Close Netflix and open this: probably the best modeling paper I've ever read

Some very good advice for thriving in graduate school dorsaamir.medium.com/modest-advic...

💖This paper has been ~11 years in the making - and probably my favorite project of all time. Thrilled to see it in @pnas.org! I'm so lucky that Zach decided to do a second PhD and join my lab @psychillinois.bsky.social back in 2014 - a fabulous scientist & human being! www.pnas.org/doi/10.1073/...

I would argue this is actually about Bayesian inference

since I first heard about LLM "silicon samples," I've been baffled by the idea. even if a group of humans were able to predict the behavior or answers of others, no one would use it instead of real data, cause there would clearly be unknown biases and situations in which the predictions would fail.

Public appreciation post for @juliabottesini.com! The community doesn't know the great lengths Julia has gone to restabilize operations at PsyArXiv. We're all incredibly lucky to have her. Many, many hours and a great deal of thought/care have gone into the task. Thank you, Julia!

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 🧵

Whoa—my book is up for pre-order! 𝐌𝐨𝐝𝐞𝐥 𝐭𝐨 𝐌𝐞𝐚𝐧𝐢𝐧𝐠: 𝐇𝐨𝐰 𝐭𝐨 𝐈𝐧𝐭𝐞𝐫𝐩𝐫𝐞𝐭 𝐒𝐭𝐚𝐭 & 𝐌𝐋 𝐌𝐨𝐝𝐞𝐥𝐬 𝐢𝐧 #Rstats 𝐚𝐧𝐝 #PyData The book presents an ultra-simple and powerful workflow to make sense of ± any model you fit The web version will stay free forever and my proceeds go to charity. tinyurl.com/4fk56fc8

This is exactly what's happening. Of course they think a LLM can do the job, when so many *already* think a statistician can be replaced with a flow chart - and they are right, if the only goal is publication. Infuriating, and harmful to patients.

McSweeney (2002) is required reading for anyone who studies or references Hofstede's individualism-collectivism--or Markus & Kitayama's independence-interdependence, for that matter. Just shoddy work up and down. doi.org/10.1177/0018...

📢 Preprint out! biorxiv.org/content/10.1... What gives rise to probability weighting, a cornerstone of Prospect Theory? We show it comes from the natural boundedness of probabilities + cognitive noise. Adding boundaries adds multiple distortions, across risky choice & perception.

i did not foresee the future in which Markdown was the machine interop format for all human readable documents blog.xeynergy.com/what-is-micr...

I'm on @nature.com's podcast talking to @nickpetrichowe.bsky.social about the role that journalists can play in combating research misconduct. Honored to appear alongside @jamesheathers.bsky.social, @abalkina.bsky.social, @richvn.bsky.social and many others: www.nature.com/articles/d41...

We often hear from reviewers: "what about demand effects?" So we developed a method to eliminate them. Something weird happened during testing: We couldn’t detect demand effects in the first place! (1/8)

We tested 18 LLMs on 37 social science annotation tasks (13M labels, 1.4M regressions). By trying different models and prompts, you can make 94% of null results appear statistically significant–or flip findings completely 68% of the time. Importantly this also concerns well-intentioned researchers!

This was a true mess with a ton of AI-generated word salad slop being submitted as preprints. The field owes these moderators a huge thanks for all the work and energy! 🙏👏

1/ Thrilled to share that my main PhD article has just been published in Nature Human Behaviour! 🎉 👉 “A randomised controlled trial on the effect of administrative burden and information costs on social inequalities in early childcare access in France” rdcu.be/eFAEA

I didn't realise how nationalistic different theories of immunology became in the 19th century, or that Pasteur and Koch hated each other.

In a new paper, my colleagues and I set out to demonstrate how method biases can create spurious findings in relationship science, by using a seemingly meaningless scale (e.g., "My relationship has very good Saturn") to predict relationship outcomes. journals.sagepub.com/doi/10.1177/...

Simon Wood, the GOAT of generalized additive models & creator of the mgcv #rstats package, has an Annual Review of Statistics essay on GAMs, available open access #statssky #mlsky www.annualreviews.org/content/jour...

“Overall, the message from this survey is positive; evolutionary biologists are readily employable outside of academia, generally well-prepared for those jobs, and report high levels of satisfaction” www.biorxiv.org/content/10.1...

Last night in Poland we saw the most serious European airspace violation by Russia since the war began, and indications suggest it was intentional, not accidental. I am in contact with NATO Secretary General Rutte and Poland FM Sikorski. The EU stands in full solidarity with Poland. (1/2)

A significant barrier to widespread adoption of registered reports is when editors and reviewers treat a Stage 2 RR as if it were a brand new paper. What are authors supposed to do with that? Is it our job to educate editors and reviewers about this format at all stages?

