Transcription
I'm federal criminal defense attorney Ron Chapman, and today on Off Air, we're going to continue our Epstein series and discuss a $30 million donation to Harvard that led to a secret Harvard lab and the rise of the viral internet.
14 years ago, Martin Novak discovered what makes things go viral. He did this based on a $30 million pledge by Jeffrey Epstein. The result: Facebook, Instagram, and Twitter. His theory: Darwin is incomplete. There's a third mechanism that's important for evolution to do its work, and that is cooperation. Economists have talked about game theory for quite some time, but game theory has never been an evolutionary concept, and we haven't really seen it in the real world work until Novak's theories and until Jeffrey Epstein's donations.
But why? Why does a billionaire want to invest in this sort of technology and research? Why was Jeffrey Epstein so interested in developing game theory and social evolution? Well, what you're going to learn during this episode of Offair is that Jeffrey Epstein was interested in something that would change our world forever. And that is viral content, artificial intelligence, and eugenics. And by the end of this episode, you're going to understand exactly what was going on at that secret underground Harvard laboratory and why a billionaire spent $30 million on research leading to overthrowing countries, the destabilization of elections, and even the destabilization of the United States of America. Stick with me here for this entire episode on Offair Air. And if you love the content so far, especially some of my prior episodes, take a minute to like and subscribe and share it. It means a lot to me and the creators of this episode.
Between 2003 and 2019, a set of ideas moved from academic theory to lived reality. At Harvard, Martin Novak's Program for Evolutionary Dynamics treated social behavior, cooperation, norm enforcement, emotional alignment, belief formation as a system that could be modeled and replicated. At MIT and adjacent institutions, researchers treated social media as a measurable engine of virality. Technology platforms scaled those dynamics to billions of users: platforms like Facebook and Instagram. Political operatives learned how to weaponize them: Cambridge Analytica. What emerged was a new landscape where contagion isn't only something biological like COVID-19, but where cooperation between individuals can be engineered or even undermined, and where hyperreality—a simulated world of narratives and symbols—can become more powerful than even the events it describes. We saw events like this during the Arab Spring. We saw events like this during Benghazi. We see events like this in Minnesota on the streets of America. Are these organic? No. What you'll realize during this episode is that many of these things are socially engineered because of virality, Jeffrey Epstein and Martin Novak's research, the theories, the models, the experiments, the platform consequences on Facebook, Instagram, LinkedIn, and the web of connections. From John Po Dexter's defense era data ambitions to Zuckerberg's global social graph, from the dual role in Facebook and Palantir to Cambridge Analytica's targeted narrative warfare, we see these tools manipulated over and over again to change the way Americans think, to change the global atmosphere. All of this research came from one tiny office complex on Harvard University's campus.
Let's begin with social networks. In my prior episode, I talked about how on February 4th, 2004, Facebook was born, the same day a DARPA project closed down. You can learn more about that on the prior episode. Coincidence? I think not. Two years before Jeffrey Epstein met with some of the biggest AI researchers in the world in the Virgin Islands, what did they discuss? The need for additional data so that they could understand humans enough to be able to create artificial intelligence. What came out of it was a new direction in artificial intelligence, a new dawn, and the decision to go out and grab as much information as humanly possible through social networks. John Po Dexter, through DARPA, decided to try to do this on his own with a government program, but that obviously got embroiled in controversy simply because Americans wouldn't have tolerated the government having that sort of data that they didn't ultimately give up themselves.
In social networks, though, ideas, behaviors, emotions, and norms can spread from person to person. We know this already. This is not new to us. But Martin Novak, on Harvard's campus, fueled by the money from Jeffrey Epstein, found out that social contagion means our contacts influence us, causing trends to propagate through communities. Classic epidemiological models like susceptible, infected, and recovered—that's typically how somebody would catch a disease and recover from it—have been adapted to social contexts. The phrase "viral" is not a mistake. The spread of information and the spread of emotional sickness as a result of what we see on social media acts very much like a virus in the real world, with one little tweak by Martin Novak.
One especially important adaptation by Novak, as a result of his research, is the SISA model. Now, I said previously, viruses work on a susceptible—you're susceptible to the virus. You get infected. You recover, and then you don't get the virus anymore. That's kind of how viruses die out. The world recovers, we get better, we get immune. Well, that doesn't really work well when it comes to information because you need information to spread and continue to reverberate through society. So, if you're somebody who's a bit of an evil genius and you want to find out how to continue to get information spread in a way that meaningfully impacts society, what you want is a SISA model. That is: susceptible—you're susceptible to the information; infected—you get infected by the information; susceptible again with automatic adoption of the information. How does that look? Well, we see it in Minnesota as we speak. Algorithms fueled by Chinese influencers invade the United States. What are they targeted at? Not real social causes, but made-up social causes that are meant to rile people up. People are bombarded with algorithms that are supposed to mean something to them, but what they really do is inflame their passions, causing them to go out on the street and do things that they ordinarily wouldn't have done, but they genuinely believe are appropriate. Susceptible: they're angry at society about something. Infected: the information reaches them. But instead of recovered, which normally happens with the virus, what we start to see is they get susceptible again with the automatic adoption of the information. Next time they see it again, they believe it as true and they adopt it. Which means you can nudge people over and over again from simply being a little bit irritated with conservatives to literally rioting in the streets, thinking that ICE is taking over their town. This isn't a mistake, ladies and gentlemen. This is social engineering, Martin Novak and Jeffrey Epstein style, and it is what is currently taking over our country.
