Graham Straus (@gpstraus) header
Graham Straus (@gpstraus) profile picture

assistant prof. @tamu political science || @ucla phd ||

Bryan, TX
Joined June 2026
Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)Photo by Graham Straus (@gpstraus)
Graham Straus retweeted
At Anthropic, we use automated red-teaming to strengthen our safeguards before deployment. Today, we're sharing our progress.
7
16
4
166
9,430
Graham Straus retweeted
Different AI models rely on different mixes of news sources to answer the same political questions -- super interesting new research from my collaborators at @freesystemslab! As we build towards political superintelligence, there are going to be a bunch of important questions about how to incorporate news content. Before we figure all that out, we need basic facts about how things are working on the ground today. This research is super helpful for that, and I'm thrilled that @danmthomp and Sho Miyazaki are doing such great work on this topic.
📰 New research: Inside AI’s election news diet What news does AI read to get up to speed on politics? And what kind of news does it reward? Following interesting recent commentary on our Japan election study from @mattyglesias, @JerusalemDemsas and @TheStalwart, we ran a bigger test on the US midterms across 35 Senate races, 18 days, and 15,750 answers. What we found: - Each model has its own favorite news outlet for political news: AP for GPT, Yahoo for Claude, PBS for Gemini, and NYT for Grok. - GPT tends to behave like a civic guide (reading more election office and campaign sites), whereas Claude behaves more like a news summarizer (reading national news, Wikipedia, polls). - Grok draws on X and Gemini on YouTube for political information, far more than the other models we studied. Gemini cites YouTube in 55% of its answers and X shows up in 59% of Grok’s. - A huge share of AI's local news is provided by large nonprofit networks that very few people will have heard of. States Newsroom, a large network of over 40 local newsrooms, appears in 60% of GPT’s answers. - For many models, prediction markets are now a go-to source for polling information. Kalshi is cited more by Claude than any poll aggregator. Full study here: freesystems.substack.com/p/i…
3
8
37
2,154
Graham Straus retweeted
📰 New research: Inside AI’s election news diet What news does AI read to get up to speed on politics? And what kind of news does it reward? Following interesting recent commentary on our Japan election study from @mattyglesias, @JerusalemDemsas and @TheStalwart, we ran a bigger test on the US midterms across 35 Senate races, 18 days, and 15,750 answers. What we found: - Each model has its own favorite news outlet for political news: AP for GPT, Yahoo for Claude, PBS for Gemini, and NYT for Grok. - GPT tends to behave like a civic guide (reading more election office and campaign sites), whereas Claude behaves more like a news summarizer (reading national news, Wikipedia, polls). - Grok draws on X and Gemini on YouTube for political information, far more than the other models we studied. Gemini cites YouTube in 55% of its answers and X shows up in 59% of Grok’s. - A huge share of AI's local news is provided by large nonprofit networks that very few people will have heard of. States Newsroom, a large network of over 40 local newsrooms, appears in 60% of GPT’s answers. - For many models, prediction markets are now a go-to source for polling information. Kalshi is cited more by Claude than any poll aggregator. Full study here: freesystems.substack.com/p/i…
1
12
1
27
4,954
Analysis of tonight’s election results (and why I don’t need to do it anymore) statmodeling.stat.columbia.e…
4
1
36
7,541
How will high gas prices affect the midterms? It's an important question but hard to answer empirically. I've written down a plan to test part of it, and I'll publish my first results on Thursday, when most of the vote is in. 1/🧵 chdausgaard.github.io/fuel-2…
6
25
6
94
20,720
Graham Straus retweeted
We built an AI ads agency and let it run autonomously on Meta for weeks, with the stated goal of netting new subscribers for the @freesystemslab newsletter. Our real goal: learn how AI is persuading people online, and what it might mean for the future of politics and our information environment. The agents did really well! They learned from their mistakes, and by the end of four weeks, they'd all brought down cost-per-subscriber by more than 50% from the first week. They brought us more than 100 new subscribers, made 70+ highly amusing online ads, occasionally ran afoul of Meta's ads policies, and taught us a lot along the way. Political campaigns are already using AI a lot in their advertising; it's only a matter of time before they embrace automated agentic systems. Check out our full piece for a ton of details on the mistakes we saw, how the agents learned, and what we think will happen next. This was one of the most fun pieces we've done, and I think it really makes it clear how AI enables totally new kinds of research. Super excited to get it out there, and huge congrats to @eigenstuffs who led this project. freesystems.substack.com/p/b…
12
28
7
242
29,206
Graham Straus retweeted
Looking for an overleaf alternative? Check out LibrePaper! It's a free writing platform for LaTeX, Typst, Markdown, and HTML. History, comments, track changes, AI, live preview, multi editors, chat, etc. Use the main server or self-host. Still very new, so I need feedback!
please someone vibecode an overleaf replacement
12
62
3
367
21,763
From our FirstView: "Public Service Decline and Support for the Populist Right: Evidence from England’s National Health Service" by Zachary P. Dickson, Sara B. Hobolt @sarahobolt, Catherine De Vries, and Simone Cremaschi @s_cremaschi. buff.ly/LZPaMR7
26
60
4,804
Graham Straus retweeted
A few notes, we give you the chance to build your own forecast going back to 1946. If you think there is another polling error like in 2016, or want to see what the model would say in 1956, try all these possibilities, make your own modeling decisions, and see the numbers change!
