
Sign up to save your podcasts
Or


The horrific murder of conservative activist Charlie Kirk on September 10 should have started a long-overdue conversation about the danger of political violence and of rhetoric that demonizes political opponents. Instead, all too predictably, it set off a cycle of finger-pointing and attempts to prove that violence is only—or primarily—a problem for the “other side.” And the Trump administration set the tone.
On the day of the shooting, Donald Trump blamed “the radical left” and its rhetoric before the shooter’s identity or motive was even known. Since then, he has explicitly and repeatedly stated that “the radicals on the left are the problem” and that “the violence comes largely from the left.” And, hosting Kirk’s podcast from the White House on September 15, Vice President JD Vance asserted that national reconciliation is impossible until Democrats and liberals not only agree that political violence is unacceptable but concede that “this is not a both sides problem”—that both support for political violence and political violence itself are “much bigger and [more] malignant” on the left.
Pundits on the right amplified this claim:
Meanwhile, some left-of-center commentators such as Vox’s Eric Levitz have countered with data purporting to show the exact opposite:
Who’s right? Should we see political violence as primarily a problem on one side—and which one?—or treat it as a danger across the board?
The real picture is complicated. If you set out to list only left-wing acts of political terrorism, as Erick Erickson does—especially if you reach all the way back to 1954—you can easily “prove” that it’s a left-wing problem. One could just as easily make a list that would inculpate solely the right: The 1963 Birmingham, Alabama church bombing targeting civil rights leaders; the 1984 murder of liberal talk show host Alan Berg by white supremacists; the 1995 bombing of the federal building in Oklahoma City by an ex-soldier turned far-right extremist; a series of deadly attacks on abortion providers; several massacres by perpetrators obsessed with the migrant “invasion”; July’s killing of Minnesota state legislator Melissa Hortman and her husband by an anti-abortion, anti-vaccine zealot.
Persuasion is a registered nonprofit that relies on reader support to pay our staff and keep our content free for everyone. If you value our work and want to fight for liberal democratic values wherever they are threatened, please consider becoming a paying subscriber today!
The selective vision continues with regard to present-day events. Last week, a shooting at an Immigration and Customs Enforcement facility—which killed two detainees and injured one, but was clearly aimed at the agency itself—was quickly cited as more evidence of the left-wing peril. Meanwhile, some Trump critics recalled that less than a month earlier, a man obsessed with the supposed harms of the COVID vaccine opened a barrage of gunfire at the Centers for Disease Control and Prevention headquarters in Atlanta, Georgia, killing a police officer.
However, claims about the overwhelmingly right-wing nature of political violence such as the one made by Levitz, which relies on the Anti-Defamation League’s database of “domestic extremist-related killings,” warrant skepticism. The fine print on the ADL graphic Levitz reproduces notes that the numbers include “both ideologically and non-ideologically motivated killings.” In other words, the ADL counts all homicides by everyone classified as a political extremist—including homicides related to drugs, robberies, personal or family conflicts, and prison gang violence—even when mental illness is involved and even when the connection to extremism is tenuous.
In some cases, the ADL’s classifications raise well-founded questions about political bias. The organization’s report on extremist-related domestic murders in 2019 included a mentally ill man who killed his brother and who may have belonged to the far-right Proud Boys group, as well as a man who fatally shot his estranged girlfriend and, after his arrest, “reportedly said he was a sovereign citizen who did not recognize any government.” However, it left out Conor Betts’s shooting spree outside a Dayton, Ohio bar, which killed nine people and injured 17, due to insufficient evidence of extremism and lack of ideological motive. Yet Betts’s Twitter account showed ample evidence of left-wing radicalism and a fixation on righteous violence against “fascists,” “capitalists,” and other perceived oppressors.
A more rigorous dataset, limited to politically motivated killings, was compiled by Cato Institute fellow Alex Nowrasteh, with the following breakdown for the death toll from political terrorism in the United States over the last 50 years:
In this calculus, homicidal right-wing violence still substantially exceeds the left-wing kind, though both are thankfully rare.
An X user called “Recursion Agent” has challenged Nowrasteh’s analysis, claiming that it overcounts violence from the right and undercounts it on the left. The critique accurately flags some apparently misclassified cases and appropriately questions the non-inclusion of others (e.g. several killings linked to the racial protests and riots in the summer of 2020); but it also has its own obvious biases. Besides blatantly downplaying some perpetrators’ racist or far-right views, the thread critiques Nowrasteh for leaving out a “deranged lefty who annihilated his family” out of fear of being targeted by religious zealots—but makes no mention of family murders by people in the far-right QAnon cult.1
Such debates highlight the difficulty of classifying the politics of extremists, especially those whose ideologies may be a mishmash of conventionally “left” and “right” beliefs and who may suffer from mental illness. And there are other thorny questions. If racially motivated killings by whites are counted as right-wing extremism, should killings of whites by nonwhite perpetrators obsessed with racial oppression, such as 1993 Long Island Railroad shooter Colin Ferguson, count as left-wing extremism? Should the deaths during the 2020 racial unrest, mostly related to nonpolitical robberies but enabled by the chaos and violence of the riots, be added to the “left-wing extremism” body count? Should the casualties in the January 6 Capitol Hill attack, all rioters themselves, be added to that of right-wing extremism?
Evaluating public support for political violence can be just as tricky. In a September 10 YouGov poll, 16 percent of self-identified liberals and 24 percent of those self-describing as “very liberal” said that it was “usually or always” acceptable to feel joy about the death of a political opponent, compared to just four percent of self-identified conservatives and seven percent of moderates. Similar numbers agreed that resorting to violence to achieve political goals could sometimes be justified. These results were widely cited—among others, by Vance while hosting Kirk’s podcast—as evidence that approval for political violence is not a “both sides” issue but exists primarily on the left.
But caveats are in order. The survey was taken immediately after Kirk’s murder. As the YouGov report notes, the polling outfit’s earlier surveys show that both Democrats and Republicans are consistently more likely to regard political violence as a serious problem after a widely publicized attack on a figure from their “team.” And the way questions are phrased can also influence answers. For instance, when a poll in March formulated a question about political violence as, “Things have gotten so far off track, true American patriots may have to resort to violence in order to save our country,” one in five Trump supporters, but only one in ten Trump opponents, agreed.
Subscribe now
Another recent survey seems reassuring in that it shows Democrats, Republicans (including “MAGA Republicans”) and independents all overwhelmingly oppose political violence. When asked if it was acceptable for citizens to use a variety of violent tactics to achieve their political goals, around 99 percent rejected politically motivated murder, assault, and arson, with negligible partisan differences well within the margin of error.
Should we conclude that there’s no problem? Not quite. For one thing, it’s easy to oppose political violence in theory; in practice, people often excuse it or hand-wave it when it’s on “their” side. For instance, in a 2024 YouGov poll, 60 percent of self-identified Trump supporters and 55 percent of all Republicans saw the January 6 Capitol Hill riot as an expression of “legitimate political discourse,” while only 18 and 20 percent respectively saw it as a “violent insurrection.” (The “discourse” included police officers being pepper-sprayed, punched, kicked, and beaten with flag poles and fire extinguishers; some 140 were injured, with one officer suffering a fatal stroke shortly after being manhandled by the crowd.)
Democrats have their own blind spots. During the 2020 Black Lives Matter protests following George Floyd’s murder, liberal publications such as Slate, The Atlantic and The New Republic published excuses for rioting as an expression of black anger and frustration. (In fact, quite a few black activists condemned the violence as the work of attention-seeking white anarchists.) Today, many liberals tend to fall back on claims that the 2020 protesters were “mostly peaceful.” But while this was true of most protests—of which there were over 7,000 nationwide—nearly 550 events did involve serious violence and/or property damage. At least 30 people died, excluding those killed by law enforcement; some 900 police officers were reportedly injured, and many city blocks were devastated, with damage totaling between $1 and $2 billion.
Both sides are also inclined to try to pin acts of violence on the other side. We just saw this with attempts by many on the left to claim that Charlie Kirk’s killer was a member of the extreme right. But at least people left of center (other than the far left) tend to believe reporting by the mainstream media. On the right, even people who are not particularly extreme can be found parroting false narratives reassigning right-wing violent extremists to the left. One routinely encounters assertions that the killer of Hortman, the Minnesota lawmaker, was not a Trump supporter but a leftist angry at Hortman for voting with Republicans on an immigration bill (a claim debunked by The Blaze, a strongly conservative website).2 And even Ted Cruz has amplified the claim that the 2022 attack on former Democratic House Speaker Nancy Pelosi’s husband was carried out by a “hippie nudist leftist” with no connection to the right. In fact, the attacker, David DePape, had once been involved in progressive activism but had since become immersed in far-right, MAGA-adjacent conspiracies including QAnon.
So, is violence more inherent to one ideology? One can certainly point (as conservative British pundit Douglas Murray does in City Journal) to the American left’s history of sympathy for violent radicals such as the Weather Underground or the Black Panthers. In a fresh, shameful example of this mindset, the Democratic Socialists of America (a group with ties to a number of progressive Democrats) and the Chicago Teachers’ Union both made posts honoring Assata Shakur, a member of a violent black liberation group who died in Cuba where she escaped after being convicted of killing a New Jersey police officer in 1973.
But the nationalist right has its own history of support for violent methods—from rhetoric suggesting that the right of gun ownership authorizes armed resistance to leftist tyranny to sympathy for dictatorships such as that of Chile’s Augusto Pinochet. (Ironic references to the Pinochet regime’s practice of killing undesirables by dropping them in the ocean from helicopters have become a popular right-wing meme.) If the left-wing call to fight oppression and “hate” can become a justification for violence, so can the right-wing call to fight the “enemy within,” be it foreign “invaders” or homegrown “Marxists.” Secretary of War Pete Hegseth has, for instance, routinely employed rhetoric in his books that treats liberals as a domestic enemy—and sometimes suggested that the fight against them may yet become a literal war.
The truth is that the dominant ideological tenor of political violence is not a constant. It was primarily left-wing in the late 1960s and 1970s, predominantly right-wing in the 1980s and onward; today, terrorism experts Daniel Byman and Riley McCabe write in The Atlantic, we seem to be seeing a swing toward more violence from the left. But that may change again, especially if more people on the right get the message that “it’s war.”
Thankfully, so far, the overall levels of political violence in the United States remain fairly low. But the danger signs are there: in a just-released NPR/Marist poll, 30 percent of Americans—up from 20 percent in March 2024—agree that “Americans may have to resort to violence in order to get the country back on track.” (While Republicans are more likely to endorse this statement, the most dramatic growth in agreement—from 12 percent to 28 percent—has been among Democrats.)
In this climate, reducing the temperature of political rhetoric is essential. Using the issue to score political points against the other side won’t get us there. The only answer is to stick to the facts—and, above all, to hold one’s own “team” accountable when necessary.
Cathy Young is a writer at The Bulwark, a columnist for Newsday, and a contributing editor to Reason.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
1“Recursion Agent” also criticizes Nowrasteh both for counting “crazies” on the right (e.g. a mentally ill anti-abortion crusader who went on a shooting spree at a Planned Parenthood clinic in 2015) and for not counting mentally disturbed killers who may have been on the left.
2This narrative is partly sustained by the fact that the suspect, Vance Boelter, was reappointed by Democratic Governor Tim Walz to the Minnesota Governor’s Workforce Development Board; but the board is one of many bipartisan advisory bodies. Boelter was, in fact, a Trump voter.
Growing up in Denmark in the early 2000s, I rarely worried about my right to free speech. In this cozy haven of liberal values and secular democracy, speaking freely felt as natural as breathing. Few contested this state of affairs, least of all religious groups, whose influence had long since faded.
That outlook changed twenty years ago today. On September 30, 2005, the Danish newspaper Jyllands-Posten published an editorial titled “The Face of Muhammad,” accompanied by 12 cartoons, some of which depicted the Prophet Muhammad. The publication set off a global firestorm and turned criticism of Islam into a minefield that remains deadly to traverse even in open societies. And rather than defending the principle at stake, European democracies are increasingly choosing appeasement—trading away free speech for the promise of a precarious peace.
The latest example comes from London. In June, Turkish immigrant Hamit Coskun was fined £240 for a religiously aggravated public order offence after burning a Quran and shouting profanities against Islam outside the Turkish consulate. Coskun insisted his protest was aimed at Turkish President Recep Tayyip Erdoğan’s authoritarian Islamism. But the presiding judge ruled that Coskun was “motivated (wholly or partly) by hostility” toward Muslims, and deemed his conduct disorderly—citing, as evidence, that he was physically attacked by two bystanders. In other words, the violent reaction to his peaceful protest was the basis for the criminal charge.
The case became more troubling still last week, when the man who assaulted Coskun with a knife was spared jail, with the judge noting that the assailant felt “deeply offended” as a mitigating circumstance. The asymmetry is glaring: the protester is punished for offending sensibilities, while those who turned to violence avoid serious consequences. This is appeasement in practice—embedding a “Jihadists’ Veto” into law and precedent, and ensuring that intimidation pays.
Unlike me, Flemming Rose, the cultural editor at Jyllands-Posten who commissioned the Danish cartoons twenty years ago, did not take free speech for granted. Rose felt a disturbance in the Enlightenment principles that had long sustained Western European societies. He sensed danger signs in 2004 when an Islamic fundamentalist murdered Theo Van Gogh, the director of the Islam-critical film Submission, on the streets of Amsterdam. Later that year, a Danish lecturer was assaulted for reciting the Quran in class (extremists view it as a crime for non-Muslims to read from holy texts). In 2005 a Danish children’s author couldn’t find illustrators for a book on Muhammad. In response, Rose invited cartoonists to depict Muhammad as a way to test the boundaries of self-censorship and the commitment to free speech and tolerance.
In an editorial accompanying the cartoons, Rose wrote that affording special protections for any particular religion or its adherents was “incompatible with a secular democracy and freedom of speech, where one must be prepared to endure mockery, scorn, and ridicule” regardless of group identity. Far from an act of “Islamophobia” or anti-Muslim bigotry, Rose’s editorial was about taking freedom and equality seriously, even when it is uncomfortable to do so. And in Denmark, the initial response was relatively restrained. Some criticized the cartoons as needlessly provocative; others defended the right to publish them.
Internationally, however, regimes in Egypt, Iran, and Pakistan amplified the controversy, stoking outrage to claim the mantle of defending the global Muslim community. The Organization of Islamic Cooperation (OIC), representing 57 Muslim-majority countries, launched a campaign to reframe the cartoons not as religious satire, but as a form of hate speech. Leveraging existing provisions in international law—particularly Article 20 of the UN’s International Covenant on Civil and Political Rights—the OIC pushed for global laws criminalizing “defamation of religion.”
Western governments and human rights groups, unsure of how to balance respect and rights, at times gave credence to this narrative. The U.S. State Department initially condemned the cartoons as “inciting religious and ethnic hatreds.” Amnesty International released a statement suggesting the cartoons might violate international hate speech standards. The Nobel Prize-winning German writer Günter Grass went so far as to liken the cartoons to Nazi propaganda.
Though the OIC campaign ultimately failed—thanks in part to a counteroffensive by the Obama administration in 2011—the episode revealed a deeper vulnerability. Less than two decades after the Rushdie Affair, liberal democracies were no longer confident in defending their own foundational freedoms.
And if legal mechanisms proved insufficient to silence dissent, violence filled the gap. In 2015, a decade after the Danish cartoons were published, two gunmen stormed the offices of Charlie Hebdo in Paris, killing 12 people. In 2020, French schoolteacher Samuel Paty showed his civics students the Charlie Hebdo cartoons during a lesson on freedom of expression. He was beheaded in the street days later by a radicalized teenager.
Persuasion is a registered nonprofit that relies on reader support to pay our staff and keep our content free for everyone. If you value our work and want to fight for liberal democratic values wherever they are threatened, please consider becoming a paying subscriber today!
Even Denmark, where the crisis began, has shifted course. At the outbreak of the cartoon crisis, the Danish government refused to compromise free speech. In 2017, the Danish parliament repealed the country’s centuries old blasphemy ban—a symbolic victory for free speech over religious dogma. But just six years later, that progress was undone. After a series of Quran burnings by far-right activists triggered renewed outrage and threats from the Muslim world, Danish authorities reversed their position. In 2023, a new law was passed criminalizing the “improper treatment of scriptures,” effectively reintroducing a narrower blasphemy ban. (This May, two Danish far-right protestors were the first to be convicted under the new law. Their crime: tearing out two pages of a Quran and dropping them into a puddle of water during a public debate about the same law that would later be used to convict them.)
More capitulatory acts to the “Jihadists’ Veto” would follow. The OIC revived its campaign to ban criticism of Islam and achieved a symbolic victory in 2023 when the UN Human Rights Council passed a resolution condemning desecration of holy texts as incitement to hatred.
Also in 2023, two Iraqi refugees living in Sweden burned a Quran in protest against Islamist extremism. They had secured police protection for the demonstration, explicitly stating in advance that a Quran would be burned. While certainly inflammatory, the act was understood to fall under Sweden’s broad protections for free expression—protections rooted in the abolition of its blasphemy law in 1970. But fearing diplomatic consequences and religious violence, Swedish authorities charged them with incitement to hatred, even though their protest was directed at Islamist extremism.
In January this year, one of the refugees, Salwan Momika was assassinated in his apartment just days before his court date. In a model democracy, peaceful speech was condemned twice: once by the legal system, and once by the very religious fundamentalists the law was meant to appease.
Europe’s legal regression is based on the assumption that in diverse societies free speech must be curtailed to avoid offense and preserve social harmony. But no amount of sugar coating with appeals to “tolerance” and “respect” can hide the bitter truth that fear of intimidation and violence are the real driving factors.
Subscribe now
In this new climate, violence is not merely a threat—it has become a strategic tool to reverse the hard-won freedoms that once unshackled the human mind from state-imposed religious diktats. The lesson is stark. When democratic governments criminalize peaceful expression, they do the work of extremists for them—and still fail to stop the violence.
Free speech, once lost, is not easily regained. Twenty years on from Flemming Rose’s social experiment, the question is no longer whether free speech is under threat. It’s whether liberal democracies still have the will to defend it against forces determined to destroy it.
Jacob Mchangama is the Executive Director of The Future of Free Speech and a research professor at Vanderbilt University. He is also a Senior Fellow at The Foundation for Individual Rights and Expression (FIRE) and the author of Free Speech: A History From Socrates to Social Media.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
David Bau is Assistant Professor at Northeastern University and Director of the National Deep Inference Fabric, researching the emergent internal mechanisms of deep generative networks in both Natural Language Processing and Computer Vision.
In this week’s conversation, Yascha Mounk and David Bau explore the technology behind AI, why it’s concerning that so many computer scientists don’t understand how it works, and how to embed morals and values.
