Heather Froehlich
@heatherfro.bsky.social
supporting researchers counting words in various ways with computers at university of arizona libraries; increasingly displaced new englander
Reposted by Heather Froehlich
Just a PSA for the DH people out there in case they're also thinking about how they could visualize their reading for fun, you can export your Libby timeline in HTML, CSV, and JSON!
#DHmakes
#DHmakes
NEW: How many books did you read this year? If it was more than 2, congrats: you read more than the median American.
Read on for way more @today.yougov.com polling from me on America's book-reading habits in 2025. 1/
Read on for way more @today.yougov.com polling from me on America's book-reading habits in 2025. 1/
December 31, 2025 at 4:53 PM
Just a PSA for the DH people out there in case they're also thinking about how they could visualize their reading for fun, you can export your Libby timeline in HTML, CSV, and JSON!
#DHmakes
#DHmakes
Reposted by Heather Froehlich
[Last blog post of 2025] "AI-powered search" hides at least 4 different things—post-retrieval features, semantic search, LLM ranking, and synthesis. Your concerns about one may not apply to another. New post unpacking what's actually under the hood. aarontay.substack.com/p/what-do-we...
What Do We Actually Mean by "AI-Powered Search"?
When we say "AI-powered search engine," we're conflating at least four different things—and your concerns about one may not apply to another.
aarontay.substack.com
December 31, 2025 at 1:47 PM
[Last blog post of 2025] "AI-powered search" hides at least 4 different things—post-retrieval features, semantic search, LLM ranking, and synthesis. Your concerns about one may not apply to another. New post unpacking what's actually under the hood. aarontay.substack.com/p/what-do-we...
“The tell isn’t that fake citations look wrong. It’s that they look too right. Too convenient. Too perfectly aligned with whatever point the AI is making“
open.substack.com/pub/cardcata...
open.substack.com/pub/cardcata...
How to Spot AI Hallucinations Like a Reference Librarian
The verification tricks that would make fact-checkers weep with joy.
open.substack.com
December 30, 2025 at 6:08 PM
“The tell isn’t that fake citations look wrong. It’s that they look too right. Too convenient. Too perfectly aligned with whatever point the AI is making“
open.substack.com/pub/cardcata...
open.substack.com/pub/cardcata...
Reposted by Heather Froehlich
We wrote a thing -- showing you don't need LLMs to model language production dynamics like the tendency for speakers to reduce predictable words. All you have to do is better model how speech rate varies depending on where a word is and how long the utterance is. arxiv.org/abs/2512.23659
Less is more: Probabilistic reduction is best explained by small-scale predictability measures
The primary research questions of this paper center on defining the amount of context that is necessary and/or appropriate when investigating the relationship between language model probabilities and ...
arxiv.org
December 30, 2025 at 1:48 PM
We wrote a thing -- showing you don't need LLMs to model language production dynamics like the tendency for speakers to reduce predictable words. All you have to do is better model how speech rate varies depending on where a word is and how long the utterance is. arxiv.org/abs/2512.23659
Reposted by Heather Froehlich
“Dozens of NSW schools teach Aboriginal languages but this is the state’s only bilingual school.
GGFS is open only to Indigenous students from kindergarten to year 8. They attend at least one language class every day and have weekly lessons on country.”
#LangSky
www.theguardian.com/news/2025/de...
GGFS is open only to Indigenous students from kindergarten to year 8. They attend at least one language class every day and have weekly lessons on country.”
#LangSky
www.theguardian.com/news/2025/de...
The Australian school spearheading an Indigenous language revival
At Gumbaynggirr Giingana Freedom School – the first Aboriginal bilingual school in New South Wales – success is judged by an unconventional metric: happiness
www.theguardian.com
December 29, 2025 at 3:08 PM
“Dozens of NSW schools teach Aboriginal languages but this is the state’s only bilingual school.
GGFS is open only to Indigenous students from kindergarten to year 8. They attend at least one language class every day and have weekly lessons on country.”
#LangSky
www.theguardian.com/news/2025/de...
GGFS is open only to Indigenous students from kindergarten to year 8. They attend at least one language class every day and have weekly lessons on country.”
#LangSky
www.theguardian.com/news/2025/de...
