While the problem of misinformation on social media has recently attracted much attention, such misinformation – whether spread unknowingly or deliberately (i.e., disinformation) – represents only one type of online threat. More broadly, foreign adversaries may seek to shape online narratives surrounding specific geopolitical issues and events to fit their agendas. These narratives often propagate alongside other spontaneously occurring (i.e., bottom-up) conversational narratives, interacting and evolving in surprising ways as they spread in distinct communities. The ability to detect, characterize, and track narratives as they change and spread across multiple social media networks is therefore key to understanding the online information environment. Equally important is the ability to connect narratives to external events, in order to understand how narratives react and change in response.
Kairos Research was recently awarded a Phase 1 SBIR contract from U.S. Special Operations Command (SOCOM) to explore possible approaches for detecting and analyzing dynamic online narratives. The project, titled “Recognizing, Analyzing, and Visualizing Evolving Narratives (RAVEN),” is led by Kairos Senior Research Scientist Dr. Locke Welborn. “While past work on narrative extraction has focused on providing useful ‘snapshots’ of narratives at a given point in time,” explained Dr. Welborn, “it generally fails to distinguish overarching foundational narratives from the concrete focal events that impact those narratives.” When asked to elaborate on the distinction, Dr. Welborn pointed to the war in Ukraine as an example. “After the unprovoked invasion of Ukraine by Russia, the predominant narratives shifted dramatically from a story of Ukrainian victimhood to a ‘David versus Goliath’ story of heroic resistance against an autocratic aggressor. This shift occurred due to a series of focal events surrounding the invasion, in particular Russian battlefield losses, which highlighted the lack of preparedness and competence of the Russian armed forces and the surprising resilience of the Ukrainian military.” To achieve the goals of RAVEN, the Kairos team is collaborating with Dr. Nitin Agarwal, a leading expert on the computational analysis of foreign influence campaigns on social media. The team plans to capitalize on recent advances in artificial intelligence, especially advanced natural language processing techniques, to identify and extract narratives and related events from multilingual social media data. If selected for Phase 2, the team anticipates developing a prototype dashboard tool that will enable information operators to track and visualize multiple narratives over time, as well as trace the flow of narratives across distinct online communities.