iiinova — where your digital self keeps evolving

Jon Green @jongreen.bsky.social · 28/09/2026

given the player and their history this makes it sound less like that’s all the information they have and more like they don’t believe him

Porzingis to miss start of Warriors camp with ‘health issue’

000

Jon Green @jongreen.bsky.social · 16/09/2026

yeah that's the problem www.nytimes.com/2026/09/16/o...

What has made the University of California exceptional is its commitment to increase access without sacrificing excellence. In recent years, California has proudly defied a national trend: On average, about 40 percent of its students are the first in their families to attend a four-year college. More than half qualify for free tuition. The percentages of underrepresented students have crept up, through a variety of initiatives. The California Dream Act gave undocumented students access to financial aid, as well as in-state tuition. Students on University of California campuses may struggle with basic math, but they are graduating at record rates.

130

Jon Green @jongreen.bsky.social · 02/09/2026

lol. lmao. datacolada.org/139

We expected this to be straightforward, but simple intuition and trial-and-error wasn’t giving us the answer. And so we resorted to an old-school AI method: linear regression. We set the dataset’s final grade as the dependent variable and the five component grades as the predictors. The coefficients tell us the average effect of each component on the final grade. Because final grades should be perfectly predicted by the grade components, we should get an R2 of 100%. Instead we got an R2 of 56.5%.

Huh.

131

Jon Green @jongreen.bsky.social · 01/09/2026

PSA your students don’t recognize this meme anymore

4261

Jon Green @jongreen.bsky.social · 18/08/2026

this seems like more work than just editing the shitty AI-generated draft

I then give that material to AI and ask it to produce a draft. The first version is often polished. Sometimes it is much too polished. It can sound confident while completely missing what I meant.

That is where the real work starts. I tell it what is wrong. I ask it to rewrite sections, restore details it has removed, add experiences it could not know, challenge the argument and reject neat conclusions that do not reflect the world as I have experienced it. This can take hours, days, weeks or, in some cases, months. I keep returning to the piece until I can recognise myself in it and until I am prepared to stand behind every claim.

120

Jon Green @jongreen.bsky.social · 10/08/2026

underrated descriptive finding here that Republican faculty aren't REPUBLICAN-Republican on average (I prefer the CFscore based estimates to the party registration based estimates for this particular RQ, though they seem to point in the same direction)

figure A1 from the paper, distribution of CFscores (a donation-based measure of partisanship) by party registration. registered republican faculty have a bimodal distribution, with many CFscores left of 0 (i.e. leaning Democratic)

1120

Jon Green @jongreen.bsky.social · 29/07/2026

"what is public opinion?" - the greatest thread in the history of forums, locked by a moderator after 12,239 pages of heated debate,

jessica tarlov: Your average Republican is pro-IVF, skeptical of six week heartbeat bills, against drporting productive, law-abiding immigrants, and for gay marriage

andrew lawrence (QT): your average republican thinks theres kitty litter boxes in classrooms and immigrants are eating your pets. your average republican cares more about a Black actress playing the little mermaid

060

Jon Green @jongreen.bsky.social · 17/07/2026

lots going on here, but I love that people who said it was 50/50 whether they'd vote turned out at almost exactly 50% bluelabs.com/dynamic-elec...

Actual turnout by self-reported prospective turnout via BlueLabs. Respondents who said they would definitely vote or had already voted have near-100% actual turnout. Respondents who say they will probably vote have ~75% turnout. Respondents who say they're 50/50 to vote or will probably not vote have ~50% turnout. Respondents who say they will definitely not vote have ~25% turnout. Respondents who say they don't know if they'll vote have ~70% turnout.

050

Jon Green @jongreen.bsky.social · 16/07/2026

late to this good post. it's a weird industry! the points about lack of accountability and lack of risk taking are related -- unless something's radically changed in the last ten years, the market rewards binary win/loss and not vote share relative to expectations www.jstor.org/stable/43862...

Martin and Peskowitz (2015): Parties and Electoral Performance in the Market for Political Consultants

Abstract: We investigate whether the hiring relationships of candidates and political con- sulting firms better resembles the predictions of the "adversarial" or "allied" models of consultant-party interaction. We find that the highest-quality consultants are not allocated to the most competitive races, consultant-candidate relationships persist even as candidates' electoral prospects change, and firms who work for challengers face a higher risk of market exit than firms working for incumbents. The market focuses entirely on win-loss records and ignores the information on consultant performance available in candidates' vote shares. These findings depict a market driven by individual candidate, rather than aggregate party,

2136

Jon Green @jongreen.bsky.social · 09/07/2026

saying you're dropping out but waiting until the absolute last moment possible to file the necessary paperwork is very consistent with this take! www.joshbarro.com/p/low-consci...

