Mind and Iron: Is political violence imminent in the U.S.? A new class of futurists say yes
There's a way to model this, and the models don't look good. Also, OpenAI's Her-gate
Hi and welcome back to Mind and Iron. I'm Steve Zeitchik, veteran of The Washington Post and Los Angeles Times and chief booker at this future-news variety show.
Every Thursday we examine how our lives will shape up in the years ahead given the social, tech and other forces at play. Please consider pledging a few dollars to our mission.
This week we're looking at the future through a slightly different lens: politics.
With assassinations and ongoing invasions in Europe, and campus conflict and candidate spectacles in the U.S., democracy as we know it seems more fragile than ever. (Plus, like some of you, I can't get "Civil War" out of my head.)
Since our mission is to gauge how we'll soon live, this week we ask how the democracy that undergirds that life will fare — whether these ominous signs portend violence and even the unraveling of freedoms. It’s a hard and scary and even unimaginable conversation to have. But also an important conversation to have.
So we’ll hear from some of the biggest experts modeling these futures — not armchair pundits but high-level thinkers who rely on models, history, data and tech to make their forecasts. (Including the godfather of this "conflict-prediction" science.) They offer some eye-opening — and chilling — findings.
And then we'll close with some more OpenAI news, because what would a week be without Sam Altman's latest grand plans to change the world/headshaking move to offend good decorum? This time with a bonus movie star.
Quick housekeeping. We will be publishing next week despite Memorial Day — with all the news we'll defer a week off to June — though it may be a slightly shorter issue given the holiday.
First, the future-world quote of the week.
“We're seeing the structural factors for political violence that are similar to the time right before the Civil War.”
—Clayton Besaw, a University of Central Florida researcher who uses tech and data to model political outcomes
Let's get to the messy business of building the future.
IronSupplement
Everything you do — and don’t — need to know in future-world this week
Is civil war in our political future?; the founder of a forecasting movement speaks; Scarlett Johansson enters the LLM chat (literally)
1. MAYBE IT’S PUNDITS’ COMPARISONS BETWEEN THE ASSASINATION ATTEMPT ON THE SLOVAKIAN PRIME MINISTER AND THE KILLING OF ARCHDUKE FRANZ FERDINAND AHEAD OF WWI.
Perhaps it's the perfect storm of campus protests, the circus outside an ex-president's trial and the fraught party conventions that await this summer.
It could be the threats to democracy in Taiwan and Ukraine, or ongoing repression in Iran.
Or maybe I've simply spent too much time thinking about "Civil War," that Alex Garland near-future dystopia about an America violently torn.
Whatever the reason, the idea of armed strife shaking up countries, especially this country, has been on my mind recently. And perhaps yours too. Will the coming months bring violence between U.S. government forces and individuals/militias? Or is this just the stuff of cinematic nightmare, its closeness an illusion?
A subset of social scientists we might call political futurists have been engaging with this subject for a while. And they have strong thoughts about whether the next year will bring a quieting or worsening of the forces that have been shaking our democracy. What do we know and how much should we fear? Their answer is both reassuring and alarming. (More the latter but don't reach for the arsenic just yet.)
Conflict-prediction science works like this. Researchers feed a model reams of data on all the civil unrest and coups that have happened around the world for decades, cross-reference them with all the factors that prevailed in those places before the coups happened, and see if a predictive pattern emerges.
Clayton Besaw, a researcher associated with the University of Central Florida, has been doing this for years at the helm of a model called CoupCast. (He had mostly been focusing on countries outside the U.S. until January 6.)
I caught up with Besaw by phone Wednesday. He gave scientific voice to what many of us instinctively feel: he’s seeing increases in several key political-unrest variables.
These include "tolerance for political violence," “democratic backsliding” (the evaporation of longstanding norms like certifying elections or just kicking major presidential candidates off ballots on technicalities) and "ideological outbidding" (the need for elected officials to one-up each other with their extremism). When combined with broader cultural factors like social-media outrage and the easy availability of weapons, toxicity ensues.
