Adam G. Hughes

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.

@skaushik100 @RealSeattleNice What’s particularly wild to me is that I learned more about the what explains the lack of good governance in this domain on a podcast than I did in YEARS of reading @seattletimes

If you’re at all concerned with public safety in Seattle, you should listen closely to the new @RealSeattleNice conversation with Scott Lindsay, former mayoral public safety advisor and (until recently) Deputy City Attorney under Ann Davision. The senior folks in the Wilson administration would be well advised to pay attention to Scott’s assessment too. Whether you fully agree with them or not, Scott’s informed takes on what’s working and what’s not to address problem an criminal behavior on our streets reminds me how shallow and polarized our civic conversation continues to be about consequential - and complicated - public safety issues, how much it’s been dominated by bumper sticker bromides from the left and the right, and how the local media needs to do a much better job of explaining the underlying realities of the public safety policy landscape in the city.

@mattyglesias @dieworkwear Alden DC is an amazing store. Highly recommend the color 8 plain toe blucher with commando sole

Of these four pairs of shoes, which do you think is the lowest quality? Make a choice before opening this thread. Then I'll tell you something about shoe quality. 🧵

@stevemur Note that the study used synthetic controls, not simulation/synthetic data. This just means pooling other empirical data to create a rough estimate of the counterfactual Oregon/Washington

Note: This study does use modeling and synthetic variables, something I've criticized before and do so again here. "Using 23 NIBRS compliant control states, we estimate a series of canonical two-way fixed effects difference-in-differences models." There's no substitute for empirical observation. Here, the empirical results seem to align with what the synthetic modeling predicted.

A new study concludes that decriminalization of hard drugs resulted in an increase in crime in Oregon and Washington. "We also found evidence that these state-level increases in crime rates were largely driven by Portland and Seattle, indicating a concentration of harm within large metropolises." crimrxiv.com/pub/heej971d/r…

that which withers in the age of mechanical reproduction is the aura of the work of art?

We just release a new analysis the highlights the potential of opt-in surveys -- those that allow people to volunteer to participate -- to provide misleading results, especially for young adults and Hispanics pewrsr.ch/49BMhqF

70mm is back in the Emerald City

New foundational research on the effectiveness of digital advertising: faculty.haas.berkeley.edu/stadelis/Learn… Proud of my colleagues who contributed to this work!

📢🚨 New paper @PatrickYWu, @j_a_tucker, @Jonathan_Nagler & me 🚨 1. We show vast knowledge embedded in LLMs (#ChatGPT) can be used for latent measurement (ideology). 2. Correlated but distinct from existing methods. 3. Extensible to new probs. twitter.com/PatrickYWu/sta…

As in many of these campus conflict stories, the villains here are the college administrators, who seem completely incompetent and maybe like actual legal liabilities nytimes.com/2023/01/08/us/…

Molly Ringwald's memoir of working with Godard on King Lear, this week in @NewYorker, is essential reading—it's cinematic in itself, with a view of an unmade film that, while reading about it, I felt like I was seeing: newyorker.com/magazine/2022/…

Our paper on community gifting groups on Facebook is out! We used de-identified data from Facebook Groups to study local gift-giving communities, specifically buy nothing (BN) groups. These communities allow people to share, reduce waste, and connect to their local community.

New working paper! "Estimating the Value of Offsite Data to Advertisers on Meta," with Nils Wernerfelt, @tuchmanna & Robert Moakler We ask, "how much would it hurt advertising efficiency if Meta couldn't use 3rd party data for ad targeting?" papers.ssrn.com/sol3/papers.cf…

Wow - awesome update/expansion of previous Pew study on the topic of political tweeters. Great to see @pewresearch continuing to work through difficult measurement problems in a rigorous way!

Can’t complain about this desk relocation

We also asked open-ended questions to capture how people defined bullying and harassment and did some simple NLP to examine overlap

One part of the study examined whether people viewed a large number of vignettes as just bullying, just harassment, both, or neither. Both terms were often applied.

Really excited to share some new public opinion research my team conducted in response to the @OversightBoard: research.facebook.com/file/993733407…