Kyle Peyton

[Marcin Wroński @mk_wronski]

Czy AI zastąpi naukowców w replikowanie wyników badań naukowych ? Na to pytanie odpowiadamy w artykule, który ukazał się na łamach @PNASNews. Dziś nie, ale skoro już wykrywa blisko 40% błędów, które dostrzegają naukowcy? https://t.co/jZLAGDw7qc

[Kyle Peyton @pylekeyton]

@seanjwestwood Tbh I'd be surprised if 9% of full professors in political science could even run that regression without help

[Kyle Peyton @pylekeyton]

@ben_golub The ranking best measures perceived insider fit to a network. Controlling for stuff and still finding elite program advantage captures additional access to same network. Cant separate value added from self-reinforcing network advantages in a closed system. What am I missing?

[Kyle Peyton @pylekeyton]

@MoreNeighbours @peter_tulip Yes, continuously declining population since ~2014 is why Chicago sits where it does. NYC and SF build almost nothing and have strong demand pressure. Houston and Austin build a lot, which absorbs strong demand and keeps ratios moderate.

[Kyle Peyton @pylekeyton]

@andrewjnorton @peter_tulip I mean, nobody takes an arts degree for income maximization. That requires a complete nonsense model of human behaviour. But an equilibrium w/ a strong Uni system and good government where human capital is optimised has huge downstream value. Australia chooses not to pursue that.

[Ryan Briggs @ryancbriggs]

Economics doesn't look better in the "robustness" paper. Honestly, econ looks worse than PS and psych but the difference is tiny and not worth obsessing over. Experimental work looks better than observational. Read the paper for details: https://t.co/KxQml5xoCD

[Dan Goldstein @dggoldst]

Nature meta-research project puts claims in social-science papers. I'm interested in Econ and Psych so I focused on that:

Econ had about the same rate of "not reproducible" analyses as Psych and a worse rate then Political Science.

https://t.co/UyIfeqt8CA

[Sean Westwood @seanjwestwood]

Nearly half of Americans (46%) report using AI to get news at least occasionally--but most are light users. Only ~14% use it 3+ times a week. Some people use AI as their primary interface to society, even if the models are not really up to the task.

[Kyle Peyton @pylekeyton]

The TL has been gushing BS from pro-AI grifters to famous profs for at least 6 months. Bulk is AI generated nonsense, but mostly harmless!

But the humanities responses have been insane. Pure vitriol/left wing authoritarianism. Foucault is dead guys. What are you doing? Plz stop.

[Ryan Briggs @ryancbriggs]

I have a new paper. We look at ~all stats articles in political science post-2010 & show that 94% have abstracts that claim to reject a null. Only 2% present only null results. This is hard to explain unless the research process has a filter that only lets rejections through.

[Yiqing Xu @xuyiqing]

1/🧵 A major update to our paper: "Scaling Reproducibility" w/ @YangYang_Leo.

We move beyond reanalyzing a single design to (almost) full-paper replication!

Paper: https://t.co/fslLN0zTQO

[Kyle Peyton @pylekeyton]

@akoustov @caltamura3 It was not *that* long ago when a sociology editor literally went after me for explaining in a response memo that a sociology paper I replicated didn't have a replication package. They blamed the ASA, of course. Thread still lives: https://t.co/ffkQpdiMxd

[Kyle Peyton @pylekeyton]

The editors were quite upset, calling this revision "petty" and "misguided"

I was "casting shade" on the original authors for not meeting replication standards that "do not *exist* [in Sociology]"

Indeed, they allege the ASA has "actively prevented their implementation"🤨10/16

[Alexander Kustov @akoustov]

Many have tried to explain discriminatory hiring and other DEI insanities academics brought on themselves with left-wing bias and self-censorship.

I think a bigger factor was simply forgetting about the taxpayer who funds us. Had we asked the public whether what we were doing in 2020-2021 was OK, they'd have said "Hell, no."

[Stefan Schubert @StefanFSchubert]

The English-speaking world is falling in rankings of life satisfaction (The Economist)

[Chagai Weiss @chagai_weiss]

⏳One month left to nominate papers for APSA's Experimental Research Section’s Best Paper Award! We’re excited to read creative experimental work tackling big political questions. Self-nominations welcome (and discussants, please nominate your panelists)!

[Kyle Peyton @pylekeyton]

@MattCowgill @dannolan @cmkusher Can confirm

[Josh Kalla @j_kalla]

Which AI is most persuasive? New working paper w/ Zhongren Chen & Quan Le, we tested 7 frontier LLMs on 19k people. Ranking: (1) Claude; (2, tied) GPT, Gemini (3) Grok. Consistent across issues and bipartisan stances

[Peter Hull @instrumenthull]

Kirill and I have revised this paper with the help of some nice referee comments. Big additions include:

1. Results on the general interpretation of recentered formula IV estimands with heterogeneous effects

2. Conditions for the asymptotic efficiency of a nonparametric split-sample estimator

Enjoy!

[Sol Messing @SolomonMg]

You can just research things. New from @j_a_tucker & me at @BrookingsInst: Coding agents like Claude Code and Codex will likely accelerate research AND undermine institutional structures we built to support it.

[Justin Grimmer @JustinGrimmer]

“Maybe if we are mean enough to people offering an opinion on AI we can avoid any disruption to our industry” a surprisingly large number of academics, apparently.

By the way assertions like “tool X can’t do Y” are difficult to provide evidence for. But if you could show it you’d have a cool paper to write and publish!