All podcasts / People I (Mostly) Admire / Summary

173. Steve Levitt Says Goodbye to People I (Mostly) Admire

2025-12-20 - source - Read full transcript

Steve Levitt (host)Stephen Dubner

Key insights

Mastery learning frees up most of the school day by letting students advance only once they've actually learned the material.

Sal Khan introduced Levitt to mastery learning on the podcast: the standard classroom model advances all 30 kids together regardless of whether any individual has learned the material, which Levitt calls terribly inefficient. Once students only move on after mastering a topic, three or four hours of the school day open up for other uses, which became the foundation for The Levitt Lab.

education-reform

Whether a student is engaged, not what AI can do, determines whether AI helps or hurts their education.

Levitt argues both sides of the AI-in-education debate are right simultaneously: for an engaged learner, AI is the best tool ever built for learning quickly; for an unengaged learner, it's the most effective tool ever built for avoiding learning anything. He says engagement is high in elementary school and drops off sharply through junior high and high school, and that drop-off, not AI capability, is what will determine outcomes.

ai-in-education

Schools should reward many different kinds of accomplishment instead of a single ladder of grades.

At The Levitt Lab, students can write and produce music, build particulate-sensing devices, or write a novella, and all get celebrated equally rather than ranked against a single valedictorian track. Levitt believes this shifts the dynamic from 'us versus them' among students to a shared culture where every student's version of success counts, which he considers the single most powerful thing the school does.

education-reform

Learning sticks when it's driven by an immediate need, not when it's taught in case it's useful someday.

Levitt adopted David Eagleman's framing of just-in-time versus just-in-case learning: schools teach content like triangle proofs on the chance a student becomes an architect decades later, and nobody remembers it. Levitt contrasts this with learning five years of math in three weeks once he was admitted to MIT and needed it, and teaching himself to program because he wanted to stop losing money at the racetrack.

education-reform

Good interviewing comes from preparation time invested, not natural talent for talking to people.

Levitt says he learned from Dubner to read every book and academic paper a guest has written before interviewing them. He argues someone with very little natural interviewing talent can still do well with enough preparation, and that guests can tell when someone has invested that time, which is what makes them open up.

interviewing-craft

Naming an unstated premise in a guest's work can break through a guarded interview.

Levitt opened his interview with Yuval Noah Harari by noting that Sapiens has no named characters, which violates what Dubner had taught him was the first rule of good storytelling. Harari, usually described as tough to draw out, visibly shifted his attitude once Levitt raised this, and the resulting conversation became PIMA's most-downloaded episode.

interviewing-craft

The hardest interviews are with guests who don't know the interviewer and who stick to a fixed script.

Levitt names his two most disappointing interviews as Arnold Schwarzenegger and Werner Herzog, both guests unfamiliar with him. Schwarzenegger refused to wear headphones to protect his hair before a TV appearance and couldn't be steered off a rehearsed life story; Herzog flatly refused to discuss his films, insisting the conversation be about his poetry instead, which Levitt hadn't read.

interviewing-craft

Imposed deadlines were valuable to Levitt precisely because his academic career had almost none.

As a tenured academic, Levitt says he essentially never had deadlines in his adult working life. The production schedule of PIMA, even though it stressed him and kept him up until 3 or 4 a.m. some nights, became a source of discipline he credits as genuinely useful. He also moved the show from weekly to biweekly because the weekly prep load was unsustainable given how much he invested per episode.

personal-transformation

The podcast shifted Levitt from producing ideas to consuming other people's ideas.

Levitt estimates he read only around 30 books in the 20 years before starting PIMA, mostly young-adult fiction he read alongside his kids. Hosting the show forced him to read broadly again and synthesize what experts were thinking under deadline pressure, which he describes as a genuinely useful pivot after years focused solely on generating his own academic ideas.

personal-transformation

Interviews are easier when the guest already respects the interviewer's own body of work.

Levitt says his easiest PIMA interviews were with guests who were already Freakonomics fans, which flattens the usual power imbalance where the interviewer sits 'below' the interviewee. Richard Dawkins, who had no idea who Levitt was going in, was a harder case; Levitt surprised him enough times over 60-70 minutes that Dawkins invited him to co-host a live event afterward, which Levitt says he then badly misjudged by focusing on Dawkins' science rather than the atheism questions the paying audience actually wanted.

interviewing-craft

Levitt sees his real legacy ambition as changing how kids are taught, not the podcast or his academic work.

Levitt says that if he could actually make a dent in how education works, it would feel like the thing he most wants to tell his grandchildren about, more than Freakonomics or his academic career. He frames engagement as the linchpin: get students engaged and results follow, fail to and the outlook is grim.

education-reform

Books referenced

Sapiens - Yuval Noah Harari - Levitt opened his interview with Harari by pointing out the book deliberately has no named characters, which violates the storytelling rule Dubner had taught him; naming it broke through Harari's guarded demeanor and produced PIMA's most-downloaded episode.
Noise - Daniel Kahneman - Cited by listener David Riedman, whose PhD dissertation built on Kahneman's 'noise audit' concept from the book, applied first to school police threat assessments and later to large language models.
Harry Potter series - J.K. Rowling - One of the only books Levitt read in the 20 years before starting the podcast, alongside his kids.
Twilight series - Stephenie Meyer - Same as above; part of the roughly 30 books Levitt read over 20 years, mostly young adult fiction he read with his kids.

