Two DigiMoD Team Members, Daniel Thiele (Freie Universität/Weizenbaum Institute) and Miriam Milzner (Freie Universität/Weizenbaum Institute), participated in a satellite event of the Association of Internet Researchers Conference (AoIR 2024) that took place at the University of Sheffield, UK.
On October 29th, Daniel Thiele presented “Suspiciously similar. An embedding approach to capturing multimodal coordinated behavior,” co-authored with Miriam Milzner at the Pre-Conference Workshop “Coordinated Sharing Behavior Detection Conference.” Our team members introduced a novel tool for detecting coordinated social media manipulation in this presentation—the R package coorsim. Coordinated social media manipulation denotes a phenomenon in which groups of accounts artificially amplify the visibility of social media content through frequent sharing of similar content. This technique has become a key tool of online disinformation campaigns. Currently, available detection software is limited to capturing the co-dissemination of identical content1 or detecting similar content based on word frequencies2. The novel tool aims to capture sophisticated, coordinated content which is semantically similar but variable in its choice of words—for instance if it has been generated by LLMs.
Following the presentation of their collaborative work, Miriam Milzner, as an accepted Participant of the 2024 Association of Internet Researchers Doctoral Colloquium, attended the AoIR 2024 from October 30th until November 2nd.
—
- Righetti, N., & Balluff, P. (2023). CooRTweet: Coordinated Networks Detection on Social Media (Version 1.5.0) [Computer software]. https://cran.r-project.org/web/packages/CooRTweet/index.html
- Graham, T., & QUT Digital Observatory (2020). Coordination Network Toolkit. Queensland University of Technology (Software) https://doi.org/10.25912/RDF_1632782596538