Vera Wilde
@verawil.de
Scientist (PhD), writer, transparency activist. Nerd-of-all-trades (research methodologist). Seeker of truth, especially wild.
Rooting around for the original of "Skinnamarink" after @dingdingpeng.the100.ci mentioned her husband's 1980s bedtime song for the kids. Turns out it's 47 years old and oh noes my ovaries are turning gray www.youtube.com/watch?v=kFFD...
Skinnamarink | Sharon, Lois & Bram
YouTube video by Sharon, Lois & Bram
www.youtube.com
December 5, 2025 at 12:05 PM
Rooting around for the original of "Skinnamarink" after @dingdingpeng.the100.ci mentioned her husband's 1980s bedtime song for the kids. Turns out it's 47 years old and oh noes my ovaries are turning gray www.youtube.com/watch?v=kFFD...
wildetruth.substack.com/p/rarity-rou...
I made an app to help people correct for base rate bias. Does it?
Pilot study data say: Maybe, maybe not.
I'm inclined to think the table visual aid format may be inferior to frequency-format trees or something else. Any thoughts? References? Anecdata?
I made an app to help people correct for base rate bias. Does it?
Pilot study data say: Maybe, maybe not.
I'm inclined to think the table visual aid format may be inferior to frequency-format trees or something else. Any thoughts? References? Anecdata?
Rarity Roulette Pilot Study
I made an app to help people correct for base rate bias. Does it?
wildetruth.substack.com
December 5, 2025 at 11:40 AM
wildetruth.substack.com/p/rarity-rou...
I made an app to help people correct for base rate bias. Does it?
Pilot study data say: Maybe, maybe not.
I'm inclined to think the table visual aid format may be inferior to frequency-format trees or something else. Any thoughts? References? Anecdata?
I made an app to help people correct for base rate bias. Does it?
Pilot study data say: Maybe, maybe not.
I'm inclined to think the table visual aid format may be inferior to frequency-format trees or something else. Any thoughts? References? Anecdata?
Reposted by Vera Wilde
I have news. Wonderful news.
New paper on #disinformation. The title is the thesis. Think of disinformation as content that functions to optimise the decisions of the audience for the disinformant. This doesn't require deception or content that induces false belief.
philpapers.org/rec/LITDIF
New paper on #disinformation. The title is the thesis. Think of disinformation as content that functions to optimise the decisions of the audience for the disinformant. This doesn't require deception or content that induces false belief.
philpapers.org/rec/LITDIF
Clayton Littlejohn, Disinformation is for Degrading the Value of Information, not Confirming Falsehoods - PhilPapers
According to a recent account of disinformation, disinformation is content that “generates ignorance” (Simion 2024a; 2024b). The view improves upon previous accounts that focused upon the potential fo...
philpapers.org
November 29, 2025 at 12:40 AM
I have news. Wonderful news.
New paper on #disinformation. The title is the thesis. Think of disinformation as content that functions to optimise the decisions of the audience for the disinformant. This doesn't require deception or content that induces false belief.
philpapers.org/rec/LITDIF
New paper on #disinformation. The title is the thesis. Think of disinformation as content that functions to optimise the decisions of the audience for the disinformant. This doesn't require deception or content that induces false belief.
philpapers.org/rec/LITDIF
Not to be confused with s-values!
search.r-project.org/CRAN/refmans...
search.r-project.org/CRAN/refmans...
November 28, 2025 at 9:33 AM
Not to be confused with s-values!
search.r-project.org/CRAN/refmans...
search.r-project.org/CRAN/refmans...
Went to a FemTech conference. Watched companies profit from: (1) conflating base rates, (2) letting babies starve, then monitoring the jaundice, and (3) selling at-home testing without statistical literacy.
1. Critique
2. Synthesize
...
4. Profit!
wildetruth.substack.com/p/femtech-fails
1. Critique
2. Synthesize
...
4. Profit!
wildetruth.substack.com/p/femtech-fails
FemTech Fails
Three technologies that mistake surveillance for care
wildetruth.substack.com
November 25, 2025 at 12:26 PM
Went to a FemTech conference. Watched companies profit from: (1) conflating base rates, (2) letting babies starve, then monitoring the jaundice, and (3) selling at-home testing without statistical literacy.
1. Critique
2. Synthesize
...
4. Profit!
wildetruth.substack.com/p/femtech-fails
1. Critique
2. Synthesize
...
4. Profit!
wildetruth.substack.com/p/femtech-fails
Oh noed!
It hath snowed
on my freshly planted windowboxes.
