Berna Devezer
banner
devezer.bsky.social
Berna Devezer
@devezer.bsky.social
Metascientist @ uidaho. I work at the intersection of behavioral sciences, statistics, and philosophy. Love thinking and talking about science. Post lots of cat and food pics. Allergic to unsolicited advice.
Pinned
I'd like to re-up this paper we published last year bec I believe it makes a fundamental contribution to theoretical metascience but it is woefully underappreciated. We address a key challenge in estimating the reproducibility of a result: The distance of a replication study from the original. 1/n
Reposted by Berna Devezer
Modeling often requires making assumptions and design choices explicit. Experimenters could benefit from such transparent thinking abt the design and knowledge of the range of experiments they are foregoing. They could also benefit from understanding how data produced is conditional on the design.
November 10, 2025 at 4:05 PM
Reposted by Berna Devezer
For everyone who replies to stuff like this with "obviously"

- it's NOT LEGIBLE TO MANY INSTITUTIONS UNTIL IT'S MEASURED
- WE NEED TO HAVE EVIDENCE FOR THE THINGS "we all know"
- ACTUALLY SEEING THE MAGNITUDE AND NUANCES OF THESE KIND OF EFFECTS MATTERS AND IT MATTERS OVER TIME TOO
*Women receive substantially lower "potential" ratings despite receiving higher performance ratings
*Differences in potential ratings account for half of the gender promotion gap
*Women’s lower potential ratings do not reflect future performance: women subsequently outperform male colleagues
Forthcoming in the AER: "“Potential” and the Gender Promotion Gap" by Alan Benson, Danielle Li, and Kelly Shue. www.aeaweb.org/articles?id=...
November 10, 2025 at 10:10 PM
could be a tiger. could be an elephant. could be a giraffe.
Trump: "Nobody knows what magnets are."
November 10, 2025 at 10:11 PM
What's alpha even? It's the first letter of the Greek alphabet. It's read 'a'. AAAAAAAAAA. Obviously a cry for help.
November 10, 2025 at 7:55 PM
My saved posts list is full of random stuff I seem to have accidentally liked while scrolling 😂😭
November 10, 2025 at 6:59 PM
Reposted by Berna Devezer
my critique is not of the first-order belief (i disagree, but whatever) but of the insensitivity to the ways in which quotation marks in english are broken. how do you write for money and not feel that in your bones
November 10, 2025 at 5:36 PM
Reposted by Berna Devezer
BREAKING: The Supreme Court rejects a call to overturn its decision legalizing same-sex marriage nationwide.
Supreme Court rejects call to overturn its decision legalizing same-sex marriage nationwide
The Supreme Court has rejected a call to overturn its landmark decision that legalized same-sex marriage nationwide.
bit.ly
November 10, 2025 at 2:41 PM
Very interesting perspective! I often have the opposite thought: "Experimenting like a modeller"
Perhaps the two could co-exist for better workflows. Extensive exploration on simulated data should be the norm before running experiments to make sure the design is capable of the intended inference.
"Validate With Simulated Truth: A first habit is to test whether an analytical pipeline can recover known conditions."

Very good advice below. So much COVID nonsense (e.g. 'immunological dark matter') basically came down to a non-identifiable model that hadn't been properly tested.
Modelling Like an Experimentalist
Dahlin et al. (2024) apply experimental thinking to a model of mosquito-borne disease transmissions.
onlinelibrary.wiley.com
November 10, 2025 at 4:02 PM
Reposted by Berna Devezer
So many nonsense ad hoc pipelines could be prevented by requiring that they work on synthetic data.

I tend to think of experiments as special cases of inference, since most of the problems I work on cannot be studied in experiments. But I get that many researchers see experiments as base analogy.
"Validate With Simulated Truth: A first habit is to test whether an analytical pipeline can recover known conditions."