Reposted by Heather Froehlich
This is an outstanding essay describing some of the important roles and influences of NIH program officers (from an insider perspective).
Read!
elizabethginexi.substack.com/p/the-quiet-...
Read!
elizabethginexi.substack.com/p/the-quiet-...
The Quiet Power of Program Officers
How invisible decisions shape what science gets done—and whether it matters
elizabethginexi.substack.com
December 28, 2025 at 11:56 AM
This is an outstanding essay describing some of the important roles and influences of NIH program officers (from an insider perspective).
Read!
elizabethginexi.substack.com/p/the-quiet-...
Read!
elizabethginexi.substack.com/p/the-quiet-...
Reposted by Heather Froehlich
This was esp. good on how different kinds of cognitive activities (grammar, inference, judgment) get classed as “next-word” problems in AI; Lit studies as especially attuned to the consequences of incorrect inferences…
"literary-critical inferences cannot be identical, but cannot be an interpretive free-for-all or just statistically approximated hearsay. AI cannot suffer the downstream consequences of faked comprehension+wrongful inferences in its body."
Nan Z Da
newleftreview.org/issues/ii155...
Nan Z Da
newleftreview.org/issues/ii155...
Nan Z. Da, Literary Criticism in the Age of AI, NLR 155, September–October 2025
Critical analyses of AI usually adopt a stance of defensive humanism. Instead, Nan Da interrogates its mode of reasoning. How do LLMs make the step from data to inference—and what does it mean when th...
newleftreview.org
December 27, 2025 at 5:02 PM
This was esp. good on how different kinds of cognitive activities (grammar, inference, judgment) get classed as “next-word” problems in AI; Lit studies as especially attuned to the consequences of incorrect inferences…
Reposted by Heather Froehlich
I wrote a little thing about "Fairytale of New York" and about hope for a better year, but mostly I wanted to share the rendition of the song played at Shane MacGowan's funeral in 2023 because it's so beautiful. Happy Christmas, I love ya baby. dansinker.com/posts/2025-1...
I've got a feeling / This year's for me and you | dansinker.com
dansinker.com
December 24, 2025 at 3:29 PM
I wrote a little thing about "Fairytale of New York" and about hope for a better year, but mostly I wanted to share the rendition of the song played at Shane MacGowan's funeral in 2023 because it's so beautiful. Happy Christmas, I love ya baby. dansinker.com/posts/2025-1...
Reposted by Heather Froehlich
A brilliant article, and one that is going to have a big impact on what I want to be doing next year.
Evaluating LLM “Research Assistants” and Their Risks for Novice Researchers | Critical AI | Duke University Press
read.dukeupress.edu
December 23, 2025 at 10:39 AM
A brilliant article, and one that is going to have a big impact on what I want to be doing next year.
Reposted by Heather Froehlich
"Moby Dick is about whales" + hi from yr 🐳 friend: printed a quote from @heatherfro.bsky.social using the 1568 whale image (bsky.app/profile/woke...) I lasercut into a bamboo printing block (bsky.app/profile/lite...):
December 24, 2025 at 2:30 PM
"Moby Dick is about whales" + hi from yr 🐳 friend: printed a quote from @heatherfro.bsky.social using the 1568 whale image (bsky.app/profile/woke...) I lasercut into a bamboo printing block (bsky.app/profile/lite...):
Reposted by Heather Froehlich
Do you enjoy my periodic rantings on Isaac Newton's alchemical citations? If so, take a look at my shiny new article on our improved #TEI encoding system of Newton's citations & what we can learn from some preliminary analyses of the data!
#BookHistory #histsci (tldr in 🧵)
doi.org/10.14434/tc....
#BookHistory #histsci (tldr in 🧵)
doi.org/10.14434/tc....
A Bibliographical Framework for Citation Analysis of Isaac Newton’s Alchemical Manuscripts: Alexandra E. Wingate
| Textual Cultures
doi.org
December 22, 2025 at 2:49 PM
Do you enjoy my periodic rantings on Isaac Newton's alchemical citations? If so, take a look at my shiny new article on our improved #TEI encoding system of Newton's citations & what we can learn from some preliminary analyses of the data!
