Lightnews — Scholar-powered news

Stefan Herzog @stefanherzog.bsky.social · 15h

🚨⬇️
___
Bald könnten ALLE Chat-Nachrichten von den Behörden durchsucht werden – selbst ohne Verdacht! Das entscheiden *heute* die Minister*innen. Sag Nein zur Chatkontrolle! Jetzt unterzeichnen ✍️ weact.campact.de/petitions/ch...

Chatkontrolle stoppen!

Die EU-Kommission will Messenger-Dienste wie WhatsApp und Signal zwingen, alle privaten Nachrichten und Fotos in Echtzeit zu scannen. Angeblich zum Kinderschutz. In Wahrheit bedeutet die Chatkontrolle...

weact.campact.de

Reposted by Stefan Herzog

Science Feedback @sciencefeedback.bsky.social · 9d

🚨 New report out: the first cross-platform, cross-country baseline on misinformation in Europe

Based on large scale data analysis: ~2.6M posts (24B views) collected on Facebook, Instagram, LinkedIn, TikTok, X & YouTube

science.feedback.org/first-measurement-disinformation-major-platforms-europe

What our first measurement says about disinformation on major platforms in Europe - Science Feedback

Science Feedback and partners have released a first measurement of Structural Indicators across six Very Large Online Platforms (VLOPs) in four EU member states (France, Spain, Poland, Slovakia). Belo...

science.feedback.org

1 12 22

Reposted by Stefan Herzog

Jamie Cummins @jamiecummins.bsky.social · 20d

Can large language models stand in for human participants?
Many social scientists seem to think so, and are already using "silicon samples" in research.

One problem: depending on the analytic decisions made, you can basically get these samples to show any effect you want.

THREAD 🧵

The threat of analytic flexibility in using large language models to simulate human data: A call to attention

Social scientists are now using large language models to create "silicon samples" - synthetic datasets intended to stand in for human respondents, aimed at revolutionising human subjects research. How...

arxiv.org

12 150 330

Reposted by Stefan Herzog

Adrienne Fichter @adfichter.bsky.social · 20d

Wir werden alle Datenfutter für die KI von LinkedIn (zu Microsoft zugehörig), wenn wir uns nicht wehren. LinkedIn nutzt das Opt-Out-Land Schweiz aus (ob EU weiss ich nicht). Und setzt den Toogle bei uns allen auf ON.

Ja, es ist verdammt frech.

Hier ⬇️aussteigen:
www.linkedin.com/mypreference...

Daten zur Verbesserung generativer KI
Darf LinkedIn Ihre auf LinkedIn erstellten personenbezogenen Daten und Inhalte verwenden, um generative KI-Modelle von LinkedIn zu schulen, die zum Erstellen von Inhalten verwendet werden?

Ihre Daten zum Schulen von KI-Modellen verwenden, um Inhalte zu erstellen

ein/aus

21 66 120

Reposted by Stefan Herzog

Mario Sixtus @sixtus.net · 20d

Eine Partei, die Frauen, die über ihren Körper selbst bestimmen, zu Straftäterinnen macht, die sich dagegen wehrt, ein Verbotsprüfungsverfahren gegen eine faschistische Partei einzuleiten, setzt sich mit einer frauenhassenden, religiös-faschistischen Terrorregierung an einen Tisch. Passt.

Migrationspolitik: Dobrindt verteidigt Gespräche mit Taliban über Abschiebungen

Bundesinnenminister Alexander Dobrindt will "regelmäßig" weitere Straftäter nach Afghanistan abschieben. Zugleich verteidigt er im Bundestag Gespräche mit den Taliban.

www.zeit.de

30 200 710

Reposted by Stefan Herzog

Patrick Breitenbach @patrickbreitenbach.de · 21d

"Free Speech" wird als taktisches Werkzeug genutzt, um gesellschaftlich eher geächtete Positionen (Rassismus, Sexismus, Klassismus) wieder in die breite Akzeptanz zurück zu holen. Es geht dabei nicht um "Free Speech", denn wir sehen ja, dass Gegenrede von den gleichen Leuten hart sanktioniert wird.

10 150 450

Reposted by Stefan Herzog

Olivia Guest · Ολίβια Γκεστ @olivia.science · Sep 6

Finally! 🤩 Our position piece: Against the Uncritical Adoption of 'AI' Technologies in Academia:
doi.org/10.5281/zeno...