Very good advice below. So much COVID nonsense (e.g. 'immunological dark matter') basically came down to a non-identifiable model that hadn't been properly tested.
Modelling Like an Experimentalist
Dahlin et al. (2024) apply experimental thinking to a model of mosquito-borne disease transmissions.
onlinelibrary.wiley.com
November 10, 2025 at 12:41 PM
If the world were a just place either changing sheets and duvet cover wouldn't be such an odious job or sleeping in clean sheets wouldn't be such an exquisite feeling. Alas the cycle continues as the cruel wheel of time turns.
November 10, 2025 at 2:03 AM
Reposted by Berna Devezer
“People are blaming republicans for food prices and flight cancellations and health insurance costs and are furious about gilded ballrooms and Gatsby parties and the president is falling asleep on tv and is the most deeply unpopular president in history, time to cave” is a hell of an argument
November 9, 2025 at 8:55 PM
so sad that there wasn't a single usable still in the whole first season of murderbot that apple had to go with this for their promo
November 9, 2025 at 6:31 PM
Making this abt his mental capacity, dementia, or stupidity attributes him some good faith that he's never earned. When has he said something bec he believes in it? He is and has always been a bad actor through & through. It surprises me that we'd think he wouldn't share this if he knew it's satire.
With a “WOW!,” President Trump posts a claim about DOGE and “royalties linked to Obamacare” that originated on a satirical website. That site’s own About section reads: “Everything on this website is fiction… If you believe that it is real, you should have your head examined.”
November 9, 2025 at 5:19 PM
Can't say I'm very good at it but so many of you are so terrible at writing alt text, I wonder, against my general stance on automatization of meaningful labor, if it would be better to automatize that function. From no alt text to one-word non-descriptions to complete lack of context.. it's dismal.
November 9, 2025 at 4:06 PM
Been craving tater tots ever since someone kept mentioning it in a movie we watched. So I finally had some tonight, after maybe 20 years or so?? Just as I remembered; not the best version of potatoes. This should last me another few decades at least.
November 9, 2025 at 4:56 AM
Reposted by Berna Devezer
That’s before I reckon with the ethics of using a thing for ghost efficiency that is pretty societally devastating.

And on a final personal note, I don’t want to become stupid before my time. I really don’t. I will keep mucking about so that I can keep mucking about, for as long as possible.
November 8, 2025 at 12:37 PM
Reposted by Berna Devezer
I'd like to learn more about how to design useful conjoint experiments. What are some resources you've found useful? #statssky
November 8, 2025 at 3:24 PM
Reposted by Berna Devezer
While my academic world is worrying about threats of AI, the conversations I hear from Vietnam is about flooding and that HCMC is sinking at rate of 3cm/year! No stopping it; this means people will need to either raise their ground or start living on floating houses.
e.vnexpress.net/news/news/en...
Ho Chi Minh City is gradually sinking - VnExpress International
With soil subsidence accelerating at 2–5 cm annually, Ho Chi Minh City faces a crisis as the sinking ground propels many low-lying areas into constant flooding.
e.vnexpress.net
November 8, 2025 at 11:18 AM
Reposted by Berna Devezer
And there are many other reasons I don't use LLMs. They're not reliable, they are built on theft, their creation incurs environmental costs & direct harm to people. The broader societal push to use AI for everything is destructive in countless ways & has horrifying political ramifications.
November 8, 2025 at 1:53 PM
Reposted by Berna Devezer
The more democratic elements of the American constitutional and economic order are actually throwing up road blocks to Trumpism. Protests and boycotts and low level courts and juries and cetera. Elite actors on the other hand are all too frequently caving. Fascinating study in who actually cares!
November 6, 2025 at 9:30 PM
Reposted by Berna Devezer
"Basic scientists consistently enhance scientific output quality, yet research and newly joining researchers have shifted toward applied orientations."

Preprint: arxiv.org/abs/2509.01530

#PhilSci #MetaSci #AcademicSky 🧪
November 8, 2025 at 11:37 AM
Reposted by Berna Devezer
#BREAKING: Justice Jackson has issued an "administrative" stay, temporarily pausing a district court order that would've required the USDA to continue using contingency funds to pay SNAP benefits.

The stay expires 48 hours after the First Circuit rules on USDA's request for a stay pending appeal.
November 8, 2025 at 2:28 AM
Reposted by Berna Devezer
We really need to stop treating betting odds (even if you call them "prediction markets") as politically meaningful data. They're not assessments of what people think (that's polls), they're assessments of what some bettors (many seemingly with more money than sense) think that people will think.
November 7, 2025 at 6:46 PM
Colleagues, I say this with utmost care and love: Using AI slop* does not improve your slides. To the contrary, they make you come across as sloppy. Please resist!

------
*I mean this very literally, as in AI-generated images that have hallucinated non-words and nonsensical details in them.
November 7, 2025 at 11:16 PM