I re-replicated and find 10 participants in all 5 average categories in the same dataset (code here github.com/petersj/big_5)

Based on preliminary findings, I should eat less whole wheat pasta (just made my levels skyrocket) but more Huel (didn’t show at all this morning) and tiramisu (was accompanied by a solid decline). Given the tiramisu finding, I’m not taking methods criticism at this point in time.

Our own @hannahritchie.bsky.social has a new book about tackling climate change coming out on September 18th! Called “Clearing the Air”, it covers everything from renewable energy and nuclear power to electric vehicles, heat pumps, minerals, carbon capture, and geoengineering.

ANSWER: 0 (yes, ZERO!) This is a result of an analysis done by a student in my grad seminar, using a large dataset (N=307,313). What this result might mean: Nobody's personality is truly "average," and people's personality profiles (at least Big 5) might be more "jagged" than we think. (🧵 1/5)

Just made this little diagram and it's so pretty! {ggdiagram} is really neat for these types of graphics #rstats (code here: gist.github.com/andrewheiss/... )

At the risk of stretching the imagination beyond the limits of human endurance, the reader is asked to pretend that the author is interested in splines #rstats

Always fascinating to read about replication before the 20th century. Here it's Pasteur doing a live demonstration of the efficacy of his animal anthrax vaccine (the third vaccine ever developed).

Blog post: github.com/sophieehill/... TL;DR: There are a LOT of errors/inconsistencies in the results reported in this paper (estimates outside CIs, sign errors, duplicates, asymmetric CIs). Even in the abstract itself! This suggests manual editing of results tables. Which is not good... 🧵

The Power of Gender Norms: In Switzerland, French-speaking cantons showed much higher support for the introduction of women's suffrage and today exhibit lower contemporary child and employment penalties for females. academic.oup.com/restud/artic... & www.lse.ac.uk/Internationa...

Research by @sergiopirla.bsky.social et al (N=60k across 30 countries) suggests a robust negative association between daily experiences of boredom and income; in lower income individuals, boredom is also more closely linked with sadness, worry and anxiety: buff.ly/S1ez2xo

Academic authors, here's a peek into the black box of journal publishing from an journal editor if you can bear it:

People forget how badass Russell was. www.nytimes.com/1961/09/13/a...

SUNDAY TIMES: “In recent months, dozens of Russian dissidents have been expelled from the US and returned to Russia… less than a fortnight after Trump’s summit with Putin in Alaska, dozens of Russians were rounded up and deported.” www.thetimes.com/us/news-toda...

oh noooo

"LLMs aren't and can never be [100%] accurate therefore they are useless" - genuinely confused about how people who say this collaborate and interact with other humans...

Hi, we wrote about this a few years ago: www.sciencedirect.com/science/arti... Essentially, we were interested in mediator-outcome overlap and what that meant for the interpretation of mediational hypotheses. We had a poster, too:

What People Think Self-Deception Is and Why Philosophers Should Care A summary of our x-phi paper (w/ @ivarr.bsky.social ) in the terrific Imperfect Cognitions blog. 🙌

Do streaming platforms trap us in cultural filter bubbles? We like to think so but the evidence says otherwise. In a new paper @abelaussant.bsky.social and I find the use of streaming platform to be associated with an increase in consumption diversity. sociologicalscience.com/articles-v12...

Among 1073 RCTs from 153 meta-analyses, RCTs designating the meta-analyzed outcome as a primary outcome produced ORs 1.27 (95% CI, 1.14-1.42) higher than RCTs designating the same meta-analyzed outcome as a secondary outcome, possibly due to higher risk of performance, detection bias #biostats

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

A case study in how Cowen's uncritical use of GPT/o3 is causing him to spread misinformation on Marginal Revolution, from @pappubahry.com pappubahry.substack.com/p/misinforma...

Beautiful illustration of the door-in-the-face technique

A remarkable colleague, whose work is the reason why we chose "Matilda" for our open bibliographical platform. Read not only her "Matilda effect" article, but her extensive work on the history of women in science. #STS #historyofscience

I’m especially proud of this article I wrote about Gaussian Processes for the Recast blog! 🥳 GPs are super interesting, but it’s not easy to wrap your head around them at first 🤔 This is a medium level (more intuition than math) introduction to GPs for time series. getrecast.com/gaussian-pro...

In general, people shouldn't treat currently high costs or a negative cost-benefit ratio as reasons to believe it's not worth tackling a problem. Costs aren't fixed or unchangeable. Innovation can be about making processes more efficient, doing more with the same resources.

Academic writing tip: if you can’t explain it clearly, just say it’s complex, layered, and theoretically rich

People are dunking on this but using ChatGPT for this type of task is a good use of the tool, especially if you can verify everything.