Media referenced

Fitzcarraldo - movie - Levitt referenced this Werner Herzog film during their interview; Herzog deflected, insisting he wanted to discuss only his poetry that day.
Freakonomics Radio - podcast - Levitt's original show with Dubner; two guest-hosting stints on it led to PIMA, and Levitt returns to it as a regular contributor after PIMA ends.
The Economics of Everyday Things - podcast - Named as another show in the Freakonomics Radio Network alongside PIMA and Freakonomics Radio.

Companies

The Levitt Lab - In-person school Levitt started, first campus at Arizona State University in Tempe, expanding to outside Boston and to L.A. in fall 2026; built around mastery learning and engagement over grades.
Khan World School - Online school Levitt urged Sal Khan to start during Covid; Khan called him back five minutes after Levitt's tenth nagging email and invited him to help build it, which preceded The Levitt Lab.
University of Chicago - Where Levitt taught; he describes watching high-achieving students there who had 'won the high school lottery' but showed no genuine interest in learning, only in checking boxes.
Arizona State University - Hosts The Levitt Lab's first campus in Tempe, Arizona.

Techniques and frameworks

Mastery learning - Framework Levitt learned from Sal Khan: students move to the next topic only once they've actually mastered the current one, rather than all advancing together on a fixed schedule; frees up three to four hours of school time per day.
Just-in-time vs. just-in-case learning - David Eagleman's distinction, which Levitt says transformed his thinking: schools teach material 'just in case' it's needed years later and nobody retains it, versus learning something because you need it right now, which sticks. Levitt cites learning five years of math in three weeks once he needed it at MIT.
Noise audit - Daniel Kahneman's method for measuring aggregate variability in human judgment across a group, cited by listener David Riedman as the basis of his PhD dissertation on school police threat assessments and later on LLM judgment variance.

Summary

In the final episode of People I (Mostly) Admire, Steve Levitt hands the interviewer's chair to his longtime Freakonomics co-author Stephen Dubner and becomes the guest of his own show. The conversation opens on why the podcast is ending after five years and roughly every-other-week production, and quickly turns into a retrospective on what five years of interviewing did to Levitt personally: it pulled him out of a decades-long stretch as a producer of academic ideas and turned him into a consumer of other people's, forcing him to read broadly again after a 20-year stretch in which he estimates he read only about 30 books, mostly young-adult fiction alongside his kids.

The bulk of the episode is about education, which has become Levitt's central post-podcast obsession through The Levitt Lab, the in-person school he started at Arizona State University that is expanding to Boston and Los Angeles in 2026. Levitt traces the idea back to a PIMA interview with Sal Khan, who introduced him to mastery learning, the practice of only advancing students once they've actually mastered a topic rather than moving a whole class forward on a fixed schedule, which frees hours of the school day for other work. He credits David Eagleman's distinction between just-in-time and just-in-case learning as similarly formative, and describes The Levitt Lab's core design choice as celebrating many different kinds of student accomplishment (music production, engineering builds, novellas) rather than a single grade-based ladder, specifically to avoid the zero-sum dynamic he watched destroy the intellectual curiosity of high-achieving University of Chicago students who had "won the high school lottery" but cared only about checking boxes.

On AI in education, Levitt takes a deliberately two-sided position: AI is simultaneously the best tool ever built for an engaged learner and the most effective tool ever built for an unengaged one to avoid learning anything at all. He argues the real variable that will determine educational outcomes going forward isn't AI capability but student engagement, which he says is high in elementary school and falls off sharply through junior high and high school.

A second major thread is Levitt's own evolution as an interviewer, someone who describes himself as having "no human connections outside of this podcast." He credits heavy preparation, reading every book and paper a guest has written, as the real driver of a good interview rather than natural talent, and singles out his interview with Yuval Noah Harari as an example: opening by noting that Sapiens has no named characters (violating what Dubner taught him was storytelling's first rule) broke through Harari's guarded public persona and produced the show's most-downloaded episode. He also recounts his two most disappointing interviews, Arnold Schwarzenegger and Werner Herzog, both guests who didn't know who he was and who each refused to leave their own script, plus a badly misjudged live event with Richard Dawkins where Levitt focused on science when the paying audience wanted atheism.

The episode closes with Dubner revealing the actual reason for the ending: Levitt is moving to occasional guest episodes of Freakonomics Radio itself, aiming to tackle policy issues, starting with AI and education, in a way PIMA's person-centered format never let him do. Four listener voice memos close out the show, naming episodes with Richard Thaler, Charles Duhigg, Daniel Kahneman, and Sendhil Mullainathan as the ones that most affected their lives.

Notable Quotes

"It is true that being an interviewer was roughly the last thing that I ever should have done." - Steve Levitt

"If you are an engaged learner and you want to learn something... there's never been a tool like A.I... If you are unengaged and you are trying to find a way not to learn anything, there has never been a tool as effective as A.I." - Steve Levitt

"I think both of those groups are exactly right... I think it's going to be the key on which everything turns - if we can get students engaged, we will have unbelievable results. And if we don't, we are facing disaster." - Steve Levitt

"There's no substitute for hard work. You can have very little talent for interviewing, but I think if you've really prepared, it can still go pretty well." - Steve Levitt

"One thing I'm glad is that I finally quit something on time because I always wait until too long, like everybody else does." - Steve Levitt