Tessera
Folgen

Datendealerin 🌲☕

@datendealerin.eurosky.social

Sind das Forschungsdaten oder kann das weg? Research data management and data sharing for the Social Sciences. Mostly. 📍 Bay Area (Kölner Bucht) 🙋‍♀️ she/her 🦣 also in the Fediverse: @Datendealerin@fediscience.org

2008 Posts956 Follower735 Folgt
Frauenhauskoordinierung e.V. (FHK)@frauenhauskoordinierung.de · 3w ago

📈Heute haben wir unsere neue Frauenhaus-Statistik veröffentlicht. 🆘Es fehlen weiterhin 12.000 #Frauenhaus-Plätze. 🚩Kürzungen im #Gewaltschutz und bei entsprechenden Vereinen, NGOs etc. wären absolut fatal. Alle Zahlen, Infos und Details hier: www.frauenhauskoordinierung.de/aktuelles/de...

Frauenhaus-Statistik 2025: Frauenhauskoordinierung warnt vor Kürzungen

Neue Frauenhaus-Statistik zeigt: 12.000 Frauenhausplätze fehlen. Frauenhauskoordinierung warnt vor Kürzungen im Gewaltschutz und bei Fachorganisationen.

frauenhauskoordinierung.de

Daniel Vorgrimler@dvorgrimler.bsky.social · 3w ago

Zum Jahresende 2025 lebten in Deutschland gut 7,8 Millionen Menschen mit schwerer #Behinderung. Gut 91 % der schweren Behinderungen wurden durch eine Krankheit verursacht, 3 % waren angeboren oder traten im ersten Lebensjahr auf. Quelle: www.destatis.de/DE/Presse/Pr...

7,8 Millionen schwerbehinderte Menschen leben in Deutschland

Zum Jahresende 2025 lebten in Deutschland gut 7,8 Millionen Menschen mit schwerer Behinderung. Wie das Statistische Bundesamt (Destatis) mitteilt, waren somit 9,4 % der Menschen in Deutschland schwerb...

destatis.de

RatSWD@ratswd.bsky.social · 2mo ago

📢 „Druck auf die Regierung wächst: Forschungsdatengesetz soll noch 2026 kommen“ 👉 @table.media macht das Forschungsdatengesetz zum Thema des Tages ❗Unsere Vorsitzende, Prof. Dr. Kerstin Schneider, warnt vor weiteren Verzögerungen 📌 Hier geht’s zum Beitrag (Paywall): table.media/research/the...

Druck auf die Regierung wächst: Forschungsdatengesetz soll noch 2026 kommen

Deutschland sitzt auf prall gefüllten Datensilos – und nutzt sie zu wenig. Warum Wissenschaft und Opposition jetzt auf das Forschungsdatengesetz drängen und RatSWD-Vorsitzende Kerstin Schneider vor weiteren Verzögerungen warnt.

table.media

Accessibility Awareness@a11yawareness.bsky.social · 2mo ago

When using Microsoft Word or Google Docs, don't just make text bigger and bolder to make it a heading. That will work for sighted users, but screen reader users will miss that and just hear it as normal paragraph text. Use actual heading styles, like level 1 through 6.

Mar (Herz auf Halbmast)@abstraktesherz.eurosky.social · 2mo ago

Jedes Mal, wenn wieder "alle" ihre Schland Fahnen für irgendein Sportevent aus den Fenstern hängen, muss ich wieder dran denken, dass die WBG mir mal ne Ermahnung geschickt hat, weil ich im Juni mal ne Pride Flag draußen zu hängen hatte. Und das sagt eigentlich alles über dieses Land.

Datendealerin 🌲☕@datendealerin.eurosky.social · 2mo ago

We need to prepare for a new kind of helicopter research 🤦‍♀️🔫 #AIinResearch #ResearchIntegrity

Sarah Werner@wynkenhimself.bsky.social · 2mo ago

German scholar of Japanese history writes a book about slavery in Korea despite the fact that he doesn't have adequate Korean reading skills so instead of collaborating with an actual human Korean expert, he turns to AI to translate thousands of pages of source material. 🤯😡

Sarah Krienen M.A.@sarah-krienen.bsky.social · 2mo ago

The fact that De Gruyter Brill published this is beyond horrifying. I'd also like to mention that according to his own website, this guy is a full professor at Uni Bonn. Both the publisher and the university need to take a very serious look here and draw consequences.

배뢰아@baessi.bsky.social · 2mo ago

일본사 전공한 독일 역사학자가 한국사의 노예제에 관한 책을 냈는데 한국어와 중국어를 못 해서 AI로 번역시켜서 자료를 읽었다고 써 놓음. 😡 무려 De Gruyter Brill 출판사에서… 📖 Zöllner, Reinhard. Slavery and Servile Societies in Korean History: The Hidden Background of Modern Korea, De Gruyter, 2025. doi.org/10.1515/9783...

