@sachnishal
HCI PhD @ Northwestern. I design human-centered AI tool to support science communication + study how AI tools might reshape people’s values, practices, relationships. she/her. nishalsach.github.io
📣New paper !
Designing collaborative generative AI is difficult precisely because most generative AI systems are built around efficiency, speed, and individual productivity.
In our conditionally accepted #CHI26 paper, we ask: what happens when those assumptions break down?
I get again find myself in the situation of having learned things via peer reviewing papers, but double-blind doesn't let me figure out who to thank (yet!).
🎉 Thrilled to share that our paper "Reporting and Reviewing LLM-Integrated Systems in HCI: Challenges and Considerations" has been conditionally accepted to #CHI2026!
A thread 🧵
Screenshot of an excerpt from a Becca Rothfeld article in the New Yorker: "There are still plenty of places to read about literature, many of them excellent. There are older and more established outlets, like the London Review of Books and The New York Review of Books; cult favorites, like Bookforum; and irreverent newcomers, like The Drift and The Point, the latter of which I edit. These magazines are delightful and, in their own way, consistently surprising; I love reading them, and I have loved writing for them. But they are produced for an audience that already knows it cares about literature. The books section of a newspaper plays an altogether different role. It does not cater to aficionados; it seeks new recruits."
"There are still plenty of places to read about literature.…But they are produced for an audience that already knows it cares....The books section of a newspaper plays an altogether different role. It does not cater to aficionados; it seeks new recruits."
www.newyorker.com/books/page-t...
nice concretizing of the core idea in the history of information overload: overload isn’t a matter of quantity alone, it requires a certain kind of subject to feel overwhelmed. (the early modern scholar who wants to have read every book in the library, the RSS reader who wants inbox zero)
the purpose of a first draft is to exist the purpose of a first draft is to exist the purpose of a first draft is to exist
Title + abstract of the preprint
Excited to present a new preprint with @nkgarg.bsky.social: presenting usage statistics and observational findings from Paper Skygest in the first six months of deployment! 🎉📜
arxiv.org/abs/2601.04253
Excited to announce the publication of our article "AI Unplugged: Exploring Pathways from Physical Simulation to Conceptualization of AI Reasoning Processes" in @acmtoce.bsky.social. We present 4 activities and explore how they support conceptualization of AI reasoning.
dl.acm.org/doi/10.1145/...
Chart showing that 70% of songs are ones almost no one ever listens to (stream count < 1000).
A chart titled "Stream Count Analysis by Popularity" showing a histogram of total streams peaking at popularity scores of 60–70, overlaid with a line graph showing average streams per song spiking exponentially at scores of 90–100. Below, a table lists the top three songs by stream count: "Die With A Smile," "BIRDS OF A FEATHER," and "DtMF."
A matrix grid with the title "Audio Features Correlation Heatmaps" showing histograms of musical traits (Tempo, Energy, Valence, etc.) along the diagonal and 2D density heatmaps showing correlations between paired features in the intersecting cells.
> We archived around 86M music files, representing around 99.6% of listens. It’s a little under 300TB. This is the largest music metadata database that is publicly available.
Many interesting 📊 charts on this page that can only be made by having this scale of data.
annas-archive.li/blog/backing...
another example of AI as tracer dye, exposing an already existing problem (and, of course, significantly worsening it)
that doesn’t solve the issue of students, journalists, etc. finding and citing fake academic papers, of course—but truly, it’s not an LLM’s fault when professional scholars do this—they are just efficient catalysts for a chemical reaction already underway
Doing this in my head after a big meal and too much beer, so I could be off, but I think Voyager 1 will reach one (1) light day from Earth in October or November of next year, and one light day from the Sun in January 2027. 🧪 🔭
thank you! this is really cool and would be nice to learn about if you're planning to expand the scope of taxonomy(e.g., other scientists, grad students)
Training LLMs end to end is hard. But way more people should, and will, be doing it in the future.
The @hf.co Research team is excited to share their new e-book that covers the full pipeline:
· pre-training,
· post-training,
· infra.
