Randomized controlled trials are a key tool to study cause and effect. Why do they matter and how do they work?

- Let’s assume the poorly paid participants carefully read and adhered to the complicated instructions - We can recruit the difficult-to-reach sample using a pre-screener asking people if they meet the criteria for the better paid main study. People on the Internet don’t just lie.

- The monkey on your back: Embodied metaphor via real monkeys impairs cognitive control - What if individual differences in brain activation via fMRI are associated with [important overt behavior]? (N = 16)

- Can this 15 second stimulus pairing intervention radically improve entrenched psychosocial problems? - What are the trajectories of [thing]: a latent growth curve analysis - How should we define replication, reproduction, verification, etc? Another taxonomy that will be ignored

Some researchers don't discuss their future research plans for fear of being scooped. Not me. I drop bad ideas for unscrupulous people to 'steal'. - What are the neural correlates of Open Science practices? - What is the role of habits in learning a new skill through repetitive practice?

The list of social media platforms that Australia is considering to ban (under the new restrictions for Aussie teenagers) includes—yes, you read that right—GitHub. www.abc.net.au/news/2025-09...

interesting bit in a Lancet article re: Australia's teen social media ban: surprisingly the amount of socialising in person hasn't dropped significantly in the past 25 years, and certainly not by the average amount of social media use.

🥳 PsyArXiv is back to normal operations 🥳 Thanks to the fantastic team of moderators (thank you all🙏), PsyArXiv has resumed normal operations. You can expect any newly posted or edited preprint to be moderated within 24-72 hours of submission. Learn more from our blog post👇 buff.ly/OtpsKhm

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 🧵

Waiting for my preprint to be accepted, so in the meantime a teaser: here's what happens when you try to estimate a between-scale correlation based on LLM-generated datasets of participants, while varying 4 different analytic decisions (blue is the true correlation from human data):

how it feels to contribute to an edited volume

How is that not sci-fi

PsyArXiv's amazing team of 100+ moderators has now approved all preprints that meet the requirements outlined in the updated PsyArXiv policies (is.gd/paxpolicy). Thank you to everyone who volunteered, this was a true community effort! #PsychSciSky

The Item Response Warehouse is a new data resource for psychometricians interested in developing methods using bigger and more diverse sets of instruments: itemresponsewarehouse.org New paper out now at BRM: doi.org/10.3758/s134...

It's very human to only double check that a process is working when you get a weird result. It's also very bad practice, because sometimes your "right" result is due to a bad process and you will be misled. Social scientists (economists) do this kind of asymmetric checking. arxiv.org/pdf/2508.20069

Deutsche Lonely Planet

As a totally separate point: the effect of choice didn't replicate in humans in a large-scale preregistered replication! And the Cohen's ds in the original paper range from 1.3 to 4.17. journals.sagepub.com/doi/full/10....

This has now been published in PNAS! www.pnas.org/doi/10.1073/... The original authors also posted a reply: www.pnas.org/doi/10.1073/... Quick thread on some additional thoughts and then I'm probably done talking about this one 🧵

Ever stared at a table of regression coefficients & wondered what you're doing with your life? Very excited to share this gentle introduction to another way of making sense of statistical models (w @vincentab.bsky.social) Preprint: doi.org/10.31234/osf... Website: j-rohrer.github.io/marginal-psy...

open.substack.com/pub/cell/p/s... Very enjoyable read

TIL about Big Nut.

Say you’ve got very good causal IDENTIFICATION. Can you say you have causality, can make a causal general claim, and can say what “THE causal effect” is? NO We explain why in new article at Journal of Causal Inference. www.degruyter.com/document/doi...

9.2% of all citations go to just 0.32% of psychology articles. To have the most impact, post-publication peer-review should focus on these influential articles with ≥30 citations per year. @error.reviews will systemically sample from such articles. Blog post: mmmdata.io/posts/2025/0...

Systematic reviewers have a 10.7% error rate when screening papers for potential inclusion. In other words, about 1 in 9 abstracts are categorised incorrectly (i.e., false inclusion or false exclusion). This really drives home the benefit of having at least two screeners

co-signed (in my 8 years of statistical consulting, clustering has not solved anyone’s problem)

Our (@malte.the100.ci, @ianhussey.mmmdata.io) commentary "cognitive dissonance in large language models is neither cognitive nor dissonant" has just been accepted for publication at PNAS! Check out the thread below for a summary of our arguments. osf.io/preprints/ps...