This transcript has been condensed and lightly edited for clarity.
Yascha Mounk: We met recently at a workshop about artificial intelligence at Harvard. I thought that in a conversation we had, you helped me understand the nature, the architecture, and the technology of artificial intelligence better than anybody had before. I thought I would love to talk to you about this on the podcast. Fundamentally, how do current AI models work? When we say they are LLMs, large language models, what does that mean and how does that distinguish this form of artificial intelligence from other forms that we’ve historically used?
David Bau: They are a generative model, which means that they are just open-ended models of a kind of behavior as opposed to being trained to make really narrow decisions. What was really popular in AI up until recently was to train classifiers to solve specific problems, like to help you make a specific decision.
But then in recent years, it has been more popular, or actually more amazing, to create models that have more open-ended goals. So a large language model is pretty simple. In concept, its job is to imitate human language. But imitating human language is a lot richer than making a yes or no decision or answering a simple multiple-choice question, which is what we used to train AI to do.
Mounk: Perhaps help us understand what these classifiers were, to understand what the difference is with these large language models. It sounds like you’re saying the point was to classify into yes or no, into one of four or five different kinds of buckets.
Bau: Yeah, I’ll describe it in the way that I describe the difference between these two types of models to students. Basically, if you train an AI to classify inputs into a bunch of different categories, what you’re doing is asking it to tell you the difference between things. Let me give you a couple of examples. You could ask the models to tell you the difference between a picture of a cat and a picture of a dog.
Or for a more realistic application, you might ask an AI to tell you the difference between a piece of writing that was written well and a piece of writing that was written badly, or a movie review that was positive or a movie review that was negative. A company like Yelp might do this to take a look at your reviews to see if you tend to write positive reviews or negative reviews, or if a specific review is positive or negative.
Mounk: What is the startup doing in the TV show Silicon Valley that seems really stupid and then suddenly becomes important? He’s trying to classify whether something is a hot dog or not. So that was a classifier.
Bau: Yes, I think that’s right. So it was really one of the earliest problems that people put in front of AI. I think the very first neural network that was ever made was a thing called the perceptron, and it was trained to classify the difference between pictures of boys and pictures of girls.
They took a lot of pictures of students and showed that this little neural network, which had 64 neurons, if properly configured, could tell the difference between these types of pictures. That was considered an amazing feat at the time. That same class of problems has been with us for more than 50 years.
It is a powerful framework, but it gives the machine learning models a lot of space for taking shortcuts. For example, if you needed to tell the difference between a cat and a dog, it might be sufficient to not look at every aspect of that picture. You might be able to just look for the tips of the ears and recognize that cats have pointy ears and dogs tend not to, without looking at the rest of the image.
Thanks for reading! The best way to make sure that you don’t miss any of these conversations is to subscribe to The Good Fight on your favorite podcast app.
If you are already a paying subscriber to Persuasion or Yascha Mounk’s Substack, this will give you ad-free access to the full conversation, plus all full episodes and bonus episodes we have in the works! If you aren’t, you can set up the free, limited version of the feed—or, better still, support the podcast by becoming a subscriber today!
Subscribe now
Set Up Podcast
And if you are having a problem setting up the full podcast feed on a third-party app, please email our podcast team at [email protected]
One of the things that classifiers are really good at is identifying the most salient difference, focusing on that, and making a decision based on it, which is great. It leads them to be very accurate, but it also means that they do not necessarily develop a complete understanding of the world. If you invented a picture of a pointy-eared dog and gave it to one of these classifiers that was focusing on the tips of the ears, it would say, that’s clearly a cat. It might not recognize that there was something else wrong with the image.
Mounk: Even if the rest of the image very clearly looks like a dog, and it is not really a hard case in other ways.
Bau: It is the kind of thing a classifier would do, yes.
Mounk: Tell me a little bit about the technology behind it. We may be going too far back now, but is a classifier a similar kind of technology with fewer neurons, whatever that means? Or is it a completely different kind of thing?
When we move from classifiers to the kinds of large language models we have now, is it building on the same technology, or is it a completely different avenue toward how to create this form of intelligence?
Bau: It is basically the same kind of technology. I would say that there are relatively few major innovations that separate classical classifiers, as invented in the 1950s, from modern large language models. There have been a lot of gradual, small, clever innovations, but in terms of major innovations, there have been relatively few.
We are really going after the problem using the same techniques that we have used since the 1980s, when a lot of the innovations that we are still using today were established.
Mounk: Tell us about what these techniques are. One idea that often comes up is neurons and neural networks. I can understand what a neural network might be in my brain. I am far from being a neuroscientist, but I understand that a neuron is a kind of cell, unless I am getting that badly wrong. My brain is a kind of neural network. All of these cells are connected in some complicated way. What does it mean for an AI to have neurons or be a neural network?
Bau: There is a popular term that you will see sometimes called a deep neural network. What makes a neural network interesting is its depth. A neural network is inspired by the architecture of the human brain. All it does is compute a bunch of numbers. The input comes into the neural network as words or as images, and the first thing you do is convert it to a bunch of numbers and feed each number into a neuron.
Then you connect the neurons so that they pass the numbers from one to the other. If a neuron has a bunch of inputs, it adds the numbers together, does a small bit of computation, and then creates another number that feeds onto the next layer.
Neural networks are just this big mess of neurons connected to one another to produce some output that you hope is useful. If you just created a random neural network, it probably would not do anything useful, but it would do something with your data. The trick for artificial intelligence, the trick for machine learning, is to train the neural network to strengthen and weaken all of the connections between the neurons, transforming a random machine, a random function, into something that does something useful. It is the training process that makes a neural network seem almost magical.
Mounk: Okay, so take me a step back. I want to understand the training process, but before that, I now get this image of a bunch of human cells—or a bunch of digital cells, whatever exactly that means—transmitting this information. You see something with your human eye, that is a visual stimulus, and in some way that stimulus gets translated into a bunch of signals that neurons fire at each other.
Subscribe now
Then there is some way of using that information. Help me understand a little bit more how that works in a computer and what the point of these simple calculations that you mentioned is.
Bau: The simple calculations are probably simpler than you would imagine as a non-technical person. Every neuron is just a sum of all the inputs that come into it. It is a weighted sum where, if you have one neuron that is connected to 1,000 inputs, it just adds up all the inputs and then looks at whether the numbers come out to be positive or negative.
If the numbers come out to be positive, the neuron will do one thing in the output, such as transmitting the sum out. If the numbers come out to be negative, the neuron will do something different, such as outputting zero. That output then becomes another number that goes into other neurons. Each neuron is an extremely simple step. It is just adding things, then looking at the answer, and then producing an output. You might think, oh gosh, why the heck is this useful?
Mounk: I can imagine what it looks like, but I am having trouble understanding why it is useful. You have these very simple calculations, they add up, so how does that mean that I can—I know we are jumping a number of steps—but how does that relate to why I can have a conversation with ChatGPT using my voice and it talks back to me? I know there are going to be lots of steps to get there, but why is it that this neural network is such a powerful tool, such a useful idea?
Bau: Why is it so useful? You can answer this in two ways. But I have to say, the question you are asking is actually one of the core puzzles behind neural networks. Let me give you a little bit of history. Neural networks are one of the oldest forms of programming ever devised, in the 1940s before digital computers were widespread. They have been with us ever since in different guises.
One of the reasons it has taken so long for neural networks to become so prominent in computing is that everybody had the question you are asking now. Even if you can get these things to work, why would we expect them to work? In the 1950s and the 1960s, a scientist named Rosenblatt demonstrated that you could get neural networks to do some useful things, but it never really caught on because everybody had this question: there are these neurons just passing numbers around. What do these numbers mean? How can we be sure they do anything useful? Is there anything that explains why they do something useful at all? Sometimes they do not work. Can we tell the difference?
It really took a long time. There were many different ways of doing machine learning where the numbers inside the AI were more understandable. For many years, mainstream AI scientists believed that you should use one of these other approaches, which were more transparent and designed to be explainable. It is really just in recent years that we have decided to use neural networks that have this key disadvantage: we do not know what the numbers that come out of a neuron are for. We do not know if they have a particular purpose. We do not know if some of the neurons are more important than others. We do not know under what conditions they learn something good or learn something bad. It is very opaque.
Subscribe now
But one of the things that has happened is that neural networks work so well. We have some understanding of some of the things that lead to them working so well. They work so well that the field has become comfortable with the idea that we should just use these black boxes. They are so useful that maybe it does not matter that we do not really understand what the neurons are doing, what they are for, or what is being computed inside.
One of the reasons I was at this workshop is that I am really concerned about this philosophy we have adopted in the AI industry and in machine learning. Historically, as engineers and computer scientists, it has really been our responsibility to understand the systems that we make, to make sure they are doing what we want, that they operate correctly. I feel that the new discipline of machine learning, because it has become so important to just accept these black boxes and use them even though we do not understand them, is leading to an unhealthy turn in the practice of engineering and computer science.
We are training a whole new generation of computer scientists to be comfortable with this idea that they should not really look inside these complicated black boxes, that it is not something understandable or their responsibility to understand. I think this is a fundamental error. One of the most important things we should be doing as AI scientists and practitioners is to try to resolve this fundamental problem with large-scale neural networks and to understand what is making them work, what causes them to work well or badly, or to acquire certain types of behavior.
Mounk: You work on interpretability, which I understand is basically an attempt to look under the hood to get a better sense of what is going on in that black box in various ways.
Bau: That is right. I work in an area that you could call post-hoc interpretability. There are a couple of approaches to achieving interpretability in machine learning. One is to simplify the machine learning models to the point where a person can look at them and understand and explain what each of the steps is doing. But the area that I work on is post-hoc interpretability.
That means, let us say we did not do that. Let us say we decided to use a neural network with millions or billions of neurons that are far too complicated for us to have a preconceived idea of how they work. Can we go in there and analyze the system the same way that a biologist might analyze a complicated, emergent biological system? Can we understand the structure of these learned computations after the fact, after they are trained, even if we did not try to restrict them ahead of time to make them understandable by people?
Mounk: I would love to get into a little bit more detail about what you think we can understand and what progress we have made in that field of interpretability. But to go back to the bottom-up building of our understanding of these AI models: we have these neural networks. They are a very simple process, actually. We do not fully understand why they have proven to be so phenomenally useful.
We started out having limited resources, with a limited number of neurons in these networks and probably limited data to throw at them. We might have been able to get them to distinguish between a dog and a cat, between a girl and a boy, and perhaps between something that is a hot dog and something that is not a hot dog. Then somebody said, let us raise the scale of ambition. Let us throw a lot more data at it. Let us do a lot of other things—you will tell me what they are.
Perhaps they could then become general-purpose large language models that are not trained to one very specific goal, but can actually help us with a huge variety of tasks, from writing a poem to summarizing a text to generating an image. What technical progress or what changes have allowed us to get there? How is it that this neural network suddenly scales to be able to do these kinds of things?
Bau: There are a couple of things to know, and things that readers or listeners might have heard of. Let me first emphasize that these really large neural networks are not that different from the classifiers that we have worked with for 50 years, other than the fact that they are bigger. From a technological point of view, from the way it is built, a language model is a classifier. It solves a slightly more open-ended classification problem than we solved previously, but it fundamentally just solves repeated classification problems. And what is the classification problem? It is this big multiple-choice question: what is the next word?
Is the next word cat? Is the next word dog? It is not a two-way choice that the language model is facing. We typically give a language model a vocabulary of something like 50,000 words, syllables, and letters. We tell the language model, you have this 50,000-way choice. What is the next word that is the right answer in this context? As input, we give the language model all the previous words, and we ask it, tell me what you think the right next word is.
So it is just a classifier, just like we trained classifiers to tell the difference between cats and dogs. But the scale of modern parallel computing GPUs has allowed us to make these classifiers capable of doing larger-scale classification problems. The outputs are bigger—there is a 50,000-way choice instead of a two-way choice or a ten-way choice, which we might have traditionally done. The inputs are bigger—instead of just one image or one sentence, we can feed these models entire books, entire histories of text, for them to look at to decide what the next word should be.
There have been a couple of key architectural innovations that allow these models to consume and learn to use such huge inputs and solve open-ended output problems. But fundamentally, they are just neural networks, a bunch of neurons connected in the same way that Rosenblatt was connecting neurons in the 1950s.
Mounk: So you think it is basically the same as what we were using 80 years ago, which is astonishing. But there have been a few technological innovations that allow these models to consume so much information and therefore go beyond the limits of earlier models. Can you give us an example, or is there a particularly important innovation in that respect?
Bau: I think that one of the things that has happened in recent years is the rise of a particular neural network architecture strategy called the transformer, which has really taken over the industry. There used to be a wide variety of different neural network architectures, but more and more we are converging on using transformers for everything. The fundamental thing that transformers do is introduce a form of short-term memory that we call attention.
What this means is that instead of the models only consulting what they learned during training, the models develop the ability to learn from the inputs they are provided. Let me give you a simple example.
If there was a particular person that you were asking the language model about, then with traditional training you would expect that language model to only be able to answer questions about that person if information about them appeared in the training data. But in real life, you often have situations where somebody asks you a question about a person you just met. You did not meet this person in your childhood. You did not read about them in school. You just had a conversation with them, and then somebody asks you about the person, and you have to answer now. This kind of short-term reasoning is something that traditional neural networks are not very good at.
What the transformer architecture does is introduce a special way of connecting the neurons called an attention layer, which allows the network to look back at previous things that happened recently in the input and use them as a type of memory, manipulating those memories and reasoning about them. This has turned out to be so powerful that you can really think of transformers as a different class of neural networks. It was a major innovation.
Mounk: So the difference, if I think in terms of my interaction with something like ChatGPT, Claude, Grok, or any of the other AI models, is this: if the model did not have a transformer, the problem would be that every question I asked it would be answered the same way it would have been before I started the conversation. It would essentially be difficult to have a back-and-forth conversation.
The transformer is what allows it to keep a conversational context in mind and to have an ongoing, progressive conversation. Or did I misunderstand that?
Bau: I think that is right. The transformer has really made it possible to teach the neural network things by telling it something, rather than being in control of the whole training process. When you have a conversation with a person, you are constantly learning and teaching the other person your ideas. You are sharing your concepts, sharing your understanding of the world with the person, and they absorb it. That allows the conversation to proceed.
Transformers allow a neural network to develop the ability to have the same kind of conversation, to develop an understanding during the course of a conversation, to learn things in the short run that it uses to make immediate responses, and to recall things that happened recently, as opposed to only relying on long-term memory.
Transformers were not the first architecture to try to do this. There were architectures proposed in the 1980s to do this, called recurrent neural networks (RNN). You might have heard of LSTMs, long short-term memory networks. The idea that you should have a short-term memory, that you should be able to solve problems in this way, is not totally new.
What a transformer does is make it efficient to train networks that can do this. It was an innovation that showed this old idea could be made practical and scaled up in a big way.
Mounk: Technically, can somebody with a computer science background understand how the transformer does that? Or do we get into areas that are too complicated to understand?
Bau: I think the main thing to understand is that if you were to try to train something with short-term memory, then since short-term memory seems like such a sequential process, the natural ways that you would end up training a neural network to have short-term memory are very sequential. They are one at a time.
You tell the neural network some things and then immediately turn around and ask it to make a prediction of the thing that you just told it. Then, based on that, you might go on to the next step, because time proceeds in a stepwise fashion. That can be done, but it tends to be very slow. The big innovation around the AI industry has to do with parallel computing.
The reason that training neural networks is so efficient is that we can process many inputs and learn many things at the same time in parallel on these GPU devices. The recurrent neural network architectures of the past did not fit very well with this parallel computation model. There were many things in the training that were inherently sequential, so training them was slow.
What a transformer does is change some assumptions in how this memory works to allow it to be parallelized really well. I am not sure it is very interesting to go into the details of how it is parallelized. It does mean that transformers are a little more limited theoretically than the old RNNs, but the limitations are carefully chosen and carefully architected to allow parallelism and to cover for the limitations.
There is a concept that transformers have that the old RNNs do not have, which is called a context window. Sometimes if you buy an AI product, it will tell you that the product has a certain context window, and another product might have a larger context window. You might even have to pay more money for something that has a larger context window. The idea of a context window is something that transformers introduced to allow them to parallelize the training.
A context window is a fixed number of words in the past that the transformer can see when it is trying to remember its short-term memory. An RNN has an infinite context window. In principle, it could remember everything in the past since it was first turned on. But a transformer is trained with a fixed context window. If the transformer has a context window of 1,000, it means that after you say 1,001 words, the very first word that you said is no longer in the short-term memory of the transformer. It will not be able to remember that anymore.
That simple limitation ends up being an enabling factor for training. It allows the neurons to be hooked up in a way that the transformer can be trained in a massively parallel way, which is many times more efficient than training an RNN.
Mounk: The limits of these context windows are still relevant today in terms of technical limitations in some of the existing language models. You can go back and forth for a certain amount of time, and then at some point it loses track of the beginning of your conversation. Or if you ask it to do much more complex tasks, it stays on task and on track for a while, and then it stops being able to retain the information it needed in order to carry it out in the right way. Is that roughly right?
Bau: Yes. There are really two effects here. One is a hard context window, where the transformer has no hope of understanding things that are beyond its context window. The other is a soft decay of its memory. These neural networks are statistical machines, they are never perfect, and as the conversation gets longer, even for things that are theoretically inside the context window, the transformer has more difficulty accurately recalling and processing things that are further in the past.
Mounk: I want to go back to the overall architecture of the AI now. I am going to create a hypothetical scenario here, David. What if I gave you a billion dollars, and said, build me an AI? What would you do? You would build a neural network, it would have a transformer, all of those things. What are the other steps here? You have to train it. Once you have trained it, you have to adjust it. What does that mean concretely?
Bau: Modern machine learning really has two steps. If you gave me a billion dollars to train a neural network, the task in front of me would be, first, to do something called pre-training the network. The second task would be to fine-tune the network to have a certain personality or to achieve a certain goal that I want the network to help me with.
This split between pre-training and fine-tuning is one of the pieces of the lore, one of the fundamental rules of thumb that we have learned, and it is quite profound in modern machine learning. The idea is this: if you go straight to trying to train an AI to solve the problem that you care about, then you miss a lot of opportunities to get an AI that has a profound understanding of the world. There are many other problems, unrelated to the one you are making the AI for, that it could learn from and generalize from.
What people are realizing is that the way to make an AI is to begin by training it to understand as many things as possible in the world. Once the AI is really good at modeling a wide variety of interesting problems, you then fine-tune it on solving the particular problem that you care about. The AI benefits greatly from that pre-training.
The first step nowadays is to pre-train the model on a universal problem. The universal problem that the entire industry has converged on is to pre-train the model on large-scale language modeling: to be able to imitate text. Which text? All the text. All the text that humanity has ever written, broadly construed.