In fact, Besaw says the data conjures the republic's darkest days. "We're seeing the structural factors for political violence that are similar to the time right before the Civil War," he said.
But he also took care to emphasize (after I picked up my jaw from the floor) that the tools to fight such violence are much stronger now and haven't dulled much even over the past few years. A metric known as "state capacity" — a kind of catch-all-term for well-established and well-funded government and law enforcement — is typically among the best defenses against these risk factors, according to the models. The U.S. is still in a very durable position in this regard, Besaw says; state capacity is quantifiably high. (In contrast, the years before the Civil War saw the U.S. barely have a standing army.)
A Donald Trump victory in November, Besaw said, would thus be concerning less due to its incitement potential and more because he could try to weaken this state capacity — say, by financially gutting law-enforcement entities. And Besaw emphasized that even if state capacity remains high, that doesn't mean political violence can be prevented, just that it ultimately won't be successful.
"Political violence is rarely solved without some kinetic violent component of action taken by the federal government,” Besaw said, citing JFK's response to the 1962 anti-integration riot at Ole Miss as an example of both the solution and its toll. (This, btw, is itself a paradox because deploying the means to repel an insurrection often means worsening it — "to stop anti-democratic forces within a democratic society you need to undertake tools that can push people further into the arms of the challengers," he noted.)
Much of the data tracking political violence in the U.S. is done by an NGO called the Armed Conflict Location and Event Data Project, or ACLED, which has been heavily active in recent years. Bracing for increases in political violence, the group relaunched a feature called "Crisis Monitor" — it offers highly granular detail on where and what kind of political violence has broken out — noting that "ongoing political uncertainty and unrest may point to a political climate that is poised for additional tensions during the 2024 election cycle."
One of the leading groups that has engaged in forecasting off such data is a CIA-convened panel of experts and data scientists known as the Political Instability Task Force. They’ve been prohibited from officially monitoring the U.S. but came up with a list of globally applicable factors —38 of them. Two were especially predictive.
One is "anocracy,” the idea that countries are at risk when they occupies the in-between state of democracy and autocracy. A little tautological — violence in a democracy happens when there's less democracy.
But the second factor is more useful, even surprising: tribalism.
According to the CIA’s model, political violence results most often when people start making electoral and political choices based on tribal identification instead of ideology.
That, for all the ways we think civil war is fought because people disagree about a country's direction, what will make them take up arms is something more primal: feeling too strongly about which group they belong to. (See under: the Balkan wars in the 90's, numerous conflicts between Sunni and Shia Muslims in the Middle East.)
Needless to say, this evokes plenty of worry for our current situation in the U.S., which as social scientist after social scientist has warned has seen tribalism explode.
The CIA task force shut down last August. But one of its members, the political futurist Barbara F. Walter, gave a TED talk that same month noting its findings. Walter wrote an acclaimed bestseller a couple years ago titled "How Civil Wars Start: And How To Stop Them,” and at the Ted talk she noted her reaction to learning of anocracy and tribalism as major risk variables. "I could see both of these factors were emerging in my country, and they were emerging at a surprisingly fast rate," she said.
Walter’s assistant said she was away from her office and wouldn't be able to talk this week; we'll bring you what she says when she does. In a Guardian piece 18 months ago, she called it a "mistaken idea that a second civil war would look like the first.”
Instead, she wrote, something plenty scary in its own right could await. “If a second civil war breaks out in the US, it will be a guerrilla war fought by multiple small militias spread around the country. Their targets will be civilians — mainly minority groups, opposition leaders and federal employees. Judges will be assassinated, Democrats and moderate Republicans will be jailed on bogus charges, black churches and synagogues bombed, pedestrians picked off by snipers in city streets, and federal agents threatened with death should they enforce federal law.”
Her assessment of the odds of this happening was not rosy. "The United States is a textbook example of a country headed towards civil war.” Cheerful stuff, what with conventions and elections approaching.
Walter's only ironclad solution: rein in social media. After all, far from illuminating minds (ah, those 1999-era ideals), online platforms cleave us into tribes, reinforce existing beliefs and provoke outrage about those we don't agree with.