The SISA framework, as I just described, matters so much to us because it doesn't treat adoption as purely social. People may catch a behavior from peers, but they may also adopt independently due to external factors. This distinction is essential for social media. A user can adopt a belief because a friend shares it, or because the algorithm places it in their feed. Back in 2006, when the feed came about through Facebook, when Mark Zuckerberg decided to create the News Feed, which was one of his most important inventions with Facebook, he did so because at that time, Novak's research and other researchers found that you would adopt beliefs from other people that you knew. And so Facebook would pick from the list of beliefs that your friends were sharing. This is how you move a social network. This is how you manipulate people using information. This is how a billionaire influences all of society through their donation. In fact, I should say a couple of billionaires.
Contagion modeling helps explain why virality on social media—how a meme or a hashtag can go viral when its basic reproduction rate exceeds one. Now, that may sound like a very complex theory, but the reality is, things would spread very slowly on social networks if the rate of reproduction is one. Somebody sees it, somebody reproduces it. Network scholars have shown that viral features and products or platforms—easy sharing, contact imports, importing your entire phone book, frictionless forwarding—can deliberately amplify peer influences. MIT's Sinan Aral demonstrated that firms can engineer social contagion by designing a viral product feature and seeding key influencers, boosting a word-of-mouth spread nearly overnight. We've seen this over and over again. In fact, some of my videos have gone viral. In short, contagion theory became foundational for understanding information dissemination and viral growth in the early Web 2.0 era. Now we're talking 2007 to 2014.
Now, there's a concept that I want to talk about that is very, very important, and I wrote about it in my last book. It's called hyperreality, and it's very important for you to understand all of this. The concept of hyperreality is associated with a guy named Jean Baudrillard. In fact, The Matrix was based off of a book that he wrote. He describes a state where simulations or representations of reality become indistinguishable from, or more influential than, reality itself. A good example here is the Arab Spring. A Tunisian market stall worker sets himself on fire in public, and he sparks a global revolution that had absolutely nothing to do with what he stood for. He became a symbol, and that symbol became more meaningful to people than the individual. In fact, I talk about the World Series of Poker in my book. There's a wonderful story in there that might help you understand hyperreality. In a hyperreal environment, people take symbols and media-constructed narratives as real, losing sight of any unmediated truth. Social media creates exactly this condition. And each user cultivates a profile—a curated simulacrum, as Baudrillard calls it—of self and consumes a personalized feed. Basically, what that means is they create a version of themselves, and other people do, and they consume the feed that the other people want them to see. Online identities and narratives begin to overshadow reality, blurring genuine experiences and digital performance. Instagram users can project idealized lives, and audiences emotionally respond to the images as if they were the thing itself. The hyperreal dynamic is fertile ground for misinformation. We see this happening in Minnesota right now. People project an image of being freedom fighters. They see a cause pop up in front of them, and they jump at it because of that projected image. Fake news and conspiracy theories thrive when this happens because, as we all know, the simulation of the truth is easier to manufacture, and in ways, it can be designed to manipulate people in a much easier way. Social media structures and algorithmic feeds, plus self-selected communities or silos that we sort of create—echo chambers, if you will—each with their own alternate reality, cause people to start becoming entrenched in their way of thinking, and false narratives can circulate and reinforce themselves very fast. In fact, I think it might have been Mark Twain who said, "A lie can get halfway around the world before the truth even puts its shoes on." I think that was a big part of a closing argument I gave recently. I love that line.