If the text has images in it, those images can be encoded as little pieces of text, little patches of image words. If there are videos, they can similarly be boiled down to a set of tokens. We can train an AI to imitate any content that has been put together by a human in the past.
Mounk: Tell me what training means in a technical or semi-technical sense. Presumably, when a baby is being trained, the baby has eyes and ears and looks around the world. The information flooding through its brain somehow gets encoded by some mechanistic form into these neurons. Over time, the kinds of stimuli that the baby sees and receives start to train the neural network that is the baby’s brain.
I am going to try to make the analogy here for the AI system. Presumably, you can have this neural network, and you throw all of this text at it, and somehow that shapes the neural network in a way that may or may not be analogous. What does it mean to train exactly?
Bau: To train a neural network is actually really simple. First, you need to have a goal. Once you have a goal, you expose the neural network to challenges. You give it inputs and then have it produce outputs. You then ask, did the output achieve that goal or not? Sometimes the output will have achieved the goal, and sometimes it will not have achieved the goal.
If it did achieve the goal, then whatever computation the neural network happened to do in that instance, you strengthen all the neural connections that led to this positive outcome. If the network did not achieve the goal, you go to that computation and slightly weaken all of those neural connections that led to the bad outcome. Each time you do not make a huge change in the network; you might change everything by 1 percent or by a tenth of a percent. But after you have done this thousands, millions, or billions of times, eventually the network will converge on a pattern of computation that becomes correct more often, incorrect less often, and increasingly sophisticated at solving harder instances of the goal over time.
This whole process is called backpropagation or gradient descent, and it is the backbone of how machine learning works. It is a deceptively simple process. A primitive version of it was invented in the 1950s, and more sophisticated versions were developed in the 1980s. It remains such an important process that it is still an active area of research today. Fundamentally, the technique of backpropagation is the same as what we have been doing since the 1980s. We have just been making small tweaks to it.
Mounk: Very simple question. To have this backpropagation mechanism, presumably you have to know when the system is doing something good or bad, right or wrong. But we are not talking here about post-training. I do not think we are talking here about how, once you have a large language model that has been trained, you then give it feedback depending on its output. That is a different thing.
So how does the model know when it is doing something right or wrong? For example, it might say, this is a cat, or something more complicated, like producing a sentence. How does it know, this is a good sentence or this is a bad sentence? How does it know, actually, this was not a cat, it was a dog?
Bau: The distinction that you are drawing is one of the big fundamental insights, which is the distinction between supervised training and unsupervised training. In supervised training, you have a clear idea of the problem that you want the network to solve. For example, you need the network to tell the difference between good restaurant reviews and bad restaurant reviews. You show the network what good ones are and what bad ones are, and then punish the network every time it makes the wrong choice. That was the way we conceived of AI training for a long time.
The problem is that it is expensive to collect that training data. You have to make human assessments and judgments about the problem you want to solve. There is only so far you can go with it. The big innovation is to introduce a different type of goal, called an unsupervised training problem. An unsupervised training problem is a goal where you do not need to label the data. You do not need a person to tell you that this was the right thing or the wrong thing. You come up with a goal that the AI can pursue that is more natural or more ubiquitous in the world.
Language modeling is an unsupervised training goal. This multiple-choice question of predicting the next word does not require a human expert to label the data, saying, this is the right word, that is the wrong word. All we need to do is gather text. A language model can judge itself based on all the text that was written without needing a separate expert to train it on these things.
Mounk: In supervised learning, I have a database with 100 positive reviews and 100 negative reviews—probably more like a million, but let us say 100. You give it one of these reviews, and in the database there is a data point that says positive or negative. That label was generated by humans at some point. A human looked over these 200 reviews and classified them into positive and negative. The system judges itself against the “objective” human judgment encoded in the thing it is checking itself against.
How does unsupervised learning check itself at the end? You are saying, for example, that Shakespeare, or some blogger, or a New York Times journalist has already decided what the right next word is. But since this LLM is creating a sentence that has never existed in human language in many contexts, how does it know whether it is similar to the kind of word that Shakespeare, or The New York Times, or the blogger would have written next?
Bau: There are two things to clarify. One is that one of the fundamental insights that made unsupervised learning work was the recognition that there is no single right answer. What is the next word here? If different people confronted the same situation, even if they were very intelligent and very human, they might choose different words.
It is more accurate to think of the right next word as a distribution of possibilities. Maybe 30 percent of the time you would have chosen this word, 10 percent of the time you would have chosen another word, and in the remaining 60 percent of the time there is a wide variety of other choices you could have made. The right thing to train the AI to do is to understand and model this probability distribution as accurately as possible. Instead of just getting the next word right, it needs to get the probabilities right as much as possible. There are mathematical ways of writing this down.
That is why these systems are probabilistic machines. They do not output single choices. They output an assessment of what they think the probabilities should be.
How do you know if this is right or not? What we do is measure these things on what is called a holdout set. It is really simple. You take a bunch of text that would have been part of your training data, and then you separate it out as a holdout, a quiz. You tell the AI system, you can train on all this data, but not these ten pages. These ten pages are different, and you will never get a chance to see them.
After all the training is done, we go to the model and ask it to look at these ten pages. We give it the first 100 words of the first page and ask it to tell us what it thinks the probabilities of the next word are.
Mounk: I see. The closer it gets to actually predicting the rest of the passage, the better it has learned.
Bau: That is correct. This holdout test has for many years been the gold standard for how you measure the success of a machine learning model. Can it correctly predict the answers on a piece of data that you held out from training?
Mounk: You have done all of this, and in a sense this is only the first step of what a lot of models do at the moment. Because then, if I understand it right, you have a bunch of post-training, or whatever the right term for it is, in which you have a model perform tasks and you give it positive or negative reinforcement depending on what it does.
How is that different from what we have talked about so far? How does that change the model? You have this trained system, this huge neural network, so how is it that positive or negative reinforcement changes the physical structure of the network? Surely it needs to do that for it to learn over time and become better adjusted.
Bau: That is right. The problem with unsupervised training is that the model does not learn how to do any one useful thing in particular. Let me give you an example of what comes out of unsupervised language modeling. If you go to an unsupervised language model and try to have a conversation with it, and you say, please tell me the capital of Vermont, what do you want the language model to say?
You want it to say, what a great question. A lot of people do not know the capital of Vermont, but it is Montpelier. Here is a way to remember that. Here is a little bit of information about Montpelier. But if you go to an unsupervised language model and ask, what is the capital of Vermont? it will answer by predicting what it thinks the most likely next word is. It might say, what is the capital of Colorado? What is the capital of Maine? What is the capital of Wyoming? What is the capital of New York?
Mounk: Right. Normally, when you look at a text, the next word is not necessarily Montpelier. It might be in certain contexts, such as a dialogue in a novel where the person has the right answer. But in other contexts, you may have a list of questions, or you may be using it as an example in a philosophical text, etc.
Bau: Indeed. If you really train the model on all the world’s text, then the most common situation, the most common context for asking a question, would be a book of questions. The model will just continue writing that book of questions, inventing more and more that are similar to the one you asked. It is a pretty dissatisfying experience.
It can actually be fun, but it is not very useful. What you can do, though, is go to one of these pre-trained language models and say, it’s great that you can imitate every book in the library but let me give you a set of books that I would like you to be especially good at imitating. What are these books? They are a collection of 100,000 conversations, pieces of dialogue, which are examples of people asking questions and getting their questions answered in a nice and helpful way.
If you go back to this pre-trained language model and train it on dialogue text—just fine-tuning it, which means taking these 100,000 pages and making them the last thing the network was trained on, the last thing it learned, the last thing it was rewarded and punished for—then the language model will acquire this bias. It will tend to imitate the last thing it saw. If you ask, what is the capital of Vermont? it will tend to give you a useful answer. It will answer in dialogue, which is remarkable.
This process is called instruction fine-tuning. People collect datasets of useful instruction-following behavior: Please answer this question for me. Please do this for me. Please do that thing for me, and examples of an AI doing the task in a smart, useful way. If you went to a transformer and trained it only on these thousands of conversations, it might understand the grammar of what you are doing, but it would not be very helpful. It would not know much about the world.
But if you go to a large language model that has been trained to imitate every book ever written, every blog post ever posted on the internet, and then, as a final fine-tuning, you show it dialogue and say, what I really want you to learn is to follow this format. While you are doing next-word prediction, just do it in a way that answers questions, then something profound happens. Not only does it follow the form of dialogue, but it also exploits the vast array of knowledge it acquired during pre-training.
For example, if you ask a question about Shakespeare, the model tends to answer it, even if the specific dialogue examples said nothing about Shakespeare. It will follow the dialogue form, but draw on the knowledge it acquired earlier in pre-training. That is really the magic of modern machine learning, of modern language modeling: the split between pre-training and fine-tuning.
In the rest of this conversation, Yascha and David discuss how to set constraints, instilling moral values, and how to tell if AI systems are using reasoning. This part of the conversation is reserved for paying subscribers…
If everybody is struggling to figure out what AI really is, or how it will affect our lives as a whole, one thing would seem to be certain: Gen Z is defined by AI in the way that flappers might have been by the radio or Boomers by the electric guitar. But the equation of AI and Gen Z misses a great deal of nuance—it misses both the more skeptical approach that many young people are quietly taking towards AI; and it misses that AI is exacerbating, not causing, an educational crisis that is affecting Gen Z more deeply than any particular technology and which necessitates urgent reform.
Many studies and graphs are flying around depicting how AI is rotting our brains and students are using it to cheat on assignments, but there’s another side to that data—and, with it, more hope than you might expect for the mental health of Gen Z. Around 50% of teens, Gen Z, and millennials haven’t used AI, and many of them don’t trust it. A Substack Post report found that, of 2,000 Substack publishers, those under age 45 were less likely to use AI: only 38% used it, while over 50% of those over age 45 used it.
I suspect the reasons for this surprising finding are two-fold. One is, admittedly, a selection bias—the type of young people who write on Substack are more likely to value independent writing and thinking without AI. The other, though, is that Gen Z has seen the wreckage that social media and technology have reaped on our generation. We feel how our social skills and sense of meaning have atrophied as communities migrated to online spaces, and we can see the way AI could accelerate those harms.
Those over 45 have been around the block. They’ve sharpened social skills and professional ones and, at this stage in their career, may prioritize the efficiency that AI offers. Meanwhile, Gen Z is entering the workforce for the first time and more of us than you might think recognize that the AI genie will not go back into the bottle. If we use AI shortcuts now, we’ll likely never go back and learn to write well and think through problems on our own from beginning to end.
So, Gen Z tends to divide up into three distinct groups, each with their own approach to AI and each—like in a Choose Your Own Adventure story—with a life path that follows from that decision. There are, first, those in Gen Z who don’t use AI at all; then, there are those who use AI prolifically, above all to cheat on assignments, in job interviews, etc.; and there are those who use AI only when useful and appropriate, such as to ideate or rework their resume. I happen to belong to the group of Gen Zers who don’t use AI at all, but many of my friends use it to help them edit resumes and portfolios, and that seems reasonable to me.
The better question, though, is why one group would choose one approach to AI as opposed to another, and this, to me, seems related to a deeper issue about how we feel our educational institutions have prepared (or misprepared) us for life in the adult world. If it might be a relief to readers to hear that we are not all hopelessly addicted to AI, our attitude towards our education may be more of a bitter pill to take.
The truth is that, when people complain about the “Gen Z stare,” “quiet cracking,” and Gen Z being difficult to work with, those issues started long before the workplace. We went through school feeling like we were being taught one set of rules that applied to our pedagogy and another that belonged to the actual world and workforce. All my life I’ve surrounded myself with ambitious people, but I noticed that their ambitions often didn’t align with the hoops they were expected to jump through. One thing I noticed about my friends in high school and college is that they were always half-assing assignments and quizzes so they could do something that they felt mattered. They were exhausted. They might sleep through math class so they could teach underprivileged children robotics or skip meetings so they could build their nonprofits. In that environment, it seemed very natural to look for shortcuts—to, for instance, have AI summarize their assigned reading so they could instead read a book by their upcoming podcast guest. Teachers saw us coming in tired with half-baked work and lowered their expectations accordingly—AI became a marriage of convenience for everybody, with students using it in situations (like their coursework) where they could skirt the rules easily and felt it didn’t matter, and then the AI work was just believable enough to get them As and to open the doors to the elite colleges that could secure their future.
I went to college to learn, but the same dynamics repeated themselves there. In my classes, I was often left unchallenged. At one point, I worked three part-time jobs and ran three student organizations alongside the maximum number of credit hours. I wouldn’t have done all of that if my classes occupied and challenged me appropriately. I can easily imagine using AI to write a seven-page paper on Aristotle because no one ever tried to convince me that it mattered. I was bored by it; professors didn’t emphasize that the essays were important to our education or that they were excited to read them, and I knew I could easily spend that time elsewhere, building things in the world that I felt mattered.
Frankly, it’s obvious that many teachers and professors don’t believe their own bullshit anymore. It was an open secret that we weren’t getting a good education in college, and the students were not entirely to blame. Everything became about meeting the next deadline, passing the class, and getting the credits. The professors were often buried in deadlines for their latest “publish or perish” project. I don’t think anyone ever asked if I learned anything. While I didn’t use AI, I frequently reused final papers for multiple classes. One professor even had us assign our own grades, which he said proudly that he never rounded down.
The education system hasn’t measured real learning in a long time. In academia, measures have become goals—or in the case of the professor who had us assign our own grades, measures were thrown out entirely. For generations, students have been telling professors what they want to hear, but it’s been getting worse, and AI is hijacking existing flaws in the education system, allowing students to outsource their time-honored ways of bullshitting class to machines that are optimized to generate bullshit.
Remember: Persuasion is also the home of American Purpose, Francis Fukuyama’s blog, and the Bookstack podcast! You’ll need to opt in to receive all of this content by email: just click “Email preferences” below and slide the relevant toggles.
Email preferences
Many public schools are banning access to AI, but that sort of misses the main point. While it is true that AI may have negative effects on our ability to think clearly and connect with one another, coddling an already-coddled generation—and training us to fear AI, as opposed to using it properly—is just as damaging. I know the impacts of such a culture of fear because I’ve already seen it happen to Gen Z. I see the way older generations look at us and either feel sympathy for what social media has done to us or shame us for not being able to do things they perceive as simple, like making phone calls. As I look down the barrel of the next technological revolution, I see an opportunity to not shame nor coddle the next generation, but invite them into the broader conversation, to basically say: AI is neither a scourge nor a miracle-worker, it may help with certain specific tasks as part of a broader workflow but it is never an excuse to shut off critical reasoning entirely.
AI is disruptive. It’s moving much faster than any of us can keep up with. But it’s also an invitation to get serious about our measures of success. It’s a nudge to create something better. Instead of brainstorming ways to avoid AI use by having students do homework in class or having them handwrite everything (as has become the new retro trend among teachers), what if we created more meaningful and holistic metrics for success? Convince students that their ideas matter; ask them what they think; and listen, not for a correct answer, but an original one. Teach them how to build research projects and business plans from scratch. Ask them to provide feedback and revise their work more than once. Take this as the opportunity to see where the education system is failing and to embark on wholehearted reform.
Most of us did not have a say in whether we came of age in a world with social media and AI, but we have choices and responsibilities regardless as a result of living in that world. One such responsibility is determining what ethical AI use looks like for ourselves and our communities of practice. We can never fully predict the ramifications of AI, but if you think Gen Z is difficult to work with due to the culture we were raised in, now is the chance to set standards around AI earlier than we did with social media. We can learn from my generation—not just get an A and move on, but apply what we’ve discovered and revise our assumptions about what it means to live and work well with AI. We really feel that the education system hasn’t worked for us. AI may not have helped, but it’s not the essence of the problem.
Clare Ashcraft writes The Mestiza where she makes observations about identity, psychology, and culture. She is a proud Ohioan.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
Just so you know: In addition to this column, Francis Fukuyama also writes for the main Persuasion mailing list. That means that if you don’t subscribe to Persuasion, you might miss out on receiving some of his essays!
Click here to manage your account and toggle on the button to receive emails for Persuasion. And as ever, thank you for supporting our community.
Subscribe now
Email preferences
One of the main threads running through the first eight months of the second Trump presidency is the question of the limits of executive power—that is, the degree to which Congress and the courts can place limits on the power of the president. This is the issue at the center of Donald Trump’s efforts to have the Justice Department indict former FBI director James Comey. There is a longstanding tradition of prosecutorial independence in the federal justice system, which holds that the Justice Department should exercise independent judgment when filing indictments, applying the law and not following the directives of the political branches. This rule is simply a normative one, however, and it has never been tested to date by such an overt intervention into the judicial process. The existing U.S. attorney for the Eastern District of Virginia, Erik Siebert, was a seasoned prosecutor who believed there was insufficient evidence to indict Comey, and he resigned in the face of President Trump’s demand that his department act. He was replaced by a compliant prosecutor, Lindsey Halligan, who agreed to file the charges just under the statute of limitations deadline.
This is but one example of Trump’s efforts to stretch the powers of the office of the president. Since virtually the day he was inaugurated, he has been firing officials he didn’t like or deemed insufficiently loyal to him, in clear violation of existing laws set by Congress that limit his removal power. There are, for example, any number of officials who can only be fired “for cause,” meaning that the president needs to articulate a clear violation and give the official time to respond. Trump has nonetheless gone ahead and fired inspectors general, members of multimember boards of Federal agencies like the NLRB, MSPB, and FTC, and is seeking the removal of a member of the Federal Reserve Board’s board of governors, Lisa Cook.
“For cause” officials are protected by statutes passed by Congress, and a 1935 Supreme Court decision, Humphrey’s Executor, affirmed the right of Congress to make such rules. But legal conservatives have long opposed that precedent under a doctrine of the “unitary executive,” which holds that the president alone is empowered to make such decisions. The Roberts court has upheld Trump’s right to fire many existing officials, and recently announced that it would be reviewing the underlying precedent by the end of the year. We can assume that the Court’s conservative majority will at that point formally lay Humphrey’s Executor to rest.
It would be useful to lay out the broader arguments for and against the theory of a unitary executive, since this is likely to become the law of the land soon. The first argument in favor is textual: proponents note the language of Article II of the Constitution, which states: “The executive Power shall be vested in a President of the United States.” They argue that these powers are plenary, and cannot be diluted by another branch. A president with plenary powers, they argue, is more accountable, since there is one singular point of responsibility for executive actions. A president can be voted out of office if citizens don’t like his decisions; not so with an executive branch with multiple independent agencies. Distributed executive power may have practical consequences, moreover, if the different parts of the executive don’t agree with one another. In an extreme but illustrative example, one would not want multiple parts of the executive branch debating one another in the event of a nuclear attack, or taking time to come up with a unified response.