It hath snowed
on my freshly planted windowboxes.
November 24, 2025 at 10:40 AM
Oh noed!
It hath snowed
on my freshly planted windowboxes.
It hath snowed
on my freshly planted windowboxes.
The FemTech Summit today in Berlin was everything a conference should be:
– big curtains around panel rooms for hiding
– speed-dating rooms w ppl shouting pitches at other ppl who were not the important ppl they wanted to pitch
– free coffee, but no decaf (cruel)
– free cookie for kid (redemption)
– big curtains around panel rooms for hiding
– speed-dating rooms w ppl shouting pitches at other ppl who were not the important ppl they wanted to pitch
– free coffee, but no decaf (cruel)
– free cookie for kid (redemption)
November 22, 2025 at 5:34 PM
The FemTech Summit today in Berlin was everything a conference should be:
– big curtains around panel rooms for hiding
– speed-dating rooms w ppl shouting pitches at other ppl who were not the important ppl they wanted to pitch
– free coffee, but no decaf (cruel)
– free cookie for kid (redemption)
– big curtains around panel rooms for hiding
– speed-dating rooms w ppl shouting pitches at other ppl who were not the important ppl they wanted to pitch
– free coffee, but no decaf (cruel)
– free cookie for kid (redemption)
Five-year-old: "They're monster cookies...
"This one is in profile."
"This one is in profile."
November 20, 2025 at 3:00 PM
Five-year-old: "They're monster cookies...
"This one is in profile."
"This one is in profile."
Finally built a working version of the Rarity Roulette simulator! verawilde.github.io/rarity-roule...
It's an app to help visualize estimated hypothetical outcomes of mass screenings for low-prevalence problems.
More: wildetruth.substack.com/p/rarity-rou...
Would love feedback + feature requests.
It's an app to help visualize estimated hypothetical outcomes of mass screenings for low-prevalence problems.
More: wildetruth.substack.com/p/rarity-rou...
Would love feedback + feature requests.
November 18, 2025 at 10:54 AM
Finally built a working version of the Rarity Roulette simulator! verawilde.github.io/rarity-roule...
It's an app to help visualize estimated hypothetical outcomes of mass screenings for low-prevalence problems.
More: wildetruth.substack.com/p/rarity-rou...
Would love feedback + feature requests.
It's an app to help visualize estimated hypothetical outcomes of mass screenings for low-prevalence problems.
More: wildetruth.substack.com/p/rarity-rou...
Would love feedback + feature requests.
New favorite example: LLMs as rare event screeners!
github.com/p-e-w/heretic
Everyone: "We'll fix alignment with more accuracy!"
Math: lol no you won't.
Heretic: "Hold my AGPL license."
Same problem as mass screenings: structure > vibes.
github.com/p-e-w/heretic
Everyone: "We'll fix alignment with more accuracy!"
Math: lol no you won't.
Heretic: "Hold my AGPL license."
Same problem as mass screenings: structure > vibes.
GitHub - p-e-w/heretic: Fully automatic censorship removal for language models
Fully automatic censorship removal for language models - p-e-w/heretic
github.com
November 17, 2025 at 6:50 AM
New favorite example: LLMs as rare event screeners!
github.com/p-e-w/heretic
Everyone: "We'll fix alignment with more accuracy!"
Math: lol no you won't.
Heretic: "Hold my AGPL license."
Same problem as mass screenings: structure > vibes.
github.com/p-e-w/heretic
Everyone: "We'll fix alignment with more accuracy!"
Math: lol no you won't.
Heretic: "Hold my AGPL license."
Same problem as mass screenings: structure > vibes.
So I was going to try to code solutions to Frederick Mosteller's *Fifty Challenging Problems in Probability with Solutions*, (I don't know, when my kids go to college?). But it looks like there is already a hipper version of out there in Project Euler
loriculus.org/blog/euler-45/
loriculus.org/blog/euler-45/
Rohan Prinja
loriculus.org
loriculus.org
November 16, 2025 at 2:48 AM
So I was going to try to code solutions to Frederick Mosteller's *Fifty Challenging Problems in Probability with Solutions*, (I don't know, when my kids go to college?). But it looks like there is already a hipper version of out there in Project Euler
loriculus.org/blog/euler-45/
loriculus.org/blog/euler-45/
Chat Control is back! Now buried in telecom regulations.
www.patrick-breyer.de/en/chat-cont...
reclaimthenet.org/the-disguise...
www.patrick-breyer.de/en/chat-cont...
reclaimthenet.org/the-disguise...