"[T]here is a potentially easy fix," she wrote. "Regulate social media, and in particular the algorithms that disproportionately push the more incendiary, extreme, threatening and fear-inducing information into people’s feeds. Take away the social media bullhorn and you turn down the volume on bullies."
Alas, Facebook and Instagram efforts to inject balance into their algorithms have been mixed at best, and X has been going the other way for a while. The social and economic forces that would require Walter's solution to happen have long since left the planet, on a Dow Jones-fueled rocket ride.
Finally, the godfather of modeling and political futurism is Philip Schrodt, a retired University of Kansas and Penn State professor who has been building these models for decades and has consulted frequently for the U.S. government and military.
Schrodt emailed me in response to my question on this topic citing a “time vaguely comparable to pre-WWI where the institutions that had worked for a long time…were, in fact, not capable of dealing with a major hit,” “massive economic inequality in developed economies,” “a major and on-going change in the media environment,” and “major and on-going changes in major military technology ($1000 locally made drones destroying $1M armored vehicles).” All of which means “the potential for serious instability is real, albeit probably localized.”
In a phone interview he was more measured — slightly.
“Compared to the late ‘60’s and early ‘70’s we’re actually in a very peaceful period,” he told me. “Back then major leaders were assassinated, cities were burning, violent groups were blowing up banks. We haven’t seen anything like that.” He paused. “We might. Just not yet.” (More from Schrodt in the next item.)
He is hopeful that a new generation of technology can help us at least partly overcome these challenges — that AI, specifically LLMs, will allow us to forecast threats we wouldn't have seen before and act accordingly. "The technology is almost a perfect match (and sufficiently available as open source) in predicting sequences using vast amounts of data,” he wrote. A host of European projects such as Oslo's Peace Research Institute are already using AI for just these purposes.
What all of this rolls up into is a picture at once both dire and slightly comforting. The nightmares that danced in my head and perhaps yours after seeing "Civil War" were quieted; nothing about what these political futurists are saying suggests such all-out war is even a remote possibility. And there are solid preventative tools in place if it was.
But to a person, the experts all found that conditions are suited to violent flare-ups, possibly chronic, in the forms of events such as riots, attacks on institutions and targeted killings. Reports Wednesday that a Supreme Court justice once had a Jan 6 symbol flying over his house also makes those preventative tools seem just a little less solid.
And while past political violence came either mainly from the left (the Weather Underground circa 1972, eg) or the right (Charlottesville circa 2017, eg), 2024 seems to bring with it potential threats from both sides. (That so many people see political violence only from one side as the “true” malignant form is itself evidence of the problem.) A culture in which a sitting Congressman invokes the Proud Boys’ call for violence and in which encounters between Gaza protesters and the police have turned violent do not, as a general matter, instill optimistic feelings.
Just a few years ago we worried about preventing "another Jan 6;" that phrase came up in more meetings at the Washington Post that I can count. The idea seems quaint in the present light — the notion of a one-off event that, for all its damage, will quickly be over and leave healing and hearings in its wake.
Instead, according to these futurists we'll get a series of smaller events — not as splashy but no less disturbing — in which democratic norms are challenged in the name of a cause, often with violent means and tragic consequences.
We can think of these experts as doctors for the health of a republic. They say Jan 6 represented a symptom. Nothing happening now, they warn, suggests we're even remotely cured of the disease. In fact it may just be starting to ravage the body.
2. LONG BEFORE THE CURRENT GENERATION OF AI MODELS — AND POLITICAL DISHARMONY — PHILIP SCHRODT WAS FIGURING OUT A BETTER WAY.
A University of Kansas and later Penn State University professor with a flair for numbers, he believed modeling political outcomes could give much more accurate pictures of where violence and wars will spring up around the world.
Schrodt has since become the force behind many predictive tools, first using math and regression models and later algorithms and technology to help us understand where conflict might happen. He created many of the systems that fed into the benchmark DARPA project Integrated Conflict Early Warning System. (He was a key member of CIA’s Political Instability Task Force as well.) Schrodt and his tools have been accurately handicapping problem areas years before journalists or pundits were talking about them.
