Securing the Vote Against Autonomous AI As Anthropic's Project Glasswing identifies thousands of vulnerabilities, election officials must race to patch the gap before November

New research from @rmichaelalvarez.bsky.social, @caltech.edu: Securing the Vote Against Autonomous AI

Read & subscribe: electionscience.substack.com/p/securing-t...

Come work with us! And please tell your students to come work with us!

#polisky #econsky

Abstract: How do voting laws impact elections? We highlight how laws targeting a specific group of citizens can have weak effects on turnout and vote shares but substantial effects on policy platforms. To parse these effects, we analyze a model of electoral competition with endogenous turnout and targeted voting costs. Each party anticipates the direct effect of raising one side’s voting costs: discouraging targeted citizens from voting. Consequently, both platforms shift towards the untargeted group. These platform adjustments mobilize targeted citizens and demobilize the untargeted, muting the net impact on turnout and vote shares—consistent with scant empirical evidence of these electoral effects. Yet, on policy they also hurt targeted citizens and their aligned party. The targeted group’s size amplifies these effects. Our results address party competition, voter participation, and representation, as well as normative and empirical evaluations of voting laws.

Now conditionally accepted at @thejop.bsky.social. Very proud of this paper. Full paper here: shorturl.at/TjEWx

Didn’t the Simpsons already do this?

After working through second and third derivatives, I am ready to abandon Bayes learning for something with an easier functional form.

Cue the john williams's jurassic park theme song

Currently in FirstView: In “Measuring the Quality of Answers in Political Q&As with Large Language Models,” @rmichaelalvarez.bsky.social and Jacob Morrier develop an approach for measuring the quality of answers in Q&A sessions using data from the Question Period in the Canadian House of Commons.

These always strike me as strong associations and I’ve always wondered how these observed rates vary by field. Eg: effect of family background of becoming an MD/physician or JD/lawyer?

If we write papers using LLM generated data, and those papers are then used to create another version of LLMs and then new datasets for new papers, will the process converge? If not, will it be stationary?

Modeling Violent Terrorism as a Tug of War Between Competing Groups Mike Gibilisco, professor of political science, and his co-author seek an understanding of how competing terrorist groups react to one another's violent attacks.

There are social scientists at #Caltech, @caltech.edu! I think they do pretty good research...at least some of the time.

www.caltech.edu/about/news/m...

This is close to stenography:
“An unnamed source told me all these things [many clearly false] so am repeating them to you”

How is this different than publishing a press release?

Next question: does monterey pdf word count pick this up? I dont know if I can sacrifice 1 of my 5 implications for this.

What Remains of U.S.A.I.D. After DOGE’s Budget Cuts? (Gift Article) The few hundred programs that survived DOGE’s purge reveal the future of foreign aid.

What remains of USAID? Practically nothing ("a shell") and millions will die because of Trump and Musk

Stand aside \bot here comes \swordcontradict

Screenshot of NSF budget cuts to various programs

Social, Behavioral, & Economic Sciences is getting hit with a 67.6% cut, worse than other programs within NSF. Funding for archaeologists comes from the division for Behavioral and Cognitive Sciences, which is facing a 77.3% cut.

It appears every post doc program has been 100% cut.

I am looking at NSF's budget request, in awe at the disinvestment in learning anything. nsf-gov-resources.nsf.gov/files/00-NSF...

A screenshot of the number of people involved in NSF activities being reduced from 330,100 to 90,000.

TO BE PERFECTLY CLEAR, THE PROPOSED NSF BUDGET CUTS NEARLY A QUARTER OF A MILLION PEOPLE FROM NSF ACTIVITIES. FOR INSTANCE, JOBS.

@caltech.edu and I both agree that international relations needs more theoretical models and fitting those models to data.

Great story. And congrats!

And everyone takes it seriously. No beer cans in sight!

To the best of my knowledge, skip day involves seniors planning a series of puzzles and challenges for the rest of the students.

Even after 8 years, I don’t fully understand caltech’s skip day, but this is the first year with an inflatable dragon outside my office.

Happy to see that this is now out. The R package associated with the project, IRTM, can now be found on CRAN. Or can you still download it from my github. We'll continue to update it over time. This is a brief thread about what IRTM can do. 🧵 (1/7)

Ha. I enjoy the wake-up emails of code or proofs at 7am

I missed the first year student in the summaries. Just wow.

This is a neat paper that illustrates the benefits of combining theory with new data. You will find this interesting!

Congrats Brad. Very well deserved. Can’t wait to celebrate at the Wallis pizza party

I am never excited for 9am meeting, but today is dissertation defense day!

Sounds great! Not as great as designing the "whats in your wallet commericials," but still pretty great!