Online Social Media in the Syria Conflict: Encompassing the Extremes and the In-Betweens

The Syria conflict has been described as the most socially mediated in history, with online social media playing a particularly important role. At the same time, the ever-changing landscape of the conflict leads to difficulties in applying analytical approaches taken by other studies of online political activism. Therefore, in this paper, we use an approach that does not require strong prior assumptions or the proposal of an advance hypothesis to analyze Twitter and YouTube activity of a range of protagonists to the conflict, in an attempt to reveal additional insights into the relationships between them. By means of a network representation that combines multiple data views, we uncover communities of accounts falling into four categories that broadly reflect the situation on the ground in Syria. A detailed analysis of selected communities within the anti-regime categories is provided, focusing on their central actors, preferred online platforms, and activity surrounding “real world” events. Our findings indicate that social media activity in Syria is considerably more convoluted than reported in many other studies of online political activism, suggesting that alternative analytical approaches can play an important role in this type of scenario.

💡 Research Summary

The paper investigates how online social media—specifically Twitter and YouTube—were used by the myriad actors involved in the Syrian civil war, and it does so without imposing any preconceived categories such as “pro‑government” versus “anti‑government.” The authors begin by exploiting Twitter’s list feature, which is often curated by journalists and experts, to gather a pool of 911 accounts that appear in 17 relevant lists. After manually filtering out non‑Syrian media, NGOs, academics, and other peripheral accounts, they retain 652 accounts that are either directly involved in the conflict or comment on it from abroad.

From these accounts they collect 1.76 million tweets posted in 2012‑2013, together with the full follower graph (≈31 k edges), mention and retweet interactions (≈203 k combined), and list‑membership data (≈22 k distinct lists). All YouTube URLs embedded in the tweets are extracted, yielding 14 629 unique channels; the authors also retrieve Freebase topic annotations that YouTube automatically attaches to uploaded videos.

To uncover the latent structure of this multi‑modal data, the study adopts the “unified graph” methodology introduced by Greene and Cunningham (2016). Seven distinct views of each account are built: (1) the set of accounts it follows, (2) the set of accounts that follow it, (3) accounts it mentions, (4) accounts that mention it, (5) accounts it retweets, (6) accounts that retweet it, and (7) the Twitter lists to which it belongs. For each view, cosine similarity is computed between the binary (or weighted) profile vectors of every pair of accounts, producing a ranked list of nearest neighbours per view. The seven rankings are then merged using singular‑value‑decomposition (SVD) rank aggregation, and the top‑k neighbours from the aggregated ranking become the directed edges of a sparse, asymmetric k‑nearest‑neighbour graph. This unified graph simultaneously encodes explicit network ties (followers, mentions, retweets) and implicit similarity derived from content and list co‑membership.

Community detection (via a modularity‑optimising algorithm such as Louvain or Infomap) on the unified graph reveals four large, coherent clusters that the authors label as: (i) Jihadist (gold), (ii) Kurdish (red), (iii) Pro‑Assad (purple), and (iv) Secular/Moderate opposition (blue). Some nodes belong to multiple clusters, reflecting the fluid alliances typical of the war. The paper then zooms in on three anti‑government sub‑communities to illustrate the richness of the approach.

The first sub‑community corresponds to “extremist jihadist” actors, dominated by ISIS‑affiliated accounts and heavily focused on combat‑related YouTube videos; its internal mention/retweet network is densely interconnected. The second, termed “revolutionary secular,” clusters around the Free Syrian Army and human‑rights NGOs; it exhibits a more diverse set of interactions and shares a mix of battlefield footage, interviews, and humanitarian content. The third, described as “moderate/center,” consists mainly of Kurdish‑Arab coalition accounts that post regional news and cultural material; although its Twitter activity is less intense, it serves as a bridge between the other groups through extensive list memberships.

All three sub‑communities display pronounced spikes in activity around real‑world events, most notably the August 2013 Ghouta chemical attack, where tweet volumes and YouTube video shares surged dramatically. This temporal alignment demonstrates that online discourse in Syria is tightly coupled to on‑the‑ground developments.

The authors argue that the Syrian online ecosystem cannot be reduced to a simple binary polarization; instead, it is a multi‑layered, dynamic network where actors occupy overlapping positions and switch allegiances. By avoiding a priori hypotheses and integrating multiple relational views, the unified‑graph approach uncovers nuanced structures that traditional hashtag‑based sampling or single‑network analyses miss.

In conclusion, the study provides a methodological blueprint for analysing social media in volatile, information‑sparse environments. It shows that (1) curated Twitter lists can serve as a reliable source of “authoritative” accounts, (2) multi‑view similarity aggregation yields a compact yet information‑rich graph, and (3) community detection on this graph reveals both macro‑level categories and fine‑grained sub‑communities that correspond to real‑world factions and events. The findings underscore the importance of flexible, data‑driven techniques for future research on online political activism in conflict zones.