John Holbein

American healthcare is the best. Call urgent care: “Do you do stitches for kids?” Them: “Yes” Wait 3 hours with kid who needs stitches. Provider pokes head in door: “We don’t do that here. Go to the ER.” 2 mins tops. They bill at a rate equivalent to paying someone $6.2 million a year.

I may be crazy, but I'm starting to think that my students put less effort into reading my painstakingly crafted emails than I put in to writing them.

Nice

johnholbein.github.ioThe LDS Apostle Political Ideology ScaleThe First Presidency and Quorum of the Twelve Apostles of The Church of Jesus Christ of Latter-day Saints, on one left-right scale. Place all fifteen yourself — then see what the words say.

Which LDS apostles speak more like conservatives? Which LDS apostles speak more like liberals? Answer three questions, play the game, see the results here: johnholbein.github.io/apostle-scale/

John Holbein@johnholbein1.bsky.social

This weekend is LDS General Conference. So I made a little game: Can you guess the LDS apostles’ political ideology from their speeches? 3 brief ?s → make your guesses → see the results Don’t know the current LDS apostles? No biggie; skip straight to results johnholbein.github.io/apostle-scale/

22 Sept

When our first kid was born 10+ years ago, I thought I could never love someone quite how I loved them. But I was wrong! When our second was born, I thought "Ok, now I can never love anyone like this." Then, our third (final) kid came, & I was proven wrong again. Being proven wrong isn't so bad.

19 Sept

I'm basically Nate Silver now.

John Holbein@johnholbein1.bsky.social

I’m regularly asked how high, or low, youth turnout will be in upcoming elections. So today, on a whim, I pulled the historical numbers and built a simple forecast. At last: my definitive prediction for how many young people will vote in 2026. You’re welcome, America.

18 Sept

There’s a lot of doom & gloom around how AI is affecting higher education. To a certain extent, I get it. AI is making some things harder. But I’ve found that AI has in some ways greatly enhanced my teaching. For one, it’s opened up applied activities that weren’t possible before. One example: 🧵

I'm also having a tremendous amount of fun with data visualizations. Here are some I made for my day on Bureaucracy.

For a discussion on what is in the Federal Register.

I say all of this not because I believe AI has no costs for higher education. It certainly does! But it also has some benefits.

johnholbein.orgThe President's BudgetA classroom federal budget exercise for LPPP 6350: draft the Trump administration's FY2028 budget proposal from real CBO options, then navigate the political consequences.

If you are interested, here are the classroom materials you would need to replicate what I did. Link to the budget simulation: johnholbein.org/teaching-act... Accompanying slides (human generated with AI feedback): github.com/johnholbein/... Replication materials: github.com/johnholbein/...

Could I have done all of this without AI? Sure. But it would have taken weeks (if not more) focusing on this one class period simulation and only this one simulation.

I was able to put it together with AI's help in a few days. I had to do lots of prompt engineering. But Claude, ElevenLabs, and D-ID made it pretty seamless.

They then had to whip on the fence members. To up the ante, I had AI make (watermarked) videos of conflicted members explaining their positions. Students could then try to persuade these members. I had AI build in noise to make persuasion difficult, but possible.

Students then got a whip count and could make choices of what to do next--e.g., revise the bill, proceed anyway, etc.

Once they felt like their budget aligned with what the Trump administration is prioritizing, I had the students pause and reflect on the decisions they had just made. We discussed which choices were the hardest. I made it sound like the simulation was over then. But it wasn't!

The draft budget then leaked to the public. An AI generated attack ad from the opposition then played. The attack ads were tailored to their group's specific budget decisions. It mentioned each group by name.

Recently, I used Claude, ElevenLabs, and D-ID to make a budget simulation for my Politics of Public Policy course. In it, the students make a federal budget that aligns with the Trump administration's budget priorities. They had researched the administration's budget positions before class.

The budget wasn't comprehensive. But they did have to make some key decisions all budgeters need to make. All of the numbers in the simulation were real and were pulled from CBO estimates. Deficit projections updated in real time based on students' choices. AI made all of this super easy.

