I strolled in and it was a bit toasty. So I turned the thermostat down, and the room cooled off. Oh. My. God. After living through 20 years of placebo thermostats in offices, classrooms, and labs, I'm flabbergasted.

Oh man, thank you CNN. It's been a heck of a day and I really needed the laugh.

All I saw was cheesecake boat and ocean for at least 5 minutes

People who think that learning statistics will help people understand statistics have a very poor understanding of statistics.

Some of the brightest lights I’ve known have been forced to eat shit repeatedly in their careers. And they were the “lucky” ones—in that they were able to stay in the game. Many more just got completely pushed out and we never know about them except as faceless statistics.

🤷🏿‍♂️

Yes, @jamellebouie.net's thread is spot on. We know from lots of research in social & in cognitive psychology that perception can become reality for many (if not all) of us in small & big ways, even when the perception is not just flawed, but not based on reality until we manifest it in some ways.

Yesterday in the orientation for my first year course on 3D form a student raised their hand and asked whether they should leave the object in the wound or remove it when someone is impaled by a projectile during the course. While unlikely, I appreciate this level of intensity in the classroom.

who are you going to believe, me or your lying eyes

"there are no gray squares, man, it's just in your mind"

Whenever I try to explain this to people who call themselves designers, I get a blank stare. But if you don't understand how your product mediates behavior, you should not be doing this job.

what an absolute and utter shambles this is, in just about every way americabydesign.gov

What is it about human brains that allows us to reason symbolically whereas most other animals cannot? And why does AI sometimes struggle to generate images of objects in statistically unusual spatial relationships to each other, when people wouldn’t struggle to imagine such relationships?

I fucking called it "the next one is coming very soon and is gonna be AGI for real this time" is the only move in their book

I too want my AI to have the disposition of a bear protecting its cubs. Seriously, though, what could possibly go wrong with an AI trying to make the set of concerns it is responsible for outcompete all other life on the planet over a narrow objective function? That's not the plot of any movies.

Large Language models (LLMs) do not simulate human psychology. That's the title of our new paper, available as preprint today (1/12): arxiv.org/abs/2508.06950

“The AI in the study probably prompted doctors to become over-reliant on its recommendations, ‘leading to clinicians becoming less motivated, less focused, and less responsible when making cognitive decisions without AI assistance,’ the scientists said in the paper.”

🤖If you're at all following AI progress on benchmarks, this Nature article is very accessible and enlightening. Lesson: Don't rush in and evaluate an AI model on some established or bespoke benchmark without critical thought. AI studies require AI expertise. Don't let #badbenchmarks propagate!

Eye movements are cheap, right? Not necessarily! 💰 In our review just out in @natrevpsychol.nature.com, Alex Schütz and I discuss the different costs associated with making an eye movement, how these costs affect behaviour, and the challenges of measuring this… rdcu.be/eAm69 #visionscience #vision

New paper: we argue that an individual's propensity to compare items and discover abstract relational rules, rather than focus on surface similarity, is stable across learning tasks, and is independent of working memory capacity. link.springer.com/article/10.3...

I've personally seen bad actors chasing prestige claim scientific credit for work they didn't do, and my observation has been that the institutions are completely unprepared mentally or procedurally to handle it. It's often so brazen that nobody can believe it could really be happening.

explains a lot why useless AI "products" are shoved down schools and universities and consumers like we are foie gras geese, the big seven and VCs need to at least recoup their investments

If you click the link, it kinda looks like someone's 2018 Flickr caption about making compound butter broke the ability of the AIs to reason about spelling. That tracks.

Wild case of ChatGPT almost killing a guy by telling him to substitute sodium bromine for table salt (sodium chloride) #medsky www.acpjournals.org/doi/10.7326/...

the scenario we all feared

In retrospect, it may have been a bad idea to let our research field be completely captured by some of the world’s most-powerful corporations

Paint the walls black, add a door and one-way mirrors, and you've got yourself a nice little psychophysics lab.

@laurennross.bsky.social is giving an important reminder of the importance of philosophy of science in cognitive science at the #cogsci2025 keynote, which I really appreciate in the current moment.

Regulating the use of AI does not imply the end of the technical development of AI. The only argument you could make there is that your tech is so crappy you need to force people to use it so you can make it memorize literally every thing a human could ever do. It would still fail.

The memes about the Coldplay people are the first real example I have seen of genAI being used at scale against people who are not famous or public officials. Maybe I missed other examples?

We thought it was a natural extension of our research ethics standards, which forbids plagiarism, fabricating data, claiming work that is not yours, etc. Since GenAI basically runs off plagiarism (albeit statistically), making up data, and work that was not done by ourselves, how could we allow it?

It's also really funny because you'd assume this quote about women being afraid that using LLMs will hurt their critical thinking skills would be followed up with all the studies that say, yep, it does that. Nope. The fact that this fear is both logical and backed up by science is never brought up.

Having a poster ready enough to be uploaded two weeks before a conference is honestly not usually how I like to roll.

My cat Seven gets it. Same, girl.

Pairs well with www.nature.com/articles/s41... “Our results indicate that individual V1 neurons do not signal how the actual visual input deviates from the animal’s predictions, as postulated within the predictive coding framework” Thrilling to see a major theory get tested!

When we design experiences and services using AI, we need to consider the risks to users, just as we would any tech. AI is just not developed enough yet for deployment at scale.

New paper with LaurenWilliams, @timbrady.bsky.social, and @violastoermer.bsky.social now out in JEP: General! "Limits of verbal labels in cognition: Category labels do not improve visual working memory performance for obfuscated objects" psycnet.apa.org/record/2026-...

1/4) I agree with a lot of this 🧵, and discussed the issues with a journalist from Science yesterday: www.science.org/content/arti... tl;dr: I think that Centaur is fascinating for a number of reasons, but I don't think it's appropriate to frame it as a foundation model of human cognition. 🧠📈 🧪

I have a lot of concerns about the Binz et al. Centaur paper, but I'll start this thread with what I like: Matching human behavioral data is critical when modeling human cognition. Trying to match neural data is interesting and possibly informative, but not sufficient for modeling the human mind. 1/

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

“Tesla’s driverless ‘robotaxis’ could launch in Austin as soon as June 22. But a demo in Austin today showed a $TSLA, manually driven to test its Full Self-Driving system, failed to stop for a child-sized dummy at a school bus—and hit it.” @cbsaustin @velez_tx

Happy 18th birthday ggplot2! #rstats

old ads used to be so cool. baffling lie. four kinds of typesetting. ends in a demand

Thinking that if I start waving my arms around like this when I am writing data analysis scripts it could potentially help with my mouse shoulder.

The eye of the Arnett, OK tornado, captured by one of our UAVs as it collected scientific data on the environment within.

"An abomination unto design", but this stuff also really matters for basic perceptual segmentation and legibility, because proximity is one of the strongest perceptual grouping cues. It's often a very fine line between an aesthetics issue and a functional issue.