My work sits at the intersection of political communication and computational social science. I organize it along three lines: what is changing in political communication, how we can measure it, and how our discipline produces evidence about it. Most of it is funded through the projects listed further below.
01
Substantive
Political communication on platforms
Who speaks about politics online, and with what effect? I study the communicators who now shape political attention alongside parties and newsrooms (social media influencers, news influencers, creators of “tradculture” content), as well as how political elites adapt their tone and style, and how generative AI is entering political debate and the information people receive.
Much of what we want to know about political communication sits in millions of posts, speeches and videos. I develop and validate computational tools to measure it: joint sentiment-topic models, multimodal emotion recognition, and, increasingly, large language models as annotation instruments, with a focus on when these tools can be trusted, and at what cost.
I also study our own discipline: whose languages our tools work for, how many defensible analytic paths a single research question allows, and how editorial rules shape whose work gets cited. This line connects methodological choices to questions of validity, openness and inclusion.
Validity and language bias in computational research
Swiss National Science Foundation (SNSF) · 2024–2028
Political Social Media Influencers: Their Role and Influence on Youth Political Engagement
Scientific lead · PI: Prof. Frank Esser I designed the project and lead it scientifically (PI: Prof. Frank Esser). Team: Eva-Maria Vogel, Morgan Wack.
Can political social media influencers re-engage young people with politics? The project maps who political influencers on Instagram, TikTok and YouTube are and what they post, traces how their frames and vocabulary travel into their audiences’ political conversations, and estimates their effects on political engagement with a panel experiment combined with donated digital trace data. Studies so far cover Austrian and German influencers and the 2024 U.S. election.
Swiss Federal Office of Communications (BAKOM) · 2025–2027
Visual Tradcultures on TikTok (VisTrad)
PI · Co-PI: Seraina Tarnutzer-Zigerli (University of Fribourg)
“Tradwife” and tradculture content on TikTok stages traditional femininity in nostalgic, aesthetically polished videos. VisTrad studies how this content depicts women, and how young people in Switzerland perceive it. It combines computational visual analysis (object detection, image-text models, multimodal LLM coding) of accounts from five countries with qualitative analysis, focus groups across the German-, French- and Italian-speaking regions, and a survey experiment on gender-role attitudes.
Does Training Fix Bias? Student Coder Bias, Training Interventions, and LLM Validation
Principal investigator
Human-coded data are the benchmark against which LLM annotations are validated. This project tests whether student coders carry systematic biases into that benchmark, and whether coder training removes them.
GRC Short Grant · 2026
Perception of Politics during the 2026 Soccer World Cup
Co-applicant
A two-wave panel survey fielded around the 2026 FIFA World Cup in the US and Canada, studying when citizens see politics in sports and how partisanship shapes those judgments.
PhD project, University of Amsterdam · 2019–2024 · completed
Blueprints and Fingerprints: Politicians' Use of Emotional Appeals in European Democracies
PhD candidate (supervised by Gijs Schumacher, Mariken van der Velden and Bert Bakker)
My dissertation measured the emotional tone of political speech across European parliaments, leaders’ speeches and manifestos over three decades, and developed the sentiment-topic models and multimodal tools used to do so.