Stephen Wild
@stephenjwild.bsky.social
I try to put straight lines through things but usually fail. Try to be Bayesian when I can. Views my own. RT/like != endorsement.
*evil cackling*
For decades, scientific research has been plagued by deep structural flaws, leading to false findings that shape medicine, policy, and public understanding. | https://bit.ly/3ZELke0
John Ioannidis explores why bad science persists, and how we can fix it.
#philsci
John Ioannidis explores why bad science persists, and how we can fix it.
#philsci
John Ioannidis on why science is broken
For decades, scientific research has been plagued by deep structural flaws, leading to false findings that shape medicine, policy, and public understanding. John Ioannidis exposed this crisis twenty y...
iai.tv
February 17, 2026 at 7:42 PM
*evil cackling*
Reposted by Stephen Wild
yep! And I do think that is where debate is heading - there are a _lot_ of problems with actually-existing-AI that not only don't rest on stochastic parrots, but that are more likely missed when you don't think about the technology's usefulness and uses.
February 17, 2026 at 5:09 PM
yep! And I do think that is where debate is heading - there are a _lot_ of problems with actually-existing-AI that not only don't rest on stochastic parrots, but that are more likely missed when you don't think about the technology's usefulness and uses.
"Stochastic parrots" is a cool phrase. Also it should be laudatory, not derogatory.
Doubling down on "just stochastic parrots" is absolutely not working but there is a lot of left-debate that is not about that and that isn't captured in this article. The what-my-Bluesky-feed-shows-availability heuristic can be misleading.
February 17, 2026 at 5:47 PM
"Stochastic parrots" is a cool phrase. Also it should be laudatory, not derogatory.
Good post
Here’s my new Substack post about the value of being curious about new fields and literatures.
open.substack.com/pub/davebrad...
open.substack.com/pub/davebrad...
February 17, 2026 at 3:26 PM
Good post
Reposted by Stephen Wild
Who do you recommend following for non-US news? Self-promotion encouraged.
February 17, 2026 at 1:28 PM
Who do you recommend following for non-US news? Self-promotion encouraged.
Reposted by Stephen Wild
In IRT modeling, shouldn't the 1PL model have one more degree of freedom than the Rasch model? In the 1PL I am estimating a common discrimination parameter where the Rasch model sets it to 1. But the mirt package in R is saying they have the same number of parameters.
February 17, 2026 at 11:32 AM
In IRT modeling, shouldn't the 1PL model have one more degree of freedom than the Rasch model? In the 1PL I am estimating a common discrimination parameter where the Rasch model sets it to 1. But the mirt package in R is saying they have the same number of parameters.
*Excellent* book that one could use the term "magisterial" to describe. But not me, because @donskerclass.bsky.social beat me to it.
One thing Elkins's LEGACY OF VIOLENCE made clear was counterinsurgency, historically, looked a lot like what's happening in MN. Arbitrary brutality, extra-legal behavior, etc. "Hearts and minds" was a PR campaign the Brits sold to the US during the Malayan emergency, where they hid the ugly bits.
February 17, 2026 at 11:48 AM
*Excellent* book that one could use the term "magisterial" to describe. But not me, because @donskerclass.bsky.social beat me to it.
Reposted by Stephen Wild
maybe helps that I don’t spend much time discussing things I’m already highly certain of on here. E.g. If someone writes a new paper showing good evidence that declining responses rates are actually lowering poll quality I’ll be all ears (it’s weird that people keep failing to find that in fact)
February 17, 2026 at 12:17 AM
maybe helps that I don’t spend much time discussing things I’m already highly certain of on here. E.g. If someone writes a new paper showing good evidence that declining responses rates are actually lowering poll quality I’ll be all ears (it’s weird that people keep failing to find that in fact)
Reposted by Stephen Wild
Yes. One thing every scientist should probably learn is that falsification can't escape the Quine-Duhem problem. In this case, if your toy statistical model of science predicts one thing, and you see another, you can't tell whether you refuted your hypothesis or your model is wrong.
