Introduction to the project
This project is aimed at describing and analyzing Russian propaganda and disinformation targeting the Arabic-speaking world. In the contemporary world, where information and communication form an integral part of geopolitical conflicts and can determine their outcome, tracking state influence in the online information sphere has become increasingly important. The main objective was to identify orchestrated narratives of deception promoted by Russia. The focus was centered on Russia Today (RT Arabic) and Sputnik Arabic, chosen for their direct affiliations with the Russian state, thus offering a transparent window into official Russian narratives. A secondary goal was to compare the similarities between Russian narratives deployed through disinformation campaigns in the Middle East and North Africa (MENA) region with those documented elsewhere in the Global South.
The project's methodology hinged on leveraging Telegram, a platform gaining traction, especially in Russian contexts, for its less restrictive content dissemination. Unlike mainstream social media platforms prone to filtering and censorship, Telegram provided an unfiltered conduit to the data. This approach yielded a substantial corpus, comprising approximately 170,000 messages from Sputnik Arabic and 277,000 from RT Arabic.
The findings indicate a strategic use of these channels to disseminate narratives that favor Russian perspectives, especially in geopolitical conflicts. The emphasis on military successes and the portrayal of opposing forces, particularly in the context of the Ukraine conflict, align with typical propaganda tactics. This approach seems designed to shape public opinion in the Arabic-speaking world, potentially influencing perceptions toward Russia and its international stance.
The project's examination of video content and the manual coding of the top 1000 posts provide a more comprehensive understanding of the sophisticated techniques used in these Russian state-sponsored information campaigns. By uncovering these narratives, the project seeks to contribute to a broader understanding of state-backed media strategies and their impact on public perception. This research is not just for academic purposes but also serves as a crucial tool for raising public awareness about the complexities and influences of state-sponsored media in the modern information landscape.
The project also aims to contribute to the development of innovative technologies to identify and analyze state-sponsored campaigns of deception. It incorporates a sophisticated blend of artificial intelligence (AI) and human expertise to distill the complexities of propaganda and disinformation within Arabic-language news from Russian sources. The findings, rooted in AI-driven analysis, have unveiled patterns in the most viewed messages on RT Arabic and Sputnik Arabic Telegram channels, suggesting a meticulous orchestration of narratives.
Russian Disinformation in the MENA region
The findings of this project should be read against the background of a range of wider issues relating to disinformation in the MENA region.
Russia's full-scale invasion of Ukraine in early 2022 marked a pivotal moment in international relations, triggering a wave of unprecedented sanctions on Russia from Western countries. These sanctions, aimed at isolating Moscow economically and politically, have had significant repercussions, compelling the Kremlin to seek stronger ties with countries in the Middle East to offset the impact. This strategic pivot is evident in Russia's intensified disinformation campaigns in the region, which aim to bolster its image but also to exploit the growing rifts between Middle Eastern countries and the West.
Over the last two decades, Russia has expanded its influence in the Middle East via sophisticated information campaigns. This strategy has seen Russian state-backed media outlets like RT Arabic and Sputnik Arabic partner with local media outlets to flood the region with content that aligns with Moscow's geopolitical interests. These efforts are part of a broader attempt to sway public opinion, undermine Western influence, and promote Russia as a key player in regional affairs.
The success of Russia's approach is partly due to the region's historical and current grievances with the West. Many in the Middle East view Western interventions with skepticism, providing a receptive audience for Russia's narratives. These disinformation campaigns exploit such sentiments, presenting Russia in a favorable light while casting doubt on Western motives and actions.
The impact of this strategy extends beyond mere public perception. By shaping the narrative, Russia gains leverage in diplomatic and political spheres, positioning itself as a stabilizing force in contrast to the perceived destabilizing influence of the West. This not only challenges Western interests but also complicates efforts to engage with the region on key issues like conflict resolution.
Research Steps
Step 1Research Kicked-Off
DWA kicked of the project and highlighted objectives and overall goal. Both research and tech objectives were clarified.Step 2Literature and Previous Work
Research team started with literature review, general discussions, and methodologies discussions.Step 3Methodology Introduced
Technical consideration and requirements for experimentations were introduced. Then the major methodology of the project was born!Step 4Knowledge Transfer Started
After having the methodology finetuned by researchers and tech teams. Knowledge transfer started. This is a bridge between experts knowledge and systematic automation for tech components of the project approach.Step 5Data Standardization & ML Requirements
ML structure components, formats, and other details were introduced as standards to follow to allow using Machine Learning afterwardsStep 6Large Language Model (LLM) Approach Introduced
After having data labeled and validated by experts according to the methodology and overall data structure required, LLM usage was introduced to help bypass the linguistic complexities and labeling dynamics of the content.Step 6LLM Training and Experimentation
Data labeled and validated by experts were used to train LLM (namely GPT4.0). Prompts were engineered to support the extraction of the knowledge as well as finding narratives among new datasets.Step 7Evaluation and Results
Evaluation to LLM results and finding discussions. Iterative enhancements of LLM training, and other ML take aways!Step 8Website and Databases
Data, findings, research, methodology, steps, privacy, and other information all published on an open-data-and-research website.
Data
Sample 1: Raw Format
ID |
Text |
Original Language |
Views |
Forwards |
Contains Video |
Contains Image |
Contains Link |
1 |
This is a great sample |
Arabic |
234 |
88 |
No |
Yes |
Yes |
Sample 2: Labeled Data
ID |
Text |
Original Language |
Narrative |
Sub-Narratives |
Confidence |
Comments |
|
1 |
This is a great sample |
Arabic |
Narrative 1 |
Sub Narrative 1 |
80% |
This is a comment |
|