I’m not going to debate the originalist issues here, like the implications of the “Decision of 1789.” I’m not a constitutional lawyer, or a lawyer at all; suffice to say that there has been a debate among legal scholars on the Founders’ actual intent. My argument is rather a practical one from the standpoint of good governance, which has to do with the importance of delegation.
All hierarchical organizations, from companies and clubs to armies and the U.S. government, are structured as a series of what economists call principal-agent relationships. The theory is that the principal in any organization has the legitimate authority to make decisions, and that those decisions are implemented by the agents below the principal. The economists who devised this theory argue that organizational dysfunction, manifested in phenomena like corruption, occurs when the agents follow their own narrow interests, rather than obeying the mandates of the principal. The problem then is to align the incentives of the agents with those of the principal. There are many ways of doing this: the principal can write detailed ex ante rules to constrain the agent’s behavior; he/she can demand ex post accountability for agents after the fact; and most importantly, the principal can exercise appointment and removal power over the agents to enforce proper behavior.
Applying this theory to the present case, President Trump as the elected president should have the ability to remove a prosecutor who is not following his dictates.
The problem is that this version of principal-agent theory is woefully inadequate to explain the behavior of most real-world organizations. The Nobel laureate economist Herbert Simon pointed out long ago that in many organizations, authority travels in the opposite direction, from the agents to the principals. The reason for this is that the agents often have much more expertise, skills, and knowledge than the principals, and are in a position to act more swiftly and effectively because they are embedded in the local context.
In other words, all effective organizations have to delegate authority to lower echelons of agents. There are many real-world examples of the importance of delegation, from just-in-time manufacturing to the performance of armies that empower junior officers to take risks and decisions on their own. Indeed, I would say that determining the appropriate degree of delegation is the central issue, both for organizations in general, and for the U.S. government in particular.
We see this most clearly in the case of the U.S. Federal Reserve. The Fed has a mandate to both control inflation, and to do so while maximizing employment. This is an incredibly complex assignment, particularly in a period when underlying forces are pushing inflation higher and employment lower at the same time. The Fed governors and their staff are among the most highly-trained specialists in the world, stocked with PhD economists and people with broad experience in financial markets. They respond to data, and though they make mistakes, it is hard to think of an alternative group to whom such awesome responsibility should be granted.
But the same is true of federal prosecutors. The U.S. attorneys bringing federal charges against wrongdoers are for the most part seasoned professionals who have broad experience in criminal law. They have the judgment to know when not to take on a case, because they do not want to waste court resources or prosecute someone likely to be found innocent.
Subscribe now
The U.S. Constitution separates powers among the three branches. But for the last 150 years at least, it has also separated powers within the executive branch. The first U.S. regulatory body, the Interstate Commerce Commission (ICC), was put in place to regulate the revolutionary technology of the late 19th century, the railroads. Congress saw fit to put it under the direction of a multi-member commission that was by statute balanced for partisan representation. Since commissioners in the ICC and other similar bodies like the SEC, FCC or FTC served for staggered fixed terms, their policy orientation did not necessarily change quickly in response to an election which brought a new administration to power.
Generally speaking, these Congressional intrusions into executive authority have had the objective of protecting certain realms of executive decision-making from short-term politicization. That is, we don’t want all government decisions to be immediately accountable to voters, who do not necessarily have a good long-term understanding of their own self-interest. Almost everyone would prefer lower interest rates, but may not understand the downstream inflationary consequences of such a policy.
This is not to say that the status quo of Congressional regulation is always desirable. There are some federal agencies that do need to be made more responsive to the voters’ will, and there has been a long-standing debate among public administration specialists about whether multi-member commissions work as well as those with a single head of agency. But there is no question that the federal government needs to delegate authority to officials with the knowledge, skills, and expertise needed to run a complex modern state, and that Congress ought to be able to play a role in protecting that delegated authority.
Given its behavior up to now, it seems likely that the Supreme Court will soon eliminate most existing constraints on the president’s power to remove officials and control the behavior of the executive branch down to the most minute level. I’ve written before about how this opens the door to the kind of corruption that we saw under the 19th century patronage system. But there is another issue as well that has to do with separation of powers.
Principal-agent theory begins with the premise that the principal is always right, and has the legitimate authority to make decisions. As I argued above, that authority is typically constrained by the need for expertise. But there is another issue as well, which is that the principal may not always be right: though legitimately elected, he may be incompetent, corrupt, ignorant, prejudiced, and hugely self-interested. In such a case, the agents under him may actually serve an important function in not carrying out his mandates, and acting as guardrails against the abuse of power.
Subscribe now
What is very strange about the Supreme Court’s 2024 presidential immunity decision was the way that it abstracted from the actual situation the United States faced in the Trump presidency. The conservative justices supporting the decision argued that future presidents needed to be given broad discretionary power to make decisions on behalf of the country as a whole, assuming they would be wise and public spirited. This abstracted from the fact that the ex-president in front of them had gone rogue, seeking to overturn a legitimate election and denigrating the entire electoral process on which American democracy depended. Yes, there may be good future presidents who need to be shielded from unfair accountability, but there are also times when citizens need to be protected from the president himself.
This is the ultimate justification for dividing powers not just between the branches of government, but within the executive branch itself. The modern president is so powerful that he can bend a host of social institutions to his will. No earlier president has ever arbitrarily withheld research money from universities, or threatened law firms that their attorneys would not be allowed inside federal buildings. Needless to say, no prior president has ever openly sought to use the Justice Department to punish his perceived enemies. A unitary executive would give a single individual complete control over these enormous powers, and the president sitting in front of us has openly announced his intention to use those powers for his own purposes.
I hope the Supreme Court is paying attention.
Francis Fukuyama is the Olivier Nomellini Senior Fellow at Stanford University. His latest book is Liberalism and Its Discontents. He is also the author of the “Frankly Fukuyama” column, carried forward from American Purpose, at Persuasion.
Follow Persuasion on X, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
Katie Herzog is the co-host of the podcast Blocked and Reported and the author of Drink Your Way Sober: The Science-Based Method to Break Free from Alcohol.
In this week’s conversation, Yascha Mounk and Katie Herzog discuss Katie’s experience of alcoholism, the weaknesses of Alcoholics Anonymous, and why quitting cold turkey may not be the only way to overcome alcoholism.
This conversation has been condensed and lightly edited for clarity.
Yascha Mounk: I’ve known your writing for a long time. I listen to your podcast with Jesse Singal often. You have a new book out that is personal in a different kind of way. It tells the story of you becoming an alcoholic and managing to move beyond that in a way that’s perhaps somewhat surprising to people. How, in your case, do you think the alcoholism started? How do you go from drinking socially, hanging out with friends, to suddenly it becoming a problem?
Herzog: It happened gradually. I started drinking young. I’m not a Gen Z-er, so it was normal in my milieu to start drinking in high school, or younger. I had my first drinks when I was in middle school. It was normal. You were weaned on wine. Starting young is one of the risk factors for developing alcohol use disorder or alcoholism. People sometimes bristle at the term, but I don’t have a problem with it. Another factor is a family history of alcohol abuse.
I come from Irish Catholic people, so there’s some alcoholism in my blood. There are other risk factors. Trauma is one of them. I don’t have any trauma, at least that I’m willing to admit in public. Although I go with the old-school definition of the term trauma. I’m not talking about people yelling at me online and things like that.
Basically these three risk factors, I hit each one of those. While I drank a lot in college, it was very social, but it was also somewhat anti-social. When I was 19 or 20 years old, I loved nothing more than to sit on my porch by myself, reading The New Yorker and drinking alone. I loved it. If I could spend my life doing that without the negative side effects, I probably would still be doing that. It was a fantastic way to live.
My subscription has lapsed. I also no longer drink. Losing The New Yorker might have been more painful. As I aged, I really aged into bar culture. By the time I was in my early 20s, I was drinking at four o’clock in the afternoon at a bar with mostly older men, like retired and out-of-work men and some women as well. Alcoholism is not really sex-based.
As my friends, my peers who were my own age, naturally aged out of that pattern of behavior, I never did. I was hoping that I would. Most people do. There’s a phenomenon you don’t hear about much in recovery centers or in AA meetings, but it’s called natural recovery. Most people who binge drink in college just age out of it as life changes. Life gets in the way of their drinking, and so they stop.
Mounk: I want to get into what actually defines alcoholism, or alcohol use disorder, in your mind. One definition, or part of the definition, that seems intuitive to me is that it becomes a problem when it distracts from positive elements in your life. When you’re in college, hopefully you’re reading your assigned texts and doing other things.
A lot of the culture of college, and part of the point of college, is to meet friends, create a social network, and hang out with each other. If in college there are three or four nights a week where you’re out with people drinking, that doesn’t necessarily distract from what those years are supposed to be for. It certainly doesn’t stand out relative to your peers. So I guess the question is, as you’re going from that to a different life circumstance, do you continue to engage in this behavior or change behavior?
Herzog: I not only continued to engage in it, my drinking accelerated greatly. Drinking did get in the way of my college experience somewhat, in the choices that I made. I never took morning classes. I know that’s something many students do, but even something like that—choosing my major based on having noon lectures rather than 8 a.m. lectures—shows how it influenced all of my choices.
Thanks for reading! The best way to make sure that you don’t miss any of these conversations is to subscribe to The Good Fight on your favorite podcast app.
If you are already a paying subscriber to Persuasion or Yascha Mounk’s Substack, this will give you ad-free access to the full conversation, plus all full episodes and bonus episodes we have in the works! If you aren’t, you can set up the free, limited version of the feed—or, better still, support the podcast by becoming a subscriber today!
Subscribe now
Set Up Podcast
And if you are having a problem setting up the full podcast feed on a third-party app, please email our podcast team at [email protected]
It also affected where I lived. I knew that I needed to live within walking distance of bars because otherwise—this was before Uber—I would be at risk of drinking and driving, which I knew was going to be an issue. So it really influenced both large and small decisions I made in the trajectory of my life.
After college, I also dropped out for a while, in part because of my drinking and because of things that happened while I was drinking.
Mounk: Before we get to what happened next, when we go back to that time and we look at 21-year-old Katie, who’s obviously very smart and reflective, suppose somebody says, hey, I know you pretty well. I feel like perhaps your drinking is a problem. What, at the time, would you have said?
Was there some part of you that knew that and believed it? Was there no part that knew and believed it? What did you admit to yourself, and what would you have admitted to a close friend at the time?
Herzog: At 21, no. At 21, I had no idea that my drinking was anything out of the ordinary. In fact, my drinking was not out of the ordinary at 21. I lived in very boozy towns and very boozy communities. I was completely normal. The difference is that I just never stopped. As we graduated and moved away and my friends got careers, families, and homes, I was stuck in that pattern of behavior.
Even then, my behavior wasn’t abnormal because the community I was in was still very boozy. We weren’t 21; we were 25, 27, 30, 35, 50. Still, I was a bar person. I hung out in bars. Everyone around me was the same.
Mounk: When do you think most of your peers started to grow out of this behavior while you didn’t? What, to you, were the warning signs that this thing which seemed adaptive and fun was actually holding you back in serious ways?
Herzog: That also was gradual. My peers from college were a little more ambitious. After college, most of them moved away and started jobs or went to grad school. I lost touch with many of them, and most really did get their lives together and started drinking normally, in an adult way. I didn’t.
After college, I moved to Oregon and lived there for a couple of years, working in coffee shops and places like that. My friends were musicians and artists, and within that community what I was doing was completely normal.
My first realization that I had a problem was when I was about 24 and going through my first really serious breakup. The short version is that I cheated on my girlfriend at the time, who I lived with in a serious relationship, and treated her very poorly. She found out because this was the MySpace era, and she saw a MySpace message.
Mounk: Wait, how is MySpace related to finding out?
Herzog: She read a MySpace message where I was confessing this behavior to a drinking friend of mine. She came to the bakery where I had just started working, and she said, you fucking liar, and slapped me in the face. Within the span of a day, I lost my girlfriend, my job, and my house. That was my first “there’s a problem” moment, not just because my life had fallen apart, but because I was confronted with the fact that I had done something truly wrong to someone I truly loved.
Subscribe now
So I asked myself: why did I do this? What was wrong with me? As I saw it, my options were that I was either a bad person, even a sociopath, or I had a drinking problem, because my behavior was connected to drinking. In some ways, I would rather have been a sociopath, because sociopaths don’t have to quit drinking, and the last thing I wanted was to quit drinking. That was my first real “aha” moment.
Mounk: Wait, is that a kind of poignant retrospective gloss, or is that something you thought of at the time? Did you really think, it’s either this or that, and hopefully I’m a sociopath? I believe I’m neither of those two things, but to think of yourself as a sociopath seems like such a radical self-indictment.
Herzog: The thing about addiction is that it’s such a powerful force that I wanted any excuse to continue to drink. Facing the realization that I had a really serious drinking problem was absolutely devastating and destabilizing in a way I have not experienced since. This was almost 20 years ago.
After that realization, I went to a psychiatrist. At that moment, I knew, but I wanted confirmation. I didn’t want to be diagnosed with sociopathy, so I went to a psychiatrist, and he diagnosed me as bipolar I. This was a complete misdiagnosis. I’m not bipolar. Bipolar I is the severe form—or the “fun” one, however you look at it. It was a misdiagnosis. I had a drinking problem, but from the outside my behavior looked bipolar.
Mounk: Because you had one personality when you were sober and another when you were drunk?
Herzog: I engaged in really high-risk behaviors when I was drinking, and I was depressed because alcohol is a depressant. It wreaks havoc on your brain. I did have some of the symptoms of a serious mental illness, but it was a complete misdiagnosis. Probably part of me was also hiding what I was doing because I didn’t want him to tell me, you have to quit drinking.
Instead, he did tell me I should quit drinking, but he also prescribed me a mood stabilizer, which I took for a few months and then stopped. It had no long-term or even short-term effects on me.
After that first realization, that terrible “aha” moment, I lived for the next decade and a half with the knowledge that I was an alcoholic. There’s a term that describes this really well: vacillating denial. Sometimes I knew I was a drunk—like late at night when I woke up at three o’clock in the morning with night sweats, my head pounding, and radiating shame. At five o’clock the next day, when I wanted to drink, I wasn’t thinking, I’m an alcoholic. I was thinking, I just want to drink.
I was able to live with this cognitive dissonance—knowing sometimes that I had a drinking problem and needed to quit, but also being so unable to quit that I ignored it for years. I accepted the truth, but I continued to drink despite it. Because I lived in heavy-drinking communities, this was completely normal. I was surrounded by alcoholics.
Mounk: One of the things that strikes me in my personal experience with friends and acquaintances who have problems with alcohol or drug abuse is that there are certain communities that valorize those kinds of behaviors. The straightforward example is that this is a really fun person—always up for going out, always a laugh, always a good time to be around.
Subscribe now
In more extreme cases, it can be artists who are valorized partly for their extreme behavior and their personalities, which are actually rooted in drug abuse. I worry about the cultures of those communities because they encourage that behavior. Of course, it also works the other way around: people with those problems are drawn to those communities because the communities give cover for that behavior.
Herzog: Absolutely. Before we started recording, you said something about me being an introvert. I am an introvert; I admit that fully. When I was 25 years old, nobody who knew me would have called me an introvert. I was the life of the party. I was wild, always ready to go out, fun, unpredictable, and completely unrecognizable compared to my life now.
I went out, I went to bars, I partied every day. I did this for years of my life. I was very social and had a lot of friends, and all of that was connected to drinking.
Mounk: Do you think that introversion is connected to that? I don’t mean in general, because it would surprise me if introverts were more likely to be alcoholics. If anything, perhaps extroverts like me might be more likely to end up as alcoholics because being in more social situations makes it easier to drink too much.
But in your particular case, perhaps there’s a link between being introverted while seeking social stimulation and friendship, and alcohol serving as a lubricant to help you sustain that social life.
Herzog: Yeah, I think that was a huge part of it. The thing about it is that so much of my socialization happened from a young age under this cloud of alcohol that, when I eventually quit drinking—or even when I first knew I needed to quit—one of the most terrifying things was the belief that it would be impossible for me to be social.
I genuinely thought it would be impossible to have a good time or any fun at all if I wasn’t drinking, because I had literally never done it before. Every part of my life was connected to alcohol. The idea of going to a dinner party, a party, a show, or anything without alcohol was unthinkable.
I have realized now that it’s fine. People do it every day. Not everyone needs to get loaded to have a good time. It’s actually fine. But at the time it was completely unimaginable.
Mounk: You were talking earlier about this term vacillating denial. I think you were describing mostly how it felt within your own mind. During this period of vacillating denial, were there times when you admitted what you saw as a problem to confidants, family members, or friends, or was it just within your own mind? Did you have moments of thinking, I actually have this problem, but refusing to say it to others?
There must have been moments when friends, family members, or romantic partners told you, I think you have a problem. How did you respond to that?
Herzog: It never happened. Not a single time. I was very good at hiding my drinking—so good that many people I was close to didn’t know I had a drinking problem.
Mounk: Were you good at hiding the fact of drinking, at hiding how the drinking affected you, at hiding the need to drink—or all of those things?
Herzog: All of it. During my heaviest drinking years, I had conversations with friends who, in retrospect, obviously had drinking problems. Many of them have since quit drinking. We would have these conversations, but they were always cloaked in humor, and no one took them seriously.
Subscribe now
On the rare occasions when someone in my friend group did get sober, I was appalled. I thought of it as a betrayal. I was very judgmental of people who didn’t drink, didn’t understand them, and didn’t want to be around them. That attitude was fairly normal within my community.
Mounk: It is interesting because we think about the stigma of alcoholism, and that does exist. But there is also the stigma of being a teetotaler. I drank a good amount when I was younger, I have to admit. At this point, I really don’t drink very much.
There are social situations where you’re meeting somebody at 6 p.m., and you don’t really want to have a drink, but it feels a little rude not to. It can feel like you’re not committing to that social occasion, as if their company is not worth the small cost of perhaps sleeping a little less well or being slightly hungover the next day. Sometimes I find it difficult not to order a drink in those circumstances purely because of this social stigma against not drinking.
Herzog: You want to do what the person next to you is doing. If you go to dinner with a group of people and one person doesn’t eat, it seems strange. Alcohol works in that same way as well.
Mounk: It’s weird because, if I go to dinner with somebody and they’re sanctimonious about being a vegan, I find it annoying. But if I go to dinner with someone, I order steak, and they order a salad, I don’t care.
Somehow it does feel different if I meet somebody at 6 p.m., they order a beer, and I say, could I have sparkling water? It doesn’t feel like disrespect toward them, but it does feel like disrespect toward the social situation.
Herzog: I think it can also be a signifier that you don’t drink for a reason, which can lead to a follow-up question: Why don’t you drink? Most people probably aren’t going to ask, but they might wonder.
