Sami Nenno

Wie viel #Desinformation verbreiten Politiker*innen? Auf welchen Plattformen & wozu? Auf dem #DigitalSociety Blog zeigt @saminenno.bsky.social mit zwei Studien, warum Desinformation kein reines #SocialMedia Phänomen ist und welche #Parteien besonders aktiv sind.

Lesen 👉 www.hiig.de/wer-verbreit...

Maybe interesting to @steuergerecht.bsky.social, @ungleichheitinfo.bsky.social

Big thanks to my co-author @lorenzspreen.bsky.social from @tudresden.bsky.social

Our take-home message: Value-based arguments are unlikely to cross the aisle and instead preach to the choir. To reach beyond their core electorate, left-wing parties should directly address economic concerns. 🧵5/5
Preprint on OSF: osf.io/preprints/so...

Supporters and opponents of stricter wealth taxation talk past each other. Economically left-wing parties use value-based arguments emphasizing inequality, injustice, and responsibility, while right-wing parties focus on negative consequences for the economy, businesses, and (housing) property.🧵4/5

However, in recent years: We observe a moderate increase in posts, mainly driven by economically left-leaning parties advocating for stricter wealth taxation. Fun fact: the business lobby ran numerous Meta ads against wealth taxation during the 2021 election. 🧵3/5

Our first finding: Politicians posted about wealth taxation less often than about other issues such as unemployment benefits or the nuclear phase-out. Along with gender-sensitive language, wealth taxation ranks among the moderately discussed topics. 🧵2/5

How do German MPs debate wealth taxation on social media?

Although wealth in Germany is distributed very unequally and taxation could address this, the topic rarely features in public debate. This was our starting point for examining politicians’ FB, Insta, and X posts from the past nine years 🧵1/5

Finally, a big thanks to my co-authors! From my team at @tudresden.bsky.social: @lorenzspreen.bsky.social and Kamil Fuławka, and from @zemki.bsky.social : @cbpuschmann.bsky.social 8/8

To end with some good news: this also means there are plenty of topics and actors where misinformation is almost absent. We should focus on high-prevalence contexts, but can remain (cautiously) optimistic about those with low prevalence. 7/8

This is a multi-panel figure that shows the probability of a misinformation posts split by party and topic. From top left to bottom right: Macroeconomics, law and crime, agriculture, energy, health, immigration, social welfare, defense, other. The party at the top is BSW, followed by AfD, CDU, CSU, FDP, Linke, SPD, and Grüne.

Overall, misinformation makes up just above 1% of posts. Yet, for some parties on certain topics, the probability of misinformation can be at 10%. So, while the overall rate is low, misinformation isn’t uniformly rare — it’s clustered in specific contexts. 6/8

Now to our study’s core idea: recent research has focused on the overall spread of misinformation. But is that the right question?
What if misinformation isn’t evenly distributed, but instead concentrated among specific actors and topics? 5/8

This is a multi-panel figure that shows the share of misinformation on posts by each party and their contribution to all misinformation. The first column shows this across all platforms, while to following columns show it for Facebook, X, Instagram, and TikTok.

The misinformation rate is highest for BSW, followed by AfD, and then CSU/CDU. This pattern is fairly stable across platforms.
However, party contributions to overall misinformation differ: on Facebook, X, and TikTok, AfD accounts for most misinformation, while on Instagram, it’s the CDU. 4/8

This is a multi-panel figure. It shows that the proportion of posts with links to news domains is at 4.17% and that of posts with texts is at 91.76% (Panel A). Panel B show the procedure of the domain-level method that detects links to untrustworthy news domains in posts. Panel C shows our text-level method that is called Retrieval Augmented Classification and uses fact-checks and community notes to match them with posts by first deploying vector embeddings and then prompting a LLM. The last panel shows the overlap of identified misinformation by the methods and illustrates that the intersection is small, the text-level methods finds more misinformation and the domain-level method is restricted to Facebook and X.