CHAT CONTROL 2.0 THROUGH THE BACK DOOR – Breyer warns: "The EU is playing us for fools – now they’re scanning our texts and banning teens!" [updated]
Just before a decisive meeting in Brussels, digital rights expert and former Member of the European Parliament Dr. Patrick Breyer is sounding the alarm. Using a "deceptive sleight of hand," a mandator...
www.patrick-breyer.de
November 15, 2025 at 7:01 AM
Chat Control is back! Now buried in telecom regulations.
www.patrick-breyer.de/en/chat-cont...
reclaimthenet.org/the-disguise...
www.patrick-breyer.de/en/chat-cont...
reclaimthenet.org/the-disguise...
Reposted by Vera Wilde
Newly released Stockholm Declaration recommends the following reforms to publishing:
1. Academia resumes control of publishing
2. Incentive systems to merit quality, not quantity
3. Independent fraud detection and prevention
4. Legislation and policies to protect science quality and integrity
1. Academia resumes control of publishing
2. Incentive systems to merit quality, not quantity
3. Independent fraud detection and prevention
4. Legislation and policies to protect science quality and integrity
Reformation of science publishing: the Stockholm Declaration | Royal Society Open Science
Science relies on integrity and trustworthiness. But scientists under career pressure
are lured to purchase fake publications from ‘paper mills’ that use AI-generated data,
text and image fabrication....
royalsocietypublishing.org
November 5, 2025 at 7:07 PM
Newly released Stockholm Declaration recommends the following reforms to publishing:
1. Academia resumes control of publishing
2. Incentive systems to merit quality, not quantity
3. Independent fraud detection and prevention
4. Legislation and policies to protect science quality and integrity
1. Academia resumes control of publishing
2. Incentive systems to merit quality, not quantity
3. Independent fraud detection and prevention
4. Legislation and policies to protect science quality and integrity
Reposted by Vera Wilde
Psychologists are also definitely guilty of confusing change with treatment response, which is why I wrote about it here:
www.the100.ci/2025/01/22/r...
www.the100.ci/2025/01/22/r...
Reviewer notes: In a randomized experiment, the pre-post differences are not effect estimates
Reviewer notes are a new short format with brief explanations of basic ideas that might come in handy during (for example) the peer-review process. They are a great way to keep Julia from writing 10,0...
www.the100.ci
September 27, 2025 at 12:40 PM
Psychologists are also definitely guilty of confusing change with treatment response, which is why I wrote about it here:
www.the100.ci/2025/01/22/r...
www.the100.ci/2025/01/22/r...
Reposted by Vera Wilde
“Landmark study” where openai comes to understand what we’ve been telling them about their own software for three years. (I’m sure many have been saying it for longer.)
"In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits."
www.computerworld.co...
#hallucinations #generativeAI
www.computerworld.co...
#hallucinations #generativeAI
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limits.
www.computerworld.com
September 28, 2025 at 12:47 PM
“Landmark study” where openai comes to understand what we’ve been telling them about their own software for three years. (I’m sure many have been saying it for longer.)
Spending some time with family in the Research Triangle area and looking to connect with folks interested in research methods, causal inferences, and AI ethics.
Ping if this is you!
I'm working on mass screenings for low-prevalence problems, science + statistics reform, uncertainty, and feedbacks.
Ping if this is you!
I'm working on mass screenings for low-prevalence problems, science + statistics reform, uncertainty, and feedbacks.
September 26, 2025 at 12:47 PM
Spending some time with family in the Research Triangle area and looking to connect with folks interested in research methods, causal inferences, and AI ethics.
Ping if this is you!
I'm working on mass screenings for low-prevalence problems, science + statistics reform, uncertainty, and feedbacks.
Ping if this is you!
I'm working on mass screenings for low-prevalence problems, science + statistics reform, uncertainty, and feedbacks.
Rejoice! Papa Bear has been found again.
His reappearance spurs me to think harder about endpoints, what really matters -- and how the liberty-security sequential-feedback continuum may extend to medical case studies.
But where fore-art thou, adequate data?
wildetruth.substack.com/p/bear-is-ba...
His reappearance spurs me to think harder about endpoints, what really matters -- and how the liberty-security sequential-feedback continuum may extend to medical case studies.
But where fore-art thou, adequate data?
wildetruth.substack.com/p/bear-is-ba...
Bear Is Back (Again)
Celebrating a reunion and focusing on what matters in science as in life
wildetruth.substack.com
August 11, 2025 at 11:57 AM
Rejoice! Papa Bear has been found again.
