Open Call project, 2022 Misinformation campaigns characteristics
Project Outcome
The SSC project led to the creation of ammico (short for AI-based Media and Misinformation Content Analysis tool), a package written in python that enables the detection and extraction of text from images, as well as visual object detection based on natural language questions. This tool made it possible to conduct multimodal (visual-text) analyses at a much larger scale than before, thereby making it possible to ask and answer a host of new questions that pertain to the current communication context dominated by very large communication volumes. This tool makes it therefore possible to track the multimodal communication framing of topics in social and traditional media, across countries and media markets.
The SSC has brought a crucial contribution to my project analyzing mis- and disinformation, by helping me employ state-of-the-art visual and textual analysis software tools to examine this type of communications. Their work is furthermore relevant for the larger community of scholars interested in media communication analysis, as the software instruments they developed for my project are assembled into the AI Media and Misinformation Content Analysis tool, which is freely available to use. I am very happy to have benefited so greatly from their expertise, and I've very much enjoyed our friendly interactions on a personal level.
Dr Delia Dumitrescu, Institute of Political Science (IPW)
Enabling new research
Without ammico, the default option would have been human coders of visual content, using much smaller datasets; this would mean that existing patterns in cross-country, cross-media communication would be impossible to detect at a large scale, but only among a few chosen cases. Alternatively, there are examples in the literature of AI models being trained to detect certain features (e.g., protest), however, this would come at a significant time cost, and would be limited for specific topics. Ammico detection performs similarly well for protest elements, but allows also for detection of many other objects.
The SSC support has made it possible for my research to explore new areas of communication, acquire and analyze significantly larger data sets that can shed light on differences and similarities in the political coverage across a large number of countries, both visually and verbally. Ammico as a tool opens up research avenues for better understanding communication on social media, ranging from normal political coverage to misinformation and hate speech.
Resulting publications
In addition to the joint publication with the SSC (link below), the following publications have resulted from this project:
- “How does War Look Like on the European Mainstream Media Facebook? A Large-N Comparison of the Media Facebook Posts Covering Ukraine and Gaza in 24 European Countries in Spring 2024” (under review)
- “Visions of the War in Ukraine across European Mainstream Media Facebook Sites: A Computational Communication Approach” (presented at EPSA Conference, Madrid June 2024)
- “The Visual-Verbal Framing of the European Farmers Protests in Early Spring 2024: a Cross-European, Computer-Assisted Multimodal Analysis with AMMICO” (manuscript in preparation for journal submission)
- “The coverage of the US Elections in November 2024 in the European Media Facebook” (manuscript in preparation for journal submission)
- “Unpacking the Visual Coverage of the 2024 European Elections on the European Mainstream Media Facebook Sites: A Computational Communication Approach”(manuscript in preparation for journal submission)

Original project outline
The goal of this open-call project was to develop a software supporting researchers to investigate the content of disinformation campaigns worldwide. Using original misinformation data that the International Fact-Checking Network collects from across the globe, the aim of the software is to extract fact- and emotion-based text content, as well as facial expressions and body posture from the data posts. Ultimately, the developed software will help researchers map out the relationship between what is being shown in images and the accompanying text, thereby facilitating research on worldwide disinformation campaigns and their impact in the political sciences.
The SSC aids in the software development with three person/months. See the project repository on GitHub.
This research is carried out by Dr. Delia Dumitrescu, Institute of Political Science at Heidelberg University.
Table
Conference / workshop | Venue |
---|
Conference / workshop | Venue | |
---|---|---|
International Communication Association Conference | Toronto | May 2023 |
International Journal of Press/Politics Conference Edinburgh | Edinburgh | October 2023 |
European Consortium of Political Research Conference | Dublin | August 2024 |
Workshop: “How Image-As-Data Approaches Can Help Analysing Protest and Its Organisation: Methods and Applications” | Berlin | September-October 2024 |
International Journal of Press/Politics Conference Edinburgh | Edinburgh | October 2024 |
European Political Science Association Conference | Madrid | June 2025 |