February 16, 2026 at 6:19 PM
Yes. One thing every scientist should probably learn is that falsification can't escape the Quine-Duhem problem. In this case, if your toy statistical model of science predicts one thing, and you see another, you can't tell whether you refuted your hypothesis or your model is wrong.
I hate that I understand every word of this tweet
Will Stancil became a hero to the left for resisting ICE, followed shortly by people coming up with increasingly esoteric reasons for Still Hating Him.
February 16, 2026 at 11:22 PM
I hate that I understand every word of this tweet
Reposted by Stephen Wild
In this post, Gelman gives 4 "sources" or "types" of priors.
Can anyone shed some light in the 4th type? "Prior as unconditional distribution"
#stats
statmodeling.stat.columbia.edu/2025/05/21/p...
Can anyone shed some light in the 4th type? "Prior as unconditional distribution"
#stats
statmodeling.stat.columbia.edu/2025/05/21/p...
Prior as data, prior as belief, prior as soft constraint, prior as unconditional distribution in a generative model | Statistical Modeling, Causal Inference, and Social Science
statmodeling.stat.columbia.edu
February 16, 2026 at 3:12 PM
In this post, Gelman gives 4 "sources" or "types" of priors.
Can anyone shed some light in the 4th type? "Prior as unconditional distribution"
#stats
statmodeling.stat.columbia.edu/2025/05/21/p...
Can anyone shed some light in the 4th type? "Prior as unconditional distribution"
#stats
statmodeling.stat.columbia.edu/2025/05/21/p...
Reposted by Stephen Wild
Or, as written up in SIGBOIVK last year:
February 15, 2026 at 3:42 PM
Or, as written up in SIGBOIVK last year:
Reposted by Stephen Wild
A new blog post motivated from this amazing discussion with @akhilrao.bsky.social and @donskerclass.bsky.social
#EconSky #RStats #Stan
jamesblandecon.github.io/posts/2026-0...
#EconSky #RStats #Stan
jamesblandecon.github.io/posts/2026-0...
February 15, 2026 at 3:44 PM
A new blog post motivated from this amazing discussion with @akhilrao.bsky.social and @donskerclass.bsky.social
#EconSky #RStats #Stan
jamesblandecon.github.io/posts/2026-0...
#EconSky #RStats #Stan
jamesblandecon.github.io/posts/2026-0...
YouTube bad
like im getting euro nazi stuff now, dont like
February 15, 2026 at 12:36 PM
YouTube bad
Reposted by Stephen Wild
Another cool one is the reflection problem academic.oup.com/restud/artic...
Identification of Endogenous Social Effects: The Reflection Problem
Abstract. This paper examines the reflection problem that arises when a researcher observing the distribution of behaviour in a population tries to infer w
academic.oup.com
February 15, 2026 at 12:23 PM
Another cool one is the reflection problem academic.oup.com/restud/artic...
These posts are great because they suggest who to mute (everyone who gets mad about putting your cast iron in the dishwasher or who tries to tell how to care for it)
All my cast iron, fresh out of the dishwasher and ready to cook
February 15, 2026 at 1:54 AM
These posts are great because they suggest who to mute (everyone who gets mad about putting your cast iron in the dishwasher or who tries to tell how to care for it)
*chef's kiss*
Famous starry-eyed idealists like Stalin and Churchill.
February 15, 2026 at 12:36 AM
*chef's kiss*
Scream it from the rooftops
True causal inference has not been tried :)
More seriously: causal inference is not about prioritizing certain methods, it’s about being very clear about effect you are actually targeting (if any) and what assumptions justify your inferences.
More seriously: causal inference is not about prioritizing certain methods, it’s about being very clear about effect you are actually targeting (if any) and what assumptions justify your inferences.
February 14, 2026 at 6:58 PM
Scream it from the rooftops
10/10 response
What do you think Claude Code is?
February 13, 2026 at 11:34 PM
10/10 response
Reposted by Stephen Wild
Social media bad
February 13, 2026 at 11:08 PM
Social media bad
Always retaliate first
The only thing necessary for the triumph of evil is for good players to not play tit-for-tat
February 13, 2026 at 10:48 PM
Always retaliate first