#BookHistory #histsci (tldr in 🧵)
doi.org/10.14434/tc....
#BookHistory #histsci (tldr in 🧵)
doi.org/10.14434/tc....
Reposted by Heather Froehlich
AI slop is getting worse. Asked an LLM to sing a Christmas song for me and after one verse about Christmas it descended into increasingly threatening demands for figgy pudding
December 22, 2025 at 6:34 PM
AI slop is getting worse. Asked an LLM to sing a Christmas song for me and after one verse about Christmas it descended into increasingly threatening demands for figgy pudding
Reposted by Heather Froehlich
i am actually working on an article right now about how the vast majority of the world's information - and so many current events - are not at all legible to LLMs, a fact that a startling number of people never seem to think about
Wait until you hear that LLMs can only train on digitized/datafied info.
Most of the FACTS scholars use are in archives/libraries.
Less than 1% of archival colletions worldwide have been digitized.
Also: lots of facts are not even in archives, but in the attics.
Most of the FACTS scholars use are in archives/libraries.
Less than 1% of archival colletions worldwide have been digitized.
Also: lots of facts are not even in archives, but in the attics.
suspect a big reason why many academics and others who work in areas where getting facts RIGHT is key are disinterested in using LLMs for research:
they’ve tried it, they keep noticing major errors in output, and they conclude that having to verify all that doesn’t actually save them time.
they’ve tried it, they keep noticing major errors in output, and they conclude that having to verify all that doesn’t actually save them time.
December 22, 2025 at 4:44 PM
i am actually working on an article right now about how the vast majority of the world's information - and so many current events - are not at all legible to LLMs, a fact that a startling number of people never seem to think about
Reposted by Heather Froehlich
I take it all back we should absolutely let AI run the vending machines
December 18, 2025 at 4:55 PM
I take it all back we should absolutely let AI run the vending machines
I served on a committee with Paul, a wonderful and fierce advocate for the modern university library. I learned tons from him about effective advocacy. A tremendous loss to UCL and all in his orbit. www.leru.org/news/in-memo...
December 18, 2025 at 4:50 PM
I served on a committee with Paul, a wonderful and fierce advocate for the modern university library. I learned tons from him about effective advocacy. A tremendous loss to UCL and all in his orbit. www.leru.org/news/in-memo...
Reposted by Heather Froehlich
A really nice intro to language models, with visualizations and links to other resources.
Learned more about LLM's under the hood from this post than my stupid $30 online course
ngrok.com/blog/prompt-...
@samwho.dev doing @samwho.dev things 🙌
ngrok.com/blog/prompt-...
@samwho.dev doing @samwho.dev things 🙌
Prompt caching: 10x cheaper LLM tokens, but how? | ngrok blog
A far more detailed explanation of prompt caching than anyone asked for.
ngrok.com
December 18, 2025 at 3:32 PM
A really nice intro to language models, with visualizations and links to other resources.
Facing Major Funding Loss? Educopia’s New Workshop Series Offers Tools, Structure, and Support educopia.org/blog/loss-of...
Facing Major Funding Loss? Educopia's New Workshop Series Offers Tools, Structure, and Support — Educopia
Loss of Primary Source of Funding: a four-session workshop designed to help teams move from uncertainty and panic to grounded, values-aligned decision making.
educopia.org
December 17, 2025 at 6:11 PM
Facing Major Funding Loss? Educopia’s New Workshop Series Offers Tools, Structure, and Support educopia.org/blog/loss-of...
i see we are talking about academic hiring again, and its almost time for zoom interviews (if they haven't already started)
so
here's my long form discussion about what to expect if you're on the market
www.dropbox.com/scl/fi/bjcpr...
so
here's my long form discussion about what to expect if you're on the market
www.dropbox.com/scl/fi/bjcpr...
www.dropbox.com
December 17, 2025 at 6:04 PM
i see we are talking about academic hiring again, and its almost time for zoom interviews (if they haven't already started)
so
here's my long form discussion about what to expect if you're on the market
www.dropbox.com/scl/fi/bjcpr...
so
here's my long form discussion about what to expect if you're on the market
www.dropbox.com/scl/fi/bjcpr...