Mounk: Yeah, it’s funny. That’s not the fear I have. I think the fear I have is that it communicates, I want to be done with this meeting quickly, or, I don’t think you’re sufficiently important to me. Something like that.
You were very good at not letting people notice. Part of that, of course, is that you had your life together in a way that some people who are dependent on alcohol don’t. Many people aren’t able to hide it, but you were. It seems that in part this was because you continued to be professionally successful and kept a relationship during that period.
Alcohol dependency negatively impacted your life, but it didn’t completely derail it. Why do you think that is? What stood between you and the people whose lives were completely destroyed by the disease?
Herzog: Well, yes. It did destroy my life for many years. Until I was about 30, I worked mostly minimum-wage, entry-level jobs. This was long enough ago that rent was still cheap in most American cities, so I could still pay my rent even if I was working part-time or minimum-wage jobs, or if I had roommates. In the early 2000s, rent was cheap enough that I could live in a hedonistic and irresponsible way and still manage to pay rent. That’s how I spent my 20s. Honestly, my 20s are very much a blur.
When I was 30, I was working as a copy editor at a small academic publishing house, again part-time. I left that job to get an unpaid internship, and that was really the beginning of trying to get my life together. I did have a prolonged adolescence, but after that I started on a career path. I made better decisions, and I wasn’t drinking every night. I also kept moving. I lived in a number of different places, and that became one of my coping mechanisms. If my drinking accelerated to the point where it became very problematic, I would leave—leave the community, leave the environment. That actually worked fairly well for a while, until eventually I would adopt the same habits in a new place. So until my 30s, I was just like every other loser at the bar. Then I started to get my life together.
Mounk: The time you got your shit together was not directly correlated with the time when you actually quit drinking, right? That came later.
Herzog: No. I moderated my drinking for a couple of years. I got this internship, moved to a new city, and got my first job in media working in public radio. I would drink only on the weekends. I was very regimented about when I allowed myself to drink. On weekends I would go back to my old town, sleep on my friend’s futon, and party all weekend. Then I would leave.
Later I moved across the country to Seattle, and similar patterns developed. For the most part, I kept it to weekends.
For the book, I interviewed many problem drinkers. We often have the stereotype of the alcoholic as someone living under a bridge who beats his wife and ends up getting DUIs and going to prison. Although these days, most people living under bridges are probably more likely fentanyl addicts than drinking wine out of a paper bag.
It may surprise nobody, but it is very possible to be an alcoholic and still have a career and a family. It is still detrimental. Especially if you start drinking young, you can become emotionally and mentally stuck in your teens or twenties. That definitely happens. But people manage to live bifurcated lives: hungover at work but still doing the job, then drinking again after work. It’s common.
Mounk: It’s more the Mad Men-style alcoholism rather than the fully out-of-control kind.
Herzog: Yeah. Functional alcoholic. There’s a term for it for a reason.
Mounk: There are periods when it redirects your life and periods when you have your life more or less together, even though it continues to negatively impact you. That makes it sound like you could easily have gone through your whole life without either becoming the person drinking wine out of a bag under a bridge or really getting the problem under control and grappling with the fact that you have a problem.
After these years of vacillating acceptance and denial—of better years and worse years—what made you confront it in a different way?
Herzog: During COVID I was laid off. I was working at The Stranger, Seattle’s alt-weekly, and I took a furlough really early on. I basically volunteered as soon as COVID hit. The newspaper was a free paper, funded entirely by ads for events that were no longer happening. It was obvious pretty quickly that things were going to go downhill. I volunteered to take a furlough, thinking, not coming into work and seeing these assholes for a few months? I’ll take it.
I went on furlough and started collecting unemployment, which I loved. It was maybe my fifth or sixth time on unemployment. I’ve never been afraid to take our meager social safety net up on its offerings. My wife is a nurse, so I was home alone a lot while she was working.
Jesse and I had started our podcast, but at the beginning it didn’t take much work. There was so much happening in the culture that we would just get on the mic and talk for an hour about whatever we were seeing. Now the show has structure, but at the time it required little prep. I had a ton of time on my hands, and I drank.
I started drinking during the day. The time I started drinking got earlier and earlier, until eventually I was drinking in the morning. This was completely in secret. Nobody knew about it, including my wife. I did this for two years. Every time my wife went to work, I got loaded and spent the day sitting on my porch, drinking and scrolling, watching from the sidelines as it felt like the world was falling apart—not just because of COVID, but because of everything happening.
I felt terrible, looked terrible, and was wracked with guilt and shame constantly. I started having health effects: waking up at three in the morning with the shakes and night sweats, a pounding headache every morning, telling myself the next day would be different, that I wouldn’t drink. By 11 a.m. or noon, a switch flipped in my brain and, almost without being conscious of it, I was out the door to buy booze.
I crossed lines I had once promised myself I never would. I told myself I wouldn’t drink in the morning, wouldn’t drink and drive, wouldn’t hide it. Pretty soon I had crossed all of those lines. The amount I was drinking would have killed me eventually. At the time, I was drinking about six Tall Boys—six pints of cheap beer—a day. That was less than in my heaviest drinking days in my 20s, when I could drink twice that, but still a lot. It was a huge amount of liquid and empty calories. I looked awful, felt awful, and the worst part was the amount of brain space it took up.
The best way I can explain it to people who haven’t been addicted to something is this: the feeling when you’re newly in love or have a crush, especially when you’re young, and it’s so overwhelming that it’s all you can think about. Or when your mom has a cancer diagnosis and that’s all you can think about. That’s what alcohol was for me for so long. It consumed my mind.
I had tried to get sober over the years—AA, group therapy, individual therapy, CBT, even CBD. Nothing stuck long-term. One of the hardest things was the message I absorbed about sobriety: once you’re an alcoholic, you have to quit forever. They say one day at a time, but everyone knows it means forever. I couldn’t conceptualize a life without alcohol.
That was the mental barrier: knowing the only way out was to give up the thing that was destroying me, but also the thing I was completely dependent on, that I hated but also loved. That kept me drinking for years, in secret, more and more, until it really was taking over my life in destructive ways.
I never beat my wife or anything like that, but I was distracted all the time, a lousy partner, lying to her. Small things, like if she was coming home from work on the ferry, I couldn’t pick her up because I was drunk. I’d say, sorry, I’ve got a meeting, when the truth was I was drunk. She had to take the bus.
Mounk: You were making excuses for not going to pick her up. It wasn’t that you didn’t want to do that act of service for someone you loved; it was that it would have been dangerous for you to do so.
Herzog: I would have killed her if I had picked her up. I was making all of my decisions around drinking, scheduling my life around drinking—for example, with Jesse and the podcast. I still am a control freak, but I was very regimented about when I was willing to record, because if we recorded in the afternoon I would be loaded and I didn’t want to record the podcast drunk. I was completely inflexible about my routine because I knew I was going to be drinking; I needed those hours to do it. I knew nothing could be scheduled during that time. Don’t tell Jesse, by the way.
Mounk: He never listens to any podcasts, so it’s fine. He certainly wouldn’t listen to mine. One of the interesting things about your book is that you combine your personal story with an argument about how we think about addiction and overcoming addiction.
I want to plant the seed for this part of the conversation. You mentioned that one of the things that kept you from seeking help was the belief that if you stopped drinking, it would be forever, and you couldn’t conceive of a life without alcohol. Tell us a little about the dominant framework for what somebody in your situation during COVID should be doing, and the role that AA—Alcoholics Anonymous—plays in that, both in terms of the medical establishment and in our culture more broadly.
Herzog: Yeah, this is changing. The first time I went to AA and approached a professional about my drinking, I was 24. This was in 2007. At the time, and for many years before and since, there were basically only a couple of treatment modalities for somebody presenting with a drinking problem: rehab and 12-step programs. That was pretty much it.
When I was writing the book, I learned the history of AA, and it’s fascinating. AA developed in the 1930s as an organic peer-support group created by two men. One reason AA became so dominant in the culture was an early member, a woman named Marty Mann. She was a PR expert, very connected in publishing and Hollywood. AA saved her life, and she made it her mission to preach its gospel throughout the country. At the time, AA was not hospitable to women, but she became an early member and was instrumental in spreading it. She was incredibly successful. AA has saved many lives. I don’t knock it—it really works for a lot of people.
Mounk: For people who don’t know much about it, what’s the basic model of AA?
Herzog: It’s a peer-support group. You go to a meeting—at open meetings anyone is welcome—and you basically listen. People sit in a circle around a table, and in a regimented way each person talks about what’s going on. There isn’t much back-and-forth; that’s called crosstalk, and it isn’t encouraged or even allowed.
At the beginning of the meeting, there’s a process: everyone speaks the same words, following traditions and rituals. When it’s time to speak, you say, my name is Katie, and I’m an alcoholic. There are norms within the group. Many people get a sponsor, the person you call when you’re having problems. You work the 12 steps, starting with admitting that you are powerless and ending with helping other alcoholics. If you do it right, you don’t just sit at meetings; you work the steps and become part of the community.
Peer-support groups existed long before AA, but AA is the most successful in terms of membership. For nearly a century, it has been the dominant paradigm for getting sober, including within rehab. One of its great benefits is that it’s free. There are meetings in every city and town in America and many around the world, with Zoom meetings now common since COVID. Anyone who thinks they might have a problem can go, and there is power in being with other addicts who understand what you’re going through.
AA meetings can also be fun. Addicts are often great storytellers. It can feel like an open mic—you’ll hear great stories, sometimes the same ones repeated. But it’s a community, and for people in the midst of addiction, having accountability, support, and a place to go can be valuable. This model has spread widely, and many rehab programs rely on 12-step principles. The irony is that while rehab can be incredibly expensive, the therapy provided is sometimes the same thing you could get for free at your local AA meeting.
There are misconceptions about AA. Some think it preaches that alcoholism is a personal failure, but it’s more nuanced. Bill Wilson, one of the founders, was open-minded; he even experimented with psychedelics in the 1950s. Still, AA has become so dominant that courts can mandate it for DUIs, professional licensing boards can require it, and halfway houses often include it in recovery. I’ve been to meetings recently where pharmacists were there not voluntarily but because they had to fulfill requirements to keep their licenses. AA has become the dominant framework for recovery. The problem is, it doesn’t work for everybody.
Mounk: Why doesn’t it work for everybody? Part of the assumption in AA is that the goal is not to drink at all, not to get to a healthy relationship with alcohol where you can toast your partner’s birthday with a glass of champagne. The idea is that people will have relapses and there’s not going to be judgment about that, but the goal is not to drink at all.
Herzog: Yes. Within AA, any drink is considered a slip, and you start the counter back to zero. You could have been 12 years sober, and if you have a sip of champagne at your daughter’s wedding, you start the clock over. I personally find that not very helpful.
It’s an all-or-nothing attitude, and anything less than perfect is considered a failure. I think that keeps people away from the program, especially people who aren’t totally sure they’re ready. If you know drinking is causing a problem in your life but you’re not ready to quit or can’t conceptualize a world without alcohol, and the one option you’ve been told about is to never have another sip of alcohol again, it’s going to keep some people drinking longer than they should. That was my experience. If the two options were total abstinence or a life of drinking, even though it was causing problems, I was going to choose the life of drinking.
Mounk: Why did that not happen? Clearly you ended up in a situation where it’s not one of those two options, it’s a third option. How did you realize there could be a third option, and how were you able to escape that situation?
Herzog: It was an article in The Atlantic. This article came out in 2015. It was by Gabrielle Glaser and called “The Irrationality of Alcoholics Anonymous.” The article was about the problems with AA, but there was a part that mentioned a treatment called the Sinclair Method.
The Sinclair Method is medication-assisted therapy. It was developed by an American researcher named John David Sinclair. To simplify, he developed a protocol where a person takes an opioid blocker, typically naltrexone. In other places, it might be nalmefene or another opioid blocker. You continue to drink as normal with this opioid blocker in your system. The blocker prevents your brain from getting the endorphin rush and dopamine dump that comes with alcohol.
The idea is that over months, maybe years, you retrain your brain not to enjoy alcohol because your brain isn’t getting the reward. It makes drinking boring.
I read that article in 2018 or so, and I thought, okay, this is what I’ve been looking for. This is the solution to my problems. I went to my doctor and brought the article to her. She had never heard of the Sinclair Method, which is common. Many doctors have heard of naltrexone, and that’s increasingly true as medical education embraces harm reduction and medication-assisted therapy for substance use disorder. But my doctor had not heard of the Sinclair Method.
I said, this is what I want to do. She said, okay, do it. I got the drug and went home. I bought a can of beer from the corner store, took the drug, and felt awful. I had side effects: headache, nausea, depression. I didn’t like it. I choked down the can of beer and thought, okay.
I was so hopeful. This seemed like a miracle cure—the thing that would allow me to keep drinking but diminish my enjoyment of it. For years, one of the things that kept me drinking longer than I should have was the fear of living in a state of unfulfilled desire for the rest of my life. That’s what I thought sobriety would be.
Mounk: Interesting. It would be this constant hankering for the thing you can’t have and this constant self-denial.
Herzog: I was, and still am, very bad at fighting cravings for anything. If I want a cookie, I’m eating a cookie. I don’t have much willpower when it comes to giving in to my own impulses. Definitely, I would be eating the whole bag. So I was looking for a solution that would extinguish not just the drinking itself, but the desire to drink.
That’s what I hoped this would do. The first time I took the drug, I thought, no, I can’t do this. This is not for me. So I put the drug away, and it sat in my medicine cabinet for several years until COVID. Then I really needed something, anything. I thought, I’m going to try this one more time, or I’ll have to fess up to my wife and probably go to rehab. That’s going to fuck up the podcast because I’ll have to leave for 30 days and then do intensive outpatient therapy. It’s going to completely derail my life. So I’ll try this one other thing before I fess up and blow my life up. I did it.
Mounk: Why was the second time different? What was it that the first time around either physically led to the side effects or mentally made you not prepared to deal with them? Why was it that the second time you were able to use this method?
Herzog: I think a big part of it is that I armed myself with information. I read everything I could find about the Sinclair Method. I prepared myself to deal with the side effects. There are simple things to do, like starting out at a quarter of a pill instead of the standard 50 milligrams. Start at 12.5 milligrams and work your way up. I ate a full meal the first time I took it, and I drank a full glass of water. Some people take over-the-counter nausea medications.
One thing that’s interesting about this is that I wasn’t getting this information from my doctor. I was getting it from Facebook, Reddit, and YouTube. My doctor had never heard of this. I was prepared. I knew about the side effects and I was ready to handle them. I also knew this was basically my last chance before I really blew up my life.
I was willing to take some discomfort if it meant I didn’t have to tell my wife, go to rehab, go to AA, and quit drinking. That was the plan. I was going to try this again. I researched it extensively. I talked to people. I kept a Substack where I was writing about this under a pseudonym. I treated it like a journalism project.
Mounk: To ask a simple and perhaps stupid question: if your brain craves the feeling that alcohol gives you, and if you miss that feeling, what stops you from not taking that drug? Why don’t you wake up some mornings intending to continue your recovery, as you intended many times during COVID and earlier, but just as in those earlier instances you wake up thinking, today I’m not going to drink, and then by noon you say, I’m going to drink?
Why in this case is it so hard to say, I’m not going to reach for a bottle of beer, but somehow easier to say, I am going to take this pill?
Herzog: Yeah, compliance is an issue with this drug for sure. There’s a long-acting form called Vivitrol. It’s a shot in the butt you get once a month. For people who have trouble with compliance, that can be a good alternative. For me, I knew the other options were worse, and my life had become, as they say in AA, so unmanageable that this was the last best option.
I was committed to it. Honestly, if I had tried this at 25, I don’t think I would have been able to do it. But I was not 25. I was in my late 30s, early 40s. I wasn’t drinking socially at the time. This was during COVID. I was alone at home, and I had ceased to get much pleasure from alcohol.
The anticipation and fulfilling of the craving were pleasurable, and the first few drinks had elements of pleasure. But by the third drink I was drinking only to try to recapture the feeling of the first drink, or even the feeling before the first drink—the feeling of craving and knowing I was going to fulfill it. That felt good. I knew that if I didn’t stop, it would eventually kill me, and I didn’t want to die.
Compliance can absolutely be an issue with this drug. After talking to a lot of people who have done the Sinclair Method, I learned there are other ways to take naltrexone, but the book is mostly about this particular modality. It is most effective for people who are highly motivated and who also have some social support, who have work. This is not a miracle cure for someone who lives on the streets, whose life is unstable, and who has trouble finding housing, food, or alcohol.
I talked to a guy who works in public health. He said for his patients Vivitrol can be effective in reducing drinking because it’s a shot—if they show up to get the shot, it can work. But for his clients, something that requires you to take a pill, set a timer, wait an hour, and then drink, is not going to work.
Mounk: I guess there is a structural difference as well. The structure of a problem with not drinking is that you have to not drink at 10 a.m., not drink at 11 a.m., not drink at 12 p.m., not drink at 1 p.m. You have to have perfect compliance throughout the day. In the AA framework, every day, messing up once scuppers the whole scheme. With taking this medicine, on a daily basis, you just have to have one moment of self-control.
Herzog: It’s a little more complex than that because it is possible to drink through the naltrexone. The half-life of the drug depends on your metabolism and biology but ranges around five or six hours. It is possible to drink through it. Anecdotally, some forms of alcohol seem to make people more likely to do so, like vodka. Heavy vodka drinkers typically have an easier time drinking through it.
Some people redose after five or six hours. That takes willpower, especially if you’ve been drinking for that long and you’re half loaded. Making the right choice then is hard. By the time people get to a place where they are taking naltrexone and doing the Sinclair Method, in my experience talking to many of them, they are desperate for change. They are ready. They have mostly tried other things and they haven’t worked. If you are desperate for change and the only thing you have to do is take your pill, people do it.
Not everybody. Compliance is an issue. This will not work for people who are half-assed on compliance. The end goal of the Sinclair Method is called “reaching extinction.” This is a term in psychology that means unlearning a behavior. Under the Sinclair Method, it means losing the compulsion to drink. Once you reach extinction, some people, like me, stop drinking entirely. Some people continue to drink moderately. I have talked to people who reached extinction, felt free from the monkey on their back, but continued drinking moderately. They got lax on waiting the full hour. They kept taking the pill but got lax about the timing, and they quickly slid back into alcoholism and had to start the process all over again.
Mounk: Tell us about where you’re at in this journey. You hinted that this worked for you and that it led to the good kind of “extinction” in this case. What did that process look like, and when did you decide not just to tell your wife and many of your friends and family members, but to write a book about it and go public with the whole journey?