0151

Jon Green @jongreen.bsky.social · 04/07/2026

blessing your timeline with conjoined twin strawberries

031

Jon Green @jongreen.bsky.social · 30/06/2026

McCarty (2004), "The Appointments Dilemma" onlinelibrary.wiley.com/doi/full/10....

As I will demonstrate, a key issue in determining whether or not the separation of powers generates inefficient policies is the extent to which the executive can commit to not changing the agency's ideal point after the budget has been allocated but before policy has been set by the agency. If the executive cannot make such a commitment, the legislature will have to take account of potential executive opportunism by further restricting the agency's discretion by lowering its budget proposal. Alternatively, if the executive could commit to a moderate agency, the legislature would pass a more generous budget which might make the executive better off and ameliorate the appointments dilemma.

Obviously, a key factor which determines the executive's ability to commit to a particular agency ideal point are the constitutional and legal provisions concerning the executive's ability to remove agency officials. If the executive lacks removal power (such as the case of independent commissioners in the United States), he will be unable to manipulate the agency's ideal point after budgetary decisions are made. However, if the agency head serves at the executive's will, ex post manipulation will be easier.

000

Jon Green @jongreen.bsky.social · 21/05/2026

when we get the autopsy report on this autopsy report

We apologise again for the fault in the subtitles. Those responsible for sacking the people who have just been sacked, have been sacked.

0121

Jon Green @jongreen.bsky.social · 21/05/2026

in fairness, if the guy I commissioned to write my "why did we lose this presidential election?" report gets stuff like "who was president when this happened?" wrong I'd distance myself from it too

Nowhere was this more apparent than in the 2022 Georgia Senate election, where Republicans supported Herschel Walker to take on incumbent Democratic Senator Raphael Warnock. Despite a series of false and unverifiable claims by Walker, as well as the near-complete absence of a comprehensive policy agenda, endorsements by Donald Trump and Mitch McConnell were enough to take the senate race to a runoff when neither candidate was able to claim 50% of the vote in the first round.

This was a blatant attempt by the Republican power base to take advantage of name recognition and tough economic conditions to push through an unqualified candidate whose job would have been little more than rubber stamping the president’s agenda[Note: in 2022, Joe Biden was president.]. Walker’s inability to muster even a shred of credibility precluded him from winning the election in one of the most polarized states in the nation.

190

Jon Green @jongreen.bsky.social · 21/05/2026

This isn't exactly the problem described in Hollow Parties, but it's not *not* the problem described in Hollow Parties. Why aren't you doing this in-house???

Disclaimer: This document reflects the views of the author, not the DNC. The DNC was notprovided with the underlying sourcing, interviews, or supporting data for many of theassertions contained herein and therefore cannot independently verify the claims presented.

2172

Jon Green @jongreen.bsky.social · 21/05/2026

what a mess www.cnn.com/2026/05/21/p...

“When I was elected DNC chair, I commissioned an after action review of the 2024 election that I wanted to be honest and transparent, and with actionable and specific takeaways for the future of the Democratic Party. When I received the report late last year, it wasn’t ready for primetime — not even close — and because no source material was provided, it would have meant starting over. I could not in good faith put the DNC’s stamp of approval on the report that was produced.”

“After last November’s massive Democratic wins, I didn’t want to create a distraction, but by not putting the report out, I ended up creating an even bigger distraction. For that, I sincerely apologize. For full transparency, I am releasing the report as we received it, in its entirety, unedited and unabridged. It does not meet my standards, and it won’t meet your standards, but I am doing this because people need to be able to trust the Democratic Party and trust our word.”

190

Jon Green @jongreen.bsky.social · 20/05/2026

real data visualization heads will need to dig into the appendix

Figure H.1: Co-listening behavior. Each line connects shows (nested within intermediate level show blocks) that respondents listed together; thicker lines represent more frequent co-listening behavior. Block color is listener-weighted show median among shows with left/right estimates;
line color averages between show-level left/right estimates (gray where either co-listed show had insufficient political episodes to be included in left/right estimates).

Figure H.2: Show-pair political discussion frequency (left) and left/right distance (right), sorted by pair averages.

