Working Papers
Working Papers
-
Far-Right Speeches and Emotional Contagion in Parliament: Sequential and Multimodal Evidence from Germany
Does the parliamentary presence of far-right parties make democratic debate more negative? Existing text-based evidence suggests a more nuanced answer: after the entry of the Alternative for Germany (AfD) into German parliaments, established parties respond with more positive emotional language. We replicate this average textual pattern, but argue that it masks two theoretically important dynamics. First, parliamentary debate is sequential: speeches respond to prior speeches. Second, parliamentary speech is multimodal: affect is expressed not only through words, but also through voice and body. We theorise three mechanisms that link far-right speeches to the responses they provoke — emotional contagion, boundary drawing, and the competition for attention — and test them with an original collection of more than 30,000 hours of parliamentary video from German state parliaments and the Bundestag. We find that after AfD speeches, established parties use more negative emotional language, especially anger, and less positive emotional language. They also become visibly more activated in posture, gesticulation, head movement, and audience interaction, while average vocal pitch remains flat.
-
Introducing Workflow Management to Computational Social Science
Replication failures are pervasive in computational social science, yet their common cause is underappreciated: pipelines are built without software infrastructure for automated execution, portable environments, scalable compute, or systematic documentation of analytical decisions. This is not a crisis of researcher intent but a crisis of infrastructure. We argue that workflow management systems (WMS) — software that represents analytical pipelines as explicit dependency graphs and automates their execution across computational environments — offer a structural solution. Drawing on a framework of six foundational properties adapted from Mölder et al. (2021) and extended with a CSS-specific property capturing analytical decision provenance, we show that a single WMS adoption simultaneously addresses the reproducibility, scalability, and transparency failures documented in the replication literature. We demonstrate the framework with Snakemake (Köster and Rahmann, 2012), applying it to a pipeline that processes more than 30,000 hours of legislative video across all German state parliaments and the Bundestag. WMS are not a niche tool for computational specialists. They are a general-purpose research infrastructure that allows social scientists to automate pipelines, scale to HPC and cloud environments, pin reproducible software environments, and document both computational provenance and analytical choices by construction. This fulfils what the replication debate has demanded for three decades: it greatly matters how research is built, not merely how it is reported.