With so much political roil in the U.S. and around the world, this seemed like a good moment to reach out to him. So I talked to Schrodt this week by phone from his home in Charlottesville, VA, about where the data was pointing us. The conversation has been edited for clarity and brevity.
Mind and Iron: A lot of us are worried about the potential for political violence in this country. You've been concerned but not alarmed, saying it's not the late ‘60's. Why aren't you unusually worried?
Philip Schrodt: A lot of what we saw from 1965-1975 we haven't seen yet. We might. But we haven't. We have these groups and their leaders that talk a big line. Particularly Trump. But we really haven't seen them do anything. Is that because post-9/11 law enforcement is better attuned? I suspect that's a lot of it.
M&I: So it's not necessarily that there aren't efforts, they've just been stymied so far.
PS: Yes I think we're seeing a lot more plots foiled before they can do damage or before we even hear of them. People are worried about the Proud Boys. But the Proud Boys are pretty thoroughly infiltrated by law enforcement now.
M&I: Of course all of that could change if law enforcement starts dropping the ball — or if individual elements within law enforcement even join an insurgency. We're only as good as the people protecting us, as so many other countries could tell you.
PS: I think that's true but then you have to look at their relationships with the people engaging in this talk. Trump has alienated the FBI and CIA. That's not a good place to start if you want a revolution — that's not how Hitler or Peron or Mussolini got going. You want the security system behind you. Don’t get me wrong — I think Trump being re-elected would be bad for democracy. We might be on the edge of becoming [modern-day] Hungary, or maybe Poland before the recent election [ie, with right-wing authoritarian leadership]. But on the edge of civil war like Russia in 1917 or Germany in 1918? I don't see that.
M&I: There’s a concern that in such a Hungary case you could get local flare-ups, where cops, say, abet or at least don’t stop a right-wing rally that turns violent.
PS: The perfect example of that is what happened in Charlottesville [in 2017, shortly after Trump took office]. The cops just stood by and let things happen. In some cases I believe they were even feeding information to the protesters. But the thing about local law enforcement is it’s really local. So there’s no single lever that could get you any large-scale conflict. Plus a lot of the people would be pissing off the cops at least as quickly as making friends. That’s not to say something couldn’t happen. But you don’t have centralized control here like you do in Eastern Europe or Russia or a lot of the Middle East.
M&I: And if Trump loses, especially in a close election, where does that leave us? Is Jan 6 just a mild preview?
PS: I don’t think so. A lot of those people have been sitting in jail for three years. Or lost their jobs — in the era of social media it happens in 24 hours. Do you really want to be doing that again? So I don’t see a major mobilization on Washington. Even local protests — you’re going out to protest the cops? They might agree with you. The reality is a lot of the country is clustered by ideology now. Maybe you can have conflict in some swing states where the result is very close. I’d keep my eye on Atlanta, where there’s a big discordance between metropolitan and adjacent areas. But many places wouldn’t be a problem. Strong right-wing support is not a hallmark of swing states.
M&I: Are state governments themselves an issue, where some governors could be egging on a Trumpian revolt?
PS: I am in fact much more worried about state governments. In Florida, in Texas, maybe Louisiana. But you have to look at the interests here. What’s in it for DeSantis? No governor is going to score points with riots breaking out and a bunch of people dead.
M&I: There’s of course also concern about protests on the left and the conflict they could bring — especially at the DNC in Chicago this summer. What is the data telling you there?
PS: To be honest I’ve been surprised by how strong the pro-Palestinian rallies have been on campus. You could easily get a repeat of the Chicago [Democratic] convention from 1968 [in Chicago this summer] or another Kent State [on campus]. But none of those really spread. It was ‘I can’t believe this is happening in America.’ And that was it.
M&I: A lot of us who spend our time thinking about the future wonder what the union will look like decades from now — or, frankly, if there will be a union. Have you done any modeling about what the concept of 50 states looks like long-term?
PS: I’ve got various scenarios. The West Coast breaking off is one. Texas breaking off is another. A place like Virginia could be in a really bad place because you have such differences within the state.