What does one wear in San Francisco in February, asking for a friend who thought 60F was chilly this morning*
*the move from the northeast to the southwest is a whole-body overhaul
*the move from the northeast to the southwest is a whole-body overhaul
December 16, 2025 at 9:30 PM
What does one wear in San Francisco in February, asking for a friend who thought 60F was chilly this morning*
*the move from the northeast to the southwest is a whole-body overhaul
*the move from the northeast to the southwest is a whole-body overhaul
Reposted by Heather Froehlich
The good thing for me is that I accepted long ago that I don’t really fit anywhere and don’t really do anything. And I wouldn’t change that to be a Real Scholar™️ But I see the toll it’s taking on people who do good work and do NOT deserve misdirected anger here or other spaces.
December 16, 2025 at 12:36 AM
The good thing for me is that I accepted long ago that I don’t really fit anywhere and don’t really do anything. And I wouldn’t change that to be a Real Scholar™️ But I see the toll it’s taking on people who do good work and do NOT deserve misdirected anger here or other spaces.
i feel really conflicted about the people who insist that the Humanities needs to be a 100% technology free space. I get you - I am very analog in my daily life, with my piles of paper and pens - but AI is also doing a lot of things you actually like and enjoy having. R's thread is right on
The collateral damage of people upset (often with good reasons) about generative AI is their digital humanities colleagues. It’s outpacing 2010 levels of vitriol, by a lot. The irony is we’re the best allies — we understand how the tech works and can translate between the doom and the hype. +
December 16, 2025 at 4:25 PM
i feel really conflicted about the people who insist that the Humanities needs to be a 100% technology free space. I get you - I am very analog in my daily life, with my piles of paper and pens - but AI is also doing a lot of things you actually like and enjoy having. R's thread is right on
Reposted by Heather Froehlich
YES! So many structures and words that have been classed as AI giveaways are what speakers of English as a second language have been taught and encouraged to do.
I try to avoid linking to Substacks but this essay is FASCINATING
I'm Kenyan. I Don't Write Like ChatGPT. ChatGPT Writes Like Me.
I'm calm. I'm calm. I promise.
marcusolang.substack.com
December 16, 2025 at 10:44 AM
YES! So many structures and words that have been classed as AI giveaways are what speakers of English as a second language have been taught and encouraged to do.
Reposted by Heather Froehlich
I always learn so much from @mariaa.bsky.social. She's really helped me think more deeply about AI models and privacy, which sometimes doesn't get enough attention.
It's like social media all over again, but now the data is even more sensitive. This clip is worth a listen!
It's like social media all over again, but now the data is even more sensitive. This clip is worth a listen!
We don't actually trust AI.
We trust the companies behind it.
As Maria Antoniak notes, every "private" chat flows through corporate systems with long histories of data misuse. If we care about AI ethics, we need to name power, not anthropomorphize models.
We trust the companies behind it.
As Maria Antoniak notes, every "private" chat flows through corporate systems with long histories of data misuse. If we care about AI ethics, we need to name power, not anthropomorphize models.
December 15, 2025 at 8:28 PM
I always learn so much from @mariaa.bsky.social. She's really helped me think more deeply about AI models and privacy, which sometimes doesn't get enough attention.
It's like social media all over again, but now the data is even more sensitive. This clip is worth a listen!
It's like social media all over again, but now the data is even more sensitive. This clip is worth a listen!
Reposted by Heather Froehlich
did you know: the American Dialect Society is the oldest "word of the year" event in the States, and the words are chosen by several hundred linguists who get raucous in a big meeting room in January. you can submit nominations for us to consider!
great activity for classrooms!
great activity for classrooms!
Submit 2025 Words-of-the-Year Nominations
The American Dialect Society’s popular 2025 words-of-the-year vote will be held during the society’s 2026 annual meeting in New Orleans, Louisiana. Nominations will be taken Thursday, January 8, 2026…
americandialect.org
December 15, 2025 at 3:46 PM
did you know: the American Dialect Society is the oldest "word of the year" event in the States, and the words are chosen by several hundred linguists who get raucous in a big meeting room in January. you can submit nominations for us to consider!
great activity for classrooms!
great activity for classrooms!