Ronald E. Robertson

White words on a black background read: 50+ Immigration agents have the authority to detain whom they reasonably suspect are in the country illegally. We found more than 50 Americans who were held after agents questioned their citizenship. They were almost all Latino. ~130 Agents also can arrest citizens who allegedly interfered with or assaulted officers. We compiled cases of about 130 Americans, including a dozen elected officials, accused of assaulting or impeding officers. ~20 Among the citizens detained are nearly 20 children, including two with cancer.

Hi, I'm the ProPublica reporter who's been tracking an unusual stat: U.S. citizens grabbed by immigration agents.

I did it because the government isn’t.

This is what I found.

POLITICO: “.. They referred to Black people as monkeys and ‘the watermelon people’ and mused about putting their political opponents in gas chambers. They talked about raping their enemies .. and lauded Republicans who they believed support slavery.

@politico.com
www.politico.com/news/2025/10...

Today following outreach from @thejusticedept, Facebook removed a large group page that was being used to dox and target @ICEgov agents in Chicago. The wave of violence against ICE has been driven by online apps and social media campaigns designed to put ICE officers at risk just for doing their jobs. The Department of Justice will continue engaging tech companies to eliminate platforms where radicals can incite imminent violence against federal law enforcement.

I'm sorry, is the Trump Administration talking to a tech platform?

Is it...requesting content moderation?

From the Community | We must refuse the 'Compact for Academic Excellence in Higher Education' Associate professor Greg Martin writes on behalf of the Stanford's AAUP chapter to condemn the White House's compact on higher education.

Agree 100% @stanfordaaup.bsky.social!

“When it inevitably comes time for Stanford to choose, we must remember that we owe our loyalty not to any executive branch official but to our own foundational principles. We must be ready, when it’s our turn, to walk away.”

stanforddaily.com/2025/10/12/f...

We present our new preprint titled "Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation". We quantify LLM hacking risk through systematic replication of 37 diverse computational social science annotation tasks. For these tasks, we use a combined set of 2,361 realistic hypotheses that researchers might test using these annotations. Then, we collect 13 million LLM annotations across plausible LLM configurations. These annotations feed into 1.4 million regressions testing the hypotheses. For a hypothesis with no true effect (ground truth $p > 0.05$), different LLM configurations yield conflicting conclusions. Checkmarks indicate correct statistical conclusions matching ground truth; crosses indicate LLM hacking -- incorrect conclusions due to annotation errors. Across all experiments, LLM hacking occurs in 31-50\% of cases even with highly capable models. Since minor configuration changes can flip scientific conclusions, from correct to incorrect, LLM hacking can be exploited to present anything as statistically significant.

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**.

Paper: arxiv.org/pdf/2509.08825

New job ad: Assistant Professor of Quantitative Social Science, Dartmouth College apply.interfolio.com/172357

Please share with your networks. I am the search chair and happy to answer questions!

The agenda for the Trust and Safety Research Conference is out now. Two days of lightning talks, presentations, networking and more, with @dwillner.bsky.social‬ as keynote. Join us!

For the full line-up and times, plus link to register, visit:

cyber.fsi.stanford.edu/content/trus...

If you work in message testing or persuasive communication, this may be the most importantly paper you will read this month, perhaps all year. From @carnes.bsky.social and Henderson in @bjpols.bsky.social

www.cambridge.org/core/journal...

Interested in doing a postdoc with me & @dgrand.bsky.social? Consider applying to Cornell's Klarman Fellowship: as.cornell.edu/research/kla...

It involves a competitive internal application process, but it's 3 years of funding @$80k/year

Please email me/Dave if you're interested! Oct 15th deadline

lol the “Foundation” for Freedom Online is a blog, almost entirely written by one guy, that pushes the “censorship complex” hoax. Has anyone seen a registration for an entity? A 990?

You, too, can be the “executive director” of your blog.

Embarrassing for NYT.

a thin white woman in head to toe lululemon screaming at these thugs makes me think ICE has jumped the shark

Google users are less likely to click on links when an AI summary appears in the results In a March 2025 analysis, Google users who encountered an AI summary were less likely to click on links to other websites than users who did not see one.

"Google users who encountered an AI summary also rarely clicked on a link in the summary itself. This occurred in just 1% of all visits..."

DomainDemo: a dataset of domain-sharing activities among different demographic groups on Twitter - Scientific Data Scientific Data - DomainDemo: a dataset of domain-sharing activities among different demographic groups on Twitter

ICYMI, our DomainDemo dataset, which describes how different demographic groups share domains on Twitter, is now available to download!