We unpick the tech industry’s marketing, hype, & harm; and we argue for safeguarding higher education, critical
thinking, expertise, academic freedom, & scientific integrity.
1/n

Abstract: Under the banner of progress, products have been uncritically adopted or
even imposed on users — in past centuries with tobacco and combustion engines, and in
the 21st with social media. For these collective blunders, we now regret our involvement or
apathy as scientists, and society struggles to put the genie back in the bottle. Currently, we
are similarly entangled with artificial intelligence (AI) technology. For example, software updates are rolled out seamlessly and non-consensually, Microsoft Office is bundled with chatbots, and we, our students, and our employers have had no say, as it is not
considered a valid position to reject AI technologies in our teaching and research. This
is why in June 2025, we co-authored an Open Letter calling on our employers to reverse
and rethink their stance on uncritically adopting AI technologies. In this position piece,
we expound on why universities must take their role seriously toa) counter the technology
industry’s marketing, hype, and harm; and to b) safeguard higher education, critical
thinking, expertise, academic freedom, and scientific integrity. We include pointers to
relevant work to further inform our colleagues.

Figure 1. A cartoon set theoretic view on various terms (see Table 1) used when discussing the superset AI
(black outline, hatched background): LLMs are in orange; ANNs are in magenta; generative models are
in blue; and finally, chatbots are in green. Where these intersect, the colours reflect that, e.g. generative adversarial network (GAN) and Boltzmann machine (BM) models are in the purple subset because they are
both generative and ANNs. In the case of proprietary closed source models, e.g. OpenAI’s ChatGPT and
Apple’s Siri, we cannot verify their implementation and so academics can only make educated guesses (cf.
Dingemanse 2025). Undefined terms used above: BERT (Devlin et al. 2019); AlexNet (Krizhevsky et al.
2017); A.L.I.C.E. (Wallace 2009); ELIZA (Weizenbaum 1966); Jabberwacky (Twist 2003); linear discriminant analysis (LDA); quadratic discriminant analysis (QDA).

Table 1. Below some of the typical terminological disarray is untangled. Importantly, none of these terms
are orthogonal nor do they exclusively pick out the types of products we may wish to critique or proscribe.

Protecting the Ecosystem of Human Knowledge: Five Principles

96 1.6K 3.1K

Stefan Herzog @stefanherzog.bsky.social · 27d

PDF available at pure.mpg.de/rest/items/i...

pure.mpg.de

1

Stefan Herzog @stefanherzog.bsky.social · 27d

🧠🤖 Ever wondered about the risks of the increasing customization capabilities of AI/AI-powered chatbots & what we can do about it? Check out our recent-ish paper:

The governance & behavioral challenges of generative artificial intelligence’s hypercustomization capabilities. doi.org/10.1177/2379...

1 2 8

Reposted by Stefan Herzog

Vito Trianni @vitotrianni.bsky.social · 29d

🌍 Join us for #HACID Webinar #4: Gender & Diversity in Hybrid Collective Intelligence 🧠🤖
🗓 Sept 29, 2025 | ⏰ 18:00 CEST

Register here: events.teams.microsoft.com/event/71775f...

Photo by cottonbro studio: https://www.pexels.com/photo/photo-of-people-using-their-smartphones-8088445/

1 7 10

Reposted by Stefan Herzog

Ana Sofia Morais @anasofiamorais.bsky.social · Sep 4

We’ve all heard it: people make 200 mindless food decisions a day. But is it true? In our latest episode, Almudena Claassen reveals why this number is misleading and offers strategies that help families and individuals make healthier choices. tinyurl.com/3k2tjw6x

2 9

Reposted by Stefan Herzog

Levin Brinkmann @levinbrinkmann.bsky.social · Jun 21

You sound like ChatGPT -- that is what The Verge said about our recent preprint on how ChatGPT is reshaping human spoken communication.

Chatbots are becoming a cultural medium and agent of cultural homogenisation. Is this inevitable, or can model diversity help? www.theverge.com/openai/68674...

