Johannes B. Gruber

Better Access: A Framework for Accessing High-Influence Public Platform Data Online platforms and services shape what we know, how we connect, and who gets heard, but the ability to study these posts and content at scale has steadily diminished. Platforms restrict researcher a...

Join us next week Friday, March 20, for a presentation by @leticiabode.bsky.social and @pfchap.bsky.social on "Better Access: A Framework for Accessing High-Influence Public Platform Data". Registration required via events.gwdg.de/event/1260/

Exciting news: @hajoboo.bsky.social will take over as Editor-in-Chief of CCR @computationalcommunication.org from June 7. Hajo is a dream candidate to take over and I have no doubt CCR will continue to flourish under his leadership. Welcome!

journal.computationalcommunication.org/announcement...

The sun is out, so 'enjoyed' a bit less internet on the weekend. How is it going in the world?! 😀

Ah, right...

📢 Study out in Research & Politics!

We (@lukasbirkenmai1.bsky.social @wurthmann.bsky.social @msaeltzer.bsky.social) find that directly elected MPs talk more about local & deprived places than list MPs on social media.

🔍 doi.org/10.1177/2053...

Some Highlights [1/3]

¯\_(ツ)_/¯

Mental model sketch to visualize the trade-off function to consider when choosing tools from the text-as-data toolbox. The plot shows each method’s capability range: the vertical span indicates the task complexity each method can handle, while the horizontal position shows the resources required to implement that method for tasks within its capability. The areas overlap because more sophisticated methods can also solve simpler tasks: LLMs can do everything dictionaries can do, just at higher cost. The “optimal method line” represents the most efficient choice: the least resource-intensive method capable of handling a given task complexity.

After showing it in several presentations already, the publication is where my mental model for choosing a #textasdata method finally found a home. Not perfect, but it has proven useful.

Newer, Larger, Better? A Critique of the Unreflective LLM Adoption in Communication Research The growing adoption of large language models (LLMs) in political communication research has prompted excitement but also concern. In this opinion piece, we offer an informed and critical overview ...

📢Publication alert 📢

LLMs are an awesome technology, but they come with trade-offs!

Paul Balluff, @justinho.bsky.social, @seanhp.bsky.social, Alexis Palmer, @lrossi.bsky.social, Irina Shklovski, Chung-hong Chan and I commented on these trade-offs in #commsky research. Out now and open access!

There is solution for the internet too 😅!

YouTube video by PureRelax4U Intercity Express TRAIN SOUND | ICE Fahrt GERÄUSCH | 4+ HOURS | ASMR Meditation - Relaxation - Sleep

Lol there totally is youtu.be/0zxpSTU3FbE

Maybe there is a ICE white noise sound. Could help you isolate one of the variables. If there is no uptick in productivity, you could put your office chair 9n a vibrating plate or buy a timer that randomly increases the remaining time.

One of my favourite events of the year! Learn about new tools/methods/people and hack on some dataset/tool/project 😍

🚀 Yaay, it's time to register for the hackathon! 🚀

Join us for the #KODAQS #Toolbox #Talk tomorrow!

Yannik Peters will present #TextPrep, a tool to assess how preprocessing methods can improve the quality of social media data.

🗓️ Thursday, 5 Feburary, 2026, 1-2 pm.
📌 Online, register here: www.gesis.org/angebot/wiss...

I was interviewed for a podcast!

I talked with Wendo King'ang'i about (large) language models and opinion mining. It was recorded some time ago, so my affiliation is old, but the topic is still relevant, and I tried to make it as accessible as I could.

Outlook dialogue box asking if the user wants to stay signed in with the box ticked to not show this dialogue box again. The dialogue appears roughly every 12 hours, and ticking the box has no effect.

I think what annoys me most about the Microsoft products I am forced to work with is that they constantly try to gaslight me into thinking I had any control over what they do.

How to precompute package vignettes or pkgdown articles As of earlier this year, we are now automatically building binaries and pkgdown documentation for all rOpenSci packages. One issue we encountered is that some packages include vignettes that require s...

Have you considered precomputing vignettes? ropensci.org/blog/2019/12...

Not the cleanest approach, but I haven't found a better one.

partycoloR is now on CRAN! Started as a simple idea 6 years ago, now it's a full-featured package. Extract party colors and logos from Wikipedia with one line of code. It's already powering ParlGov Dashboard.

install.packages("partycoloR")

Mitarbeiter*in für vertrauenswürdige Forschungsumgebungen (DSS-25) | Jobs bei GESIS – Leibniz-Institut für Sozialwissenschaften GESIS ist eine der weltweit führenden Infrastruktureinrichtungen für die Sozialwissenschaften und steht Forscher*innen mit Expertise und Infrastrukturangeboten auf allen Ebenen ihrer Forschungsprojekt...

There is a permanent position available in my team! You will help researchers gain access to highly sensitive data and analyse them in a secure environment: gesis.jobs.personio.de/job/2495658?...

Go for it and see you in Cologne ❤️🤍

3 rejections this year 💔. But (mostly) very helpful comments and the chance to further develop these projects (maybe for #ica27 in Glasgow?). (And still one accepted presentation 🥵)

I will teach Web Scraping and Data Management at @essexsummerschool.bsky.social this year! Learn to:

- manage research data (in tables, files, databases etc.) 📂
- Systematically get data from the internet (APIs, scraping, browser automation) 🤖
- wrangle your data in #rstats to make it useful 📈

Talking to someone recently provided me with a new perspective on LLM usage: if you don't give a f*ck about the quality and accuracy of the output/product, using LLMs is great 😅

If you do research with/about LLMs, image or audio models, or classic NLP methods, you should check out this call 👇

#textasdata #NLP #llm #commsky #polsky

🧵 PhD position (75%) in Political Behavior / Political Communication / CSS
📍 LMU Munich | ⏳ 3 years | 🗓 start March–May 2026

We’re hiring for DemocraGPT, a @bidt.bsky.social-funded project developing an AI-based training for difficult conversations in times of growing polarization

AAAAAAAAAAAAAAAAAAAAHHHHHHHHHHH

Not long ago I discussed with people what our red lines would be before not traveling to the US anymore. We landed on illegal occupation of another country, like Russia is attempting in Ukraine. We're there now.

…Really undermines the sanctity of the FIFA Peace Prize, doesn’t it?

AAAAAAAAAAAAAAAAAAAAAAAHHHHHHHHH

😍 how did you get Arch on your phone? Is this the native terminal or termux?

Applications via the EU's DSA Data Access Portal: An Introduction Researchers can submit requests to the EU's "Data Access Portal" to obtain non-public data from very large online platforms in certain cases. What information do researchers need to provide and what d...

January 16, 2026: Julian Jaursch from the German Digital Service Coordinator @bnetza.bsky.social will introduce the EU's #DSA Data Access Portal during the series "Show & Tell – Social Media Data in Research Practice". Register here: events.gwdg.de/event/1259/ #digitalservicesact #platformdata

The Call for Papers and Panels for #COMPTEXT2026 in Birmingham (23-25 April) is out; feel free to circulate: shorturl.at/gRg0p!
Deadline: January 16!