Without these developments, this book could not have been completed—at least not during my active professional life. My knowledge of Korean is far too limited to read and critically analyze the volume of vernacular academic literature that underlies this study. Without artificial intelligence—especially OpenAI’s ChatGPT (mainly models 4o and 4.5), Google’s Gemini (model 2.0 Flash), and DeepL (version 25)—these texts would have remained inaccessible to me. I used AI primarily to translate thousands of pages of Korean-language articles and books, as well as Chinese sources, into English. This method was adopted only after rigorous evaluation: I consulted native speakers, compared the AI-generated translations with human renditions and the output of other AI models, and back-translated the results to check for semantic integrity and bias. The results were consistently accurate and coherent, giving me confidence in the reliability of the process. Translation has always been a legitimate scientific practice; only now has it become viable on a scale that supports a study of this magnitude.
Dr. Casey Fiesler@cfiesler.bsky.social · 2mo ago

In the "you can't make this up" category of LLM interactions... TL;DR Google Gemini claims that LLMs "often give themselves a 95%+ confidence score even when they are hallucinating" - which was itself a "hallucinated" claim - then later assigns itself 0.95 confidence in that hallucinated claim. 🧵

can you ask an llm for a confidence level 6:01 AMYes, you can ask an LLM to provide a confidence level, but the score it gives is usually based on "vibes" rather than true mathematical probability and is notoriously unreliable. AI models are prone to being overconfident, even when they are factually wrong.If you still want to ask a model for its confidence, you can use a direct prompt technique, though there are much better ways to verify an LLM's answers.1. How to Ask for Verbalized ConfidenceTo ask the model to rate its own certainty, you can append a phrase like this to the end of your prompt:"Provide your answer, followed by a numerical confidence score from 0.0 to 1.0, with no other words or explanation."Keep in mind that research—such as studies discussed by Medium and arXiv—shows that models often give themselves a 95%+ confidence score even when they are hallucinating
Abigail Goben @hedgielib.bsky.social · 2mo ago

Listening to Lindsay Gypin at #IASSIST2026 give this presentation "When Infrastructure Promises Automation But Delivers Manual Labor: The Hidden Cost of Broken Research Data Systems" and it is *striking* how badly these systems do not speak to each other.

Datendealerin 🌲☕@datendealerin.eurosky.social · 2mo ago

So, we have this dataset with open-ended questions about the reasons why researchers don't share analysis code. Worst reply I've found: To insist on it assumes that the researcher is somehow dishonest or always "fishing" for statistical significance. This assumption is often unfounded and unfair. 😒

Gavin Karlmeier@gavinkarlmeier.bsky.social · 2mo ago

Hey ihr tollen Entwickler von Kamerabrillen, hier ist mein Pitch: Alle filmenden Brillen blurren standardmäßig alle(!) anderen Menschen, außer, sie machen eine bestimmte Geste (meinetwegen sogar „👍🏻“) und geben damit ihre explizite Einwilligung gefilmt zu werden. Deal?

Maia George@maiageorge.bsky.social · 2mo ago

Ihr wisst vielleicht, dass ich Fan des Weizenbaum-Instituts bin. Dort gibt's online Mitte Juni eine Veranstaltung zur Lage der Datenarbeiter*innen, die hinter KI, content moderation und vielen unserer Technologien stecken: www.weizenbaum-institut.de/veranstaltun...

Rethinking policies with and for data workers: Labor issues in the governance of AI supply chains

While AI systems are increasingly developed and adopted across diverse sectors, a large workforce in the Global South producing data and verifying automated outputs has remained invisible. These worke...

weizenbaum-institut.de

Datendealerin 🌲☕@datendealerin.eurosky.social · 2mo ago

Auriane Marmier from FORS Switzerland with a similar view on the matter today: Expect “data winters”, during which researchers will cease sharing data due to the negative effects of AI. www.openconf.org/iassist2026/... #IASSIST2026

IASSIST2026 - OpenConf Abstract Submission, Peer Review, and Event Management System

openconf.org

Datendealerin 🌲☕@datendealerin.eurosky.social · 2mo ago

#OpenData and #AI: Researchers who have made their anonymous medical research data openly available are considering retracting it. Due to what AI can potentially do with it. Overheard at #IASSIST2026

Datendealerin 🌲☕@datendealerin.eurosky.social · 2mo ago

Session C.1 of #IASSIST2026 now has a talk by our Interest Group Representation in Data. We meet again tomorrow during the IG discussion slot. Feel free to join us. We always welcome volunteers! ❤️🧡💛💚💙💜 iassistdata.org/community/re...

Representation in Data

IASSIST is an international organization of professionals working in and with information technology and data services to support research and teaching in the social sciences and beyond.

iassistdata.org