Media monitoring can play an important role in tracking, preventing, responding to and understanding violence and unrest. It is often used to identify, track and map incidents of insecurity and patterns in escalation and de-escalation in ongoing crises. Monitoring can draw on either ‘old’ media, such as newspapers, and/or ‘new’ social media, such as Twitter or Facebook. To date, however, there has been limited empirical research comparing the profile of violence represented in these different sources.
A new paper, co-authored by ACLED Director of Research & Innovation Dr. Roudabeh Kishi along with DCU researcher Caitriona Dowd and IDS researchers Patricia Justino and Gauthier Marchais, published this month in the journal of Research and Politics, assesses the comparative advantages of ‘new’ and ‘old’ data sources for crisis monitoring. The paper compares reports of political violence and demonstrations published on social media (Twitter) and traditional media (drawn from ACLED data) during the Kenyan elections in August and October 2017 and finds relevant differences across ‘new’ and ‘old’ media sources in the geography and timeliness of violence reporting.
What ‘New’ and ‘Old’ Media Tell us about Crisis
The research scanned public posts to Twitter and reports drawn primarily from traditional media to analyze the profile of violence and unrest reported, and compare the geography and timeliness of those reports.
The paper finds that the profile of violence and unrest varies significantly by source. Records of unrest from Twitter are more geographically concentrated, particularly in the capital city and wealthier areas, and are timelier in the immediate period surrounding elections. By contrast, records from traditional media collated by ACLED have a wider geographic reach, and are relatively more numerous than those from Twitter in rural and less wealthy areas. They are timelier and more consistent in the run-up to and following elections. The table below summarizes key findings from the new study:
Table I. Key differences in reporting by data sources (Dowd et al. 2020, p. 7)
| Category | ‘Old’ Media Reports (ACLED) | ‘New’ Media Reports (Twitter) |
|---|---|---|
| Geography: Physical coverage | More coverage in rural areas outside the capital city | More coverage in densely populated areas and capital city |
| Geography: Socio-economics | More coverage in economically less developed areas | More coverage in economically more developed areas |
| Temporality: Timeliness | Faster reporting overall, but relatively less timely immediately surrounding elections | Slower reporting overall, but more timely in period immediately surrounding elections |
| Temporality: Sustained reporting | More reporting in the pre- and post-election period | More reporting in period immediately surrounding elections |
The findings suggest that contrary to initial, optimistic accounts, social media and digital technologies (SMDTs) present both opportunities and limitations for violence monitoring and crisis response. There are significant trade-offs, which should be carefully weighed by practitioners, policymakers and researchers relying on either ‘new’ or ‘traditional’ media sources to understand the nature and dynamics of unrest.
While neither source can reveal the ‘true’ violence that occurred, one way to mitigate against these limitations is to triangulate and supplement monitoring systems using a constellation of sources tailored to the unique crisis context, in order to reduce biases or gaps in any individual system.
Social Media and Digital Technologies in Kenya’s Unprecedented Elections
The new Economic and Social Research Council (ESRC)-funded study builds on several important themes in recent research on political violence and unrest. First, SMDTs have become increasingly important sources for crisis monitoring – adding to, and often transforming, the information landscape and facilitating the dissemination of (often real time) details of crises as they unfold. The study builds, therefore, on a growing body of research documenting the impact of digital inequalities and the changing role of social media in violent contexts (see Roberts and Marchais, 2017).
In the specific context of Kenya, the country has a long history of cyclical violence surrounding elections. SMDTs have also played an important role in past election cycles in crisis monitoring (see Mutahi and Kimari, 2017). The 2017 elections were unprecedented in that the initial election in August was annulled following allegations that the system had been hacked, with a re-run held in October of that year. Both elections were accompanied by significant violence and unrest.
Against this unique backdrop, the research team gathered and analyzed real-time reports of violence and unrest. The paper is among the first to map the specific differences in the profiles of violence produced through reliance on different data sources, and makes a strong case for the importance of drawing on a constellation of sources to leverage complementary advantages of different source types.
A Multi-Disciplinary Research Team and Approach
The research project was a partnership between multiple institutions, including researchers from Dublin City University and the Institute of Development Studies, alongside ACLED’s Director of Research & Innovation, Dr. Roudabeh Kishi. ACLED is the world’s leading source of real-time disaggregated data collection on political violence and unrest across a growing number of regions.
Across the over 150 countries that it now covers, ACLED relies on numerous sources of information, including traditional media (with an emphasis on local reporting) in over 50 different languages, reports from international institutions and NGOs, data and information from partnerships with local conflict observatories, and select new media (i.e. only using targeted and verified accounts, not crowdsourcing). For each context that it covers, ACLED develops a tailor-made sourcing process to ensure the data being collected are most reliable; this is because there is variation in both the conflict and media landscapes of each context.
One important lesson from the research is that reliance on Twitter in the Kenyan context, for example, can aid in timeliness around crisis periods, for example, while less so outside of such periods where traditional media might be more reliable. However, in other contexts where Twitter use is negligible and social media penetration is low, Twitter might not be useful even in crisis contexts. On the other hand, in contexts where press freedom is low and the media environment is quite closed, new media sources like Twitter may become even more integral in capturing patterns of disorder. Accounting for such variation is important to ensure each unique environment is captured most reliably. (For more on ACLED’s sourcing strategy, see this methodology primer.)
At the time of data collection for this project, ACLED relied largely on published media (newspapers, newswires and published reports) in the Kenyan context. However, building on the study’s findings, the research has already informed changes in ACLED’s sourcing strategy. Namely, select new media accounts – including Twitter accounts, as well as Facebook accounts – operated by trusted sources are now included in ACLED’s regular coverage of Kenya.
The Centre for Human Rights and Policy Studies (CHRIPS) in Nairobi was also a partner in the project. CHRIPS is a leading international African research centre based in Kenya that conducts high quality policy relevant research on human rights, security, terrorism and counter-terrorism, violence, crime and policing. As part of the project, researchers from CHRIPS produced a working paper on the impact of social media in Kenya’s elections, which in turn, inspired a later publication by CHRIPS researchers Patrick Mutahi and Brian Kimari on fake news in Kenya’s 2017 election.
Looking Forward
While this new research focuses on Kenya, the findings point to the value of drawing on a constellation of sources to leverage complementary advantages of different source types in different contexts. Efforts to leverage the potential of new technologies for effective and timely crisis prevention and response benefit from tailoring information gathering to the relevant information landscape, its associated (digital) inequalities, and the nature and profile of unrest. Further research into the generalizability of the findings concerning trade-offs in coverage over time, and in rural and urban areas, for example, and the verification and underlying accuracy of reporting across source types, would be valuable as these initiatives continue to grow.
Read the full article here.
Read related working papers and blogs here.
© 2020 Armed Conflict Location & Event Data Project (ACLED). All rights reserved.
