12. Here's the key insight though: as labor-augmenting technologies LLMs make researchers more productive and therefore they *raise the opportunity cost of researcher time*. As a result researchers will have incentives to invest less time in polishing a current project before moving on to a new one.

not literally about cocaine

A cocaine skeptic starts an unwise habit – Notes from a data witch

You will be shocked to learn that this is not actually a post about cocaine

blog.djnavarro.net

New paper finally out in @NatureComms with E. Leib, D. O’Shaughnessy, C. Gallardo, @sferrigno.bsky.social, and @spiantado.bsky.social. 📝Children across cultures discover the latent algorithms that structure what they see, even without instruction, feedback, or formal schooling.🧵 tinyurl.com/4a238m2d

1. How common is LLM use in scientific publishing, and how does it vary across field, publisher, journal prestige, author demographics etc.? @kylesiler.bsky.social has new paper in PNAS that addresses this question on a massive scale: 7.3 million papers from Elsevier, PLOS, MDPI, and Frontiers.

The diffusion of large language models in published academic articles | PNAS

Large language models (LLMs) are rapidly changing academic research, raising questions of who is adopting these tools and under what conditions. Th...

pnas.org

As a PhD student who has to both write her dissertation and grade papers, I would like to express my resentment at the addition of unprompted "Hey sis, wanna cheat?" messages from Gemini to Google docs.

If these effects accumulate over months and years of daily AI use, by the time they're visible they may be very hard to reverse aka the "boiling frog" dynamic. Each individual act of using AI feels costless but the cumulative effect may not be.

is a balloon more conscious than a piece of rock? i think it is. consciousness refers to the capacity to respond meaningfully & interestingly to external stimuli. a rock doesn't do much when we kick it. the balloon, however... is so 'fun' so play with. so unpredictable. it has a real personality

📄 Long overdue paper announcement: Core Vocabulary in Language Representation and Processing was published late last year in Cognitive Science. The work explores novel approaches of quantifying vocabulary centrality and tests them in an empirical task. 🔗 onlinelibrary.wiley.com/share/author...

i dunno, seems bad arxiv.org/pdf/2603.18161

Yay! www.npr.org/2026/04/14/n...

Millions of people are pretending to be AI chatbots — for fun

Websites like youraislopbores.me have become playgrounds for people looking for light relief in a bot-heavy world.

npr.org

We found that 10% of participants endorsed this statement (i.e., they selected “Definitely true” or “Probably true”) and that this was a strong predictor of endorsing six pre-existing conspiracy theories, including two that that directly contradicted each other.

Humans have a remarkable capacity for efficient, intelligent reasoning. In a new working paper, @perfors.net and I discuss how our ability to reason from *social* information is one way that we achieve this capacity. osf.io/preprints/ps...

Hey, I wrote a thing about AI in astrophysics ergosphere.blog/posts/the-ma...

genuinely fascinating that we’ve hit a point where the full monetization of the internet and ai bot takeover has led to a majority of people wanting tech minimalism

🧵 New preprint led by @bingbrunton.bsky.social, @elliottabe.bsky.social, @lawrencehu.bsky.social We gave a worm brain control of a fly body and it walked What did we learn? Nothing, other than deep reinforcement learning is effective We call it the digital sphinx www.biorxiv.org/content/10.6...

NEW study found chatbots supported violence 80% of the time—providing key details & weapon advice. Here’s the rub: there is currently no legal requirement for U.S. chatbot creators to alert law enforcement when their systems flag potential violence. The result is real-world harm. trib.al/XRqylye

I always had the vague feeling that Scientific Reports and Nature Communications are mainly APC business models. A paper estimated the total APC for gold/hybrid Open Access per journal 2015–2018: doi.org/10.1162/qss_... Surprise, surprise - there are 2 outliers at the top😐

🔥 🔥 🔥

Richard McElreath 🐈‍⬛@rmcelreath.bsky.social · 5mo ago

Statistical Rethinking 2026 is done: 20 new lectures emphasizing logical and critical statistical workflow, from basics of probability theory to causal inference to reliable computation to sensitivity. It's all free, made just for you. Lecture list and links: github.com/rmcelreath/s...

What they want is some path out of that, and someone who is proud and unashamed of being sincere -- who in fact points out that rejection of meaning and failure to stand for anything is the coward's way out -- provides that path. We can't be apologetic. Standing proudly is its own proof of concept.

scenes from the MIT philosophy department

Nature Publishing Group finding more dastardly ways to lock us in, well beyond what I complained about in my letter declining reviewing for them

1/10 So here's a puzzle. Efficient communication perspectives argue that speakers tailor their messages to make their listener’s visual 👁️ (and auditory?👂) search faster and more accurate. However, even when visual search is tightly controlled, people really over-use 🌈colour🌈 terms. Why? 🧵

Thinking about this quote from Persepolis creator Marjane Satrapi again.

I'm only partway through, but it's clear that this book by Chater & Lowenstein is one of the most important books to ever come out of psychology, cognitive science, and behavioral economics.

Bullshit

Writing is thinking Outsourcing the entire task of writing to LLMs will deprive us of the essential creative task of interpreting our findings and generating a deeper theoretical understanding of the world.

first empirical paper from the Aulet lab! What does learning to count do to a brain? We trained neural networks on different tasks and only counting (i.e., predicting the number of dots in a dot array) reorganized representations around number. New preprint:

PsyArXivBot@psyarxivbot.bsky.social · 6mo ago

Counting Training Creates Number-Biased Representations in Deep Neural Networks: https://osf.io/356xd

Bit of a life update: After 4 wonderful years in Melbourne, yesterday I arrived in the US to start a postdoc at UC Berkeley with @wdt.bsky.social

I just created a series of seven deep-dive videos about AI, which I've posted to youtube and now here. 😊 Targeted to laypeople, they explore how LLMs work, what they can do, and what impacts they have on learning, well-being, disinformation, the workplace, the economy, and the environment.

Part 1: How do LLMs work?

YouTube video by Andrew Perfors

youtube.com

It's a struggle, but it's gratifying once things start coming together. What I have on paper tonight is a clean expression of idea I didn't have well-formed when I began this morning. Some people think we should hand off this painstaking form of labor to LLMs and get back to doing science.