New 2026 US House & Senate election forecasts, using all political science knowledge we could systematize, & our statistical methods from “If a Statistical Model Predicts That Common Events Should Occur Only Once in 10,000 Elections, Maybe It’s the Wrong Model” with @DannyCEbanks and @Jonathan_N_Katz. GaryKing.org/electit We make these forecasts available (including all the input data, methods, and procedures), not only for their possible public interest, but because academics learn by making themselves vulnerable to being proven wrong. We will make no changes to our post-primary forecasts (which do not use polls) but will update two other sets on this site incorporating poll results as new ones appear. Comments on the paper or forecasts are welcome!
1
6
2
28
2,123
Graham Straus retweeted
if you still think LLMs suck at math, at this point it really is a skill issue
25
13
2
260
9,361
Graham Straus retweeted
Are prediction markets biased by white men? Republicans? Young people? Dartmouth professor Eric Zitzewitz analyzed 145 years of election prediction market data, spanning multiple exchanges and countries. His findings: • No systematic evidence of political bias across different eras, countries, and offices. • No systematic evidence of pricing bias based on candidates’ race, gender, or age in the recent elections studied. • Traders weren’t always representative of the voting population, yet that didn’t translate into systematic bias in election forecasts. The paper points to a powerful mechanism: people who spot a mispriced outcome have a financial incentive to correct it. Even if some participants trade on personal preferences, informed traders keep prices grounded in reality. Open competition incentivizes research, challenges consensus, and gives the best ideas a chance to surface. People are rewarded for being right, regardless of their background or preferences. The paper does find some pricing biases, including slight overpricing of longshots and underpricing of favorites. Its central finding on election markets remains: no systematic evidence of political or demographic bias. In an increasingly polarized world, these markets give us something valuable: an impartial source of information shaped by people with an incentive to get it right.
28
68
31
336
194,625
Congratulations to @PBucchianeri, @craigvolden, and Alan E. Wiseman, winners of the Best Journal Article Award for "Legislative Effectiveness in the American States," published in the APSR. buff.ly/n9cCOrH
9
1
43
4,277
Graham Straus retweeted
Come work at Tufts. Great students and colleagues!
we’re hiring an assistant professor in American politics. 🐘 apply.interfolio.com/194599
4
21
1,973
Graham Straus retweeted
I'm excited to share 'City Limits and Partisan Sorting', available online @JPIPE_journal! It's cool to see this chapter of my dissertation in print. doi.org/10.1108/JPIPE-05-202… Using geocoded election results, registered voter data, and city shapefiles in the U.S., I show that...
2
6
1
33
1,492
Doing one of these for Political Science but it’s just coming up with a scaling of legislators that is 0.84 correlated with NOMINATE but is not actually NOMINATE
In light of the progress in mathematics, we at Edison Scientific and FutureHouse have assembled a set of Millennium Problems for Biology. They are chosen to be very hard to solve but very easy to validate in a simple laboratory environment. Any of these, if solved, would mark a major advance in biotechnology, and most of them would contribute materially towards curing disease. These are, in some sense, the “last reasonable eval” for AI in biology. This was work primarily by @MichaelaThinks and myself, with contributions from many others. Short descriptions below. The full descriptions of the problems with acceptance criteria are at the Bio Millennium Problems website, linked in the next post. Share more if you have ideas. If they meet our criteria, we’ll add them to our list (with attribution and permission).
4
21
2,029
Graham Straus retweeted
Congratulations to @Spencer_Goidel and @Polling_Station on their new book from Palgrave Macmillan, which draws on Verasight survey data. "The Trump Realignment: How Donald J. Trump Redefined the American Party System" examines how the Republican coalition changed under Donald Trump, including the party's relationship with corporate America. Chapter 6, "Corporate America and the GOP: How the Republican Party Fell Out of Love with Big Business," draws on a June 2024 Verasight poll measuring how Americans view companies described as "woke." Read the book: doi.org/10.1007/978-3-032-15…
3
11
275
Graham Straus retweeted
Here is how suburban Texas moved toward Democrats by income. From 2012 to 2024, the results barely moved, by only 1.2 points, but the coalitions have substantially transformed. One of the driving factors in this was income.
4
13
63
4,066
Graham Straus retweeted
I built DeepSky to see how much of American political geography you can recover from satellite embeddings. Using Google's AlphaEarth, I estimated how 62,145 precincts voted in 2024. A few maps and findings: 🧵
4
6
39
1,841
Graham Straus retweeted
👵 Do pension expansions mobilise older voters? ➡️ @GanslmeierM, M. Klymak & T. Vlandas find that a 36% pension increase in Illinois had no effect on turnout, challenging expectations that material gains increase political participation doi.org/10.1017/psrm.2026.10…
9
1
33
2,255
New paper with @johnmmurph just conditionally accepted at @apsrjournal: Can people guess how majorities of each party in Congress voted on bills? Mostly yes. Can they guess how their own representatives voted on bills, above what is accounted for by party? Mostly no. 🧵
3
18
4
59
8,274