These are some of the things on my to-do list, and from next week, I'll be working full-time at Works in Progress as an editor on science & global health, working with authors, and writing my newsletter Scientific Discovery, co-hosting the podcast Hard Drugs, to cover these topics and many more.

✨ ERROR (@error.reviews) is a bug-bounty program for science that seeks to estimate the prevalence and nature of errors. error.reviews

We are very pleased to announce the 2025 Commendations for the Society for the Improvement of Psychological Science! Join us in celebrating the exceptional projects that have revolutionized the field of psychological science. buff.ly/rzTeLYL

We are very pleased to announce the 2025 Awards Winners for the Society for the Improvement of Psychological Science! Join us in celebrating the exceptional projects that have revolutionized the field of psychological science. buff.ly/rzTeLYL

If schools were actually preparing kids for white-collar work they wouldn't be phone-free, you'd have to have your phone on your desk all day for the authenticator apps

Taking inspiration from @dingdingpeng.the100.ci et al.'s Loss of Confident Project, I've posted my own Loss of Confidence statement on my blog. tl;dr: I don't have much confidence in my early empirical publications conducted before I embraced open science practices.

if you think about it oedipus was the first guy to really comprehensively lose the game we now call fuck marry kill

I mostly liked this, except the part about twin studies' assumptions: Almost all the assumptions also apply to heritability estimates from GWAS (including diagnostic misclassification, no interactions, sample representativeness, no assortative mating, and even the equal environments assumption)..

Sad news: The dataset behind this famous chart on the decline in costs of genome sequencing has had its NIH funding cut. I loved this chart because it was the first that made me appreciate the impact of dataviz. But it also tracked progress towards an ambitious goal ($1000 genome) that succeeded.

I recently taught a course on minimizing mistakes in science. The most convincing argument to my masters students was that you never check mistakes that support the expected outcome, you only check if the outcome is unexpected. That's a hugely important bias!

I thought I'd add a guessing parameter for my multiple choice items, but now I realise the model cannot handle people systematically choosing distractors more often than guessing would predict (think cognitive reflection task). Now tempted to simplify again? Reading advice? Maybe @pgmj.bsky.social?

Tomorrow I will talk to an academic publisher regarding the possibility of a book 👀 Any questions I should absolutely ask? I already got e-book pricing on my list; any and all pointers are welcome!

A post on @error.reviews

I thought I'd add a guessing parameter for my multiple choice items, but now I realise the model cannot handle people systematically choosing distractors more often than guessing would predict (think cognitive reflection task). Now tempted to simplify again? Reading advice? Maybe @pgmj.bsky.social?

Where was this high horse when we learned that "wine makes you live long", "video-games are heroin" or "we can feel the future"? Who is eroding trust in science, and who are we placing responsibility with? And if primary research is not pointless to begin with: Improving it seems pretty useful.

Reckon I can submit this as a figure for a paper?

Anyone got any reading tips for blogs or papers on a) generalizability theory with brms b) G theory + IRT (I have Choi's 2017 dissertation)?

Anyone got any reading tips for blogs or papers on a) generalizability theory with brms b) G theory + IRT (I have Choi's 2017 dissertation)?

It's now happened to me a few times that I fit a brms model that turned out a little, erm, big, which made a bunch of post-processing annoying. So, I had Gemini help me write a function to thin the model after fitting. gist.github.com/rubenarslan/...

I’ve given up on deutsche bahn being on time and would now settle for them telling me the root cause of a delay rather than a fucking mediator.

I cannot be convinced there exists a better title than "ItJustAintDopeToDropTheSlope.pdf" (which is how the .pdf on OSF is called).

*Another* blog post about @posit.co's Positron! Its Remote Explorer feature lets you connect to other computers via SSH, including locally-running Docker containers, which means you can write and run code in version-locked #rstats environments! www.andrewheiss.com/blog/2025/07...

“What is the relationship between LLMs and human cognition?” We’re concerned by the prevalence of conceptual slippage of articles on this topic. We’ve written a commentary on a specific article, but our concerns are general.