Hendrik Meyer

Polarised media framing of climate movements: A comparative mixed-methods analysis of Australia and Germany - Katharina Esau, Hendrik Meyer, Mike Farjam, Axel Bruns, Helena Rauxloh, Michael Brüggemann... This study investigates the media framing of climate movements in Australia and Germany. Media frames are powerful tools for shaping public perceptions of socia...

New mixed-methods study out in Media International Australia! We compare 🇦🇺 & 🇩🇪 news framing of climate movements — revealing media polarisation: right-leaning outlets often frame activists as threats, while centre/left ones portray them less negatively. 👉 doi.org/10.1177/1329...

Beyond anti-elitism and out-group attacks: how concerns shape the AfD’s populist representation on German TikTok during the 2024 European elections TikTok is a pivotal platform for political communication, especially among younger users. This study examines how the German rightwing populist party Alternative for Germany (AfD) used TikTok to di...

🔗 Full #OpenAccess Article:
www.tandfonline.com/doi/full/10....

🌄 Limitations & Outlook (follow-up studies are in the making)

– Analysis focused on elites; the fringes of far-right TikTok may look quite different.
– Since EU elections, platform dynamics may have shifted, with other parties (e.g., Die Linke) catching up ahead of the 2025 federal election.

⏩ Overall, the AfD generates engagement through divisive frames yet sustains attention — while veiling its radical and extremist positions — by blending established far-right narratives with everyday concerns.

– Out-group and migration-related themes appeared less often but still generated strong engagement.

This may reflect ...
(a) a strategic shift toward themes that resonate with citizens, and/or
(b) entrenched exclusionary tropes of 'thick populism' that remain implicit in much of the discourse.

❗/💡 Discussion / Interpretation
– Beyond identity-based attacks, the AfD strategically foregrounded real-world concerns that resonate with potential voters ... while still intertwining them with anti-elite cues.

– Horizontal protectionism (e.g., migration critique, gender/wokeness) was least frequent (~30%).
– AfD dominance: higher overall engagement rate and far greater output (from accounts with ≥100k lifetime likes) than all other German parties combined.

📈 Results: What Was Posted & What Drives Engagement?
– Anti-elitism and out-group attacks generated higher per-video engagement.
– Yet most content leaned toward anti-elite messages or concern-focused themes (economy/inflation, security, rights/freedoms).

🧩 Populist Themes & Types (theory-grounded)
We identified 12 themes organized into 3 populist types:

1. Horizontal protectionism (identity/out-group: migrants, “wokeism”)
2. Vertical protectionism (anti-elitism/anti-institutionalism)
3. Concerns of the people (economy/inflation, security, freedoms)

🔍 Data & Methods
Timeframe: Mar–Jun 2024
– Content analysis: LLM-enhanced topic modeling (based on 'Concept Induction' by Lam et al., 2024) on 1,271 AfD video transcripts from 54 AfD accounts.
– Engagement comparison: Platform metrics from 109 politicians (5,590 videos in total)

❓ Research Interests
– How did AfD politicians use TikTok to communicate populist content during the run-up to the EU elections?
– How did these communication strategies relate to user engagement?

Excited to share our new #OpenAccess article in Information, Communication & Society:
“Beyond anti-elitism and out-group attacks: How concerns shape the AfD’s #populist representation on German #TikTok during the 2024 #EU #elections”
/w @julialenz.bsky.social, L. Rodeck & @drfell.bsky.social
(⬇️ 🧵)

From disruptive protests to disrupted news frames: Comparing German news on climate protests - Hendrik Meyer, Mike Farjam, Helena Rauxloh, Michael Brüggemann, 2025 This paper examines how news frames on disruptive climate protests differ from coverage of conventional protests in news outlets with different political leanin...

🛜 Link to full article (Open Access):
doi.org/10.1177/1464...

3️⃣ This also calls for journalistic self-reflection: Is the broader media sphere amplifying narratives set by right-wing populist actors—even when attempting to deconstruct them?

2️⃣ Crucially, when covering LG, the media sphere converges on narratives used mainly by right-wing populist outlets for FFF, even when some outlets aim to critique/deconstruct those narratives.

❗ Main takeaways
1️⃣ The protest paradigm intensifies with disruptiveness: event-, criminality-, and extremism-focused coverage crowds out substantive questions of climate justice.

- The debate about protest demands recedes when protests are disruptive; extremism/criminality take center stage.
- Qualitative evidence shows that, based on newsroom ideologies, outlets still differ in evaluation and in who the emotional language targets—but the topical focus nonetheless converges.

🔎 What we find
- For FFF, frame use and anger align with newsroom ideology: right-leaning outlets stress criminality/extremism; other outlets reference climate justice more often.
- For LG, coverage is more emotionally charged and dominated by criminality/extremism frames across newsroom ideologies.

🧪 What we did
- Combined Word2Vec semantic mapping, an anger classifier, and qualitative close reading.
- Identified three salient frames: Global Climate Justice, Criminality, Extremism.
- Modeled outlet differences across the ideological spectrum.

➡️ We analyzed 6,632 news articles from 21 German outlets (Jan 2022–May 2023) to compare coverage of Fridays for Future (FFF) and the more disruptive Letzte Generation (LG).

📢 Happy to see this #OpenAccess paper out in #Journalism!
"From disruptive protests to disrupted news frames: Comparing German news on climate protests"
with @mikefarjam.bsky.social, Helena Rauxloh & Michael Brüggemann

A Practical Guide and Case Study on How to Instruct #LLMs for Automated Coding During #ContentAnalysis |@mikefarjam.bsky.social, ‪@hendrikmeyer.bsky.social‬, Meike Lohkamp

journals.sagepub.com...

#GenerativeAI

Okay, puh --- also kein falscher Link von mir :D

Super, dann sind die technical issues auf der Seite wohl gelöst & danke! :)

If you’re working with LLMs for text analysis, this framework might serve as a foundation for building your own coding pipeline around LLM-based classification (?) :)

LLM disagreement can potentially serve both as an indicator of model limitations and as a diagnostic tool—revealing deeper issues in the operationalization of coding categories.

➡️ When reliability drops, it may not always be the model that fails—in our use case, these shortcomings may result from the coding category being too vague, the documentation too superficial, or the concept evoking different understandings.

📉 We also reflect on the limitations of our approach when applied to particularly complex coding categories: For constructs like “efficacy,” our results suggest that we may need to rethink the categories themselves—not just the tools we use to code them.

Humans remain in the loop—ensuring reliability and coherence between theory and classification.

We show that this setup outperforms unstructured prompting—particularly for more complex constructs—achieving significantly higher intercoder reliability, comparable to that of human coders.

Rather than relying on a single model run, our pipeline combines:
- an Instructor LLM that generates optimized prompts,
- multiple iterations of a Classifier LLM that codes texts with explicit reasoning,
- and a Judge LLM that adjudicates between outputs when inconsistencies arise.