000

Jon Green @jongreen.bsky.social · 20/05/2026

but if you first infer metadata for a wide range of shows, you can aggregate across similar shows (i.e. even if we can't characterize the audience for show X, we can say "shows like X have audiences like Y") and get more precise estimates of what people are actually consuming

Figure 5: Correlates of block-level listening behavior among podcast listeners, for blocks with at least 100 listeners. Point ranges reflect logistic regression coefficients with 95% uncertainty intervals. Black point ranges are statistically significant.

Figure 8: Political knowledge by exposure group. Groups defined by average self-reported show-level political news frequency and average show-level political discussion frequency relative to median (0.17).

100

Jon Green @jongreen.bsky.social · 20/05/2026

new WP (w/ @stecula.bsky.social): we infer show-level metadata for a large sample of popular politics and politics-adjacent podcasts, then merge it with survey data to characterize political information exposure for their audiences link: osf.io/ycujf/files/...

title: A Framework for Studying Podcasts and their Audiences at Scale
authors: Jon Green and Dominik Stecula
abstract: A large and increasing share of media consumption in the United States takes place on podcasts. However, two of the most theoretically interesting features of this development also make it difficult to study empirically: the boundary between political and non-political shows is porous, and the audience is highly fragmented in ways that frustrate standard survey research. We address these challenges by first inferring political discussion frequency, relative left/right
orientation, and network cluster in a sample of roughly 7,600 popular shows that are either categorized as political or share at least one guest with a show categorized as political. We then demonstrate how these show-level metadata can be incorporated into survey research to
dramatically broaden and deepen our understanding of the political information consumption in this developing ecosystem.

Figure 2. Hierarchical stochastic block model results. Colors reflect terminal (lowest-level) block pairs.

Figure 4: Audience profiles for show blocks with at least 100 listeners. All values for show blocks reflect averages across respondents who listen to at least one show in the given block unless otherwise noted. Respondents can list up to three shows.

Figure 7: Left/Right distributions of shows, overall (a) and weighted by audience (b). Both distributions are limited to shows with at least ten episodes classified as likely to feature political discussion.

2114

Jon Green @jongreen.bsky.social · 19/05/2026

new WP (w/ @ntd.bsky.social): we use the case of Christian nationalism to make a broader point about "niche ideologies" in mass opinion research, and why you shouldn't scale them the same way you'd scale more general attitudes Link: osf.io/wbdut/files/...

Title: Niche Ideologies in the Mass Public: The Case of
Christian Nationalism
Authors: Jon Green, Nick Davis
Abstract: It is common in survey research to measure political attitudes by inferring latent dimensions along which all respondents can be scored. We argue that this approach is ill-suited for the study of “niche ideologies” that organize the beliefs of concentrated minorities within specific political domains. We support this argument with the case of Christian nationalism, which is theorized as an ideology but typically measured as an attitude. Across two survey samples, we show that response patterns are better explained by a process in which individuals endorse, reject, or are unfamiliar with clusters of beliefs associated with Christian nationalism than by processes in which responses are dimensional. We then demonstrate that measuring Christian
nationalism as an attitude can distort its relationship with theoretically relevant policy preferences. Our results clarify distinctions between attitudes and ideologies and offer suggestions for how to investigate other niche ideologies in mass opinion research.

Figure 1. Distributions of Posterior Probabilities by DGP. Group-based is strongly preferred (higher probabilities for vast majority of respondents) in both BRS and PRRI samples.

Figure 3: Ideological identification by DGP. The Rejecter group is majority liberal across both samples with few conservatives; the Adherent group is majority conservative across both samples with few liberals.

Figure 6: Comparing group- versus dimensional-based estimates of the relationship between Christian nationalism and abortion attitudes (PRRI 2022). All models control for age, education, race, gender, party identification, and ideological identification.

When Christian nationalism is measured as a single dimension factor, it has positive relationships with both support for overturning Roe v. Wade and for more restrictive abortion policy. But when Christian nationalism is measured via a group-based mixture, being in the Adherent (vs. Innocent) group is only positively associated with more restrictive abortion policy, but not with support for overturning Roe v. Wade.

24926

Jon Green @jongreen.bsky.social · 26/04/2026

men in their 30s when they see another man in their 30s with a copy of Dungeon Crawler Carl

020

Jon Green @jongreen.bsky.social · 03/04/2026

I stg this is the type of headline it seems like people on here want to read

edited headline: FEMA Official Says He Teleported to Waffle House He Was Obviously Just Blackout Drunk. We Are Exactly the Same Type of Progressive as You.