M&I: A split just seems so inconceivable — it goes against everything we were founded on, the ‘whole more perfect union’ that was meant to stop each state from going its own way.
PS: It does. But we should remember that if it did come to that it could happen relatively peacefully. Post-Cold War that happened all the time with countries like Czechoslovakia; they split up quite peacefully. And then you have places like Sudan and South Sudan or Eritrea and Ethiopia where it’s been very violent. But in most of my scenarios I don’t see a split happening. This country has seen much worse in terms of violence and that didn’t split it up.
M&I: You've been an advocate for using the latest tools — at the moment that would be AI and LLMs — to help us predict this stuff. Are we doing that?
PS: We're seeing it in Europe with programs like the Peace Research Institute in Oslo and a program out of Uppsala College in Sweden. But a lot less in the U.S. You have some DARPA projects but like all DARPA projects they’re short-term; you can’t get any continuity. The closest we have is the Global Terrorism Database out of the University of Maryland. But that's just terrorism.
M&I: So weird. Why do you think that is?
PS: I don't know what's happened. It feels like we've dropped back 20 years. Maybe it's just the advantage in Europe that you have 20-odd countries that have resources available to fund these projects. But we need to do more.
M&I: Do you feel confident LLMs will mark a quantum leap in improving the work you began decades ago — that we could really forecast conflict like we forecast the weather?
PS: I do. We need to make sure we customize the models — where we get ‘all of Wikipedia but leave out the fan fiction’ kind of thing. That stuff takes a lot of money so we’ll see how fast it comes. The problem is also that the weather doesn’t know it’s being forecast. These are actors who will respond to what’s happening — who might make a different choice because they know what people are expecting. That makes it harder to predict than a tornado. You can’t rely on it the same way.
M&I: Is there also any concern you have about how these predictions could be used? The potential for policy weaponization or at least policy excuses seems high.
PS: There is. If the model says a developing country is at real risk, then the IMF decides not to invest in it, then the economy craters and then it becomes a real risk whether it was before or not.
M&I: The self-fulfilling prophecy.
PS: The self-fulfilling prophecy. I worry about that a lot, especially with developing countries. But economists have been dealing with this idea [of forecasts influencing the reality] for years and it’s mostly fine. There’s always a resistance to predictions. There was a backlash to hurricane predictions when they started — critics said they could panic people. And who would oppose a hurricane-prediction system now?
3. SCARLETT JOHANSSON IS FAST GAINING A REPUTATION AS SOMEONE WHO CALLS OUT LARGE COMPANIES for their digital exploitations. She's the Black Widow of Big Tech.
A few years ago she sued Disney for taking her Marvel movie straight to streaming. And on Monday, as you almost surely have heard by now, she went public with the back-and-forth that preceded Open AI releasing GPT 4o.
The recap: As we wrote last week, OpenAI released its new product with enhanced voice capabilities. One of those voices, known as Sky, sounds like Johansson in “Her” — I mean, until someone pointed it out it didn't really sound like that to me, but I'm hardly going around listening for such things. Anyway, it did sound like that to a lot of people. Combine that with Altman himself noting "her" when posting on X about the release, and it felt strongly like he was drafting off her appearance in the movie.
That all seemed cringe but basically fine, a tech bro nerding out over a tech movie. Except it may not have been a coincidence at all. According to the statement Johansson released Monday, Altman had actually been negotiating with Johansson over using her voice a while back. She says she rejected the overtures, and that was that. Then she heard from him again last week, just before the GPT 4o release, asking her to reconsider. She says she hadn’t even had a chance to respond when the product dropped 48 hours later.
Johansson suggests the act was both bad faith and possibly allegedly illegal, since a company that seemed to think it needed permission was now marching ahead without it. And also, one might add, kind of brazen, because if you didn't secure permission for something you may have needed, the last thing you'd want to do is call attention to it by building it the same way anyway. And then posting about the connection.