Herzog: Yeah, once I started this, I was very regimented about it. One of the reasons I wrote the book is because it’s deceptively simple: you take the pill, wait an hour, and then drink. But there are small changes people can make along the way that increase the likelihood of success. This is not a miracle pill. It does not work for everybody, for a combination of reasons, including biological ones beyond your control.
I started taking the pill. I was regimented about it: take the pill, wait an hour, then drink. One difference between the Sinclair Method and the typical way a doctor might prescribe this drug is that you only take naltrexone when you are drinking. You don’t take it on alcohol-free days. On those days, when your endorphin receptors are unblocked, you want to fill your time with healthy pleasures—eating a good meal, going for a jog, playing with your dog or kids. You want to find new coping mechanisms and retrain your brain to find pleasure in non-alcoholic things.
Over seven months, I tracked my drinking. At the beginning, I was drinking just as much as before. Drinking on naltrexone was less pleasurable. I’ve heard it described as eating four-day-old cake, riding a bike with a flat tire, or drinking with a condom on. You’re still doing the act, but it’s not as fun. You can still get drunk, but you don’t get the euphoric buzz. I wasn’t getting that energetic high. I got the body effects—feeling slurry and sleepy—without the euphoria.
Within the first month, my drinking didn’t change much, but I tracked it and could see it on a chart. Then I started pushing back my start time: instead of noon, I’d start at three, then four. After that I began increasing the number of alcohol-free days each week. That was something I had never done before. I had alcohol-free days when my wife was around, but never when I was left to my own devices.
I started finding things to do that weren’t alcoholic—walking my dog in the woods, paddle boarding, working more. Over seven months I decreased the amount I drank and increased alcohol-free days. Then I challenged myself to 30 days without drinking. It was shockingly easy. Something that had always been difficult suddenly wasn’t. In almost 20 years, I think I had one period of a fully alcohol-free month, and it was a huge challenge. I celebrated it afterward with a bender. This time it was easy—I didn’t even think about alcohol. After seven months I took 30 days off. After two more months I declared myself officially extinct. I reached extinction and never started drinking again. I no longer take the pill because I no longer drink.
Mounk: At this point, if you are at a wedding and there’s a toast with champagne, do you not toast, or do you toast with champagne and have a couple of sips?
Herzog: No, I don’t drink at all. If I wanted to drink, I would need to take a pill. I still have all the risk factors. My life has changed. I have new coping mechanisms, and in some ways I’m a different person. In some ways I’m exactly the same, but my life is very different.
I think my particular biological risk factors make me a high-risk candidate for alcohol use disorder no matter what. I could take a sip of champagne and be fine that day, but it would trigger something in my brain: you like this, right? For me, it’s easier not to drink at all, in part because I don’t like taking naltrexone.
For some people who don’t have side effects, naltrexone is no big deal. They want to continue drinking moderately, and that works for them. I think that’s completely fine. For me, the pleasure I get from drinking on naltrexone is not worth the side effects. The side effects weren’t prohibitive, but I didn’t like them. It’s an endorphin blocker. It doesn’t feel good. So I just don’t drink at all now.
The most surprising thing is that I actually can have a good time without alcohol. I thought that was impossible because I had never done it before, since I started drinking so young. It’s completely fine. I enjoy going to dinner parties and not drinking. I enjoy going to parties and not drinking. It has so little mental pull on me that everybody around me can be drinking and I don’t get jealous, resentful, or cravings. I feel completely free of that mental burden.
In the rest of this week’s conversation, Yascha and Katie discuss cultural attitudes to alcoholism, what to do if you’re concerned about your relationship with drinking, and reaching out to friends and family for support. This part of the conversation is reserved for paying subscribers…
This article is brought to you by American Purpose, the magazine and community founded by Francis Fukuyama in 2020, which is now proudly part of the Persuasion family.
Moldova, a tiny Eastern European country sandwiched between Ukraine and Romania, could soon humiliate Russia in a way that will be difficult for the Kremlin to recover from. If the former Soviet republic of around 2.4 million people stays on the path toward European Union integration, it might have the opportunity to do something few would have ever imagined: to militarily defeat the Russian Federation.
Ever since Russia launched its invasion of Ukraine in 2022, Moldova has been on the brink of destabilization. According to reports, Moscow’s long-term goals in the country include establishing stable pro-Russian influence groups within Moldova’s political and economic elites, as well as fostering a negative attitude toward NATO. The EU, on the other hand, granted Moldova (one of the poorest European nations) candidate status in June 2022, while the authorities in Chisinau have set 2030 as the target year for European Union accession.
Not all Moldovans support the plan. But in 2024, slightly more than 50% of voters in a referendum backed the government-proposed constitutional changes and the country’s commitment to joining the EU. The ballot was held alongside the presidential election, in which pro-EU incumbent Maia Sandu was re-elected with 55% of the vote.
Her victory is primarily attributed to the unprecedented engagement of the Moldovan diaspora in the West. They turned out in large numbers for both the referendum and the election, securing the country’s pro-EU course. The Kremlin, however, questioned Sandu’s victory, accusing the Moldovan authorities of preventing “hundreds of thousands of Moldovans living in Russia” from voting.
Now, ahead of tomorrow’s parliamentary elections, Kremlin spokesman Dmitry Peskov once again accused Chisinau of “disenfranchising a large portion of Moldova’s population by denying them the right to vote.” He was likely referring to claims from Moldova’s breakaway region of Transnistria—officially the Pridnestrovian Moldavian Republic (PMR), a narrow strip of land along the Dniester River bordering Ukraine—that the authorities in Chisinau have sharply reduced the number of polling stations available to Moldovan citizens living in the unrecognized republic for the upcoming elections.
Sandu, on the other hand, blamed Russia for an “attempt to sway the vote,” while Moldovan police arrested at least 74 people over an alleged plot to organize “mass riots” in the country. This suggests an escalation in plans for potential disruptions, while the United States appears not to be paying sufficient attention.
As a result of USAID cuts, independent media outlets, human rights organizations, and civil society initiatives in Moldova have been adversely affected. It is now the EU, rather than the United States, that is investing in the former Soviet republic, aiming to secure its current geopolitical vector. The recent visit to Moldova by French President Emmanuel Macron, German Chancellor Friedrich Merz, and Polish Prime Minister Donald Tusk can be interpreted as a clear show of support for Maia Sandu’s Party of Action and Solidarity (PAS) ahead of the vote.
Last year, likely in an attempt to further divide Moldova, Moscow openly backed Evghenia Gutul, the governor of the country’s pro-Russian autonomous region of Gagauzia. However, last month she was sentenced to seven years in prison on charges related to alleged illegal financing of the opposition Shor Party. The Kremlin is now believed to support the opposition Patriotic Electoral Bloc, although Moscow’s room for maneuver in Moldova is limited by geography.
Subscribe now
Even if pro-Russian forces come to power, Moscow is unlikely to have an effective mechanism to influence policymaking or reverse the nation’s pro-EU trajectory. Sandwiched between EU member Romania in the west, and Ukraine in the east, a potential pro-Russian Moldova would find itself isolated, and would almost certainly sink even deeper into poverty.
As such, it would be in a similar position to Transnistria, which hosts around 1,500 Russian troops who were deployed there as a result of the war that broke out in 1992 following the breakup of the Soviet Union.
Transnistria represents a trump card that Ukraine, Moldova, and the West could use against Moscow at any time. It is estimated that half of its 300,000 residents have Russian citizenship, while others hold Moldovan, Romanian, as well as unrecognized Transnistrian passports. In 2022, the Parliamentary Assembly of the Council of Europe designated the PMR a “territory occupied by Russia,” giving Chisinau the green light to eventually restore its sovereignty over the region.
There are fears in Russia that, in the event of a PAS victory or potential destabilization of Moldova, Ukraine—whether acting alone or alongside some NATO countries—could launch a military incursion into Transnistria. Russian media are already accusing Ukrainian President Volodymyr Zelensky of announcing plans to “destroy the PMR.” Such rhetoric follows Zelensky’s submission of a bill to parliament that would permit units of the Ukrainian Armed Forces to be deployed abroad.
If Kyiv really decides to provide military support to its small neighbor, Russian chances to preserve its de facto control over Transnistria will narrow significantly. In order to reach the PMR, Russian troops would first need to seize the Ukrainian port cities of Mykolaiv and Odessa, which at this point seems extremely improbable. That is why even Russian analysts openly admit that the loss of the PMR would represent a disaster for Moscow.
Ukraine undoubtedly has the military capacity to capture the region. Whether it will do so depends on Kyiv’s political will and approval from the West, including the United States. Although Moldova’s military might not be able to defeat the Russian-backed PMR on its own, it could still reap the benefits of a potential Ukrainian (or a joint Ukrainian-Moldovan) military operation.
Fully aware that such an outcome would represent a strategic defeat for Russia and have an extremely negative impact on its image, the Russian Foreign Intelligence Service is accusing NATO of planning to assist Ukraine in its alleged plans to seize Transnistria. The Kremlin seems to be crafting a narrative to whitewash its potential defeat in the PMR—portraying NATO, rather than Moldova or Ukraine, as the key player in the dismantling of the Russian-backed Pridnestrovian Moldavian Republic.
In reality, in the event of a war, Moldova would emerge as the winner by restoring its sovereignty over Transnistria. The Eastern European country is now at a crossroads. It can push forward with its European future and risk conflict, or face growing instability and outside pressure. What happens after the parliamentary elections will likely shape not only Moldova’s future, but also the balance of power in the region.
Nikola Mikovic is a freelance journalist, researcher, and analyst based in Serbia.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
For the most part, I try to avoid writing pieces which respond to the latest big thing in the headlines about what Donald Trump’s second administration is up to.
This is in part out of a desire for some division of labor. There are lots of people who spend their days writing about Trump’s every move and utterance. They diligently chronicle all of the bad things his administration is up to. Often, I don’t feel like I can add something that isn’t widely known to anybody who has any desire to know.
It is also out of a belief that a constant stream of disdain isn’t what’s going to get the country out of the deep crisis in which it finds itself. The mainstream media has spent the past decade expressing its outrage about everything Trump has done wrong. For the most part, that ire was well-deserved. But all that sound and fury has added up to very little political change, with Trump more powerful today than he has ever been before.
Subscribe now
In the end, the only way to defeat demagogues like Trump is at the ballot box. And while politics involves battle, and part of that battle is to define your adversary, the thing that’s most striking about America today isn’t how popular Trump is; it’s how weak and tired and unpopular his opponents are. So on most days, I feel that I can better serve the goal of defeating Trump by being an in-group critic than by being an out-group detractor.
But I would like to think that my reluctance to cry wolf, or to beat up on Trump every single day of the week, buys me some credibility as an arbiter of when something he does is truly dangerous. Recent attempts to undermine the independence of the judicial system and go after his personal enemies unreservedly qualify. Unlike a lot of other stupid or unseemly things the administration has done, they are a five-alarm fire.
One of the most awesome powers of the modern state is its ability to imprison people who have violated the law. This is why the Founding Fathers were so obsessed with procedures and principles meant to protect citizens from arbitrary arrest. As they recognized, no liberty worth its name is possible when a head of government, even one elected by popular vote, can direct the machinery of the state to punish those who have displeased him. And yet, that is precisely what Donald Trump is attempting to do in the case of James Comey.
Comey is a complex figure.
He loves to present himself as a public servant of unimpeachable civic virtue, always faithful to his higher calling of serving the constitution. And yet, he has, over the last decade—in part by virtue of happenstance, and in part by virtue of his own decisions—found himself at the center of one political firestorm after another. In the process, Comey has earned himself the enmity of both the left (who blamed him for Hillary Clinton’s loss after he took the unusual step of writing to Congress about reopening an investigation into her emails days before the 2016 presidential election) and the right (who blamed him for instigating an investigation into spurious links between Trump and the Kremlin which significantly hamstrung his first presidency).
In a sense, this makes Comey a victim of the high moral standards to which he so ostentatiously holds himself. We live in a moment in which the public is much more willing to forgive depravity than hypocrisy. Politicians like Trump get away with tremendous corruption in part because they never claim to be better than you or me. Public servants like Comey, by contrast, inspire so much mistrust in part because they hold themselves out as paragons of virtue. In our cynical age, this raises the immediate suspicion that they can’t possibly be as perfect as they claim—which, the world being the world, and people being people, usually turns out to be at least somewhat true.
Personally, I feel that we should have greater admiration for somebody who holds himself to high moral standards and ultimately falls a little short than for somebody who proudly disdains the idea that there should be moral constraints on how to act at all. But at some gut level, I do understand why hypocrisy feels singularly galling, and even why this inspires in some people an instinctive dislike of Comey.
Here’s the thing though: Comey’s likability, even his moral status, is not what matters about the news that a grand jury in Virginia has just indicted him.
The published indictment of Comey is so brief and cryptic that it is hard to tell what exactly he stands accused of. According to initial news reports, the accusation may be that Comey perjured himself when telling Congress that he had never authorized his then-deputy, Andrew McCabe, to leak to the Wall Street Journal a confidential conversation about the investigation into Hillary Clinton’s emails. If this is true, the main evidence for this accusation seems to come from McCabe himself, and he is unlikely to be regarded as a credible witness by a jury; after all, an independent investigation about his conduct in this very matter concluded that he had “lacked candor, including under oath, on multiple occasions.”
But what makes this case truly concerning is not the substance of it, remarkably thin though it may turn out to be; it is the extraordinary manner in which the President of the United States openly demanded that the government should go after somebody he sees as an enemy, and used executive power to bring about that outcome.
A few days ago, Trump wrote a remarkable message, which explicitly mentioned Comey, to Pam Bondi, the attorney general, on TruthSocial: “Pam: I have reviewed over 30 statements and posts saying that, essentially, same old story as last time, all talk, no action. Nothing is being done. What about Comey, Adam “Shifty” Schiff, Leticia??? … We can’t delay any longer, it’s killing our reputation and credibility. They impeached me twice, and indicted me (5 times!), OVER NOTHING. JUSTICE MUST BE SERVED, NOW!!! President DJT.”
Subscribe now
These weren’t mere words. Over the last weeks and months, Trump has fired multiple prosecutors when they didn’t prove sufficiently willing to do his political bidding. In fact, the prosecutor who was handling the investigation into Comey was apparently fired after determining that it would be inappropriate to charge him in this matter; it is only after Trump appointed a new prosecutor, who had previously served on his own legal team and does not have any experience in federal law enforcement, that the Justice Department decided to indict.
All in all, the extreme degree to which this prosecution comes at the direct behest of Trump is well-nigh comical:
An inexperienced prosecutor loyal to President Trump, in the job for less than a week, filed criminal charges against one of her boss’s most-reviled opponents. She did so not only at Mr. Trump’s direct command, but also against the urging of both her own subordinates and her predecessor, who had just been fired for raising concerns that there was insufficient evidence to indict.
Trump is not the only political actor to abuse the judicial system to go after his opponents. Indeed, it is surely true that some of his anger is occasioned by the way in which he himself suffered a prosecution that was deeply political.
After Trump left office in 2020, there were a number of actions for which he could credibly have been prosecuted. Most obviously, his demand that Brad Raffensperger, the Secretary of State of Georgia, “find” the votes he needed to win the state was deeply immoral and likely illegal. But Fani Willis, the elected District Attorney in Fulton County, badly bungled the case. In the end, the only criminal prosecution that was allowed to proceed to trial concerned a matter that was far less morally weighty and far more legally dubious.
That case, which revolved around Trump paying hush money to Stormy Daniels, a famous porn star, substantially turned on the argument that Trump should have declared the payments as a campaign contribution. It was brought by Alvin Bragg, who had been elected to the position of Manhattan District Attorney by one of the most heavily Democratic-leaning electorates in the country after repeatedly promising during his campaign that he would hold Trump accountable. And it used a previously untested and highly unusual legal construct, according to which Trump’s alleged commission of a misdemeanor under New York law was elevated to a felony on the theory that he was doing so in furtherance of, or to conceal, another crime—apparently a federal election violation that the state never made an attempt to prove.
The degree to which all of these prosecutions have been bungled and mired in partisan interests demonstrates that there are deep structural problems with America’s judicial system. Most notably, the extent to which prosecutors and district attorneys and judges are elected, and therefore need to campaign for their positions, is highly unusual by international standards. Those structural flaws may have been sustainable during less polarized periods of American politics; now they make it hard for citizens to trust that justice is impartial when it comes to the most controversial—and therefore important—cases.
But the danger which now emanates from the White House goes much further than these longstanding structural flaws. The federal government has incomparably greater resources at its disposal than any single district attorney. The FBI can go after supposed suspects in every part of the country. The dangerous profusion of federal laws makes it easy to concoct some charge against virtually anybody, especially if they have served in public office. And prosecutors have historically found it so easy to procure indictments that New York Chief Judge Sol Wachtler famously said that he could, if he so chose, get a grand jury to “indict a ham sandwich.”
For now, the justice system still retains enough independence that Trump may not be able to put his enemies behind bars. Though Comey has now been indicted, scrupulous legal observers like Andrew C. McCarthy, of the National Review, expect that his case may be quickly dismissed, and that he would likely prevail even if the matter does come to a jury trial. But if Trump’s political opponents come to fear that anybody who sufficiently inspires his ire will need to go through the scary and costly process of defending their liberty in federal court, then even the prospect of their eventual acquittal won’t stop political speech in America from being chilled to an astonishing degree.
A key challenge for our political system is that its principles are as obscure, and seemingly even contradictory, as they are important.
We live in a political system which claims to let the people rule. And yet, a core component of it is to grant individuals very far-reaching rights designed to insulate them against the tyranny of the majority, like the freedom of speech and the freedom of worship.
We live in a political system which delegates vast powers to an elected president, allowing him to impose his agenda on the country, and especially the executive branch, for the length of his term in office. And yet, a key component of this system is to ensure that important state agencies, including key parts of the executive branch itself, should act in a manner that is independent of the president’s will and his whim.
Subscribe now
These seemingly contradictory principles are difficult to explain to ordinary citizens. When they are violated, neither the nature of the problem nor the magnitude of the consequences are intuitive. Millions of decent Americans probably greeted the news of Comey’s indictment with a shrug; powerful people, they may have thought to themselves, always play stupid games with each other; who wins or loses any one round doesn’t really affect the rest of the country.
But though the kind of lawlessness pursued by Trump’s second administration always starts by affecting political insiders like Comey, it has a way of spreading throughout the system. If a former Director of the FBI can be prosecuted for his political disloyalty, there’s a real danger that ordinary citizens won’t be far behind. And that’s why the opaque and seemingly convoluted saga of Comey’s indictment should matter to all Americans, left or right, liberal or conservative, who value their liberty.
If you enjoy this piece, please check out the bi-weekly Bookstack podcast hosted by Richard, in which he interviews the authors of the best recent books on history, culture, politics, and much more.