You sound like ChatGPT

AI isn’t just impacting how we write — it’s changing how we speak and interact with others. And there’s only more to come.

www.theverge.com

1 2 8

Reposted by Stefan Herzog

Ezequiel Lopez-Lopez @eloplop.bsky.social · Jul 7

New paper out! 📣
We explore GenAI’s hypercustomization capabilities, the behavioral and governance challenges they may introduce in the foreseeable future, and potential strategies to address them.

Christoph Abels @cabels18.bsky.social · Jul 7

Generative AI (GenAI) can do more than just answer questions - it can tailor its responses based on users’ preferences, habits, and even emotional tone.

We call this capability hypercustomization - explored in our new paper in Behavioral Science & Policy. journals.sagepub.com/doi/10.1177/...

The governance & behavioral challenges of generative artificial intelligence’s hypercustomization capabilities - Christoph M. Abels, Ezequiel Lopez-Lopez, Jason W. Burton, Dawn L. Holford, Levin Brink...

Generative artificial intelligence (GenAI) is changing human–machine interactions and the broader information ecosystem. Much as social media algorithms persona...

journals.sagepub.com

1 6

Reposted by Stefan Herzog

Iyad Rahwan | إياد رهوان @iyadrahwan.bsky.social · Jul 15

Great coverage by @sciam.bsky.social of our research on how ChatGPT is influencing the words we use in conversation!

Work led by Hiromu Yakura & @levinbrinkmann.bsky.social

www.scientificamerican.com/article/chat...

ChatGPT Is Changing the Words We Use in Conversation

Words frequently used by ChatGPT, including “delve” and “meticulous,” are getting more common in spoken language, according to an analysis of more than 700,000 hours of videos and podcasts

www.scientificamerican.com

6 12

Reposted by Stefan Herzog

Christoph Abels @cabels18.bsky.social · Jul 28

🔍 “I just want a second opinion.”

More people are turning to ChatGPT for health advice. Many would use it for self-diagnosis.

But here's the problem: These tools don’t just answer, they align. And that’s where things get risky.

🧵 on GenAI, health, and confirmation bias

Ezequiel Lopez-Lopez @eloplop.bsky.social · Jul 28

New research out!🚨

In our new paper, we discuss how generative AI (GenAI) tools like ChatGPT can mediate confirmation bias in health information seeking.
As people turn to these tools for health-related queries, new risks emerge.
🧵👇
nyaspubs.onlinelibrary.wiley.com/doi/10.1111/...

NYAS Publications

Generative artificial intelligence (GenAI) applications, such as ChatGPT, are transforming how individuals access health information, offering conversational and highly personalized interactions. Whi...

nyaspubs.onlinelibrary.wiley.com

1 7 13

Reposted by Stefan Herzog

Ezequiel Lopez-Lopez @eloplop.bsky.social · Jul 28

New research out!🚨

In our new paper, we discuss how generative AI (GenAI) tools like ChatGPT can mediate confirmation bias in health information seeking.
As people turn to these tools for health-related queries, new risks emerge.
🧵👇
nyaspubs.onlinelibrary.wiley.com/doi/10.1111/...

NYAS Publications

Generative artificial intelligence (GenAI) applications, such as ChatGPT, are transforming how individuals access health information, offering conversational and highly personalized interactions. Whi...

nyaspubs.onlinelibrary.wiley.com

1 9 15

Reposted by Stefan Herzog

Claes de Vreese @claesdevreese.bsky.social · Jul 14

The discussions about big platforms are intense. The systemic risks are real.

Now the #DSA Delegated Act is finally here to ensure data access for researchers. Time to put this system to work.

W @lkseiling.bsky.social, we show how @techpolicypress.bsky.social

www.techpolicy.press/time-to-deli...

Time To Deliver: Stakeholder Roles In The EU’s Delegated Act On Data Access | TechPolicy.Press

The Delegated Act on data access under the DSA provides much-needed clarity about both the access process and expectations towards the actors involved in it.

www.techpolicy.press

13 28

Reposted by Stefan Herzog

Kathrin Kühn @kathrinkuehn.bsky.social · Jul 15

Meinungsbildung über Social Media. Eigentlich wissen es alle - was hier läuft, ist nicht repräsentativ, im Gegenteil. Das Bild ist verzerrt, das Extreme bekommt mehr Raum. In den Dlf Systemfragen hat @lucarehseknauf.bsky.social die Lage gebündelt: www.deutschlandfunk.de/social-media... 1/

Social Media: Wer bricht das Schweigen der stillen Beobachter?