As it turned out, OpenAI didn’t take Johansson’s voice. On Wednesday evening my former colleague Nitasha Tiku broke the story that the actress that was actually used for Sky just happens to sound that way (presumably this was part of the appeal to Altman given his love of the movie) and was signed even before he even made his first entree to Johansson. (Tiku reviewed the documents and recordings and concluded all was kosher.)
But it still was an odd set of moves on Altman’s part. If you were going to land someone who sounds like Johansson, then why go to her afterward? And then if you’re not going to get her, then why continue to use the voice that sounds like her?
And even if you did do all that, why go out of your way to cite “Her” when the product was released? Like, at least let the fans do it. Then you could have plausible deniability. But why put an idea in people's heads when that idea is the very thing you wouldn't want a spotlight on? Maybe you don’t care about the negative attention. But you might care about a lawsuit?
Also and most oddly, wth is up with contacting Johansson two days before the release? Would you have delayed it to get the recordings done if she said yes?
It’s all really strange. Now OpenAI is saying that they won’t use Sky for the time being. But what a needless mess.
In an irony that may or may not have been known to Altman, Johansson herself was not the original voice in “Her” — Samantha Morton was. Morton was the one who shot the whole part on set. Johansson only replaced Morton in editing, because director Spike Jonze liked her voice better. There’s some kind of weird history-inverting-itself thing going on here.
Anyway, as with so many moves Altman has made, it's hard to know if this was four-dimensional chess, an ill-conceived troll or just simple sloppiness — someone, somewhere, is working on a research paper comparing his low-key provocations and Musk’s noisy ones. (Btw the release timing of GPT 4o was itself a troll, dropping a day before Google was scheduled to make its big AI Assistant announcement.)
What we do know if that it caused an avoidable backlash for the company, which is heavily backed by Microsoft. Satya Nadella could practically be heard making one of his “What are you doing, man?” calls to Altman he has no doubt made before during one of these escapades.
This especially hits a nerve because we’re all pretty worried — justly — that AI companies are going to hoover up our creativity and identity to put in their giant profit-making machine without any notification, let alone compensation.
It could, some might even think, turn the consumer tide against OpenAI and its GPTs.
Not so fast. For all the social-media ScarJo support, the reality is that Altman will have the last laugh. Because while the incident temporarily casts him and OpenAI in a bad light (again) and surely didn't give an easy weekend to the publicity team at Microsoft, it will have very little effect on adoption — which is the only thing Open AI (and, as a result, we) care about.
Sure, it makes the company a sneering Big Bad Wolf, an Aaron Rodgers-y ‘we owned Hollywood’ villain. But for all the handwringing over incidents like this in Big Tech, the relationship between cultural negativity and market backlash is almost always thin. The sooner we realize this the better.
“The Social Network” and its prickly Mark Zuckerberg portrayal did not make people stop using Facebook, and all the negative Amazon stories in the world won’t make us give up our Prime memberships. Listening to policymakers, we worry about TikTok and Chinese spying — and then go right back to checking out that hilarious Blinding Lights vid. What we think of the people running these companies has never been a mass hindrance to devouring their products.
The tech moguls have long understood what we’ve only recently begun to figure out: convenience and addictiveness is why we come to them. Conscience or that icky feeling? That isn’t anywhere near the list. And no voice — even Scarlett Johansson’s — is ever enough to convince us otherwise.
The Mind and Iron Totally Scientific Apocalypse Score
Every week we bring you the TSAS — the TOTALLY SCIENTIFIC APOCALYPSE SCORE (tm). It’s a barometer of the biggest future-world news of the week, from a sink-to-our-doom -5 or -6 to a life-is-great +5 or +6 the other way. Last year ended with a score of -21.5 — gulp. Can 2024 do better? So far it’s been pretty good. This week it’s a….not great.
THEY HAVE SEEN THE FUTURE, AND IT IS VIOLENT: U.S. political turmoil ahead -4
MAYBE THERE’S A LITTLE REASON FOR HOPE, SAYS THE GODFATHER OF THOSE PROPHETS: +1.5
WHAT IF OPENAI GOT A WHOLE BUNCH OF BAD PRESS AND NO ONE CARED: -2