iiinova — where your digital self keeps evolving

Joe Bak-Coleman @jbakcoleman.bsky.social · 1h

For preprint servers I like moderation priority being a function of cumulative papers published by the authors. First paper of the year in December? Front of the line. 1000 papers across the four of you? lol good luck getting approved.

000 0

Joe Bak-Coleman @jbakcoleman.bsky.social · 1h

Then divide h-index by papers published…. If we’re gonna calculate metrics like that. An h-index of 30 with 30 papers is someone who produces straight gas. H-index of 30 with 120 papers is 🚨

210 2

Joe Bak-Coleman @jbakcoleman.bsky.social · 1h

An easy first step is to start evaluating candidates on cvs containing no more than two papers a year.

150 5

Joe Bak-Coleman @jbakcoleman.bsky.social · 7h

On the flip side, the folks who actually go live in the woods off-grid that I've met tend to be *very far* from having defined political beliefs.

000 0

Joe Bak-Coleman @jbakcoleman.bsky.social · 9h

If you’re interested in false memories, study false memories rather than take the broad features of an old design and assume the way humans engage with car crashes is similar enough that you can render verdict on the original study.

120 2

Joe Bak-Coleman @jbakcoleman.bsky.social · 9h

Surprisingly enough most cars do not get filmed and become stimuli. If you read her thread it makes a pretty unarguable point that all of the *real world not stimuli* features of that claim have changed such that the word replication is conceptually absurd here.

120 2

Joe Bak-Coleman @jbakcoleman.bsky.social · 18h

It’s a mix. They’re mostly oversold and overly complicated but one or two have been interesting and the rest kinda duds. I hadn’t seen ABMs in a while, especially in these venues.

010 1

Joe Bak-Coleman @jbakcoleman.bsky.social · 20h

Old school, but I think generally coded by agentic stuff.

130 3

Joe Bak-Coleman @jbakcoleman.bsky.social · 22h

Yup, and that Lakens shared it, then retweeted his sharing it...

130 3

Joe Bak-Coleman @jbakcoleman.bsky.social · 23h

There's some highly replicable effect of dudes hallucinating your thread talking about the stimuli.

230 3

Reposted by Joe Bak-Coleman

Berna Devezer @devezer.bsky.social · 16/10/2026

so ok when your work and posts are critical of a field, pushback is expected so i'm not gonna act scandalized. however, here is something that happens repeatedly in metascience: i get verbally insulted by dudes who ultimately agree with all of my arguments. and *that* is a gendered dynamic.

2729 72

Joe Bak-Coleman @jbakcoleman.bsky.social · 23h

Anyone else seeing a proliferation of agent based models to review in their inbox?

230 3

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

This exchange is emblematic of trying to critique metascience: Berna: "A Lot changes in 50 years, what are you hoping to get from a replication?" Zwaan: "Berna is lying and saying we're using 1970s stimuli!" Berna: "??????"

1193 19

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

You can legitimately read her thread, it's still up. It has nothing to do with Zwaan's complaint: "Yes, that would be silly, wouldn’t it? Using car footage from the early 1970s. Shame on us for failing to realize this blindingly obvious fact!" No stimulus mention at all... I'd delete and reread.

020 2

Reposted by Joe Bak-Coleman

Berna Devezer @devezer.bsky.social · 16/10/2026

it is only fair that i share this post where i am accused of sharing misinformation. i stand behind my assessment. no information provided in this post changes my take. nowhere in my thread did i say anything about a video?? how very weird!! maybe best to read my research in long form.

rolfzwaan.substack.com

A case of misinformation on academic Bluesky

When the mood strikes, I like reading the comments under news articles.

3408 40

Reposted by Joe Bak-Coleman

Dr Abeba Birhane @abeba.blacksky.app · 16/10/2026

"Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power. We urge mathematicians to discontinue their work with OpenAI and to return to a vision of science that centers human understanding." www.ahmath.org/statements

111177365 1.2K

Reposted by Joe Bak-Coleman

Effin' Birds @effinbirds.com · 16/10/2026

Great thread. This right here is why almost all of my appearances are at comic cons or other ticketed events — that’s who can afford to fly me out. (and buy Delilah’s new book)

020132 201

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

Years later bring it back Remember ADJECTIVE NOUN

010 1

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

It sure beats a great name for a beer that’s awful!

110 1

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

I could easily see myself running a brewery into the ground with obscure stats named beers.

110 1

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

With rare exception you might have a choice or two that isn’t obvious and makes a nontrivial difference in a statistical analysis, for a well posed question. If you have more than that you probably need to understand your domain or stats better and should be doing descriptive work.

060 6

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

In the context of these AI specification curve nightmares, knowing the above cuts out any cause for most of the model formulations. It’s not some subjective researcher flexibility but just models that are flat wrong.

180 8

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

There’s a certain flavor of scientist who somehow thinks we need data to support all claims. We really don’t need data to know that changing your statistical model will result in a different estimate (because it’s a different estimator for a different estimand).

3302 30

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

“Who is more insufferable and less often correct?”

160 6

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

These interviews spare them from having to hire and editor of a fact checker or a journalist to cut the bullshit. The robots are gonna get me another bedroom in Queens?

2121 12

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

Replacing all physical labor with robots in five years? This is the kind of shit people say who haven’t done physical labor.

1191 19

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

Gotta hit that recursive self improvement beat. Can’t be the cult of yud without it.

