Benjamin Gagl

"Digitale Auslöschung": Wie US-Sanktionen einen europäischen Richter lahmlegen Nicolas Guillou ist von beinahe allen digitalen Diensten dieser Welt gesperrt – von Amazon bis Paypal. Europa wirkt dagegen völlig machtlos

Diese Geschichte ist unfassbar.
Weil der Trump-Regierung die Arbeit des Internationalen Strafgerichtshofs nicht passt, wird ein französischer Richter digital de facto ausgelöscht und bekommt keine Kreditkarte und kein Bankkonto mehr:

TODAY! In case you are interested ;)

🚨Now out in *Open Mind* 🚨

"Can we utilize Large Language Models (LLMs) to generate useful linguistic corpora? A case study of the word frequency effect in young German readers"

direct.mit.edu/opmi/article...

with
@jobschepens.bsky.social @hannawoloszyn.bsky.social @nicolekmarx.bsky.social

Meaningful Work with Sign Language and Other Factors for Job Satisfaction - Pamela Perniss Professor Dr Pamela Perniss, Chair of Sign Language Interpreting: German Sign Language - German, talks about her love for her research and educational work, her colleagues and students and the special...

Professor Dr Pamela Perniss, Chair of Sign Language Interpreting: German Sign Language - German, talks about her love for her research and educational work and the special contribution sign language makes to people's inclusion in society, education and life👇
andhowaboutyou.podigee.io/28-pamela-pe...

Fun project with much much more nuanced experimenting and discussion of the effects.

Interestingly, over the ten corpora we derived in the study we found a correlation showing that the corpora with lower richness generally generate the more adequate word frequency measures.

However, when using derived frequency to estimate the word frequency effect in the child word recognition behavior model, the fits indicate that the frequency measures based on the LLM corpus captures the effect more adequately than frequency based on the children's book corpus.

Investigating the characteristics that LLMs generate when asked to produce text for children.

We focus on lexical richness and word frequency, showing that the text is generally less rich compared to children's books.

🚨Now out in *Open Mind* 🚨

"Can we utilize Large Language Models (LLMs) to generate useful linguistic corpora? A case study of the word frequency effect in young German readers"

direct.mit.edu/opmi/article...

with
@jobschepens.bsky.social @hannawoloszyn.bsky.social @nicolekmarx.bsky.social

In the contribution by Elif Ecem Çalışkan, we compare human ratings for semantic variables to those produced by LLMs in the Turkish language, which has not been studied a lot to date.

Looking forward to being part of this symposium at TeaP2026. I will present our TRUST guidelines for reading research and the German Lexicon Project (GLP)

Congratulations to our #GeWoNN members on their accepted symposia at the #teap2026 Conference! 

See titles below:

Welcome to join our upcoming Journal Club meetings in Oct & Nov! Happy to have @cyhsieh.bsky.social & Dr. Cosimo Iaia as our guest speakers!

Register to get the Zoom link or sign up as our next guest speaker:
forms.gle/TY6Cuxut87Ej...

More exciting talks coming up. Please share & see you there!

Eye movements during text reading align with the rate of speech production - Nature Human Behaviour Speech is produced and perceived at a rate of 4–5 Hz. Gagl et al. show that the temporal structure of our eye movements during reading is similar (3.9–5.2 Hz), suggesting a temporal link between readi...

This 2021 paper by @benjamingagl.bsky.social and colleagues is full of important findings and ideas about the temporal structure of reading and speech. I keep coming back to these data.

Eye movements during text reading align with the rate of speech production
www.nature.com/articles/s41...

Binz et al. (in press, Nature) developed an LLM called Centaur that better predicts human responses in 159 of 160 behavioural experiments compared to existing cognitive models. See: arxiv.org/abs/2410.20268

Privates Waffenverbot jetzt! Der Amoklauf an einer Grazer Schule zeigt, dass das Waffengesetz in Österreich viel zu locker ist. Das muss sich ändern! Fordere jetzt mit uns ein Waffenverbot für Privatpersonen.

Wir brauchen ein sofortiges #Waffenverbot für Privatpersonen – damit sich eine schreckliche Tat wie #Graz nicht wiederholen kann. Jetzt Appell an Innenminister Karner und die Bundesregierung unterzeichnen: actions.aufstehn.at/waffenverbot...

Privates Waffenverbot jetzt! Der Amoklauf an einer Grazer Schule zeigt, dass das Waffengesetz in Österreich viel zu locker ist. Das muss sich ändern! Fordere jetzt mit uns ein Waffenverbot für Privatpersonen.

