FWIW, the secret to finding long lasting love is just being a human to other humans til one human and you start hanging out a lot.

Someone at the grey lady greenlit a story on incel tinted eugenics dating with AI. Given current events, it’s an odd use of key strokes.

Silicon Valley is a disorder. www.nytimes.com/2026/01/06/s...

This paragraph is a pretty good exercise in spot-the-bullshit. Can you see it? Reading this paragraph, you'd walk away with the conclusion that mere FDR correction can address replicability issues. The authors certainly try to make this claim. www.nature.com/articles/s41...

The author's should ask ChatGPT to provide feedback on the risks of nonresponse bias in their design. If you're trying to design-bias a study towards LLM-friendly findings, this is how you do it.

I'm teaching Statistical Rethinking again starting Jan 2026. This time with live lectures, divided into Beginner and Experienced sections. Will be a lot more work for me, but I hope much better for students. I will record lectures & all will be found at this link: github.com/rmcelreath/s...

who wants to see baby howlers? also look at that mom's face! she looks exhausted and exasperated because the babies don't stop moving for a second! unless they're asleep in ridiculous poses.

I’ve been in compliance

What makes this remarkable to me is that we have many published papers relying on metas data, and they’ve demonstrated a willingness to alter data being externally evaluated. Unlike Pruitt or Gino, we have no way of checking the integrity of data they’ve provided in the past.

Must read reporting

New reporting from @jeffhorwitz.bsky.social shows meta actively altering shared data to produce rose-tinted results in searches of the as library. Can we reasonably trust the data they share with academics? www.reuters.com/investigatio...

Every week, Nature publishes yet another breathless puff piece about some AI startup, based only unpublished claims from the company and interviewing only those who work there. How can the leading scientific journal publish piece after piece that would make Kevin Roose blush? I think it's that...

It's strange how rarely journalists covering this space can ask "Hey, but is it bullshit?". CEO's whose companies are over-leveraged, bubbly, and unprofitable promise a major breakthrough that will justify their fragile position and no one seems to *consider* they may be lying. It's Theranos 2.0.

700 miles driven in a day is a bit much.

*in expectation*

It sticks out in this article that #openscience top brass quickly pivoted from critiquing “gold standard” science to embracing it. Helping the very same nih director who is slashing science in the us, politicizing it, and shared conspiracy theories about me and my colleagues work.

Agree with the quoted statement by @richarddmorey.bsky.social and disagree with Gelman on that point of asymmetry. Much of what we call metascience today has gained popularity and momentum from scientists/policy makers/funders is by making major positive claims about how science works/doesn't work.

This smells distinctly like collider bias and/or selection bias and/or regression to the mean... You simply can't select teen prodigies, and world class athletes rom databases, and go run regressions without serious consideration of the selection process!

More thorough write up here bsky.app/profile/mich...

Michel Nivard is on it. bsky.app/profile/mich...

Reading the structured abstract alone should set off a half dozen statistical alarms in any scientists head. The fact that this found its way into Science makes me wonder how we drive home those stats 101 lessons and make them intuitive. www.science.org/doi/10.1126/...

These claims about what AI can do truly need citations. Checking statistics isn’t routine work, nor is whether citations to the literature are sensible. Whether statistics are appropriate and render reasonable inferences is inherently a question of human judgement.

@jbakcoleman.bsky.social and I had some thoughts about this a while back.

Gonna be rejecting any paper that draws any inference about people from this approach.

This is pretty much how making coffee for my wife goes every morning.

at this point no one cares, but i was invited to talk about the high-rep retraction saga, and once again, i find myself perplexed at the exclusion of confirmatory results in any calculation in a study making a key claim about the inclusion of confirmatory studies making results highly replicable.

Going well over on the aixiv, guess the vibe code didn’t think to render latex.

Christmas came early this year! Very happy to see our paper out in Science Advances. Led by @lfoswaldo.bsky.social, we ran a unique collective field-experiment on Reddit, to better understand who is participating in online debates and why. Paper: www.science.org/doi/10.1126/... And more below 👇

I can be picky about stats and methodology but this is just a fantastic on platform experiment top to bottom.

Really great thread to read if you use regression. A key point of awkward in peer review for me is pointing out that a preregistered choice to make causal salad yields and analysis that can’t answer the questions.

Fantastic and blistering comment about the need for transparency in industry-academic collaborations. From @georgiaturner.bsky.social, Ian Anderson, and Luisa Fassi www.pnas.org/doi/10.1073/...

Flytrap flowers might as well be saying “sorry I ate your buddies”

This a pretty funny encapsulation of ai scientist work. The authors hand it some softball software tasks, then it comes back with very different results than what it’s replicating and the authors shrug at the data being different www.nature.com/articles/s43...

I honestly would have expected the retraction of High Replicability paper would have taught the metascience community some things abt design and inference. But ofc due to the authors reframing it as "a minor honest mistake of forgetting to preregister", important lessons have not been learned.

I love the absolute bogeyman that Bayesian stats has become for some folks.

Heads up for colleagues from abroad planning travel to the us. www.reuters.com/world/us/tru...

The proposed terminology is not helpful for our theoretical work because it loses a lot of important distinctions that we often have to make in our work for the precision of arguments. Here are some examples:

Pro tip: you can use em dashes and not get accused of llm use if you’re bad at spotting typos.

Sign me up for anything with honeyguides.

A curious feature of this challenge is that it doesn’t provide information about the nature of the replications (sample size etc…) and the predictive models are a coin flip. This is a pretty straightforward prediction of our piece, which argues replicability is largely an artifact of rep N.

Artists: “ai can never replace what we do, we get the value we infuse into the world” Scientists: “what if the computer went brrr and spit out science”

I’m glad to see on platform ranking algorithm work done without platform scientists in the mix. I think it’s quite telling that, by the only means we have to verify Metas election collaborations results… we can’t confirm their claims of little to no effect.

www.science.org/doi/10.1126/... An interesting read from Alondra Nelson regarding A.I. research.

Reviewed a lovely little paper tackling a big question with little more than counting. Truly nothing I love more than using the least math and statistics possible to answer a question.

making chicken pozole verde. the smell of roasting poblanos and tomatillos to make the salsa from scratch was worth the whole ordeal. i have a huge pot of it simmering on the stove. i love the soup season!

With all the rancor over file drawers and selective publishing it astounds me that journals and the scientific community are consistently comfortable with social media companies having troves of data indicating harms and then occasionally putting out papers to the contrary. #metascience

This is gonna be a shitshow

It would, of course, be great if the underlying results were made public but it looks like meta in 2019 ran withdrawal experiments to examine polarization and well being ahead of forging collaborations with academics. Seems like more than enough to design bias the collaborations towards nulls 🧵