Emory Richardson

With very large numbers of n’s you don’t need randomization, and with LLM’s we can generate very large numbers of n’s, so I think all of science is solved by now. I don’t see any problems with this.

Claude Code 27: Research and Publishing Are Now Two Different Things Some Claude Code fan fiction about the economics of publishing with AI agents set in the very near future

This post really put together the pieces in a way that floored me. Everything is about to change and we have to confront that reality causalinf.substack.com/p/claude-cod...

A related point is that confounding can both increase and decrease effects sizes. When people say "there could be a confounder," it is equally likely a priori that the confounder results in bias *in a conservative direction.*

I have almost never seen this discussed in published research.

Our new short piece in TiCS on intuitive theories of truth: how people judge whether statements could be true, whether statements are true, and whether to assert them as true. A great collab with @keremoktar.bsky.social
@ihandleyminer.bsky.social @kevinzollman.com @lianeleeyoung.bsky.social

I'm a cognitive scientist with an interest in epistemic vigilance, and this essay that's been going around gave me pause.

I don't think it's straightforward to apply the concept of epistemic vigilance to interactions with LLMs, as this essay does.

🧵/

sbgeoaiphd.github.io/rotating_the...

A new preprint, co-authored with @johnwkrakauer.bsky.social:

The Deliberation Taboo

Cognitive science is, nominally, the science of thinking. We argue that the field has no theory of what thinking is and, even worse, that the topic has largely dropped out of focus. 1/

osf.io/preprints/ps...

lots of good points have already been made on using AI Agents for cheating (e.g. the latest Canvas-bot), it degrades learning, etc.

One additional thing I'd like to point out: if you use this stuff, you're not being clever, you're just an asshole.

to explain:

> and while under some conditions universities could react to save the value of the diploma, they’re already under duress in other ways, and reputation is lost much more quickly than gained so it would probably be too little too late > (*some sort of unspecifiable collapse of H.E. as we know it*)

Probably a catastrophizing spiral, but: if we go from 3-5% cheaters to 10-20% cheaters, and the cheating itself is really hard for profs to verify in a way admin will trust > admin will stop punishing cheaters > employers will see diploma’d students lacking skills > perceived diploma value plummets

yes. tbc i'm super friendly to the idea that part of the product is the scientists themselves. maybe even most of it. you want to have the smart physicists around, you need to produce them, which means you need to give them physicist-y things to do.

From the guy that gave you "Agricultural revolution happened because of changes in Earth axial tilt" & "Serfdom in Russia happened because defending low-yield steppe from nomad attackers steppes requires it", take a sit for "bacteria made WWII victory possible".

I promise it will be worth the ride.

Controls for selection by examining movers between places (some similar studies have not) and even compares movers between similar kinds of places (no effect). Seems well designed.

What’s this cool workout app the infers load across exercises?!

Also, I wouldn’t trust a workout app that infers a person’s front squat from their back squat as much as that person’s own judgement or a general rule (eg try 60% and add 10% until you can’t). The limiters on FS are waaay different

An LLM-backed "socratic tutor" to replace reading responses Blog about fatherhood, langauge, developmental psychology, and cognitive science.

New blogpost about a teaching experiment I'm doing this quarter: a "socratic tutor" bot to help students gain understanding of specific reading assignments.

babieslearninglanguage.blogspot.com/2026/02/an-l...

It's not worth debating "does the MVT predict Dems should moderate." The MVT says that *if* the issue space is continuous & 1-dimensional you *must* moderate to win. It's a mechanical relationship.

We don't rely on the MVT because elections are never about a single and easily-divisible issue.

Ahhh, I’m thinking I read this differently than it was intended - “trust others” from a neutral position vs when we have reason to believe otherwise

Really? Because of the “only”, or because of the “evidence”? I thought I’d read the higher order evidence lit relatively closely and would have expected most to accept it with qualifications

Institutions are how we scale up cooperation among millions | Aeon Essays Good institutions are social technologies that scale trust from personal relations to entire nations. How do they work?

Institutions are the social technologies that power our world, allowing us to rely on complete strangers every day of our lives. But how do we ensure that this trust isn’t misplaced? In this Essay, the game theorist Julien Lie-Panis explores what makes institutions function @jliep.bsky.social

Very happy to see "Pretending not to know reveals a capacity for model-based self-simulation", a collaboration with @chazfirestone.bsky.social and @ianbphillips.bsky.social, out in Psych. Science!

journals.sagepub.com/doi/10.1177...

🧵

Bluesky Map Interactive map of 3.4 million Bluesky users, visualised by their follower pattern.

I made a map of 3.4 million Bluesky users - see if you can find yourself!

bluesky-map.theo.io

I've seen some similar projects, but IMO this seems to better capture some of the fine-grained detail

Oscillating between "wow LLMs are really progressing into good products" (Claude Code) and "These seem like they suck and people are acting like they're great" (Clawdbot, SWARMS, etc )

whybot prototype for kids

turing test I made for class

I am flabbergasted I am by how much vibe coding has expanded my capacities as a scientist and teacher.

In the last few weeks, I've mocked up class demos of a live turing test, generated cross-references for an encyclopedia, and prototyped new tablet tasks for developmental psych.

It's wild.

🚀 dashboardr is out!

An #rstats package for building interactive dashboards with tidyverse-style syntax.

Launched with a fun hackathon at @ascor.bsky.social (🍕 included).

📦: favstats.github.io/dashboardr/

Big thanks to the dashboardr team, Digicomlab, and @vivifabrien.bsky.social for the logo 🎨

dplyr 1.2.0 dplyr 1.2.0 fills in some important gaps in dplyr's API: we've added a new complement to `filter()` focused on dropping rows, and we've expanded the `case_when()` family with three new recoding and re...

dplyr 1.2.0 is out now and we are SO excited!

- `filter_out()` for dropping rows

- `recode_values()`, `replace_values()`, and `replace_when()` that join `case_when()` as a complete family of recoding/replacing tools

These are huge quality of life wins for #rstats!

tidyverse.org/blog/2026/02...

I regret to inform you that “1979” by Smashing Pumpkins was released in 1996, which is now 30 years ago. Meaning a similar song released in 2026 would be called “2009”

R Color Palettes Refresh Total rewrite of R Color Palettes and new release of paletteer.

I decided to give one of my most well liked project a bit of a make over, and I wrote about it too!

See how I updated the visuals of r-color-palettes

emilhvitfeldt.com/post/r-color...
#rstats

Really cool new project from @urvi.bsky.social that finds that kids are much better at temporal reasoning than previously reported, if we test them with REAL passing time, rather than hypothetical past or future events and differentiate past and future at 3 years old.

Bayesians Commit the Gambler's Fallacy The gambler's fallacy is the tendency to expect random processes to switch more often than they actually do—for example, to assign a higher probability to heads after a streak of tails. It's often ta...

Great paper giving a rational explanation for the gambler's "fallacy".

so what is the best way to share our shiny apps? @posit.co connect? or shinapps.io?

#rstats

From Debate to Equilibrium: Belief-Driven Multi-Agent LLM Reasoning via Bayesian Nash Equilibrium Multi-agent frameworks can substantially boost the reasoning power of large language models (LLMs), but they typically incur heavy computational costs and lack convergence guarantees. To overcome thes...

A fascinating (for me) paper about BNE in multi-agent systems:
arxiv.org/abs/2506.08292