Only about 4% of posts on FB, Insta, X, TT by German politicians include news links but over 90% contain text. Our text-level method, which matches fact-checks and community notes with posts, detects about ten times more misinformation than the news-domain approach. 3/8

Previous studies identified misinformation via link sharing to news domains. However, news sharing has always been just a fraction of social media activity, has declined in recent years, and isn’t possible on all platforms. This is reflected in the numbers… 2/8

Happy to share our new preprint “Content-based detection of misinformation expands its scope across politicians and platforms.” We analyzed misinformation at the text level in posts by German politicians on Facebook, Instagram, X, and TikTok.
osf.io/preprints/so... 🧵1/8

We have a new preprint: osf.io/preprints/so...

What have we learned about social media - the constantly moving target of empirical research - over the past decade?

Separate worlds of misinformation. An explorative study of checked claims in German public broadcast news and talk shows Research on misinformation has mostly focused on social and alternative media. By contrast, we know little about the extent of misinformation in mainstream media. While many argue that it acts as a...

#OutNow in #iCS
This study maps fact-checked claims across German public broadcasting, showing that misinformation largely remains separate from social media content. Yet in political talk shows, unverified statements often persist unchallenged.
www.tandfonline.com/doi/full/10....

I’m happy to share my first publication after my PhD, during my move from @hiigberlin.bsky.social to the CSS group led by @lorenzspreen.bsky.social. 🧵10/10

This study has many limitations. That is why I call it exploratory rather than representative. But take a look yourself! 🧵9/10

But the difference lies in targets: CDU/CSU use misinformation mainly against political opponents. AfD uses it to undermine democratic institutions. 🧵8/10

I also found a pattern on the supply side of misinformation. In quantity, politicians from AfD and CDU/CSU spread about the same amount (though conservatives are invited more often).
🧵7/10

That matters: research shows that failing to challenge radical claims makes audiences believe they are publicly accepted. This may also apply to misinformation.
🧵6/10

The talk show results aren’t surprising. In live, heated debates it’s hard to stop guests (often politicians) from spreading misinformation. Sometimes moderators challenged the claims. But in many cases, they didn’t. 🧵5/10

Result: about 12% of talk shows contained at least one match, compared to just over 2% of news programs. 🧵4/10

I analyzed subtitles from four news and six talk shows over one year. Too much material to check manually. So I used vector embeddings and LLMs to match fact-checks with subtitle segments. If a claim could be debunked with a fact-check, I counted it as misinformation.
🧵3/10

I examined German public broadcasting news (e.g., Tagesschau) and talk shows (e.g., Markus Lanz). Are they gatekeepers or amplifiers of misinformation? Spoiler: the truth might be somewhere in between. 🧵2/10

Separate worlds of misinformation. An explorative study of checked claims in German public broadcast news and talk shows Research on misinformation has mostly focused on social and alternative media. By contrast, we know little about the extent of misinformation in mainstream media. While many argue that it acts as a...

Excited to share that my paper "Separate Worlds of Misinformation" has just been published. When we talk about misinformation, we usually think of social media. But what about other information environments, even the most trusted ones?
www.tandfonline.com/doi/full/10....
🧵1/10

Postdoc Computational Social Science Am Leibniz-Institut für Medienforschung | Hans-Bredow-Institut (HBI) in Hamburg ist ab dem 1. Mai 2025 folgende Stelle zu besetzen: Wissenschaftliche*r Mitarbeiter*in (Postdoc) (m/w/d) im Bereich Comp...

Our team at the Leibniz-Institut has two 100% vacancies! Deadline is 23.03.2025!

@dgpuk-meth.bsky.social

leibniz-hbi.de/hbi-news/ste...
leibniz-hbi.de/hbi-news/ste...

Elon Musk wants far right AfD to win the German election - here’s how he’s doing it Germany’s leading political figures say Musk will face consequences for his endorsement of the country’s far-right party, Tom Watling and Alicja Hagopian write

And another report on our work (with @saminenno.bsky.social), very pleased with the first output of our brand new group (css-synosys.github.io).

www.independent.co.uk/news/world/e...