Hendrik Erz, PhD
@hendrik-erz.de
Political Sociologist, studying discursive dynamics and democratic lawmaking. Program Chair MetaCRiSP 2026 (IEEE S&P) Website: https://hendrik-erz.de I develop @zettlr.com, a free academic writing app.
+++ NEW ARTICLE +++ Entitlement and Culture in Open Source Projects, amplified by LLMs. www.hendrik-erz.de/post/entitle...
Entitlement and Culture
Open source always had issues with entitled people. Many people don't make a distinction between paid software and free software, and as such it is …
hendrik-erz.de
What studying U.S. Congress for six years does to your political stance.
An accurate depiction of the pain of every text analysis scholar.
New-Cleckit Dominie@ncdominie.bsky.social · last wk.This is just to say I have OCRed the text thet was in tne arch:ve and whicb yov were prabaldy seamhing for "breahtast" Fargme ms fheg mem debo:w5 sb ?mccf mmb 5c oc1b
+++ NEW ARTICLE +++ Your weekend read is here: I illuminate the relationship between code and data and explain what LLMs are made of. Also, of course, some tips for social scientists and our R code! hendrik-erz.de/post/llms-an...
LLMs and the Blurry Distinction Between Code and Data
There is still a lot of confusion about what LLMs actually "are." A very helpful mental model to understand LLMs a bit better is to recognize the …
hendrik-erz.de
The fact that the currently most unhinged take on the improvements in Chinese LLMs comes both from the biggest racketeering company and, more specifically, an important Heritage Foundation fellow, tells you everything you need to know that OpenAI is scared sh**less. Which is good.
Oh sure, because the politics academia subordinates itself to is wokeness. Famously the most anti-scientific politics of them all. Jesus Christ, why are people so incredibly tasteless?
I agree that peer review has its issues, but this take is just insane. “Experts point to good ones” is not how science works. The solution to AI is not to throw one’s hands up in the air and give up, but by enforcing ethical standards. Defeat is not an option.
Dark times are upon us. I just received a bunch of perfectly valid and useful PRs for Zettlr from what I could gather, but the descriptions don't follow the template, and reek of LLM-speak. I really don't want to close them prematurely, but I really also don't want to talk to an LLM.
But ethical considerations don't stop with impacts on the people being directly studied (even if IRB concerns usually do). Among other things, I worry about this study legitimizing using review-shaped blocks of synthetic text as anything of value in the actual process of peer review. >>
Also, @emilymbender.bsky.social just made this point much more sternly: bsky.app/profile/emil...
Hendrik Erz, PhD@hendrik-erz.de · 3w agoIf Claude can write your paper, why do you still work that job and take the grant money? If you don’t have to think about results yourself anymore, why don’t you let an agent do it for you? LLMs are causing a deep, nihilistic trend where science stops being a purpose.
Lots of thoughts on this piece, but one dimension seems to be amiss right now: a moral and science logical dimension that pertain to the “what” and “why” of doing science in the first place. www.science.org/doi/10.1126/...
AI in scientific publishing: Slower, worse, and more expensive
There’s a saying in the management world, popularized by NASA administrator Daniel Goldin in the 1990s, that the goal of technological improvements is to make products faster, better, and cheaper. Alt...
science.org
This article ends with a good conclusion, but it is extremely misguided along the way. It is also a rich text for exploring how clear language helps make the point better. A thread: www.science.org/doi/10.1126/... >>
AI in scientific publishing: Slower, worse, and more expensive
There’s a saying in the management world, popularized by NASA administrator Daniel Goldin in the 1990s, that the goal of technological improvements is to make products faster, better, and cheaper. Alt...
science.org
"AI slop in the scientific record and generally on the internet is making science less trustworthy ...AI enthusiasts are telling the world that the remedies should all be easy to automate...That’s not the reality for scientific publishing right now or likely in the future." @holdenthorp.bsky.social
AI in scientific publishing: Slower, worse, and more expensive
There’s a saying in the management world, popularized by NASA administrator Daniel Goldin in the 1990s, that the goal of technological improvements is to make products faster, better, and cheaper. Alt...
science.org
I’d counter that the dominance of a Democrat-Republican system is more the exception rather than the norm. Once you take a look at Congress in the 19th century, you suddenly have insanely volatile parties. My guess is that the Reps will split and one group becomes their successor.
Josh Zingher@zingherpolisci.bsky.social · 3w agoAmerica has weak parties, that's true, but i don't think we've really wrestled with just how much of the Republican infrastructure is Trump specific now. It is unclear to me how Republican politics will function without Trump. He's the institutions at this point. It's not clear they're transferable.
If you don’t follow German politics closely, this is how polls look like since a few months. The two new major parties are center right and right-wing. Poland and Hungary may be back on a more democratic track, but Germany is heading the other way.