200+ pages of what worked and what didn’t. ⤵️
Halloween blog post: Italian Giallo Horror Films
thisismattmiller.com/post/giallo/
- Using vision language model to analyze a 70 film corpus (🧟) / 80,000 frames
- Build and plot “trope clusters” across movies
Probably the longest eye acting supercut you've seen: youtu.be/cGrmkOwut6k
A rectangular Northwestern University flyer titled “Call for Study Participants.” The top banner is pink with bold black text reading “CALL FOR STUDY PARTICIPANTS.” Below it, highlighted in yellow, is the question: “Are you a journalist writing about technology and/or computing?” The flyer explains that researchers at Northwestern University are seeking professional journalists who cover technology or related topics, write news in English, and are based in the U.S. to participate in an online study. The study involves using a tool for two weeks, providing feedback, and receiving $100 compensation (IRB Study #STU00224608). It includes a yellow box that says “Fill out our eligibility form if you’re interested!” with the link https://tinyurl.com/newscompass and contact emails: Nick Diakopoulos (nad@northwestern.edu) and Sachita Nishal (nishal@u.northwestern.edu). The Northwestern University seal appears in the top-right corner.
📣 Attention science + tech journalists!
Trying to keep up with the flood of research papers that come out every day? Attempting to track what research has already been covered by others, and what may benefit from deeper exploration?
You are invited to participate in our research study!
[1/3]
Please join our 2-week online study to help us evaluate this tool: tinyurl.com/newscompass
Contact @ndiakopoulos.bsky.social or me on here or via email with any questions you may have!
#ScienceJournalism #TechJournalism #Journalism #NewsDiscovery #ScienceWriting
[3/3]
We're researchers at Northwestern University
@nuschoolofcomm.bsky.social developing a tool to help journalists navigate the landscape of tech/computing research papers + their news and social media coverage, to discover newsworthy stories and fresh angles (IRB Study # STU00224608)!
[2/3]
A rectangular Northwestern University flyer titled “Call for Study Participants.” The top banner is pink with bold black text reading “CALL FOR STUDY PARTICIPANTS.” Below it, highlighted in yellow, is the question: “Are you a journalist writing about technology and/or computing?” The flyer explains that researchers at Northwestern University are seeking professional journalists who cover technology or related topics, write news in English, and are based in the U.S. to participate in an online study. The study involves using a tool for two weeks, providing feedback, and receiving $100 compensation (IRB Study #STU00224608). It includes a yellow box that says “Fill out our eligibility form if you’re interested!” with the link https://tinyurl.com/newscompass and contact emails: Nick Diakopoulos (nad@northwestern.edu) and Sachita Nishal (nishal@u.northwestern.edu). The Northwestern University seal appears in the top-right corner.
📣 Attention science + tech journalists!
Trying to keep up with the flood of research papers that come out every day? Attempting to track what research has already been covered by others, and what may benefit from deeper exploration?
You are invited to participate in our research study!
[1/3]
One of the coolest video game preservation stories I have ever covered arstechnica.com/gaming/2025/...
“Editing according to traffic is not editing; it’s engineering,”
nymag.com/intelligence...
What are your favorite recent papers on using LMs for annotation (especially in a loop with human annotators), synthetic data for task-specific prediction, active learning, and similar?
Looking for practical methods for settings where human annotations are costly.
A few examples in thread ↴
Check out the camera-ready version of our ACL Findings paper ("Taxonomizing Representational Harms using Speech Act Theory") to learn more!!! arxiv.org/pdf/2504.00928
CONGRATULATIONS!!!🥳
columnar joints in a japanese waterfall. these are rocks carved out when water eroded through the cooled lava.
happy birthday! it is also my birthday! here is a photo of columnar joints in a japanese waterfall that I saw last week--rocks formed when water cut through the cooled lava from a volcano!
1. On Saturday, @mnlee.bsky.social will present our work on designing for agency in LLM-infused writing support tools for journalists at the Tools for Thought workshop.
Lead author @sachnishal.bsky.social will be nearby at her own workshop on news + HCI!
ai-tools-for-thought.github.io/workshop/
sometimes when i'm writing a sentence i really like it feels like i'm playing the piano
Remember this @neuripsconf.bsky.social workshop paper? We spent the past month writing a newer, better, longer version!!! You can find it online here: arxiv.org/abs/2502.00561
Screenshot of Google docs generative AI agent asking me to Rephrase, Shorten, Elaborate, More formal.
Holy moly. I'm trying to write an academic paper, and nearly every application I'm using is not only offering Generative AI as an option for writing, but *pushing it* -- pervading the design to the point where a simple misclick would make my content AI-generated. Here's why that's a problem. 🧵