140

Jon Green @jongreen.bsky.social · 03/04/2026

extremely happy to share that "Political Pundits and the Maintenance of Ideological Coalitions" is conditionally accepted at @polbehavior.bsky.social osf.io/8c3fr/files/...

Title: Political Pundits and the Maintenance of Ideological Coalitions
Authors: Allison Wan and Jon Green
Abstract: Political ideologies help groups advance diverse sets of interests under common agendas. However, it is unclear how these groups maintain their norms regarding what it means to be
a member in good standing in a dynamic information environment. Building on theories of “long” political coalitions, we hypothesize that ideological elites’ rhetorical influence on one another will tend to be concentrated and specialized with respect to specific concepts. We
find support for this expectation using an original dataset of over 1,000 prominent political pundits in the United States, in which we infer coalition membership and the diffusion of novel language over a range of specific concepts. While pundits may discuss many concepts,
they tend to “send” language to other pundits in relatively few, resulting in both concentration and specialization of influence within coalitions. These results clarify conceptual distinctions between political ideologies and political philosophies, and demonstrate real-time dynamics of contemporary ideological coalitions.

Figure 1: Stylized expectations. Each node reflects a hypothetical coalition member, colors denote concepts, and arrows indicate members’ influence on each other. For example, member (a) influences members (b), (c), and (e) with respect to one concept, such as immigration, while member (d) influences members (b), (c), and (e) with respect to a different concept, such as health care. Member (e) acts as a broker, both sending and receiving influence across the coalition. Every member is influential, but influence tends to be concentrated (for any given concept, one member accounts for 3/4 of influence) and specialized (4/5 members are influential on only one concept).

Figure 3: Concentration of influence within concepts, observed network compared to permuted networks. Degree, Community, and Concept denote structural features either are or are not preserved in network permutations.

Figure 4: Specialization of influence within concepts, observed network compared to permuted networks. Degree, Community, and Concept denote structural features either are or are not preserved in network permutations.

02611

Jon Green @jongreen.bsky.social · 22/03/2026

this will be the sixth time that we have destroyed woke and we have become exceedingly efficient at it

061

Jon Green @jongreen.bsky.social · 19/03/2026

quite a statement about how improbable this score was that it still only implied a 37% chance of the upset

at 17:30 in the second half siena is winning 47-34 against duke but only has a 37% win probability

200

Jon Green @jongreen.bsky.social · 16/03/2026

folks I regret to inform that Newt Gingrich is in on the meme

Newt Gingrich on X: Instead of fighting over a 21-mile-wide bottleneck forever, we cut a new channel through friendly territory. A dozen thermonuclear detonations and you’ve got a waterway wider than the Panama Canal, deeper than the Suez, and safe from Iranian attacks.

Community note: The open letter ends with:
"The views expressed above do not necessarily represent those of anyone with brain cells"
https://www.chinatalk.media/p/its-time

182

Jon Green @jongreen.bsky.social · 23/02/2026

lot of QTs (rightly) focusing on the "your professor will never find out" FAQ but I found "if we get stuff wrong that's actually still your problem" pretty funny

What if Einstein gets an answer wrong?
You can review every solution before he submits it. Einstein also improves over time as he learns your course material and professor's expectations.

250

Jon Green @jongreen.bsky.social · 18/02/2026

always breaks my students' brains when I have them read the Catalist report this stat comes from, which shows that Black voters contributed more net votes to Biden than Clinton despite Clinton winning the higher % because Biden got higher turnout (and the lower propensity voters are less partisan)

From 2012 to 2024, Catalist showed a steady drop in the Black Democratic presidential share, from 96 percent in 2012 to 93 in 2016 to 89 in 2020 to 85 percent in 2024. The falloff is most acute among Black men, who went from 95 percent Democratic in 2012 to 79 percent in 2024.

0162

Jon Green @jongreen.bsky.social · 18/02/2026

X's algorithm is in fact doing what you think it's doing. www.nature.com/articles/s41...