His reappearance spurs me to think harder about endpoints, what really matters -- and how the liberty-security sequential-feedback continuum may extend to medical case studies.
But where fore-art thou, adequate data?
wildetruth.substack.com/p/bear-is-ba...
His reappearance spurs me to think harder about endpoints, what really matters -- and how the liberty-security sequential-feedback continuum may extend to medical case studies.
But where fore-art thou, adequate data?
wildetruth.substack.com/p/bear-is-ba...
Reposted by Vera Wilde
but that’s fine.
It served a useful function at the time in getting people to think about potential informational value in P-values - it just was too crude an approach, so stimulated development of other methods.
Can we have a “z-curve for dummies” explainer to improve take-up?
It served a useful function at the time in getting people to think about potential informational value in P-values - it just was too crude an approach, so stimulated development of other methods.
Can we have a “z-curve for dummies” explainer to improve take-up?
R.I.P p-curve. Even the authors of p-curve no longer believe in the method.
replicationindex.com/2025/08/09/d...
replicationindex.com/2025/08/09/d...
Datacolada Has Given Up on p-Curve - Replicability-Index
When p-curve debuted in 2014, it was billed as a powerful tool for detecting publication bias and estimating evidential value from the distribution of statistically significant results. Its creators —...
replicationindex.com
August 10, 2025 at 5:28 AM
but that’s fine.
It served a useful function at the time in getting people to think about potential informational value in P-values - it just was too crude an approach, so stimulated development of other methods.
Can we have a “z-curve for dummies” explainer to improve take-up?
It served a useful function at the time in getting people to think about potential informational value in P-values - it just was too crude an approach, so stimulated development of other methods.
Can we have a “z-curve for dummies” explainer to improve take-up?
Reposted by Vera Wilde
Arguably the first and most important use of language: getting others to do things.
(Conveying factual information, on the other hand: not universally acquired and unstable without training.)
(Conveying factual information, on the other hand: not universally acquired and unstable without training.)
July 31, 2025 at 10:47 AM
Arguably the first and most important use of language: getting others to do things.
(Conveying factual information, on the other hand: not universally acquired and unstable without training.)
(Conveying factual information, on the other hand: not universally acquired and unstable without training.)
Pediatrician: Does your six month old say "baba"?
Me: My baby says "hi," "hey," "Mama," "I go," and ("Are you gonna go get your brother?") "I'm gonna"!
Me: My baby says "hi," "hey," "Mama," "I go," and ("Are you gonna go get your brother?") "I'm gonna"!
July 31, 2025 at 9:53 AM
Pediatrician: Does your six month old say "baba"?
Me: My baby says "hi," "hey," "Mama," "I go," and ("Are you gonna go get your brother?") "I'm gonna"!
Me: My baby says "hi," "hey," "Mama," "I go," and ("Are you gonna go get your brother?") "I'm gonna"!
I got a free copy of *The Precipice: Existential Risk and the Future of Humanity* from 80,000 Hours. Glancing at the table of contents, it fails to recognize perspectival limits including bias and error in the evidence on which it's based. But, on p. 6 so far, it's pretty good.
Ruin it for me.
Ruin it for me.
July 22, 2025 at 10:21 AM
I got a free copy of *The Precipice: Existential Risk and the Future of Humanity* from 80,000 Hours. Glancing at the table of contents, it fails to recognize perspectival limits including bias and error in the evidence on which it's based. But, on p. 6 so far, it's pretty good.
Ruin it for me.
Ruin it for me.
Reposted by Vera Wilde
So much this. Except I started five blog posts in the midst of a mild autoimmune flare, childcare going away for a while, and grandparents flying in from the States for a summer visit. Now I am not sure which if any of us will survive the next two weeks.
I started a blog post but now I have to pick up the kids and then daycare will be closed for two weeks. So I'm just gonna leave that here as a cliffhanger
July 18, 2025 at 9:19 PM
So much this. Except I started five blog posts in the midst of a mild autoimmune flare, childcare going away for a while, and grandparents flying in from the States for a summer visit. Now I am not sure which if any of us will survive the next two weeks.
Day 5 with my mother.
Please send help.
Chocolate, flame-throwers --
every bit helps.
Please send help.
Chocolate, flame-throwers --
every bit helps.
July 15, 2025 at 2:38 PM
Day 5 with my mother.
Please send help.
Chocolate, flame-throwers --
every bit helps.
Please send help.
Chocolate, flame-throwers --
every bit helps.