Matt Grossmann
banner
mattgrossmann.bsky.social
Matt Grossmann
@mattgrossmann.bsky.social

Michigan State political scientist & IPPSR Director; Hooked bookstore/cafe Co-owner; Science of Politics Podcast; New book: Polarized by Degrees

Political science 59%
Business 20%

Niskanen also released their website redesign today:
www.niskanencenter.org
Niskanen Center - Niskanen Center
Niskanen Center works to advance U.S. public policy through engagement on issues like immigration, climate, and social policy.
www.niskanencenter.org

My new personal website is ready:
mattg.org

Which groups win policy under each party?

Whites under Republicans, minorities under Democrats
New #ScienceOfPolitics podcast/transcript with Agustin Markarian
www.niskanencenter.org/which-groups...
Which groups win policy under each party? - Niskanen Center
Agustin Markarian finds that different groups see their policy preferences better represented depending on which party is in power.
www.niskanencenter.org
Correct link here dynamicdemocracy.us
I'm pleased to report that my forthcoming book is now available for pre-sale. What's more, if you become a paid subscriber to Tusk, I'll send you a complimentary copy of the book!
More about the book below (1/4):
smotus.substack.com/p/my-new-boo...

Reposted by Matt Grossmann

#AI may indeed exacerbate our bad academic habits:

#LLM science summaries were nearly five times more likely than human-authored science summaries to contain broad generalizations (95% CI [3.06, 7.70], p < 0.001).

And newer models over-generalized more; not less!

doi.org/10.1098/rsos...

#philSci

Because AI models are trained on current research practices, they may exacerbate our tendencies like overselling results from limited data, overinterpreting statistical significance, & quickly claiming causality, especially because they are even better at generating interpretations of any results

We can progress across-the-board if we attend to these trade-offs, but that may mean lowering relative standards for generalizability & causal inference for hard ?s we still want to advance. We already do this implicitly, but may need to formalize these expectations.

AI-enhanced research acceleration is likely to supercharge these trade-offs, helping exhaust what can be learned from current mass public data, geographic relationships, & text & voting data, but diverting from less-tractable but key historical, elite-level, & international ?s.

Social science has long faced trade-offs between data availability (& causal answerability) and relative question importance. For example, survey experiments on convenience samples have grown even in international relations, where they cannot advance key ?s but may inflate others

Reposted by Matt Grossmann

Reposted by Matt Grossmann

Cowburn & Conroy test their novel theory of asymmetric partisan opportunity for women in primary elections. They are less likely to win highly competitive Democratic primaries. In Republican primaries, women are least likely to win in more conservative districts.

Full article: shorturl.at/lVJVD

Reposted by Michael McFaul, Timothy Snyder, Mark Galeotti , and 472 more

Reposted by Matt Grossmann

Reposted by Matt Grossmann

Reposted by Christopher D. Manning, Alberto Alemanno, Torben Pedersen , and 477 more

Reposted by Michael McFaul, Timothy Snyder, Mark Galeotti , and 470 more

Reposted by Christopher D. Manning, Alberto Alemanno, Ian Goodfellow , and 474 more

Data center construction is facing much more political opposition. If NIMBYs win, what does that imply for the AI "bubble"? Is it inflated by announcements of financing & builds that won't happen or will slow? Are companies over-announcing, anticipating delays & cancelations?