iiinova — where your digital self keeps evolving

Andrew Little @anthlittle.bsky.social · 07/10/2026

As always Pr(anti-zionist|anti-semite) is not the same as Pr(anti-semite|anti-zionist)

030

Andrew Little @anthlittle.bsky.social · 05/10/2026

Write as if you know someone will be running it through Pangram.

040

Andrew Little @anthlittle.bsky.social · 02/10/2026

If beardos, weirdos, and radicals were bad warriors you think winning a war with Iran would be pretty easy?

060

Andrew Little @anthlittle.bsky.social · 30/09/2026

So if you get a review that lights up pangram and is terrible on content my guess is it’s an LLM writeup of bad notes; not an obvious policy violation!

010

Andrew Little @anthlittle.bsky.social · 30/09/2026

Two tricky things here. 1 it’s very hard to tell the difference between someone inputting their own notes to an LLM and “one shot” asking for a review based on a paper upload. 2 one shotting will topically produce a mediocre but not terrible review on content.

120

Andrew Little @anthlittle.bsky.social · 25/09/2026

Studying bias with math (stats, formal, etc) is good at forcing you to always think “biased with respect to/compared to what”?

050

Andrew Little @anthlittle.bsky.social · 30/08/2026

Excuse me while I nervously check if I have sent you an ms recently 😜

030

Andrew Little @anthlittle.bsky.social · 28/08/2026

Also an extremely high “normal American first name”/“normal American last name” ratio

110

Andrew Little @anthlittle.bsky.social · 28/08/2026

You and Przeworski share many traits, apparently including Claude not knowing what you look like

120

Andrew Little @anthlittle.bsky.social · 18/08/2026

Not disagreeing with you but “subscribe to my substack to fix it” is pretty hard to distinguish from parody

3210

Andrew Little @anthlittle.bsky.social · 12/08/2026

Yeah my interests are more academic but knowledge of the form “how likely is a party to win a race with candidate x or y“ is obviously useful too!

100

Andrew Little @anthlittle.bsky.social · 12/08/2026

Prediction markets are messed up in many ways but the common claim here that they don’t do a good job of assigning probabilities to events is just plainly false.

121

Andrew Little @anthlittle.bsky.social · 12/08/2026

All of the serious evidence shows that prediction markets are in fact predictive and human experts are the ones who are bad at prediction. (More work on this soon 😉)

151

Andrew Little @anthlittle.bsky.social · 14/07/2026

UC system may be a counterexample to this rule

000

Andrew Little @anthlittle.bsky.social · 14/06/2026

Total speculation but I would not be surprised if the main use case for slow-working perfectionists is repeatedly asking multiple LLMs to check their papers/code to suggest ways to improve and then do a bunch more work in response, further reducing productivity.

180

Andrew Little @anthlittle.bsky.social · 13/06/2026

Yeah I read your post as speculative, and it’s not a crazy hypothesis. Enough red flags in the study that I give it ~0 weight though.

020

Andrew Little @anthlittle.bsky.social · 13/06/2026

Another high T move: do multiple comparisons, assert you did not do multiple comparisons

150

Andrew Little @anthlittle.bsky.social · 13/06/2026

Maybe you need high T to report a one sided t test on non-preregistered and nonobvious comparison between two groups on a 5 category variable

290

Andrew Little @anthlittle.bsky.social · 11/06/2026

Tangential but reminds me of this www.everywhereist.com/2021/12/bros...

020

Reposted by Andrew Little

Brad LeVeck @bradleveck.bsky.social · 30/05/2026

If you're UC faculty, I'd encourage you to sign this letter in support of the University of California reinstating the SAT in admissions. ucstudentsuccess.org

ucstudentsuccess.org

UCStudentSuccess

1135

Reposted by Andrew Little

Thomas Zeitzoff @zeitzoff.bsky.social · 27/05/2026

Thrilled to share that my book, No Option But Sabotage, was reviewed in @prospectmagazine.co.uk Roberts engages with my book and highlights the dilemma facing eco activists in a world of rising climate pressures and policy failures. www.prospectmagazine.co.uk/culture/7363...

prospectmagazine.co.uk

The great schism in eco-activism

Some want the movement to have broader appeal. Others want a return to the violence of the past. Will one side win out?

193

Andrew Little @anthlittle.bsky.social · 05/05/2026

(By “denigrating conservatives I mean as a broad group, obviously denigrating particular conservatives for their words or behavior is perfectly reasonable)

010

Andrew Little @anthlittle.bsky.social · 05/05/2026

So you agree with the words in Matt’s post just not the fact that he quoted Jake in saying it? FWIW I don‘t think Jake was denigrating conservatives in that post (or any other one I can remember) but it is a common phenomenon worth calling out.

