08 Oct 2025
The University of Parma team recently presented groundbreaking methodological research at the 38th Annual Conference of the Italian Political Science Association (SISP) in Naples, Italy. The paper, titled “From Manual to Automated Coding: Challenges and Opportunities in Computational Protest Event Analysis” showcased the innovative approaches being developed within INTERFACED’s Work Package 5, which deals with the interactions between social movement organisations and news media institutions since the onset of the pandemic, through protest event analysis, social media analysis, and interviews with activists and journalists.
Presented at the Political Communication section panel “Tracking the Hybrid Media System: Old and New Tools for Measuring Communication Behavior,” the research addresses one of the most pressing methodological challenges in contemporary social movement studies: how to systematically analyse protest events across multiple countries and languages using artificial intelligence.
The research outlined how INTERFACED is analysing protest events across five countries—Hungary, Italy, Romania, Tunisia, and the United Kingdom—from 2020 to 2025. This ambitious scope reflects the project’s commitment to understanding how democratic participation manifests across diverse political and media landscapes, from consolidated democracies to hybrid regimes experiencing democratic backsliding.
For each country, three newspapers representing different ideological orientations will be carefully selected: a high-quality conservative outlet, a progressive publication, and a tabloid or alternative voice. This triadic sampling strategy ensures that multiple perspectives on protest are captured, helping to address the well-known problem of media bias. By analysing how the same events are framed differently across ideological divides, the project reveals not just what protests occur, but how they are constructed as meaningful public events.
The research team will make an innovative use of Large Language Models (LLMs) through zero-shot prompting, a methodological choice that ensures greater transparency and cross-linguistic scalability. This approach allows the project to process vast amounts of multilingual news content while maintaining analytical consistency across different national contexts.
A significant portion of the presentation focused on the practical challenges inherent in AI-based protest analysis. The researchers discussed issues ranging from linguistic complexity in morphologically rich languages to the risks of algorithmic bias and the technical difficulties of event deduplication when the same protest generates multiple articles. Language-specific validation is conducted by national partners who possess deep knowledge of local political contexts. Human-in-the-loop checks are implemented at critical stages—including keyword translation, codebook validation, and sample verification—ensuring that computational efficiency does not compromise accuracy.
Rather than relying on extensive lists of protest-related terms—which often generate overwhelming false positives—the team adopted a minimalist strategy using just two core stems (protest* and demonstrate*) combined with sophisticated AI filtering. This focused approach achieves high recall rates while dramatically reducing noise in the dataset.
What distinguishes this research is its ambition to analyse not only structural variables (who protested, when, where) but also interpretive dimensions such as media frames and legitimacy cues. The project investigates questions like: Are protesters portrayed as legitimate political actors or as disruptive threats? Which grievances receive sympathetic coverage and which are delegitimised?
By integrating both quantitative event characteristics and qualitative framing analysis at scale, INTERFACED is pioneering a more comprehensive understanding of how protest is constructed and represented across different media and political environments.
The SISP presentation generated considerable interest among conference participants, sparking productive discussions about computational social science methods, the ethics of AI-driven research, and the challenges of cross-national comparability. The feedback received will inform the ongoing refinement of the project’s analytical pipeline as the team moves toward full-scale data analysis.
This presentation represents just one example of how INTERFACED partners are actively contributing to scholarly debates and methodological innovation in the study of democratic participation. The project’s commitment to transparency and methodological rigor continues to position it as a reference point for researchers exploring the intersection of social movements, media representation, and computational methods.
Copyrights © 2025 | INTERFACED Cookie & Privacy Policy - University of Parma Presents Cutting-Edge Computational Protest Analysis at SISP 2025 INTERFACED Webdesign & Development by ALYS