Re-thinking conflict early warning: rapid context data in Syria and Libya

[This post is part of a series on re-thinking conflict early warning.]

In March 2011, the United Nations Office for the Coordination of Humanitarian Affairs activated the Standby Volunteer Task Force (SBTF) to provide live crisis mapping support for its operations in Libya. Over the course of four weeks, SBTF volunteers collected and organized context data, mainly drawing on social media and traditional media sources. At a time when few people were on the ground, the data provided important situational awareness for OCHA and other responders to the crisis.

The Libya Crisis Map was an experiment – and as such had some significant flaws in design and implementation. You can read a summary of the impact it had here. To me, it was also a proof of concept about the potential of using social media to gather rapid context data in hard-to-access situations. Which is why I was excited to learn recently about the work that the Carter Centre is doing to understand the current situation in Syria by mining social media. The Syria Conflict Mapping Project has been tracking posts on YouTube, Facebook and Twitter by the main actors in the Syrian conflict. With this input and the interpretation of the Centre’s analysts, the project produces detailed, insightful reports on the dynamics unfolding on the ground.

It all started when the first defectors from the Syrian military began posting their defection statement on YouTube. The first defectors gave a large amount of information on who they were and what part of the army they were defecting from. Many others, who also employed YouTube to spread their message and confirm their defection, quickly followed this method of defecting. Some analysts wondered whether the defections would cause the military to collapse. Christopher McNaboe, at the time an intern at the Carter Centre, thought he might be able to look through the social networks of these defectors to identify how they related to each other and to other main actors, perhaps shedding more light on the dynamics that would unfold. By tracking the divisions from which soldiers were defecting, he was able to reverse-engineer the structure of the Syrian military (or at least those sections of it that were defecting), and use this information to conclude that the military was unlikely to collapse as a result of these defections.

In the process, he began to uncover a rich source of context information. Defectors would often form armed groups, and their YouTube posts stated or showed not only names and locations, but also details of the newly formed group (size, armament, links to other existing groups, etc). Chris and the Carter Centre team began to curate a database of videos of defectors – each video was watched by an analyst and tagged to extract as much information as possible. Running network analysis on this dataset shows the structure of armed opposition in each part of Syria at any given time. The groups are very consistent at keeping tabs of activity and posting information about successful operations, making it possible to track who controls what and how that’s changing. The network analysis can show how armed groups in a particular region of Syria relate to each other – in strength, affiliation to each other and affiliation to outside bodies or trends (e.g. Islamists). Most impressive is to see a network of all armed groups, and watch it dynamically on a timescale. Just looking at this network expand and unfold provides insights into the relationship between critical actors – showing for example how the relationship between the Military Council and Liwa Al-Tawhid has changed over time.

Analysing the network of armed groups is a first step to scope the dynamics of the conflict. From the armed group YouTube videos, the Syria Mapping Project analysts can also track down the social media (Twitter, Facebook) profiles of key individuals – and who they relate to – allowing for some more complex network analysis. For example, one analysis shows how the twitter account of an armed group, connects most closely with followers in Saudi Arabia or Kuwait that identify as extremists. (The analysis embeds some assumptions about what is a close relationship, i.e. not just following / followed but also pattern, volume and type of communications to measure the strength of connection.)  These followers are mostly in charge of charities and therefore able to provide funds; many of them have been jailed by their own governments. The same network analysis using Twitter accounts can show the connection between two armed groups – and show also that they are not funded by the same people. Similar analysis can also be run on the connections between armed groups and politicians or civil society actors, showing overlapping (or not) networks between these groups.

The insights that can be gleaned from this analysis are critical to better understand dynamics in the Syrian conflict. Chris makes an excellent point about the importance of injecting this type of complex analysis into public discourse, so that decision makers can better understand what they are dealing with: “People talk of funding the opposition. But you can’t control where arms go in a network with so many connections and so much fluidity.” The Syria Mapping Project shines a light into this complexity. It is already a great resource to policy makers, and will be a critical input when peace negotiation start.

There are of course some organizations that are doing excellent work collecting data on the ground in Syria. And yet, as the Carter Centre team explain, you can’t really say anything about Syria unless you also look at social media – so much of the conflict dialogue is unfolding there. Chris points out that a recent report by the PEW Research Centre found that people in Arab countries are three to four times more likely to share community related content online – including political views. The volume and regularity of content sharing by armed groups is in line with this finding. In fact, the Carter Centre team find consistently that groups with a stronger online presence also have a stronger offline presence.

The kind of network analysis of actors using social media content that this project is undertaking is more useful than most conflict early warning systems focused on incident reporting. And why limit it to Syria? The tools that the Carter Centre team have so far used are free and available to anyone: social media content manually curated in a spreadsheet, data cleaning using GoogleRefine, network analysis run using Gephi. Right now, undertaking this kind of data curation and analysis is quite time consuming, but new tools are appearing all the time. QCRI recently put out the Artificial Intelligence for Disaster Response platform, which uses machine learning to automatically identify informative content in Twitter during natural disasters, cutting down the time spent on manual curation. I look forward to a time when these and other tools will make social media curation and analysis for conflict contexts more systematic and user friendly.

Advertisements

One thought on “Re-thinking conflict early warning: rapid context data in Syria and Libya

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s