Introduction
Dinsinfo-Demasked is a research and technology initiative that focuses on using Large Language Models (LLMs) to discover disinfomration trends among the global south.
Project Overview
The “Disinfo Demasked” project addresses the global threat of information disorders affecting countries in the Global South. This initiative focuses on analyzing propaganda and disinformation targeting the Arabic-speaking world. By leveraging AI and LLMs, the project unveils the underlying narratives in these disinformation campaigns. Although Russian propaganda serves as a study case, the methodologies and findings can be extended to other sources of propaganda and disinformation
Key Features
- Collaboration and Expertise: The project was initiated by DW Akademie and implemented in collaboration with Deutsche Welle’s Research & Cooperations Unit (DW ReCo) and SocialLab as the technology partner.
- Data Collection and Analysis: Data was collected from the Telegram channels. The data was analyzed using both human and machine labeling to identify key narratives.
- Narrative Detection: A rigorous methodology was developed for narrative coding, involving the selection and manual labeling of the most engaged posts.
- Machine Learning Experiments: The project utilized machine learning for automated labeling, demonstrating the potential of AI in narrative detection even with limited datasets. The results were validated through human review, ensuring accuracy and reliability.
Target Audience
- Journalists and Fact-checkers: By providing tools and insights to better understand and counteract state-sponsored disinformation.
- Academics and Researchers: Offering a comprehensive dataset and methodologies for further study on disinformation and media narratives.
- Policymakers and Government Organizations: Assisting in developing strategies to counter disinformation and safeguard public discourse.
- General Public: Enhancing media literacy and promoting critical consumption of news and information.
Technology
The technology stack for this project includes:
- Social Listening Technologies: Used to monitor and analyze social media content.
- Large Language Models (LLMs): Leveraged for their potential to analyze vast amounts of textual data and detect underlying narratives.
- Proprietary and Open-Source Tools: Employed for data collection and processing, ensuring comprehensive and accurate analysis.
Partnership
The project was spearheaded by DW Akademie, Deutsche Welle’s center for international media development, in collaboration with Deutsche Welle’s Research & Cooperations Unit (DW ReCo). SocialLab, a multinational AI service company, served as the technology partner, bringing in their expertise in AI, machine learning, and data analysis.
Disclaimer
This project was implemented in partnership with DWA in Germany. If you have questions about the initiative and the topics chosen, please get in touch with DWA team at disinfo-demasked@dw.com
More Information
Link: https://disinfo-demasked.com
Launching Date: April 2024
Department: SocialLab Innovation Factory