Article: The political effects of X’s feed algorithm

Abstract: Feed algorithms are widely suspected to influence political attitudes. However, previous evidence from switching off the algorithm on Meta platforms found no political effects1. Here we present results from a 2023 field experiment on Elon Musk’s platform X shedding light on this puzzle. We assigned active US-based users randomly to either an algorithmic or a chronological feed for 7 weeks, measuring political attitudes and online behaviour. Switching from a chronological to an algorithmic feed increased engagement and shifted political opinion towards more conservative positions, particularly regarding policy priorities, perceptions of criminal investigations into Donald Trump and views on the war in Ukraine. In contrast, switching from the algorithmic to the chronological feed had no comparable effects. Neither switching the algorithm on nor switching it off significantly affected affective polarization or self-reported partisanship. To investigate the mechanism, we analysed users’ feed content and behaviour. We found that the algorithm promotes conservative content and demotes posts by traditional media. Exposure to algorithmic content leads users to follow conservative political activist accounts, which they continue to follow even after switching off the algorithm, helping explain the asymmetry in effects. These results suggest that initial exposure to X’s algorithm has persistent effects on users’ current political attitudes and account-following behaviour, even in the absence of a detectable effect on partisanship.

Figure 2. ITT estimates of feed-setting changes on engagement and political attitudes. ITT effect estimates of switching the algorithm on and off (in s.d.). Left, effect of moving from the chronological to the algorithmic feed for users initially on the chronological feed. Right, effect of moving in the opposite direction for users initially on the algorithmic feed. For each outcome, the results of two specifications are reported. Blue, unconditional estimates with robust s.e., controlling only for the initial feed setting and, where applicable, pre-treatment outcome levels. Orange: conditional estimates, controlling for pre-treatment covariates using GRFs; 90% and 95% CIs are reported. Numerical effect sizes and P values correspond to the conditional estimates (all tests are two-sided). The unit of observation is respondent. From top to bottom, sample sizes are n = 4,965, n = 3,337, n = 4,965, n = 4,965, n = 4,596, n = 4,596 and n = 4,850. Tests are described in Methods. Supplementary Information Table 2.16 reports the exact numerical point estimates, s.e., CIs and sample sizes for every specification. All outcomes are standardized. Additional results are presented in Supplementary Information section 2. PCA, first principal component from principal component analysis.

311888729

Jon Green @jongreen.bsky.social · 10/02/2026

in-partisans are much better at parsing in-partisan ideological heterogeneity than out-partisans

silver bulletin chart: how political figures are perceived by the left, right, and center. shows that Republicans rate Democrats from Alexandria Ocasio Cortez through Chuck Schumer as similarly extreme (while Democrats differentiate her, Mamdani, and Sanders from establishment Democrats) while Democrats rate Republicans from Donald Trump through Robert F. Kennedy as similarly extreme -- only rating Mitt Romney, Susan Collins, and George W. Bush as meaningfully more moderate -- while Republicans perceive more variation among these Republicans.

110

Jon Green @jongreen.bsky.social · 31/01/2026

It’s not like DoorDash hides the option to order for pickup and save on the delivery fees www.nytimes.com/2026/01/30/d...

There’s pasta in the pantry and jarred sauce in the refrigerator. So what compels Kiely Reedy to keep having spaghetti with marinara delivered from the restaurant down the street, for several times the cost of cooking the dish herself?

It’s not that the restaurant dish is particularly good, she said. “It’s the instant gratification.”

000

Jon Green @jongreen.bsky.social · 28/01/2026

takes. these are called takes. www.avclub.com/bari-weiss-c...

“We are going to put huge emphasis on scoops,” Weiss continued of her plans. “Not scoops that expire minutes later. But investigative scoops. And, crucially, scoops of ideas. Scoops of explanation. This is where we can soar, and where we’ll be investing.” Per THR, she described CBS News as “the best-capitalized media startup in the world,” but added, “But startups aren’t for everyone. They’re places that move at a rapid speed. They experiment. They try new things. They sometimes create noise and, yes, bad press. If that’s not your bag, that’s ok. It’s a free country and I completely respect if you decide this is just not the right place at the right time for you.”

1243

Jon Green @jongreen.bsky.social · 23/01/2026

ok let me stop you right there...

Zaid Jilani on Twitter: I don’t think Tucker is antisemitic but in his interview with Peter Brimelow he says there has never been a real antisemitic movement in America. Today America is not very antisemitic but in the past, Jews did face widespread discrimination and legal repression.

The first line of the tweet: I don’t think Tucker is antisemitic but in his interview with Peter Brimelow

050

Jon Green @jongreen.bsky.social · 09/12/2025

also just fun to see Perriello running for something again. here’s 19 year old me from when I worked on his (unsuccessful-but-way-above-replacement!) reelect in 2010

Me in 2010 wearing a “Tom Perriello U.S. Congress” shirt

0150