1110 11

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

Naturaly they land at the exact same regulatory solution as AI companies are suggesting: have their incestuous EA research groups check in on them, or maybe the consulting firm they have a giant financial entanglement with.

1132 13

Joe Bak-Coleman @jbakcoleman.bsky.social · 16/10/2026

Can anyone explain to me why @nytimes.com is having handing Silicon Valley Doomers and EA types air time at this point? Did they decide just to cut the middle man when Roose left?

6394 39

Reposted by Joe Bak-Coleman

Berna Devezer @devezer.bsky.social · 15/10/2026

abstract: We show that fully autonomous AI analysts built on LLMs can, cheaply and at scale, produce the analytic dispersion observed in human many-analyst studies limitations: the analytical paths AI analysts explore need not coincide with those human analysts would traverse 🤔

pnas.org

PNAS

Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...

2111 11

Reposted by Joe Bak-Coleman

Matthew Salganik @msalganik.bsky.social · 14/10/2026

Happy to share a new working paper: "Machine Learning Models Trained on a Longitudinal Survey Dataset Poorly Predict Hundreds of Life Outcomes" arxiv.org/abs/2610.17115 You can hear Emily Cantrell present it Thursday Oct 15 at Oxford: metrics-and-models.github.io (online participation possible)

arxiv.org

Machine Learning Models Trained on a Longitudinal Survey Dataset Poorly Predict Hundreds of Life Outcomes

Understanding the predictability of life outcomes is important for public policy and science, but existing studies focus on a limited number of outcomes and therefore struggle to describe the broad pa...

07029 70

Reposted by Joe Bak-Coleman

Berna Devezer @devezer.bsky.social · 15/10/2026

"Empirical conclusions depend not only on data but also on analytic decisions" Apparently we know this from... wait for it... many analyst studies. Pro tip: If you don't open a book or talk to a statistician, you can experience the joy of discovery every day 🤯

3434 43

Joe Bak-Coleman @jbakcoleman.bsky.social · 15/10/2026

Sometimes it is.... ($250M in compute!) money.usnews.com/investing/ne...

money.usnews.com

110 1

Joe Bak-Coleman @jbakcoleman.bsky.social · 15/10/2026

One thing I'm noticing looking at AI funding is that there is a class of researcher who relies almost solely on industry funding and *doesn't* report it anywhere; rarely having a CV. You just find it in like university announcements.

4478 47

Reposted by Joe Bak-Coleman

Alex Wenzel @alexwenzel.bsky.social · 15/10/2026

Hopefully after the crash it will become impossible for journalists who put their names on this nonsense stenography to find reporting work again.

0243 24

Joe Bak-Coleman @jbakcoleman.bsky.social · 15/10/2026

If the Dems whiff messaging when “the billionaires are polluting your back yard and raising your energy bill in hopes they can steal your job” is on the table, it’s gonna be sad.

2100 10

Joe Bak-Coleman @jbakcoleman.bsky.social · 14/10/2026

Dario Amodei: "What if a worm but needlessly complicated? FEAR OUR TECH AND PUMP THE IPO"

160 6

Joe Bak-Coleman @jbakcoleman.bsky.social · 14/10/2026

A massive loss for science journalism and infectious disease communication. Truly a legend.

0191 19

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

And the wildest thing is that @adambecker.bsky.social makes a good enough podcast to fit the name.

140 4

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

Not fully out of the woods (and no cures etc..) but it all bodes well for the steroid sparing meds keeping my immune system in check.

030 3

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

Aka the immunological gunk is off my cranial nerves.

4210 21

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

Rad thing to read: “Specifically, the previously noted enhancement along the cranial nerves on a study from May 2, 2026 is no longer visualized.”

1190 19

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

The push to just chuck LLMs at papers in hopes of improving them is *wild* and is going to create so many headaches.

150 5

Reposted by Joe Bak-Coleman

Ian Hussey @ianhussey.mmmdata.io · 09/10/2026

Our reflections on the recent "AI and Research Integrity" meeting in Montreal: We are somewhat worried that many AI trustworthiness assessment tools are not trustworthy. @lhdjung.bsky.social @jamiecummins.bsky.social trustworthy.scientific.claims/posts/trustw...

trustworthy.scientific.claims

Trustworthiness assessment tools must be trustworthy

17934 79

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

Good post. There's been a *ton* of groups pushing this, almost ubiquitously funded in part/wholly by Coefficient giving or one of the other EA outfits. It's great to see some pushback from folks in metascience.

050 5

Reposted by Joe Bak-Coleman

Sophie E. Hill @sophieehill.bsky.social · 12/10/2026

Scientists at Newcastle & Durham published a study supporting the talking points of the grouse moor industry, funded by an "anonymous donation". After a year-long investigation, the authors now admit that the money came from... a group of grouse moor owners. onlinelibrary.wiley.com/doi/10.1002/...

onlinelibrary.wiley.com

Correction to “The Influence of Different Aspects of Grouse Moorland Management on Nontarget Bird Assemblages”

Click on the article title to read more.

46441 64

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

I feel like this is the kind of graph you get when you’re vibe data sciencing and don’t know what things are called.

2100 10

Joe Bak-Coleman @jbakcoleman.bsky.social · 13/10/2026

I like thinking about how they’re passing data around that x changes between the jitter and the box.

020 2