Bis zu 90 Prozent der User auf Social Media sind stille Beobachter. Extreme Meinungen werden so sichtbarer, Diskussionen polarisierter. Wie wäre das zu ändern?

www.deutschlandfunk.de

4 9 27

Reposted by Stefan Herzog

Tobias Galla @tobiasgalla.bsky.social · Jul 16

Review "Opinion dynamics: Statistical physics and beyond"
arxiv.org/abs/2507.11521

Lab experiments, data, models, analytical/computational tools. 93 pages, >1k references.

With Fabian Baumann, David Garcia, Gerardo Iñiguez, Márton Karsai, Jan Lorenz, Katarzyna Sznajd-Weron. Led by Michele Starnini

Opinion dynamics: Statistical physics and beyond

Opinion dynamics, the study of how individual beliefs and collective public opinion evolve, is a fertile domain for applying statistical physics to complex social phenomena. Like physical systems, soc...

arxiv.org

1 8 24

Reposted by Stefan Herzog

Stephan Lewandowsky @lewan.bsky.social · Jul 14

Fully transparent and open AI coming out of Switzerland soon. Something worth monitoring. ethz.ch/en/news-and-...

A language model built for the public good

ETH Zurich and EPFL will release a large language model (LLM) developed on public infrastructure. Trained on the “Alps” supercomputer at the Swiss National Supercomputing Centre (CSCS), the new LLM ma...

ethz.ch

2 32 110

Stefan Herzog @stefanherzog.bsky.social · Jul 8

Brodeur, A., Lé, M., Sangnier, M., & Zylberberg, Y. (2016). Star Wars: The Empirics strike back. American Economic Journal: Applied Economics, 8(1), 1–32. https://doi.org/10.1257/app.20150044

Using 50,000 tests published in the AER, JPE, and QJE, we identify a residual in the distribution of tests that cannot be explained solely by journals favoring rejection of the null hypothesis. We observe a two-humped camel shape with missing p-values between 0.25 and 0.10 that can be retrieved just after the 0.05 threshold and represent 10–20 percent of marginally rejected tests. Our interpretation is that researchers inflate the value of just-rejected tests by choosing “significant” specifications. We propose a method to measure this residual and describe how it varies by article and author characteristics.

1 5

Reposted by Stefan Herzog

Jamie Cummins @jamiecummins.bsky.social · Jun 23

Social scientists should not use chat interfaces when using LLMs in their research: they are impressively inefficient, and obscure/impose important methodological decisions that require thought.

THREAD🧵

3 29 70

Reposted by Stefan Herzog

Jamie Cummins @jamiecummins.bsky.social · Jul 7

New preprint commentary from me, @malte.the100.ci, and @ianhussey.mmmdata.io.

Cognitive dissonance in large language models is neither cognitive nor dissonant.

THREAD BELOW 🧵

osf.io/preprints/ps...

OSF

osf.io

4 37 74

Reposted by Stefan Herzog

Christoph Abels @cabels18.bsky.social · Jul 7

Generative AI (GenAI) can do more than just answer questions - it can tailor its responses based on users’ preferences, habits, and even emotional tone.

We call this capability hypercustomization - explored in our new paper in Behavioral Science & Policy. journals.sagepub.com/doi/10.1177/...

The governance & behavioral challenges of generative artificial intelligence’s hypercustomization capabilities - Christoph M. Abels, Ezequiel Lopez-Lopez, Jason W. Burton, Dawn L. Holford, Levin Brink...

Generative artificial intelligence (GenAI) is changing human–machine interactions and the broader information ecosystem. Much as social media algorithms persona...

journals.sagepub.com

1 6 7

Reposted by Stefan Herzog

Center for Adaptive Rationality (MPIB, Berlin) @arc-mpib.bsky.social · Jul 3

We wrapped up the 22nd Summer Institute on Bounded Rationality at the Center for Adaptive Rationality at the Max Planck Institute for Human Development (@mpib-berlin.bsky.social).

ECRs from around the world joined us to share their insights on rationality, cognition, and decision-making.

1/🧵👇

1 4 19