Wir brauchen ein sofortiges #Waffenverbot für Privatpersonen – damit sich eine schreckliche Tat wie #Graz nicht wiederholen kann. Jetzt Appell an Innenminister Karner und die Bundesregierung unterzeichnen: actions.aufstehn.at/waffenverbot...

Understanding the Neuro-Cognitive Mechanisms of Orthographic Learning in Humans and Baboons: A Comparative Study Using Domain-Specific Mechanistic and Domain-General Connectionist Models Script is a key technology for humans, as mastering reading is essential for successful social participation. Hence, understanding the neuro-cognitive mechanisms underpinning the processes of learning...

We're happy to share a new preprint by our members, @jnpauli.bsky.social & @benjamingagl.bsky.social: "Understanding the Neuro-Cognitive Mechanisms of Orthographic Learning in Humans and Baboons: A Comparative Study Using Domain-Specific Mechanistic and Domain-General Connectionist Models"!

🧠✍️ How do humans and baboons learn orthography?

We simulated their orthographic learning using CNNs (CORnet-Z, ResNet-18) vs. neuro-cognitive model (Speechless reader model, SLR).

🔍 SLR not only outperformed CNNs—it revealed how both species learned to differentiate between learned and novel words

🧠✍️ How do humans and baboons learn orthography?

We simulated their orthographic learning using CNNs (CORnet-Z, ResNet-18) vs. neuro-cognitive model (Speechless reader model, SLR).

🔍 SLR not only outperformed CNNs—it revealed how both species learned to differentiate between learned and novel words

An example of T1w image before and after defacing. Defacing is typically implemented by zeroing the voxels around the face (left-hand side, panel “T1-weighted”). The “Background noise” panel shows two visualizations extracted from the MRIQC visual report, in which a window is applied to select the lowest intensities and then they are inverted to enhance patterns in the background. The red arrows indicate aliases along the anterior–posterior axis—which had the lowest bandwidth in the example—produced by eye motion. These aliases are straightforward to notice in front of the ocular globes in the “Background noise” panel of the “nondefaced” image because of the absence of signal sources. This aliasing also spreads in the opposite direction, overlapping brain tissue. However, this overlap is often very hard to notice against the signal of interest within the brain. The corresponding “defaced” version of the “Background noise” panel shows how defacing eliminates valuable information for quality assessment.

To protect an individual's #privacy, facial features are removed before data-sharing of human #neuroimaging data. @oesteban.bsky.social &co show that this procedure affects the quality ratings given by human raters but not automated image quality metrics @plosbiology.org 🧪 plos.io/3Gtr3lo

Some time ago, I asked ChatGPT for a title for my paper. It suggested a title containing 'Unveiling'—a word I didn’t know but liked. Title found! A few weeks later, I came across another paper with 'Unveiling' in the title. A doubt crossed my mind… I checked and found this:

Join us on site and if to far away from Cologne online!

Science Under Threat in the United States: How scientists and institutions should respond Individual researchers and university leaders need to make the case for science to their elected representatives and to the public at large.

Science is under threat in the US. @elife.bsky.social have commissioned a series of articles discussing the implications and what we can do. The first three articles are now live. More to follow:
elifesciences.org/articles/106...
elifesciences.org/articles/106...
elifesciences.org/articles/106...

Join us on site and if to far away from Cologne online!

🚨 TRUST Open Lecture – Join Us! 🚨

Is #standardisation in research boosting #replicability or slowing scientific progress? 🤔

📅 March 31 (Mon), 2025
⏰ 16:00–18:00 CET
📍 Hörsaal H112, IBW-Gebäude, University of Cologne

💻 Live-stream available! Registration info here:
sites.google.com/view/gewonn/...

YouTube video by DW Documentary How Germany's elite research institution fails young scientists | DW Documentary

www.youtube.com/watch?v=n5nE...
Important and painful

Yes, #TeaP2025 was great this year, thanks to the organizers!

Come join our symposium at #teap2025! 😊

Topic: From Babies to Semantics: Leveraging #LanguageModels for #Psycholinguistic Research

March 10 | 11:00 - 12:30

Chair: @benjamingagl.bsky.social
Speakers: Fritz Günther, Katharina Menn, @hannawoloszyn.bsky.social, @boevesam.bsky.social, Marco Marelli

Welcome to join our open lecture on #standardisation & #replicability in Cologne on March 31, organised by the #TRUSTNetwork, led by @benjamingagl.bsky.social, @janahasenacker.bsky.social, @troembke.bsky.social & @xeniaschmalz.bsky.social!

See 🧵below for details 👇...