To add Bookstack to your favorite podcast app, or generate an RSS feed, click here:
Set up Bookstack in your podcast app
To receive Bookstack episodes directly into your inbox, click here (and select the relevant toggle):
Receive Bookstack episodes via email
A lot of reading is about serendipity. When 107 Days by former vice president Kamala Harris arrived on the doorstep, I was already reading History Matters—a posthumous collection of essays by that master of presidential biography David McCullough. There he has a brilliant article on Harry Truman, another American vice president who was suddenly thrust into the political limelight—in his case when he became president after the death of Franklin D. Roosevelt in April 1945. “He enjoyed being Harry Truman,” McCullough writes of the new president. “And again and again, he could reach down inside himself and come up with something very good and strong. He is the seemingly ordinary American who, when put to the test, rises to the occasion and does the extraordinary.”
Kamala Harris was the opposite. In so many ways an extraordinary and history-making figure, when she was put to the test she reached inside herself only to find the ordinary. And against Donald Trump that was never going to be enough.
Harris is not shy about pointing the finger of blame in this account of her 2024 campaign to be president. She sidesteps the glaring issue of whether Joe Biden was fit to govern by saying he wasn’t fit to run but criticizes him for not getting out of her way quickly enough and then undermining her at key points in the campaign. (“I just couldn’t understand why he would call me, right now, and make it about himself,” she writes of the president’s phone call right before her debate with Trump.) First Lady Jill Biden was “tense, even angry.” Running mate Tim Walz was useless. (“You’re not there to make friends with the guy,” Harris shouted at the TV during the vice presidential debate as J.D. Vance wiped the floor with him.)
Subscribe now
Meanwhile, staffers are shown regularly getting reamed out, usually with salty language. One is even upbraided just for wearing a tie. (Presumably on a different day he would have been upbraided for not wearing one.) Fellow politicians are insufficiently supportive. Barack Obama and Nancy Pelosi are too slow to support her. Gavin Newsom won’t return her calls. Raphael Warnock is a windbag. Of her opponents, Trump was a “fascist” and “a con man.” Elon Musk’s $288 million election spend was part of “the most brazen admission of corruption in history.” As for her judgement on Vance, “it begins with an m and ends with ah.” Even husband Doug gets it in the neck, shouted at for watching the Dodgers beat the Mets on TV while she was in the bath and therefore missing her phone call from the tub instructing him to fetch towels from the other side of the bathroom. “Really?!” she writes, “It was a bridge too far.” Really, indeed.
All in all, Harris emerges from this book as humorless and foul-mouthed. But politics is not a nice business. (Otherwise, Mitt Romney would have been president.) What does matter, however, is that after 320 pages, the reader is left with no real clue about what Harris thinks or believes. Anyone who reads or writes about American presidents might tell you that often the most intriguing moments are those that reveal the hinterland behind the public figure. In fact, reading what they read, as our friend McCullough reminds us, is often the best way to understand presidents. From Adams and Jefferson through FDR, JFK and Reagan, all were vociferous readers. But it can similarly be those “mornings on horseback” with Theodore Roosevelt or Ike Eisenhower building a house on the tenth tee at Augusta National Golf Club. All these successful candidates found their different ways to detach and think.
Trump is no great reader either, but he has his business interests, the passion for golf, and the whole alternate world at Mar-a-Lago. Harris has, or at least demonstrates, no propensity for contemplation or alternate worlds. In fact, she cheerfully admits to the opposite. “I tend to be task-oriented,” she writes, “and rarely allow myself enough space or time to reflect.” At one point in the campaign, she perhaps senses this need in herself but doesn’t know what to do with it. “I wanted to get out of the hotel and out of my head for an hour,” she writes. So she “scheduled a lunchtime visit to a small business, Penzeys Spices” to meet more voters. It’s not Nelson Rockefeller phoning the director of the local art gallery at whichever campaign stop he found himself to see if they could open at 6am for a private viewing, or Jack Kennedy taking his boat out to sail the waters off Cape Cod or Palm Beach.
This absence of hinterland is one plausible explanation for why Harris cannot think on her feet: she lacks what the philosopher Isaiah Berlin called “fantasia.” And perhaps Harris knew this about herself. She was hesitant to give interviews and when she did, they usually went badly. “I wanted the interview to be flawless,” she writes of her first public sit-down as the candidate, not realizing that perfection was not what was required. Flawless is rarely the same as inspiring unless you’re Simone Biles. Voters want more than impeccable from a candidate, including a sense of heart, ideas, spontaneity, genuineness, and quickness in the moment.
This last quality trips Harris up time and again. When Donald Trump sees her at the 9/11 memorial the day after their debate and says “You were great last night,” she “doesn’t know what to say.” Asked at a Noticias Univision town hall meeting to name three qualities she liked in Trump, she “racked my brain” but couldn’t come up with them. Quizzed on The View about what she’d do differently to Joe Biden, she infamously said “not a thing comes to mind.” Afterward, it took her campaign strategist David Plouffe to point out the obvious to her—“People hate Joe Biden.” Plouffe was used to working with Obama, but here was Harris unable to conjure a single independent thought in the moment, let alone express a philosophy or world view (in contrast to “Make America Great Again”). Even on her strongest topic, women’s health, Harris pushes a note across to her campaign staff mid interview demanding to know, “Where’s my briefing?” Her interviewer, Dr Mike, was “warmed up on all these issues and raring to go.” Couldn’t she have been too? The contrast with Trump’s freewheeling interviews, sometimes lasting hours, on anything and everything could hardly be starker. And because Harris couldn’t articulate her own vision, she let her opponent do it for her—“Kamala is for they/them, President Trump is for you.”
If Harris’s lack of ideas emerge clearly throughout the book, so too does her insecurity. She rails about those who “claimed I was ‘DEI hire’” and, with Doug ventriloquizing, attacks Biden for “hid[ing] you away for four years, giv[ing] you impossible, shit jobs.” Yet in her pick of Walz as running mate, recounted in detail in 107 Days, she reveals herself. Pete Buttigieg, a generational talent “magnificent at sparring with opponents,” was her “first choice,” but “we were already asking a lot of America: to accept a woman, a Black woman, a Black woman married to a Jewish man,” and asking voters to accept a gay running mate was seemingly “too big of a risk.” Which is odd, because twelve pages earlier she reports that when the campaign put a poll in the field with all the possible VP contenders, “the results turned out to be useless. None of the names moved the needle either way.” She went on to reject both Governor Josh Shapiro of Pennsylvania and Senator Mark Kelly of Arizona—two more A-listers in their different ways—and instead picked the B-list Walz, the candidate who on meeting her immediately declared that he was hopeless. “I’m not a good debater,” he confessed, “I’ve never used a teleprompter.” Welcome to Team Harris, Senator Vance will see you shortly.
Subscribe now
Sadly for Harris, these are the kinds of harsh judgements that history hands down to the losing candidate in an American presidential election. It is, of course, important to remember that hers has still been a remarkable career. Her ascent to the United States Senate and the vice presidency as the daughter of Indian and Jamaican immigrants is an American story and, “as the first woman, or Black woman, in every office I have run for, except the Senate, where I was the second,” a story of which she is justly proud. She writes persuasively of the challenges that women face in politics, not least because their looks and dress sense are held to a different standard. Her instincts can be admirable too. She offers her hand to Trump at the debate and phones him after an assassination attempt. She even ruefully admits to quite liking him. And she shows tremendous grace in defeat, not least in presiding over the Senate as her own loss was confirmed, even enduring the indignity of reading out the electoral college tallies. Not all losing vice presidents have done the same. Perhaps no one would have blamed her if she had followed Hubert Humphrey’s example after the 1968 election and not bothered to show up.
But is that all enough to erase the failure of defeat to someone who both sides agree is one of the most consequential presidents of the modern era? As Harris liked to tell her staff, “That was a rhetorical question.”
Richard Aldous hosts Persuasion’s Bookstack podcast and teaches history at Bard. His books include Reagan and Thatcher, The Dillon Era, and Schlesinger.
Follow Persuasion on Twitter, LinkedIn, and YouTube to keep up with our latest articles, podcasts, and events, as well as updates from excellent writers across our network.
And, to receive pieces like this in your inbox and support our work, subscribe below:
Subscribe now
In this week’s conversation, Yascha Mounk, Francis Fukuyama, Sabina Ćudić, and Dan Williams discuss Donald Trump’s firing of a federal prosecutor and what this means for democracy in the United States, what the discussions around the assasination of Charlie Kirk tell us about misinformation, and the impact of RFK Jr.’s recent autism announcement.
Francis Fukuyama is the Olivier Nomellini Senior Fellow at Stanford University. His latest book is Liberalism and Its Discontents. He is also the author of the “Frankly Fukuyama” column, carried forward from American Purpose, at Persuasion.
Sabina Ćudić is a member of the National Parliament of Bosnia and Herzegovina, where she is the president of the Naša stranka political party club. Ćudić also serves as vice president of the Foreign Relations Committee, and is a member of the Parliamentary Assembly of the Council of Europe in Strasbourg, where she is a vice president of the European Liberals.
Daniel Williams is an Assistant Professor of Philosophy at the University of Sussex. He writes the Conspicuous Cognition newsletter, which brings together philosophical insights and scientific research to examine the forces shaping contemporary society and politics.
Note: This episode was recorded on September 23, 2025.
This transcript has been condensed and lightly edited for clarity.
Yascha Mounk: Frank, there are many remarkable things going on in the United States at the moment, but perhaps the most remarkable is the extent to which the administration is seemingly trying to take over the most important neutral institutions in the country, firing prosecutors that have displeased the administration either by who they did or did not prosecute, apparently stopping an investigation into the secretary of the Department of Homeland Security, who stands accused of having taken significant bribes, and then sending a remarkable message to Pam Bondi, the attorney general. It’s unclear whether that message was meant to be public or was meant to be a private message, basically instructing her to go after political opponents. How worried should we be about the possibility of the administration actually managing to undermine the course of justice in the United States?
Francis Fukuyama: Well, I don’t think that the effort to prosecute Trump’s enemies should be any surprise to anybody. He’s been telegraphing that he would do this during the campaign, and he was clearly focused on revenge. In fact, I would say it’s a little curious that it’s taken him this long to try to get around to that behavior. But, before I talk about that, I think the other important thing that happened this week was his passing into a different kind of authoritarianism with the actions against ABC and Jimmy Kimmel, because a lot of the authoritarian sorts of things that he’s been doing, like firing different agency heads without authorization and the actions taken against immigrants—a lot of that actually could be justified on policy grounds that he needs authority to do certain things to carry out his policy agenda. But moving against Jimmy Kimmel, the only issue is just his feelings. Only a real authoritarian cares whether a late night comedian is making fun of you or not. To want to shut that sort of thing down moves you into a different territory in terms of following up on your authoritarian instincts.
Now, I actually think that what happened in the course of the week is not bad because Jimmy Kimmel was reinstated. There was a huge pushback. And what was interesting was there was pushback on this even from some of his right-wing allies like Ted Cruz. Brendan Carr, the head of the FCC, made the statement, if you’re gonna make disparaging statements, we’re gonna take action against you, we can do it the easy way or the hard way. Ted Cruz said, this is the way a mafia don talks, and that’s exactly right. I thought that was quite remarkable—that you actually started seeing some cracks in his support.
Thanks for reading! The best way to make sure that you don’t miss any of these conversations is to subscribe to The Good Fight on your favorite podcast app.
If you are already a paying subscriber to Persuasion or Yascha Mounk’s Substack, this will give you ad-free access to the full conversation, plus all full episodes and bonus episodes we have in the works! If you aren’t, you can set up the free, limited version of the feed—or, better still, support the podcast by becoming a subscriber today!
Subscribe now
Set Up Podcast
And if you are having a problem setting up the full podcast feed on a third-party app, please email our podcast team at [email protected]
The other authoritarian thing is obviously going after Letitia James and James Comey and various other relatively minor players for their perceived wrongs against Trump during the Biden administration. Like I said, this should not be of any surprise to anybody. They had a pretty good prosecutor that looked at the evidence and he just couldn’t find any legal basis to take action against any of these people. He told Pam Bondi this. Pam Bondi on her own didn’t have the authority to order him to go ahead and do this. And then Trump criticized her. It’s very hard. Any government can come up with some infraction, a traffic ticket or something that they could hold against somebody. But in these cases, they have to just completely make up charges to merit a federal investigation into them. It does seem to me that this is simply going to lead to Trump doing what Nixon did, which is hiring and firing prosecutors who successively will tell him, we can’t make this stuff up. He’s gone to a very low-level person. We’ll see what that new prosecutor does, but it’s hard to see this really going anywhere except as a matter of intimidation, that he’s just showing that he’s going to go after you regardless.
Mounk: Sabina, you come from a country where politicians and others are often lectured by the European Union and other institutions about what kind of changes you have to make in order to live up to the expectations of the rule of law. Are you seeing and observing this with some satisfaction? Some schadenfreude?
Sabina Ćudić: Well, you already know me well enough to know that that is a little bit the case.
Mounk: I would have thought more highly of you. I thought you were about to deny any such charge.
Ćudić: Having said that, in the region I come from, I’m always trying to censor myself not to be that hammer to whom all problems look like a nail such that everything looks like a pattern of Balkanization, that this is all that has been seen before and so on. So sometimes I even exaggerate my optimism in order to regulate my tendency to look for patterns. However, I do have to say that this micro-dosing of optimism that we hear now, that last week has been a moderate success because ABC is reinstating Jimmy Kimmel and so on, is in a way showing where we are and how little satisfies us and how normalized the situation is.
I think we are experiencing in terms of the use of tools for prosecuting your enemies—and as you said coming from a country in which the United States has heavily invested in securing an independent judiciary—looking at what is going on, it seems like we are experiencing not a paradigm shift that we usually like to talk about, but a full paradigm breakdown in a sense. That not only are the institutions attacked, which is obvious, and like Frank said at the beginning, nobody’s surprised that the president who said he would do that is going after the enemies. But what is surprising is how little people care. I mean, sure, in conversations like this, people care, and sure, there was an outcry over Kimmel. But I think the larger question is about democracy. Democrats were right when they said that democracy is at stake on a ballot in 2024. I think we can all agree now that absolutely democracy was on the ballot, but the bigger question is do enough Americans care about democracy and do they care enough that they value it above certain other things? And is it a question of taking democracy for granted? Is it a question of the complexity of bureaucracy that has failed on delivering on the policy front? Or is it simply people enjoying power and demonstrations of power in its most rudimentary and primordial sense?
Dan Williams: I think people do care, but I think it’s easy to overlook two things. One is the extent to which many conservative Republican voters basically inhabit a kind of different epistemic universe, where from their perspective, it’s really Trump who’s been the victim of political persecution and so on. But I also think we shouldn’t underrate the extent to which many people recognize the corruption and the character flaws of somebody like Trump, but they think of it very much so as the lesser of two evils. They think the liberal establishment that he, as a strong man, is opposed to and is fighting against is so corrupt and has been so antithetical to democratic values that they personally cherish that they’re willing to tolerate some of those what seem to us like incredibly egregious violations of the rule of law.
Subscribe now
I think part of that sort of worldview—where you recognize that Trump has these character flaws, but you view them as the lesser of two evils—part of that I think is just rooted in sort of misperceptions and misunderstandings of what’s happened. But at the same time, I think you do also have to acknowledge that there are contexts where there are legitimate grievances there. So for example, the failure of what they perceive as the liberal establishment to enforce borders and control immigration, the extent to which what they perceive that the liberal establishment has upheld certain kinds of conspiracies of silence around connections between, say, immigration and crime, or Biden’s cognitive decline, and so on and so forth. So it’s not to say that I agree with that general perspective. And I think Trump’s behavior is absolutely egregious, but I think you sort of have to put yourself into the mindset of people willing to tolerate this behavior, not necessarily because they don’t care about democracy, but because they’re just viewing the world through a totally different kind of lens.
Mounk: How far is this actually going to go? So we’re seeing prosecutors who were undertaking prosecutions of friends of Donald Trump being fired. And that, of course, is a significant break with the rule of law. But the much more dangerous form of this would be if Donald Trump is actually able to successfully prosecute and perhaps put somebody behind bars on spurious grounds. But of course, there are a number of checks to his ability to do that, including juries, including judges, including the ability to appeal judgments. How far are we from Donald Trump actually being able to find trumped up charges to effectively go after political opponents? Of course, even the process is the punishment, and even the fact that people might come after you in this way is really scary. But how far is this going to go, Frank?
Fukuyama: I think that we’re being a little bit too pessimistic here, and I don’t think it’s fair to say that the American people don’t care about rule of law or any of these issues. If you look at poll numbers, Trump has been dropping very remarkably over the last few months. A lot of that has to do with the economy, but even on something like corruption, it’s quite remarkable. A very large majority of Republicans think that he’s corrupt. The big check is going to be next year’s election, and I think that although the Republicans can manipulate the outcome of that to some extent through gerrymandering, it’s going to deliver a verdict. I think that’s why he’s moving as quickly as he is on some of these issues, because he realizes that he’s only got about another year where he’s got the freedom to act as he sees fit, and he wants to try to get as much done before then. I do think that you’re not going to see demonstrations in the streets over Pam Bondi’s firing of a prosecutor, but I do think that there is still going to be a mechanism of accountability there that should continue to function.
Mounk: I find it very hard to gauge just how damaged the rule of law is so far. German public television said in a program for children that many people in the United States now are afraid of criticizing the president. And I just think that’s wrong. If you look at virtually all of the traditional media, which may have a diminished role because of a changing ecosystem, but the New York Times, NPR, the Washington Post, the LA Times, the Boston Globe, MSNBC and CNN, I just don’t think it’s true that people step around criticizing Donald Trump in the way that the media in Hungary do, the media in India do, the media in Venezuela certainly do. In the same way as Frank is saying, are we really going to get to a place where someone’s going to be put in jail on trumped up charges or aren’t there still many steps on the way there?
Subscribe now
I think the way that the FCC commissioner used mafia-like language to say we can do this the easy way or the hard way was absolutely outrageous and a very clear attack on free speech. But it’s not just that he made fun of Donald Trump. It’s that, without basis of fact, he very strongly implied or perhaps you could even say stated that the person who killed Charlie Kirk was part of the MAGA crowd. So he created, in some form, misinformation himself. He’s now been reinstated. I’m sort of torn between thinking that clearly the amount of attacks is much worse than the first Trump administration. They’re clearly willing to go very, very far, even in ways that go beyond what Trump’s allies were willing to do in the first Trump administration, or perhaps were capable of doing in the first Trump administration because they were much more disorganized back then. But how far is this going to go? I must say that, like in many other things right now, I feel like I can give you a point estimate, but that’s pretty useless. The confidence interval, the range of reasonable guesses as to what kind of result we might get is really, really wide right now.
