This project examines how geopolitical relationships shape international news coverage, focusing on how political alliances influence media attention, tone, and narrative framing. I argue that global reporting is not driven solely by events on the ground, but also by shifting diplomatic alignments, which systematically shape how countries are portrayed.
Using large-scale data from the Media Cloud API, I apply computational text analysis to track changes in sentiment, framing, and coverage volume over time. The analysis combines dictionary-based sentiment methods and topic modeling to identify patterns in how narratives evolve alongside geopolitical change.
This project examines how visual media can help address gaps and distortions in text-based conflict event data. While photographs and video footage are often treated as supplementary to written reporting, they frequently contain critical information that is absent from or underemphasized in text.
Building on work by Andrew Shaver (2023), which demonstrates that image captions can reveal unreported or underreported events using large language models (LLMs), this project moves beyond captions to analyze the visual content itself. Using computer vision and vision-language models (VLMs), I assess whether images capture additional signals, such as the presence of security forces, violence, or context cues.
The broader goal is twofold: first, to evaluate whether visual data systematically reveals conflict indicators overlooked by traditional sources, and second, to assess the capabilities and limitations of current AI systems in extracting meaningful information from complex visual environments. By doing so, the project contributes both to improving measurement in political science and to advancing interdisciplinary work at the intersection of AI and social science, with applications extending to disaster response, climate monitoring, and public health.
This co-authored project with the Political Violence Lab examines how global media systems systematically underreport deaths and disappearances in maritime migration. By constructing a novel dataset that combines incident reports from Securewest and the IOM , it provides one of the most comprehensive global trackers of maritime migrant incidents to date .
The analysis reveals substantial and systematic underreporting by major international news media, with approcimately two thirds of incidents, particularly those involving smaller numbers of casualties, going unreported. As a result, most deaths and disappearances are effectively invisible in global coverage, despite representing the majority of cases.
This co-authored project with the Political Violence Lab examines why some regions within countries receive significantly more international media attention than others. It examines how both resource-driven constraints and narrative framing bias shape patterns of coverage across space.
Using subnational data from global sources (e.g., WorldPop, Ookla, NASA, World Bank, Google Maps) combined with large-scale media data from the Media Cloud API, the project analyzes how infrastructure, geography, and political constraints influence where journalists report. Factors such as connectivity, transportation access, terrain, and proximity to major cities are used to explain systematic gaps in coverage.