Googling these days

14 Sept

Generative AI makes replication research even more important. Academia should take this opportunity to boost the incentives for replication. It’d be a huge missed opportunity if we don't.

It’s 2023. 4 people are finishing a political science PhD. Same subfield. Same PhD program rank. Same publication record. 2 are female, 2 are minorities. Their predicted chances of a tenure-track job by fall 2024? Non-minority man: 27% Non-minority woman: 44% Minority man: 47% Minority woman: 45%

www.cambridge.org/core/journal...

cambridge.orgWhere You Go, What You Do, Who You Are: Predictors of Political Science Job-Market Outcomes, 2023–2024 | PS: Political Science & Politics | Cambridge CoreWhere You Go, What You Do, Who You Are: Predictors of Political Science Job-Market Outcomes, 2023–2024 - Volume 59 Issue 4

APSR acceptance! Woot!

It's for this paper. We'll release it soon.

It's for this paper. We'll release it soon.

John Holbein@johnholbein1.bsky.social

APSR acceptance! Woot!

8 Sept

Make academics bartenders again

8 Sept

Source:

5 Sept

But what happens when AI companies finally run out of generic, vaguely profound names for their models?

johnholbein.orgWhere Is Christ in the Canon? 1.22 Million Verses of DataFind Jesus in holy writ. Every verse of 22 scriptural canons, coded for whether it refers to Christ — and the answer moves by a factor of eleven depending on the rule.

5 Sept

Where does Jesus appear in scripture? Data: every verse in 22 canons. 36 Bibles. 11 Qur'ans. LDS standard works. etc. I coded 1.22M verses for Jesus mentions by name, by title, or through speech. I ran it multiple ways to try to (imperfectly) account for disputes. johnholbein.org/christ-menti...

3 Sept

Getting closer to releasing estimates of LDS general authorities political ideologies based on data from General Conference speeches...

John Holbein@johnholbein1.bsky.social

working on something...

3 Sept

working on something...

2 Sept

You can literally see how Ariely* manually changed data to produce the results he wanted Truly remarkable level of fraud from one of the most famous social scientists *I bet Ariely will use the Gino dog-ate-my-homework defense that a mysterious RA selectively manipulate the data did it. I doubt it

datacolada.org[139] Artificial Deadlines (Part 2): Retrograding Ariely & Wertenbroch (2002)As described in our previous post (.htm), back in 2006 ariely@mit.edu sent three Excel spreadsheets, each “last saved” by user “Dan Ariely”, to then-PhD-student-but-now-full…

2 Sept

"Artificial Deadlines (Part 2): Retrograding Ariely & Wertenbroch (2002)" datacolada.org/139

2 Sept

"We have direct evidence that, yes, in fact, this falsified analysis was conducted and included as a response to a request from reviewers."

We should give Ariely some credit: he’s a great storyteller. Of fiction.

John Holbein@johnholbein1.bsky.social

More evidence out today that Dan Ariely likely fabricated data.

Academics are biased against null effects. As it turns out, so is the broader public.

John V. Kane@uptonorwell.bsky.social

🧐 What do citizens think about "null results" in scientific research? To find out, @johnholbein1.bsky.social & I fielded a large-scale online experiment. We gave respondents straightforward info about a study, asked them some Qs, showed them the results, then asked them Qs again.

More evidence out today that Dan Ariely likely fabricated data.

datacolada.org[138] Artificial Deadlines (Part 1): Evidence of Fraud in an Influential Study About ProcrastinationA new paper in Psychological Science (.htm) reports a failure to replicate Study 2 of Ariely and Wertenbroch’s influential article entitled, “Procrastination, Deadlines, and Performance: Self-Contr…

“Artificial Deadlines (Part 1): Evidence of Fraud in an Influential Study About Procrastination” datacolada.org/138

28 Aug

lol

25 Aug

We need a word that describes the simultaneous exhaustion and excitement that comes from teaching.