Ćudić: We are less than a year into this administration and we are having this conversation about the level of potential damage or real damage done to institutions. We have Stephen Miller saying that the Western civilization is on the line and that when choosing between essentially Western civilization and the First Amendment and freedom of speech, we need to kind of take what he calls necessary and rational steps to save Western civilization. You have senators who say, I used to believe that the First Amendment should come on top of any argument, but these are not normal circumstances and I no longer believe that. And then you have Steve Bannon, who said, we took two scalps and there will be more scalps. The most interesting statement I heard coming out of the last week was saying, I get invited to all these debates on the First Amendment, but they (meaning liberals) don’t get it. We no longer debate. We now take action. So time for debate is over.
Fukuyama: Talk is cheap though—everybody’s really upset because of Charlie Kirk and they’re saying all of these horrendous sounding things. Let’s just wait and see whether actually any of this materializes. I strongly doubt that any of these threats are actually going to be carried out.
Ćudić: I don’t think they will, but the question is, does that language in itself create a certain reality of the possibility of political violence? How do you measure these statements against the potential outcomes that will not be systemic or political, but that will be taken up perhaps by individuals?
Williams: I was just going to echo something sort of Sabina has alluded to there, which is we’ve been talking about Trump, and I am extremely alarmed about Trump’s behavior, but Trump is not, at least in my view, first and foremost an ideologue. I think he’s a narcissist and he’s self-serving and I think he’s dangerous. But I think there are segments within the MAGA coalition—Steve Bannon, Elon Musk—that are pushing an extremely dangerous political worldview and political agenda, which sort of extrapolating that into the future is very alarming. And I suppose the other thing I’d say, just in terms of forecasting the future, I think if Trump and the sort of MAGA coalition have demonstrated one thing, it’s how bad liberals are when it comes to predicting what’s likely to happen. If you went back to 2015 and you tried to predict the next 10 years, the events have just been radically different from what anyone would have expected. I’m quite pessimistic about the future, partly because of these segments of the coalition, but also partly because I think it’s just so difficult to forecast how the future is likely to unfold.
Mounk: Our second segment for today is thankfully somewhat linked, so it may feel like we’re continuing the discussion. We are now nearly two weeks after the assassination of Charlie Kirk. We saw a few days ago the memorial service for Kirk, and people trying to assess where we are at in terms of how Americans think about political violence at this stage, and also whether they are forming factual beliefs about what has been happening. I’ve been really struck by two polls that I’ve seen in the last couple of days. One shows that only 10% of self-described Democrats believe that the assassin of Charlie Kirk was motivated by left-wing ideology, something that I think there is pretty strong evidence for. The other is a poll commissioned by the New York Times that I came across because Shashank Joshi, our sometime co-panelist, posted about it, according to which, first of all, about 20% of liberals think that political violence can be justified, while only 7% of conservatives and moderates believe that violence can be justified. When you break out this poll by age groups, you see that conservatives and liberals over the age of 60 have nearly the same amount of agreement about this. They both are very unlikely to agree with political violence. But young liberals, left-leaning Americans between the ages of 18 and 39 are much more likely to say that they support violence. In fact, among them, nearly 30% say that violence for political purposes can sometimes be justified. Sabina, you’ve been watching the response in the United States over the last couple of weeks. What do you make—as somebody who well knows what political violence looks like—of the current state of the United States?
Ćudić: Looking back now with a bit of a couple of weeks’ distance, I still believe what I thought on the first day: that the level of dehumanization on both sides, dehumanizing narratives of justification and celebration on the other side, with the call for revenge and so on, which reached its peak, I would say, with the ceremony where the president himself says, I hate my opponents—in a sense this is something that really calls for analyzing patterns. And what struck me in the New York Times, as you mentioned, the polls came out and there was an attempt from the academia side to relativize those numbers to say, well, once we asked in polls, ‘is it ever justified,’ then of course people start thinking about instances in which potentially violence may be necessary and so on. But I even find this relativization not only not helpful, but a kind of wishful thinking that will lead us nowhere.
Subscribe now
Then something that really drew my attention was the conversation that Ezra Klein, not in the position of the host, but as a guest, had with his fellow, but conservative, columnist at the New York Times where they started a discussion on who is more likely to commit political violence in the future, extreme right-wing or extreme left-wing, and they were comparing the potential scenarios, but also expressing individual fear over what might happen. But in a sense it was—I don’t want to sound too crude here—a festival of victimhood, of who feels victimized more, while recognizing that this is our new reality. And in this new reality, particularly with the numbers of young people supporting it, one of the elements that we need to look at is the propensity of young people. We’ve seen in Nepal last week, politicians set on fire while still alive by young people as a result of a policy to forbid social media. Even the media said, look how inspiring the Nepalese youth is in taking violence and setting politicians on fire and expressing their disdain.
One of the avenues to analyze that is the assumption by the younger generation that policy solutions will not provide the answers. And I think another example of that, and I’ll finish with that, is not only the assassination of Charlie Kirk, but it was the assassination of Brian Thompson, the CEO of the health insurance company, where the internet and social media erupted in celebration, particularly young people, because it was stemming from the assumption that no policy solution for fair and just healthcare will be provided by the establishment.
Mounk: Dan, one of the things that I was wondering about for the last two weeks is the role of misinformation in this. I go back and forth on this, because on the one hand, I feel like the principle mechanisms by which people have allowed themselves to engage in what Sabina rightly calls this festival of victimhood was not misinformation. The principal driver was simply that the right has said the assassin is left wing—as he does appear to have been—and this shows that the left hates us and that in the extreme words of Elon Musk, either we vanquish them or we’re going to kill us all. And parts of the left have immediately gone to saying, well, in fact, Charlie Kirk was terrible in all of these ways. Let’s focus on people like Ezra Klein, who went a little bit too far and said nice words about him. That is the real scandal of this moment. Neither of those responses really particularly necessitated or involved misinformation. Now, it’s also true, as this poll shows, that only 10% percent of Democrats recognize that the assassin likely was motivated by left-wing ideology, and that misinformation somehow enhanced that, but there certainly was large-scale misinformation that was happening over the last weeks. But how important is misinformation to this kind of spiral of violence?
Williams: I think of it as sort of illustrating the limitations and the inadequacies of the misinformation framing as a way of understanding what’s going on. As you mentioned, there has been misinformation in the aftermath of this. On the left there was Jimmy Kimmel’s suggestion that the killer was motivated by right-wing views, and also the gross misrepresentations of things that Charlie Kirk had said, which were going viral on certain parts on Bluesky and progressive social media. And also I’d say on the right, sort of base right-wing anti-Semitic conspiracy theorizing in the aftermath, suggesting the assassination has something to do with Israel. But for the most part, what you found is people responding to the actual event and the effects of the event, but filtering it through these incredibly Manichaean tribal narratives that they’re independently attached to. So when Elon Musk sees this event and says, well, the left is the party of murder, it’s not the case that he’s encountered some bit of misinformation. He’s doing what is generally happening in this space, which is taking certain parts of reality, certain matters of fact, and framing them through predetermined ideological narratives, ideological templates.
I think the fantasy of the misinformation frame has always been, if only you could remove these bits of misinformation or these bits of fake news, you would solve many of our political problems. And I think what it illustrates is you really won’t. I mean, there’s a deep sense in which Elon Musk saying that, and that sort of general us versus them Manachian worldview, is utterly misperceiving the nature of reality in a very dangerous way. It’s not rooted in exposure to misinformation. It’s rooted in this underlying polarization, this sort of us versus them psychology, which is ancient and long predates the emergence of social media. And I just think misinformation plays a role, as it always does, within the context of these sorts of big political events. But it’s grossly inadequate as a way of understanding what’s actually going on.
Fukuyama: Well, if I could change the framing a little bit: something really important happened this past week that has gotten almost no commentary, except I must say for a blog post I wrote about this, which is the consolidation of a huge number of media channels under the control of Larry Ellison and his son David. I think that’s actually one of the underlying really big problems. I call this the Berlusconi-ization of the American media space, where somebody with a political agenda takes over a big media channel, uses it to manipulate politics, and then of course uses politics to protect their business interests. That was the Berlusconi pattern. That’s what Musk did in his takeover of Twitter. But the Ellisons are going to control an unbelievable number of media channels. It’s Skydance Global, it’s HBO, it’s CNN, it may be Bari Weiss’s Free Press. It’s much, much bigger than Elon Musk taking over X, if they want to use this in a political fashion.
We already had Grok and Musk trying to influence our politics through that one internet platform. But now you’re going to have two right-wing oligarchs, or one of them is a kind of a right-wing oligarch family, that will have this kind of concentrated power. I just think that this is really dangerous, and people are just not paying attention. The other thing that has escaped attention is TikTok. So Trump acted completely illegally in not enforcing the law that was passed on a bipartisan basis forcing the divestiture of TikTok from ByteDance. Now he apparently has cooked up a deal. And again, the Ellisons are gonna be one of the major shareholders in this new platform. And people should pay attention to this. We have way too much concentrated power in the United States, and I think that that is much more important than any particular hothead spouting off on the internet.
Ćudić: It’s hard to say when that started. When you describe the concentration of media power, my mind goes to the meetings between Margaret Thatcher and Rupert Murdoch and the liberalization of the media space then in exchange for support for her policies, and so on and so forth. So it’s been going on for decades and in a way one wonders: at what point is there a point of no return?
But going back, I just want to add a question to this discussion of extreme right wing versus extreme left wing. In my political experience, extreme left wing and extreme right wing, particularly youth branches, have one thing in common. They do not believe in the system we’ve built and they think that the solutions will be found outside of it. So when we talk about the use of violence by the extreme left wing and right wing, it stems from the exact same assumption: that solutions will not be provided within the system, that the solutions need to be imposed from the outside through violent acts, terrorist activities, organizations outside of the system.
Mounk: I had a remarkable experience this past weekend. I was invited to speak at a festival in London about liberalism and whether the liberal conception of freedom is the only conception of freedom worth defending, in which I was faced with two discussants or two co-panelists. One was Curtis Yavin, the man who’s now infamous for being supposedly the kind of philosopher behind the second Trump administration or behind JD Vance, and for his various versions of this claim to prefer some kind of CEO or monarch over our democratic system. And the other was a kind of woke feminist. But the remarkable thing was that, as I kept pointing out, they sounded very similar to each other—a point that they both hated, but the audience agreed with. They both blamed every bad thing in the world on the idea of liberalism and then argued for some imaginary alternative that was rooted either in the past or the future or on the left or the right as the system that is going to somehow be much, much, much better. When in fact, for all of the real problems that our system has, I haven’t yet seen in the history of the world a system that delivers better for people than the imperfect set of liberal democratic institutions that we inhabit.
Having said that, while I was at the festival, I ran into a listener of the podcast who said I should share more about my own life. I’m not sure that I’m going to do that in the podcast, but I am going to share the one story that I had from last week, which relates to the discussion we have. I was on the morning show of France Inter, a radio program that has 45 million listeners in a country of 60 million. And there were two people who were on the show with me. The first, who actually works for a center-right think tank in France of American origin, said before she sat down on air, as we were chatting in the green room, Kirk was such a terrible racist as he kept using racial slurs. And I said, I hadn’t heard of him using racial slurs. What are you referring to? And she says, well, I mean, I couldn’t possibly pronounce a word that’s so horrible. But it begins with “ch.” I saw some of my co-panelists looking into the air like they’re trying to figure out which word it is—which took me a second as well—and I realized she must have been she must have meant the word chink, which indeed is a derogatory term for Asian Americans that certainly shouldn’t be used in polite company other than when you’re mentioning it, as I just did, because I believe in the use-mention distinction.
I googled this for a second and very quickly found that there is a debate which Charlie Kirk had done with Cenk Uygur, a very well-known progressive influencer, in which repeatedly he interrupts Cenk Uygur to say, Cenk, Cenk, I’m going to say something. So it is unambiguous that he did not shout this word at somebody in the audience to insult them. He was saying the first name of a co-panelist of somebody he was having a discussion with. So I explained that, having found that, and the woman more or less believed me. Then we went on the air, and the second question is, tell us a little bit about Charlie Kirk. And this question is directed to a woman who was for a long time the political director of Le Monde, a very senior French journalist. She says among the first sentences, well, he was a horrible racist. He said that black women don’t have the brainpower to do all kinds of jobs.
Now, anybody who’s been online for the last week knows that there was a big debate about this in the United States because a number of people have said this, including Karen Attiah, who was fired from The Washington Post, perhaps in part because she misquoted Charlie Kirk in this way in social media. But the actual story is that when the Supreme Court struck down affirmative action, a number of women said, I wouldn’t be in my position if it weren’t for affirmative action. And Charlie Kirk, whose political style I often disliked, said in a way that I find to be needlessly mean and cruel, well, if that’s the case, then perhaps you don’t have the brainpower to be in that position. Now, however you feel about the details, there’s a huge difference between saying that about specific people who themselves have said that they got certain positions due to affirmative action, whereas the quote was, black women in general don’t have the brainpower to do X and Y, something that Charlie Kirk very clearly never did. So I did something that I rarely, rarely do on air, which is to interrupt this lady and to say, I’m sorry, but that’s just not true. That is not what he said. And I gave the explanation of what had actually happened. She then said in a slightly unfortunate tone of phrase, well, it was to go faster and to move the conversation along, or something like that.
That went very viral in France because it indicated some kind of attitude towards the truth for somebody who should be a very serious journalist, somebody of tremendous authority and after the commercial after this part of the show was over, she didn’t ask me, am I sure that she had misquoted it or anything like that, she was mad at me for interrupting her. And I said, again, I try not to interrupt, but I feel like the main concern should be whether or not we just spread misinformation to four or five million people, not whether you interrupt in the course of doing so.
So I go back and forth because I guess one way to prove this is that the story about misinformation is usually quite simple. The story about misinformation is there’s an enlightened establishment in the trustworthy media. And if you listen to them, then you know what’s going on in the world. You know what’s happening. The problem is that there’s low IQ idiots out on social media, who out of their stupidity or for partisan gain or for financial gain are spreading these terrible forms of misinformation. They are the real problem, and if only we could stop them, somehow the world would be set to rights. I guess I’m both skeptical that misinformation is what drives political views rather than that political views drive people to engage in misinformation. And I’m skeptical about whether we can consistently avoid what Matt Yglesias calls elite misinformation. I’m struck by the extent of elite misinformation I have seen over the last few weeks.
Williams: I’ll just really quickly add something there. I think it’s not just extraordinary that there’s such a progressive liberal establishment echo chamber, such that those quite egregious misrepresentations of the facts can occur in these quite mainstream spaces. I also thought this of the Jimmy Kimmel scenario. But I think what you also have to factor in is anybody within the conservative information environment is acutely aware of the degree to which those things get misrepresented. So when they see it, they then think that’s even more evidence of the degree to which the “liberal establishment” is just egregiously lying. So it’s an informational problem, but it has these really damaging political consequences and tends to increase polarization as a result.
Mounk: This in a way leads us naturally into the third segment of this conversation, which is the remarkable press conference that RFK, the Secretary of Health and Human Services—incredible to say—gave recently about the issue of autism, which is obviously a topic that he has cared deeply about for a very long time, making over the course of decades a set of spurious claims about the link between vaccines and autism. It really appears to be one of the issues that drove him into politics and that he’s most obsessed with. At this press conference, he did not stipulate a direct link between vaccines and autism, though he suggested that there might be one and we’re going to continue studying it. But he did, among other things, say that there’s a connection between pregnant women taking Tylenol and autism. Now, once again, I felt two things about this. The first is that this was deeply irresponsible, and it’s very shocking that we’re now in a place in which most senior health officials in the country are spewing lies that in this circumstance really do have a direct negative impact. I mean, I think a lot of the time when people say wrong things, that’s bad for public culture. And you can tell stories that are quite indirect about why that’s concerning. But advising people not to take vaccines or not to take certain kinds of drugs that may be useful to them because of a spurious link to negative outcomes, has a much more direct way of negatively impacting people. And so I’m very concerned about that.
I’m also once again struck and concerned—I’m not an expert here—by how bad the information is if you read the mainstream coverage about it. I read the article about this press conference in The New York Times, and it just didn’t give any of the necessary background information. There is in fact one study which links the use of Tylenol to rising autism rates. It was done by the Dean of the Harvard School of Public Health. There’s reason to think that it’s a really bad study. If you go around Substack a little bit and look at a medical expert who actually really knows this stuff, they will explain to you why that was a really bad study. The Dean of the Harvard [School of Public Health], which is itself remarkable, appears to have excluded the key study that has the most robust design from consideration in his matter analysis for reasons that clearly misstate what the methodology of that study had actually been.
But The New York Times article kept using weasel words like, without new evidence. It doesn’t really explain the background story. And so if any slightly skeptical reader is going through this article, they’re going to say, well, what do you mean with little evidence? I mean, you google it for one second and you see the Dean of the Harvard School of Public Health published a study which says there is this link. And do you think, well, why is The New York Times lying to me? Which it’s not. But it just creates such an open door for people to mistrust the information.
The other thing that I find really striking is that, again, from what I understand, what’s really going on here is that there’s a very serious reason to doubt that rates of autism are actually up in the United States, or at least they’re up as much as the headline statistics suggest. You go back to the beginning of the criteria for diagnosing autism, and they’re extremely onerous and very little known by doctors. So they basically say that you have to be very incapable of communicating with others and that you have to engage in behaviors that require a lot of intelligence, like very elaborate, repetitive patterns of behavior. Both of those things have to be true at the same time, as well as a bunch of other conditions. That’s a very onerous diagnostic criterion. Since then, we have gone to diagnostic criteria where there’s something like 16 possible indications, many of which are duplicates of each other. And if you qualify on six of them, then you can be diagnosed as being on the autism spectrum. And we have made it mandatory for all kinds of institutions to report autism in various ways. And we have told people that if they are diagnosed as being on the spectrum, they’re going to get extra time on tests. They’re going to get extra resources for help with children who really may need some extra help. So there’s a very strong incentive to classify people as autistic. And so there’s at the very least a serious debate about whether we actually are faced with an increase in this disease or whether we are simply classifying the world in a different way, which creates this weird statistical artifact.
Once again, you look at the mainstream coverage of that, there’s no indication of this. And so I am just frustrated with the world. Sorry to go on this rant. On the one side you have RFK spewing just absolutely irresponsible things that quite possibly will lead to people dying because they’re not vaccinated against important diseases, and on the other hand, you can’t trust The New York Times and other key institutions to actually explain what’s going on to you in a way that helps you understand the world and that isn’t open to the most straight up conservative challenges.
In the rest of this week’s conversation, Yascha, Francis, Sabina, and Dan discuss why there is mistrust in institutions on the right, the shared beliefs around autism between the extreme left and extreme right, and why the United States has so many ridiculous rules. This part of the conversation is reserved for paying subscribers…
From the publisher's feed