LLM safety work often reasons over high-level policies (be helpful & polite), but must tackle on-the-ground cases (unsolicited money advice when stocks are mentioned). This can feel like driving on an unfamiliar road guided by a generic driver’s manual instead of a map. We introduce: Policy Maps 🗺️

There is a lot of talk and effort to figure out how genAI is different (I am also guilty of this!) -- the reality is that genAI is not that different and genAI is not that new either; it was hard to evaluate in the past, and it is still as hard to evaluate now #facct2025

Check out the camera-ready version of our ICML position paper ("Position: Evaluating Generative AI Systems Is a Social Science Measurement Challenge") to learn more!!! arxiv.org/abs/2502.00561

(6/6)

Somewhat oddly, the Trump regime's initial moves to cut science funding went very broad. This is catalyzing solidarity & advocacy around the value of science (see comms featuring cancer cures & tech innovation). But their next move may be to try to drive a wedge b/w "good science" & "bad science".

How's fascism going for you? medium.com/bits-and-beh...

Securing your phone before entering the USA. 🧵
At CDC we travel around the globe to confront health threats and have never worried about returning home. Yet more researchers, including US Citizens, are having their phones searched at Customs.

security.ucop.edu/resources/tr...

Unidentified men grabbing someone off the street and putting her in a car because she wrote an op-Ed. This as flatly authoritarian as anything we’ve seen in this country in a very long time.

1. @alisongopnik.bsky.social, Cosma Shalizi, James Evans and myself have a new piece in Science on "AI" Large Models, pushing back against much of the collective wisdom about what they can and can't do. Official below, unpaywalled at henryfarrell.net/large-ai-mod... . So why this now?

How can we help communities collaboratively shape policies that impact them?

In our #CHI2025 paper, we present PolicyCraft, a system that supports ✨collaborative policy design✨ through case-grounded deliberation.

(🧵/11)

How can we better think and talk about human-like qualities attributed to language technologies like LLMs? In our #CHI2025 paper, we taxonomize how text outputs from cases of user interactions with language technologies can contribute to anthropomorphism. arxiv.org/abs/2502.09870 1/n

"18F works just how Musk + team pretend that they want gov't to work. But when his team found it, they destroyed it. 18F’s work is evidence that gov't works well, which undermines their msg that it doesn’t. 18F’s parent agency, the [GSA], turns a profit as an agency. So it has to be destroyed too."

#Pittsburgh friends, Join me on Friday #standupforscience2025. Our city will be devastated if our universities lose critical funding. Our universities are top 10 employers, the entire city’s economy is dependent on science funding. www.eventbrite.com/e/stand-up-f...

🏛️ How can we ensure that AI enhances, rather than harms, public services?

In our policy memo, published with the Federation of American Scientists (@scientistsorg.bsky.social), we provide recommendations for how the fed gov't could help state & local agencies make better AI procurement decisions.🧵

Joseph Ura: “A president refusing to abide by the law or the Constitution and ignoring court orders to stop his illegitimate actions would be a constitutional crisis like a bank robbery is a cash flow crisis.” www.theatlantic.com/newsletters/...

Friendly reminder to get your free Covid tests before they throw them away.

Covidtests.gov

wapo.st/42VLGQh

for pity’s sake, I am BEGGING university leaders to stop framing this issue in terms of what this will cost the university. people dont care!

EXPLAIN WHAT THIS WILL COST THE PUBLIC: closed hospitals and clinics, skipped treatments, loss of access to experimental drugs, unemployment, recession

"AI ... becomes a tool for replacing politics. The Trump administration frames generative #AI as a remedy to 'government waste.' However, what it seeks to automate is not paperwork but democratic decision-making."

Scientists and journalists need to figure out right quick how to explain to the average person how a massive change in research indirects will impact the medical care they and their children get (eg at the local children’s hospital), the education their children will get, the price of tuition, etc.

What I should have done in the beginning - here's a starter pack of 20-odd experts who journalists or others might want to read/talk to to understand the gutting of the administrative state (as opposed to broad decline of democracy etc). I've focused on academics go.bsky.app/MaC7vnv

First off, it's critical to note:

NSF is CONGRESSIONALLY MANDATED to evaluate proposals via broader impacts, as defined by this law below.

Hence, if the admin wants NSF to deviate, that's called **breaking the law** (important to remind Congress of that)

www.law.cornell.edu/uscode/text/...

2/5

I'm an adventurous eater, but even I draw the line here

The moderation approach with custom labelers etc in Bluesky is pretty clever and worth understanding when you use this platform: bsky.social/about/blog/0....

Bitter Lesson¹ maximalists seem to ignore the distinction between AI capability improvement and AI-based product development. When it comes to the latter, the Bitter Lesson has never been true, and it would be silly to suggest that it is.

¹ www.incompleteideas.net/IncIdeas/Bit...

Alright it's here! Where we push back against the exploratory-confirmatory distinction that's become canonized in metascience and talk about what we think exploratory research (including exploratory experimentation and modeling) means and why our view is incompatible with 'unplanned data analyses'.