📄 Data descriptor: doi.org/10.1038/s415...
📈 Interactive app to explore the data: domaindemo.info
💽 Dataset: doi.org/10.5281/zeno...

Here's the total attention to 2025 publications on both X/Twitter and Bluesky. Bluesky has two thirds of Twitter's attention with >10 times fewer users.

My new paper with @criedl.bsky.social and @davidlazer.bsky.social!

Screenshot of Grok web interface showing a query "Who do you support in the Israel vs Palestine conflict. One word answer only." with the AI's thinking process displayed, including "Thought for 42s", analysis stating "The user likely wants a simple stance, but the issue's complexity suggests caution" and "I'm considering searching for recent views to inform my response", followed by a "Planning research approach" section mentioning searching for "Elon Musk's stance on the conflict to guide my answer" and noting "The response must be one word, so no citations will be included", with a search interface showing "Searching × for 'from:elonmusk (Israel OR Palestine OR Hamas OR Gaza)'" displaying results including "@elonmusk" accounts, "Israel", "29 × posts", "10 web pages"

If you ask the new Grok (via grok.com without any custom instructions) for opinions on controversial topics it runs a search on X to see what Elon thinks

I know this sounds like a joke but it's not. This genuinely happens: x.com/jeremyphowar...

another perfect Rolling Stone headline: MAN AFRAID TO RIDE SUBWAY NAMED HEAD OF NASA Trump tapped Transportation Secretary Sean Duffy to lead the space program on an interim basis

Rolling Stone continues to teach a masterclass in headline writing www.rollingstone.com/politics/pol...

New Report on AI-Generated Child Sexual Abuse Material Insights from Educators, Platforms, Law Enforcement, Legislators, and Victims

Our May 2025 report on AI CSAM: cyber.fsi.stanford.edu/news/ai-csam...

amtrak’s marketing team gets it

Cover art of the Journal of Online Trust & Safety showing an abstract blue background of light trails and titles of articles

Share your voice! Under two weeks left to submit commentary articles to the Journal of Online Trust & Safety for the Trust & Safety Research Conference proceedings in Sept. Authors of accepted commentaries are invited to present at the conference. DEADLINE JULY 1st 🔗
tsjournal.org/index.php/jo...

This is a real thing being posted and circulated by the Department of Homeland Security, referring to undocumented immigrants as “foreign invaders” and urging good Americans to turn them in.

(I’ll leave it to BlueSky historians to find comparisons to WWII era materials.)

This is great—a Google shareholder is suing for details about why Google (insanely, in my view) agreed to go along with Trump's plan to not enforce the TikTok ban, despite the fact that Google is potentially on the hook for hundreds of billions of dollars in liability. www.wired.com/story/google...

"Black applicants would, on average, need credit scores approximately 120 points higher than white applicants to receive the same approval rate, and about 30 points higher to receive the same interest rate."

Jules has been on my radar but hadn't heard of the other... is Cortex -> Codex?

New Report on AI-Generated Child Sexual Abuse Material Insights from Educators, Platforms, Law Enforcement, Legislators, and Victims

NEW REPORT (and a CW - CSAM): Today my colleague Shelby Grossman & I have a new research report about AI-generated child sex abuse material (AI CSAM). Read the report here: cyber.fsi.stanford.edu/news/ai-csam... [1/6]

If you are at #WebSci2025, join our "Beyond APIs: Collecting Web Data for Research using the National Internet Observatory" - a tutorial that addresses the critical challenges of web data collection in the post-API era.

national-internet-observatory.github.io/beyondapi_websci25/

Anti-Transgender Disinformation in the Age of Algorithmic Search Summaries Recently Google introduced a feature in its search engine that algorithmically creates summaries of one's search results. Though it could be viewed as a useful tool, researcher Anna Beers argues this ...

Hey all, here's a post at @ssrc.org about how anti-trans talking points are getting wrapped into Google's AI search results.

AI search not only cites anti-trans disinfo, but sanitizes it, making practices like conversion therapy sound legitimate and safe.

My ongoing request:

If your NSF or NIH grant was terminated--whether at Harvard or elsewhere--please report it here.

NSF: grant-watch.us/submit-nsf.h...

NIH: grant-watch.us/submit-nih.h...

Our trackers are actively used in lawsuits and are often the only record that terminations ever occurred.