Ben Radford
banner
benradford.bsky.social
Ben Radford
@benradford.bsky.social
Computational social science, machine learning, Bayesian stats, political conflict, forecasting, cybersecurity, international relations.

www.benradford.com
The War on Christmas escalates
Fox guest on how to save money during the holidays:

"Remember, adults don't need gifts. Focus on the people in your life who are aged 3 to 18. Grandma doesn't need slippers. If they don't live by you, don't get them a gift."
November 17, 2025 at 10:35 PM
Reposted by Ben Radford
Fun fact. Trump destruction of soybean exports (rightfully) gets lots of attention. But you know what we sell even more of to China? Education.
2024 soybean exports to China: $13B.
Higher Ed: $14B.
Yes, all those internat’l students count as *exports.* Trump is STILL actively destroying that market.
October 27, 2025 at 12:22 PM
Ok it was the first one
Even odds between complaining about unclear gender signage on latrines versus surrendering NATO command to Russia
October 1, 2025 at 4:12 PM
Even odds between complaining about unclear gender signage on latrines versus surrendering NATO command to Russia
September 25, 2025 at 4:33 PM
Reposted by Ben Radford
Im begging the NYT to spend time talking to people who actually teach and work at universities. Especially ones that don't have billion dollar endowments.

My issues are AI use, crumbling infrastructure, vanishing staff, unfunded state mandates, a customer service model of education...
September 25, 2025 at 4:08 PM
Please please please let all the university presidents see Disney and learn how easy it is
September 22, 2025 at 11:06 PM
Reposted by Ben Radford
Very excited that my paper with @katakeith.bsky.social is now out in @polanalysis.bsky.social. We investigate whether LLMs actually follow the instructions/definitions provided in codebooks, propose some diagnostics, and release a new evaluation dataset.
www.cambridge.org/core/journal...
Codebook LLMs: Evaluating LLMs as Measurement Tools for Political Science Concepts | Political Analysis | Cambridge Core
Codebook LLMs: Evaluating LLMs as Measurement Tools for Political Science Concepts
www.cambridge.org
September 19, 2025 at 1:45 PM
Reposted by Ben Radford
what are the most widely read digital newspapers by web traffic across the US and EU? do you know of any current repositories?
September 16, 2025 at 4:38 PM
I swear If I ever find out who's telling my students they need to test models with ViFs...
August 22, 2025 at 3:09 AM
Reposted by Ben Radford
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!
August 21, 2025 at 6:50 PM
Skills & tools that nobody teaches you but that pay for themselves many times over in social sciences:

1. regex
2. Linux
3. Selenium & beautifulsoup
4. Git
5. SQL
August 21, 2025 at 6:13 PM
Learn to love regex. Don't let ChatGPT steal the joy that could be yours
August 15, 2025 at 5:40 PM
There needs to be a public forum for rebutting and shaming bad reviews in political science. Just got rejected on measuring ideology of LLMs bc reviewer "sincerely hopes we're never in a world where LLMs make policy so [doesn't] know why [they] should be interested"

I've got bad news, buddy...
July 31, 2025 at 5:53 PM
Mario Rabbids best series ever
a close up of a mario cartoon character
Alt: a close up of a mario cartoon character
media.tenor.com
July 31, 2025 at 5:24 AM
regex is my favorite (and most useful) skill that nobody ever teaches
July 31, 2025 at 3:51 AM
Uh oh #polisky Microsoft thinks we're more AI-replaceable than basically everyone

arxiv.org/abs/2507.07935
July 28, 2025 at 9:53 PM
Reposted by Ben Radford
41. Computational social science is a bizarre concept, creating by non-social scientists. You don't call all the people who do math in chem, bio, and physics "numbers scientists."
July 28, 2025 at 12:27 AM
Get on my level
July 5, 2025 at 1:00 AM
🔥 take: Nine and California are Blink-182's best albums
July 3, 2025 at 1:24 PM
Statistics is 🔥
May 3, 2025 at 5:26 PM
Reposted by Ben Radford
Thank you for inviting me to collaborate on the ideas and writing of the brief wit your great students!

For anyone curious about the arguments but short on time, their Harvard Cyber clinic blog post condensed our contributions to the Massachusetts court.
April 21, 2025 at 3:45 PM
Reposted by Ben Radford
Meet `azllm` by our team – an easier way to run multiple LLMs:

- Many prompts on one model (batch)
- One prompt on many models (parallel)
- Custom settings per run
- Supports propietary + local models (@ollama)
- Parses responses (structured outputs)

pypi.org/project/azllm/
azllm
A Python package that provides an easier user interface for multiple LLM providers.
pypi.org
April 30, 2025 at 5:29 AM
100%
I think people have a kind of Gell-Mann amnesia towards LLMs — like, they can think "AI is bad at things I am good at and know about, but good at things I can't judge because I'm bad at them"

Tech bros think AI is good at everything, including coding, and that's because they are also bad at coding
April 30, 2025 at 5:22 AM
I don't think we have to concede the premise: there ARE conservative professors. They just also believe in science, which means modern MAGA conservatives don't consider them pure
April 21, 2025 at 3:22 PM
If there's one thing voters just don't have any real world sensitivity to, it's prices
Trump: "Prices are coming down, not going up. Only the fake news says they're going up. The only thing that's even are interest rates and if we had a Fed chairman that understood what he was doing, interest rates would be coming down. He should bring them down."
April 18, 2025 at 10:49 PM