Tracking social media may uncover potential electoral violence

By Jillian Slutzker

July 6, 2015

As more than 70 percent of Guyana’s population headed to the polls to vote in the 2015 national elections, a team of election and technology experts in Washington, DC, were busy tracking thousands of social media feeds from the country, searching for threats of pending electoral violence.

Guyana has experienced violence in many of its elections since 1992. Voting blocs have tended to fall along ethnic lines, aggravating existing cleavages and igniting protest over contested elections results.

With the proliferation of social media enabling real time reporting by citizens, the opportunities to track these types of electoral violence outbreaks have multiplied in recent years. Amidst the social media noise, citizens are sharing valuable real time information that can help identify and mitigate electoral violence.

“Using real time social media tracking to improve electoral process in practice has grown significantly in the last few years,” says Luther Jeke, a Technology for Development Fellow at Creative. “If violence erupts in a particular election, the hours or minutes saved by having identified the situation through social media posts could make a significant difference in response.”

In the innovative pilot study to monitor Guyana’s 2015 presidential election in May 2015, Creative Associates International’s Electoral Education and Integrity team and Technology for Development experts used two social media data mining software programs to analyze Guyanese Facebook pages, Twitter feeds, blogs and online news sources, searching for key words that could be red flags of impending electoral violence.

Mining social media for keywords and hashtags from #Votelikeaboss (a voter mobilization campaign of the Guyana National Youth Council) to “Guyana assault” and candidates’ names, Creative’s team tracked the entire elections cycle—looking for threats in the pre-elections phase, polling station vulnerability on Election Day and potential protests and flash mobs after results were announced.

The team communicated their findings throughout the process to on-the-ground electoral observers with the Carter Center and the Organization for American States.

Forecaster of electoral violence

In recent years, social media has emerged as a conveyor of threats during election cycles, says Jeke—for example, citing incidences of intimidating text messages being sent to voters of certain tribes in Kenya’s 2007 elections. In Haiti’s 2010 presidential election, Jeke recounts, a video was posted online showing the shooting of a supporter of presidential candidate Michel Martelly in order to intimidate other supporters from hitting the polls.

Citizens have also used social media to document electoral violence.

In the aftermath of Iran’s 2009 presidential elections, a video of the killing of 26-year-old protester Nedā Āghā-Soltān by the state-sponsored Basij militia provided evidence of a brutal post-election crackdown to those outside the country.

As a forecaster of electoral violence, social media is a relatively new tool.

“This is an emergent field of interest and our pilot project was representative of the kinds of examinations on the topic currently underway,” says Jeff Fischer, Senior Electoral Advisor at Creative.

New tools, new challenges

As electoral observers like Creative’s team explore the capabilities of social media to track threats and outbreaks of electoral violence, new challenges are emerging. The main challenge, Creative’s observers say, is verifying the veracity and authenticity of the information gathered.

“Tweets often don’t provide accurate descriptions of important details surrounding an event such as who, what, where and when,” says Christopher Smyser, an Electoral Education and Integrity Program Specialist at Creative.

To fill in the missing details and triangulate data, Smyser says that after initially registering an incident, team members searched for second references to the incident in news sources and used social media to communicate with the original author of the tweet or post.

“Had we seen something that appeared to be what we felt was a real-time threat we would have done our best to verify with other tweets or another source and then would have proceeded to pass the information along [to on-the-ground observers],” says Smyser.

Relatively peaceful, not incident-free

In the wake of the May election, the citizens of Guyana, who have witnessed repeated outbreaks of electoral violence over the last two decades, are breathing a sigh of relief. Creative’s team reported that there were no major threats or violence.

However, says Fischer, while the elections cycle was relatively peaceful, it was not entirely incident free.

“While we were able to produce alerts to possible threats, the relatively low number of incidents inhibited our making many correlations between threats and incidents” says Fischer. “But, I think that we created an early warning environment where there was sensitivity to the types of incidents which could emerge.”

Moving forward, Fischer says, large teams of electoral observers could use similar technology to create early warning systems and monitor electoral violence, ensuring a consistency in approach to deliver the most accurate results—an outlook Fischer’s team members share.

“I certainly think that the ability to monitor social media for the purpose of predicting incidents of electoral violence will be an increasingly useful new tool for electoral observers,” says Smyser. “The ability to access real-time predictions of violence or unrest could certainly help improve the safety of everyone involved in the electoral process.”