I am so far into the Breiman camp that the idea of looking at coefficients seems so obviously flawed to me. At the very least report held out R^2s (on a data fold that wasn't used to tune hyper parameters).
Very excited to have finally found a reason to actually watch a TED talk!!!
cool post, but re:
This adequately controls for confounding but the data requirements to do this for a large number of traits are onerous, and as I might discuss in a future post, there is a good reason not to examine outcomes like relationship satisfaction trait-by-trait.
Why not do LASSO?
I found docs.google.com/document/d/1... pretty persuasive as far as polling opinion goes.
Isn't this exactly why Deciding To Win did issue polling by asking actual political consultants to frame the issue from the left/from the right? In competitive politics, you get to decide how you talk about an issue, but, the other side will not pick your framing to discuss it.
Teachers and nurses are paid less because historically they were one of the only socially acceptable professions for women, so you could get brilliant women as teachers “for cheap” because they couldn’t go into other careers.
Aside from me personally, EAs co-organized the only real-life protest against the cuts, phone-banked to get Congress to revoke the cuts, and donated between seven and eight figures to help African clinics affected by the moves (for example, CTRL+F "GiveWell" in www.nytimes.com/2025/12/28/h... )
Completely agree with this, and of course it will play well with the mouthbreathers on this site. For more examples: www.vox.com/future-perfe... benthams.substack.com/p/the-upcomi... etc etc
The elites are definitely guiltier because they have more agency but there is plenty of blame to go around. It’s never been easier to be well informed and lots of focus groups showed that voters knew about the coup but thought that inflation was a bigger deal.
Figured I’d re-up this with an excerpt.
Sure, more useful to call it significant cognitive decline than dementia, but, I don't think the label is what's particularly relevant.
The single most obvious piece of evidence is that Biden did significantly less than the bare minimum level of public interaction that a president is expected to do: his staffers were aware of what he looked and sounded like when speaking in public.
There is a cool paper showing that modern LLMs “know” that p-hacking is wrong: however they can be coaxed with some creative prompting: andrewbenjaminhall.com/asher_et_al_...
It's a genuinely extremely difficult problem because VPs need a number they can track to see how their teams are doing and it's very difficult for them to appreciate how the context around a number changes.
For example: if you have an OKR to increase the number of the customers at the top of the funnel, and an objective to increase total sales - by pushing hard to increase the number at the top of the funnel, the fraction that will convert into sales will almost certainly decrease.
I largely agree, the problem, imo, is that you *inevitably* end up running into Goodhart's law. OKR's start off as highly correlated with the objective, but, by treating them as a target, you end up breaking the correlation.
The most effective ad against Harris was Trump reminding people of a stance she took under pressure from the ACLU in 2019!
I don't think I've ever seen gears shift this fast before.
What happens if you persuade a candidate to take on very unpopular positions in the primary as a price of winning progressive support? Even if the candidate tries really hard to distance themselves during the general, the opposition will do everything to remind the public of the stances they took
Bluesky in a nutshell.
Some context: BLS released its jobs numbers today, which were better than expected, and resistance libs would rather stay mad/conspiratorial than listen to actual on-the-ground experts that the data is still reliable.
Way to erase the serious labour of your dog & your cat.
What’s a multiverse good for anyway? Julia M. Rohrer, Jessica Hullman, and Andrew Gelman Multiverse analysis has become a fairly popular approach, as indicated by the present special issue on the matter. Here, we take one step back and ask why one would conduct a multiverse analysis in the first place. We discuss various ways in which a multiverse may be employed – as a tool for reflection and critique, as a persuasive tool, as a serious inferential tool – as well as potential problems that arise depending on the specific purpose. For example, it fails as a persuasive tool when researchers disagree about which variations should be included in the analysis, and it fails as a serious inferential tool when the included analyses do not target a coherent estimand. Then, we take yet another step back and ask what the multiverse discourse has been good for and whether any broader lessons can be drawn. Ultimately, we conclude that the multiverse does remain a valuable tool; however, we urge against taking it too seriously.
New preprint! So, what's a multiverse analysis good for anyway?>
With @jessicahullman.bsky.social and @statmodeling.bsky.social
juliarohrer.com/wp-content/u...
Richard Dawkins and Carl Sagan (especially the latter) are the best at conveying the beauty of a godless non magical universe. Sagan in particular had a very loving tribute to his wife: "In the vastness of space and the immensity of time, it is my joy to share a planet and an epoch with Annie"
An interview by Chotineer with the cinematographer behind Melania, it went exactly as you'd expect.
www.newyorker.com/culture/q-an...
A video of Alex Pretti reading out the final salute of an unnamed veteran he cared for until the end of his life in the ICU, posted to Facebook by his son.
Amazingly balanced and down-to-earth discussion on the future of AGI.
Very good, knowledgable, informative host with great questions driving it, referring back to relevant past comments.
However you feel about the topic, this is worth watching. 1/4
www.youtube.com/watch?v=02YL...
I think probably in buildings, fellowships (it's fully tuition free iirc), paying professors, etc. The people that probably gained the most were the people that were cancelled and were hired there as staff, but, they probably don't make outrageous amounts.
I don't think anyone made a lot of money from this? It was a very stupid ideological project that was obviously doomed from the start but I don't think anyone lined their pockets.
I originally read that as Terence Tao graduating there when he was 6 months old, and was like, dang didn’t realize he was that precocious
Got confirmation she knows.