Wahlrecht.de@wahlrecht.de · 3w agoSonntagsfrage zur Bundestagswahl • INSA/BamS: AfD 29 % | CDU/CSU 21 % | SPD 13 % | GRÜNE 13 % | DIE LINKE 10 % | FDP 4 % | BSW 3 % | Sonstige 7 % ➤ Übersicht: wahlrecht.de/umfragen/ ➤ Verlauf: wahlrecht.de/umfragen/ins...
+++ NEW ARTICLE +++ Stop taking "smart notes." Start doing whatever brings you closer to the goal of having a finished paper. I argue that the online productivity discourse is snake oil and a distraction from reflecting on your workflow. www.hendrik-erz.de/post/how-to-...
How to Smartly Take Notes
Taking "smart" notes is a business. Dozens of influencers on YouTube, Instagram, LinkedIn, and the likes, try to sell you the next best workflow, …
hendrik-erz.de
+++ NEW ARTICLE +++ Some reflections on the INAS 2026 conference in Oxford last week, and the future of political sociology in analytical sociology. www.hendrik-erz.de/post/some-re...
Some Reflections on INAS 2026
Last week, the 2026 edition of the INAS conference convened at the birthplace of analytical sociology - Nuffield College at Oxford University. It was …
hendrik-erz.de
SocArXiv is 10 years old today! On July 7, 2016, PNC said: "We are building the future of social science scholarly communication. It’s past time for social scientists to bring their work out into the open, to make it better, faster, more accountable, and more transparent.” Now we host 21,341 papers.
Announcing the development of SocArXiv, an open social science archive
It’s past time for social scientists to bring their work out into the open, to make it better, faster, more accountable, and more transparent.
socopen.org
Most companies will send you an email a few weeks in advance when they update their terms. Apple, on the other hand, does not send emails. They send HTTP error codes. (P.S.: the terms update that triggered this error goes into effect TOMORROW.)
Unsere neue Ausgabe mit dem Titel ,,(Un-)Sichtbarkeiten'' ist erschienen. Diese Ausgabe ist eine besondere, denn es ist ein Doppelheft mit einer großen Vielfalt an Themen soziologieblog.hypotheses.org/soziologiema...
Soziologiemagazin
Das Soziologiemagazin ist unsere Hauptpublikation. Das Soziologiemagazin erscheint seit 2009 und seit 2011 immer halbjährlich Mitte und Ende des Jahres. Es handelt sich um eine Open Access-Publikation...
soziologieblog.hypotheses.org
Not gonna lie: the left hand driving in the UK messes with some hard wired connections in my brain more than I’m willing to admit.
It was so hot today that I had to actively cool my watch with a frozen stone for it to actually charge.
+++ NEW ARTICLE +++ Exploring solid use cases for LLMs is hard because most treat their “prompts” either as a trade secret or as not important. Both stances are bad. In this article, I change this by sharing useful academic “prompts” with you. www.hendrik-erz.de/post/how-i-u...
How I use Generative AI in My Work
Locally-running generative AI has made considerable jumps in quality in the past three years. I think it is finally time to evaluate such models in …
hendrik-erz.de
I'm currently saving for a new MBP. Appears I will have to add three more months of saving before I'll be able to afford this. AI companies are garbage and need to be bankrupted.
The modern day book-burnings. It's ironic how Google Books now looks so benign because they cared enough for books not to destroy them. Unlike Anthropic. I'm all in for digitizing books and good LLMs. But not like this.
Carolin Amlinger@camlinger.bsky.social · last mo.Dass die KI nun wortwörtlich Bücher frisst, ist auch ein Symptom der zynischen Vernunft unserer Zeit www.literaturcafe.de/kaufen-ki-un...
Short update: I made it … at 5am and after an entire 22 hours on the train. Just to give you a comparison: A plane from Europe to the East Coast of the U.S. takes ca. 9hrs and makes about 10,000 km. I was stuck on the final train for the same amount of time but only made 300 km.
Hendrik Erz, PhD@hendrik-erz.de · last mo.Something new that I didn’t have before: apparently the train radio ACROSS THE ENTIRETY OF GERMANY has just failed, meaning that for the first time in my life (ever?) there is NOT A SINGLE TRAIN in the entirety of Germany going because they … just can’t. Jesus.
It’s Reddit, so not at all authoritative, but allegedly the first engine operators are able to register in the radio network again. We’re grasping for straws here
Hendrik Erz, PhD@hendrik-erz.de · last mo.Something new that I didn’t have before: apparently the train radio ACROSS THE ENTIRETY OF GERMANY has just failed, meaning that for the first time in my life (ever?) there is NOT A SINGLE TRAIN in the entirety of Germany going because they … just can’t. Jesus.
Something new that I didn’t have before: apparently the train radio ACROSS THE ENTIRETY OF GERMANY has just failed, meaning that for the first time in my life (ever?) there is NOT A SINGLE TRAIN in the entirety of Germany going because they … just can’t. Jesus.
