DEEN
02 — Research

Lab

Research you can get your hands on: my projects from university, many with data to click through yourself.

Filter by topic
Master’s · since 2025

M.A. Journalism & Communication Science

Universität Hamburg
2026/27
Analysis underway
Project report · group work
Research project “Political Communication Using Short-Form Video” · Summer 2026 & Winter 2026/27

Personalised feeds, polarised publics?

What do first-time voters see on TikTok ahead of the state election in Mecklenburg-Western Pomerania?

Which political content does TikTok serve young people before their first state election, and how much depends on their own profile? For a sock puppet audit, we steered ten test accounts with different interests through the “For You” page by hand until the day before the election and recorded every video we were shown.

10sock puppet accounts
5profile groups
6weeks of data collection up to election eve
15 minof use per account and week
The profile groups
Apoliticalonly apolitical interests & creators2 accounts
Progressiveprogressive media, parties, social & environmental issues2 accounts
Conservativeconservative & liberal media, economy, taxes, migration2 accounts
Right-wing populistright-wing media, AfD circles, anti-migration topics2 accounts
Controlno follows, no targeted interactions2 accounts

One female and one male profile per group, one with an urban and one with a rural home town in the state. Coding is still underway, but here’s a first look at the raw data.

Try it yourself: how much politics lands in the feed, week by week

Preliminary: 9,208 videos served across 7 collection waves, 8,216 of them with a transcript of the audio. Political reference means a party, an election, a parliament or a leading politician appears in the caption, hashtags, on-screen text or speech (automatic keyword search, not a substitute for our coding). Click the legend to hide or show lines.

Try it yourself: what it’s about when it gets political

Share of all videos with a political reference in each group, based on the party, issue and level variables in our codebook. A video can cover several parties, issues and levels. The lighter the cell, the more frequent.

First observations
  • Roughly one in seven videos served has a recognisable political reference (13.5 %). Politics remains the exception on the “For You” page.
  • The right-wing populist profile sees the most politics (33 %), followed by the progressive one (27 %). The apolitical accounts see almost none (just under 3 %).
  • When a party shows up, it’s usually the AfD: it appears in around half of all political videos, and in 60 % for the control accounts without any follows.
  • Despite the state election, politics in the feed is mostly federal: federal politics comes up in 38 % of political videos, state politics in only 11 %. The most common issue is migration.
Algorithm auditSock puppet methodQuantitative content analysis
2026
Completed
Course paper
Seminar “Advanced Digital Methods” · Summer 2026

Between opportunity and criticism

(Changing) frames in Hamburg parliamentary speeches on the city’s Olympic bids

How do the parties in Hamburg’s parliament justify their stance on the Olympics, and how has that changed between the first bid and the 2026 referendum? It’s based on every plenary speech mentioning the Olympics since 2001, which I processed with a custom Python parser.

473plenary speeches
81plenary records
2001–2026period of study
4debate phases
How the arguments shift
CDUincreasingly frames the bid as an opportunity for the future
AfDfrom support to criticism of procedure and legitimacy
GREENSmore support, but sharper criticism at the same time
THE LEFTconsistently critical of the costs
Try it yourself: how the arguments shift

Share of each argument type among all coded words of a party in that phase (dictionary analysis). Net = opportunity share minus criticism share in percentage points. Pick two phases to compare above. Hollow dots rest on few hits; the FDP was not in parliament in 2023–26.

Framing analysisAutomated content analysisPythonR
Bachelor’s · 2021–2025

B.A. Political Science

Universität Hamburg
2025
Grade 1.0
Bachelor’s thesis
First supervisor: Prof. Dr. Judith Möller

Climate change denial on TikTok

A content analysis of TikTok videos tagged #Klimawandel

Is climate change denial favoured by TikTok’s algorithm? I coded videos posted under #Klimawandel (German for “climate change”) and looked at how denial narratives and news factors relate to reach and engagement.

117videos analysed
29 %with a denial narrative
+153 %more views with visible emotionality
Try it yourself: with vs. without feature, average per video

Means from 117 coded videos. On the right: how much more or less videos with the feature reach on average. Hollow dots rest on fewer than 10 videos; the thesis’ regression models account for more factors.

Key findings
  • Videos with denial narratives get clearly fewer views on average, so the idea of blanket algorithmic favouritism doesn’t hold up.
  • Even so, they’re shared just as often, so their “shareability” is high.
  • Classic news factors still count in the age of algorithms, above all emotionality and a personalised presentation of facts.
Quantitative content analysisNews value theoryRegression analysisStata

More papers from the bachelor’s

Project paperGrade 1.0

Agenda setting in the GOP primaries

Project seminar “Coverage of the Primary Elections” · Summer 2024

A content analysis of CNN’s and Fox News’ online coverage after the fourth Republican TV debate in 2023. The result: the two networks differ less than expected in how often they mention candidates and issues, with no significant differences.

Quantitative content analysisAgenda settingRegression analysisR
Try it yourself: Fox News vs. CNN

29 online articles after the fourth TV debate on 6 December 2023: 8 from Fox News, 21 from CNN. The regressions show no significant differences, so with numbers this small, read with care.

Research designGrade 1.0

Populist parties and parliamentary discipline

Seminar “Comparative empirical populism research” · Summer 2023

A research design asking whether populist parties change the culture of debate, measured by calls to order in the Bundestag across legislative terms 1 to 19 and comparing before and after the PDS (1990) and the AfD (2017) entered parliament.

Research designPopulism researchLongitudinal analysis
Results report · group workGrade 1.3

Visibility bias in sh:z election coverage

Reading course “Political news and voting decisions” · Summer 2022

We analysed 58 online articles by the Schleswig-Holsteinischer Zeitungsverlag on the 2022 state election. Compared with the election result, the FDP, SSW and Greens appeared too often in the headlines and the CDU clearly too rarely.

Quantitative content analysisMedia biasSignificance testsStata
Try it yourself: visibility vs. election result

58 sh:z online articles from 11 April to election day on 8 May 2022. Each arrow runs from the election result to the party’s share of all party mentions in the headlines. The CDU clearly gets too little attention.

Research project · group workGrade 1.3

Multilingual advertising for the Thalia Theater

Introductory course “Methods of Empirical Social Research” · Partner: Thalia Theater Hamburg · Summer 2022

Can advertising in several languages make the Thalia Theater’s audience more diverse? We ran an online survey, partly at the “Embassy of Hope” café. Around two thirds of respondents whose first language isn’t German found posters with the title in their own language more appealing, and almost 63 % preferred posters mentioning surtitles.

Standardised surveyQuestionnaire designAudience analysis
Try it yourself: what respondents said
Arabic n=8
87.5 %
English n=3
33.3 %
Turkish n=3
33.3 %
Other language (title in English) n=24
62.5 %

Online survey from 2022 with 94 participants, broken down by first language. Some groups are very small, so read these as tendencies.

Term paperGrade 1.0

Is the CDU still a catch-all party?

Introductory course “Introduction to Political Science” · Winter 2021/22

I looked at the CDU through Otto Kirchheimer’s party typology. In terms of content, its manifesto meets the criteria of a catch-all party (“Volkspartei”), but structurally it fits better with types such as the cartel party.

Party researchDocument analysis