200

Andrew Little @anthlittle.bsky.social · 05/05/2026

As someone on this list I’m a bit confused. Should I have responded by expressing general disdain for conservatives (including my students)?

200

Andrew Little @anthlittle.bsky.social · 01/05/2026

Seems like the natural next step in the gerrymandering wars is targeting specific members of congress and breaking up their district. Why shouldn’t NY target Stefanik?

140

Andrew Little @anthlittle.bsky.social · 01/05/2026

Also considered "(in logit space)" but we are in mixed company

130

Andrew Little @anthlittle.bsky.social · 01/05/2026

Parallel updating (plus floor effects) undefeated

2156

Andrew Little @anthlittle.bsky.social · 30/04/2026

Yeah all valuable work, but back to @benansell.bsky.social‘s initial point present bias/”coding harsher for recent times” is a widespread and well documented phenomenon across a bunch of domains, so I find the resistance to admitting it’s likely at play here too kind of puzzling

110

Andrew Little @anthlittle.bsky.social · 30/04/2026

If you are willing to make such assumptions you get pictures like this which to my (biased?) eye seem best explained by coders getting harsher over time

130

Andrew Little @anthlittle.bsky.social · 30/04/2026

Maybe I’m missing something but it seems like if you aren’t willing to make an assumption about the mapping between observables and expert coding there is a basic observational equivalence problem where coders getting harsher (on average) over time looks the same as things actually getting worse

130

Andrew Little @anthlittle.bsky.social · 28/04/2026

Getting a PhD is a weird career choice if you aren‘t a curious person, and most curios peoples’ interests don’t lie nearly within a subfield of one discipline. (Not a knock on Paul’s answer I just find it odd that people ask this question.)

071

Andrew Little @anthlittle.bsky.social · 23/04/2026

Yeah I too would like clarity on that. Definitely a heavier lift to ask editors to figure out local models but maybe worth it...

010

Andrew Little @anthlittle.bsky.social · 23/04/2026

Not sure about other publishers but Wiley would not allow this under current policies, which state "manuscripts cannot be uploaded to AI tools by editors or reviewers." www.wiley.com/en-gb/publis...

wiley.com

Editor FAQs AI guidelines | Wiley

Wiley’s AI Guidelines for research authors, editors, and reviewers

211

Andrew Little @anthlittle.bsky.social · 23/04/2026

Yeah might say something about their strategy that the content seems reasonably consistent and non-crazy but the headlines go full “provoke the line”

120

Andrew Little @anthlittle.bsky.social · 22/04/2026

The headline is bad but the one about TX isn't actually that crazy and correctly predicts it might not end up helping Rs.

120

Andrew Little @anthlittle.bsky.social · 18/04/2026

”the people on the bus go up and down” does sound kinda unsafe

120

Andrew Little @anthlittle.bsky.social · 18/04/2026

Dunno what’s more surprising this or conservatives getting shrooms

060

Andrew Little @anthlittle.bsky.social · 17/04/2026

Great to see some fun coverage of @carleton.edu, but "BREAKING" may be a somewhat exaggerated way to describe a house where students have been making cookies for 40+ years.

020

Andrew Little @anthlittle.bsky.social · 17/04/2026

On your second Q, you spurred me to poke around and countries with that score do have a nontrivial amount of HOG turnover in elections (~10-20%?) though in my read a lot of those are cases where protest was required to get the incumbent to step down. So still rare all things considered.

010

Andrew Little @anthlittle.bsky.social · 17/04/2026

Fair questions, on the first it's hard to say since I'm not a V-dem coder 😉. I was reacting more specifically to more general writing about the case; don't agree with everything in this article but it has some good examples of what I mean: mikepesca.substack.com/p/orban-elec...

mikepesca.substack.com

What Kind of Autocrat Loses an Election?

How to Beat Trump is How Hungary Beat Orban. Just Vote Him Out.

110

Andrew Little @anthlittle.bsky.social · 15/04/2026

That said I do see the general unwillingness to grapple with the question of whether Orban losing means the experts at least partially misread this case (again, a if not the banner case for many!) makes me even more wary of relying too much on expert surveys to answer these questions.

110

Andrew Little @anthlittle.bsky.social · 15/04/2026

Well you are the first to answer so you get extra credit :) Definitely fair to take a principled stance to look at global averages with different methods and not overreact to individual cases.

110

Andrew Little @anthlittle.bsky.social · 15/04/2026

Saying “Hungarian democracy was fine!” is obviously dumb but what about “Hungarian democracy was not as damaged as the conventional wisdom stated.” Many experts said Orban losing was near impossible. When near impossible things happen beliefs should change!

000

Andrew Little @anthlittle.bsky.social · 14/04/2026

Which should have contributed to the tilting of the playing field! Maybe taking over traditional media matters less than we thought…

110