APIs, as well as strategic lessons on social media strategy and digital organizing.

Sources:
1:Background on the Kenyan Electoral Violence
http://www.haguejusticeportal.net/index.php?id=11604 
2: Uchaguzi Deployment
https://wiki.ushahidi.com/display/WIKI/Uchaguzi+-+Kenyan+Elections+2013
3: Uchaguzi Overview
http://reliefweb.int/report/kenya/uchaguzi-kenya-2013-launched