Dean Eckles
banner
eckles.bsky.social
Dean Eckles
@eckles.bsky.social

networks, contagion, causality
faculty at MIT

Mathematics 19%
Physics 19%

Yeah not sure... Changes in the relative cost of manufacturing with standardized parts vs not, thereby making this more appealing for manufacturers?
The U.S. military is being used inside the United States. There's a lot we don't know about how, why, and under what authorities.

Today, Lawfare is launching a new project–which includes a tracker and a map–to follow where and how the military is being domestically deployed.

Reposted by Dean Eckles

the "tech debt" metaphor primes people for bad intuitions

debts have a strict lending schedule, and refinancing if you're in a pinch

debtors don't wake up Tuesday at 3am and set part of your business afire unless you pay them off

the only part that really fits is the carrying cost
ICE agents in Chicago violently detain a non verbal autistic man for being “non compliant”

He couldn’t “comply” because of his disability…they don’t care

This is not the first time they’ve gone after a disabled person

They target the most vulnerable because it’s an easy way to meet their quota
Haha, this from the New Yorker is getting passed around the math dork community. I did a comic about this kind of thought a few years ago: www.smbc-comics.com/comic/commut...

Reposted by Dean Eckles

Feels like the opening to a very bad sci-fi film.

Reposted by Dean Eckles

🧠⚙️ Interested in decision theory+cogsci meets AI? Want to create methods for rigorously designing & evaluating human-AI workflows?

I'm recruiting PhDs to work on:
🎯 Stat foundations of multi-agent collaboration
🌫️ Model uncertainty & meta-cognition
🔎 Interpretability
💬 LLMs in behavioral science
The aesthetic of an authoritarian not befitting the civic traditions of the United States
With large portraits of himself, 24-karat golden adornments and a triumphal arch, President Trump is asserting his power through what might be called an imperial aesthetic, or visual cues designed to project personal command and grandeur.
The imperial aesthetic at the heart of Donald Trump’s presidency
In addition to his accumulation of political power, Trump has embraced visual cues designed to project personal command and grandeur.
www.washingtonpost.com
This is quite a story: the US employed every strongarm tactic available, including personal threats against delegates and their families, to derail an agreement on reducing shipping emissions that was almost finalized. on.ft.com/47lq3dZ

Not a public link?

An early analysis I did at FB was revisiting the market research team's clustering of users — for which they got help from the data science team, so it was a complex variational Bayes thing. It basically just sorted them into 6 clusters, with clear ordering by how active they were.

Confusingly, I gather in biostat "completely randomised" means Bernoulli randomization

Reposted by Dean Eckles

Nothing to worry about, totally normal
“100% ANTI TRUMP, WHICH IS PROBABLY ILLEGAL.”

Well I switch back and forth between WMBR and Emerson's WERS :)

Reposted by Dean Eckles

Voting is sexy

Reposted by Dean Eckles

Why analysis informing a decision shouldn’t try and predict the decision it’s informing: kucharski.substack.com/p/why-do-peo...
What if people appreciate having an abundance of content and communication, more than they feel overloaded by it?

@annisch.bsky.social et al decided to have a look. Their results? "We found that appreciation for abundance was about twice as common as overload".

Paper: journalqd.org/article/view...
There still seems to be a lot of confusion about significance testing in psych. No, p-values *don’t* become useless at large N. This flawed point also used to be framed as "too much power". But power isn't the problem – it's 1) unbalanced error rates and 2) the (lack of a) SESOI. 1/ >
But here's, the thing, p values and significance become useless at such large sample sizes. When you're dividing the coefficient by the SE and the sample size is in the tens of thousands, EVERYTHING IS SIGNIFICANT. All you're testing is whether the coefficient is different than zero.

Reposted by Dean Eckles

Journalism tip: If someone says drop the story, that’s a sign to keep reporting.

Back reporters who won’t flinch. Support our independent journalism ➜ propub.li/43MugoG
But here's, the thing, p values and significance become useless at such large sample sizes. When you're dividing the coefficient by the SE and the sample size is in the tens of thousands, EVERYTHING IS SIGNIFICANT. All you're testing is whether the coefficient is different than zero.

Reposted by Dean Eckles

Also new in v1.109: custom feed operators now receive more interaction events.

If you run a custom feed, now you'll know when people like, quote, repost, reply, and see posts in your feed. All valuable signals that can help you drive higher-quality recommendations in your feeds.
🎇New package alert @wired.com! This one has been in the works for months. If WIRED was going to tackle AI -- something we cover daily -- we had to go big. So here are 17 different stories about the way AI is changing us, even as the technology itself keeps moving www.wired.com/ai-issue/
AI of a Thousand Faces
What happens now that AI is everywhere and in everything? WIRED can’t tell the future, but we can try to make sense of it. Behold: 17 readings from the furthest reaches of the AI age.
www.wired.com
Appendix A7 most striking graph for me:
At a certain point in Turkey, virtually every non-public person I interviewed refused to give their last name. Eventually I stopped even writing "______ declined to give his/her last name for fear of government reprisal," because it became redundant, and editors understood. But here's the thing...
An embarrassingly obvious point, but...

This. Is. Illegal.

If anything like the rule of law were in force, Trump couldn't arbitrarily increase a tariff because he is annoyed.

In important and increasing ways, the rule of law in the US no longer exists.
www.cnn.com/2025/10/25/b...
Trump says he’s increasing tariffs on Canada by 10% after Ontario’s Reagan ad | CNN Business
President Donald Trump said Saturday he is increasing the tariff on Canada by 10% over current levels, further escalating trade tensions over what he called a “fake” ad that featured parts of an anti-...
www.cnn.com

Reposted by Dean Eckles

For the second time, Cato has sued DoJ & FBI for internal records regarding the Bureau’s use (or misuse) of FISA Section 702 surveillance powers. The de facto destruction of PCLOB & Patel's elimination of an internal audit unit make such oversight litigation imperative.

www.cato.org/blog/has-tea...
Cato Sues FBI Over FISA Records, Again
Two things are certain. The first is that the FISA Section 702 program is set to expire in late April 2026. The second is that the kind of public interest FOIA litigation Cato is engaged in on this ca...
www.cato.org

So what do people think of using (quasi-)Poisson regression for a binary outcome?

How about in discrete & continuous time hazard models?

Crazy or actually sensible?
Trump: "I don't think we're necessarily going to ask for a declaration of war, I think we're just gonna kill people that are bringing drugs into our country. We're going to kill them. They're going to be, like dead."

Reposted by Dean Eckles

I was interviewed on the Kill Switch podcast about @iceblock.app, the mutual aid group @norcalresist.bsky.social, and the insane situation we’re in where Apple and Google are collaborating with Trump to censor apps at will podcasts.apple.com/us/podcast/k...
maybe ICEBlock was 'activism theater,' but is banning it protecting us?
Podcast Episode · kill switch · 10/22/2025 · 35m
podcasts.apple.com

Reposted by Dean Eckles