Hyperreality amplifies social contagion. That's why I wanted to bring it up. Not only do behaviors spread, but beliefs and distorted realities spread as well. A recent analysis of online political discourse described coordinated buzzers or influencers for hire, creating a hyperreal identity of politics, flooding feeds with manufactured narratives that inflamed polarized people into social conflict. Again, we saw this in the Arab Spring. We see this in Minnesota happening in real time. And we've seen it across the country with protests related to George Floyd and other scenarios. Now, some of those are a little closer to the truth. Minnesota: very far from the truth. George Floyd: it depends on who you are and what you believe. But the reality is, people are using these strong emotions to bend other people's thinking into certain types of action. As contagion and hyperreality take hold, polarization absolutely flows from it. Online communities begin to segregate. We see silos in places like Reddit. Individuals connect with like-minded peers across the country, but then they start to disregard the people around them, becoming more entrenched in their way of thinking. One great example is the rise of the incel movement that I write about in my last book. Over time, this narrows exposure to opposing perspectives and produces intellectual isolation. Studies indicate that filter bubbles intensified by personalized feeds have accelerated opinion polarization within insular communities. Confirmation bias becomes dominant. Confirmation bias is the prospect of only believing those things which are very similar to the things that we already believe. People encounter posts validating their narrative while contrary evidence is filtered out or dismissed. In fact, we probably flick through our feeds 100% of the time looking for things that are consistent with our worldview and rejecting everything else. This environment fosters more extreme viewpoints than any other environment we could have.
Now, where does this come into Novak and Epstein? Well, the fake news epidemic of the mid-2010s showed us a pattern. Fabricated stories spread within partisan communities with very little correction, boosting a lot of mistrust and anger, which entrenches both sides. A landmark MIT study done by one of the labs that Jeffrey Epstein fostered in 2018 found that false news spreads further, faster, deeper, and more broadly than true news, especially on Twitter, largely propagated by humans and polarized networks eagerly sharing and confirming their own falsehoods. The architecture of social media networks combined with human psychology fragments the public into polarized communities, each inhibiting its own hyperreal narrative reinforced by contagious information flows.
Let's get to Novak. Now, I talked about him a bit in my last episode, but here we are going to take a deep dive. Martin Novak was a mathematical biologist at Harvard, and he launched the Program for Evolutionary Dynamics. I'll refer to it as PED. He started it in 2003 to study the mathematics of cooperation. He studied networks and the evolution of social media networks and biology. Novak and his colleagues, many of whom later moved into positions at Harvard, Yale, MIT, and beyond, focused on cooperation and social dynamics in an increasingly interconnected world. Their core question was simple and vast: How can cooperation emerge and persist in a competitive environment? The answer matters in biology and in human society, but it also is very, very important for the development of social media networks.
Novak identified five fundamental mechanisms for the evolution of cooperation: First, you have kin selection; then you have direct reciprocity; then you have indirect reciprocity; spatial network selection; and then finally, group selection. Novak's research essentially told us that sometimes it benefits you to cooperate with the group, even when it doesn't necessarily serve yourself. Prior to that, most researchers believed that almost everybody acted in their own self-interest all the time. But here, there was strength in social groups in humanity, and that cooperation within social groups might allow certain people to drag other people along when making decisions. This is how social networks, this is how confirmation bias of the things inside of your feed can pull you along. If a lot of people in your social group believe something that is absolutely wacky, and you see it come up on Facebook, you are more likely to agree to it because of the social consequences of ignoring them. Let's take, for instance, the fact that the moon is made of cheese. If I tell you this on Off Air, and all of your friends start sharing information about the moon being made of cheese, if you actually believe that the moon is not made of cheese, you are probably going to stay silent. This means you're going to cooperate with the view that the moon is made of cheese, or at least you're not going to reject it. We then get an echo chamber of the moon made of cheese until everybody starts to believe the moon is made of cheese. Switch "moon made of cheese" to global warming, and you start to see what I'm talking about. The social consequences of saying that global warming is not real are very, very severe. In many circles, you might just get outright canceled for saying something like that. And so the reality is, is that cooperation sometimes occurs even when a person is sacrificing a bit of their own worldview.
Now, Novak at PED used mathematical models and computer simulations and lab experiments to really understand this concept. One of his major contributions was formalizing social contagion models and human behavior. He came out with a paper in 2010 that really described how social media could explode. Novak and his other researchers effectively studied how to get things to spread even more on social media, and their findings suggested that there was a way to achieve a sort of overall contagion with information. His results were very important, and they challenged simplistic notions of how we share and believe information.
Now, PED and its collaborators ran a lot of experiments. David Rand, one of Novak's close collaborators and later a professor at MIT Sloan, helped lead pioneering network experiments on cooperation. In a 2013 PLoS paper, Rand and his colleagues had volunteers play work public goods games to test whether cooperation spreads. They found clear evidence that cooperative behavior can be contagious in certain network conditions, but network structure changed the outcome. In static networks, cooperation and selfishness both spread. In fluid, dynamic networks where people could rewire ties, selfishness remained contagious. What does this mean for the casual viewer? Well, you have your news feed. You have your social circle. Where that social circle can change quite a bit, it's very, very difficult to get people to move along with cooperation. But where that social circle for the most part stays the same, where you are already entrenched in your worldview with a lot of people, cooperation with the group matters a lot. How does this translate into something like Facebook? Make sure you don't have too much flowing in and out of your social group, and make sure that you are bombarded with the same signals and messaging over time. Don't let a lot of new messages into your algorithm, and you can effectively shape the way people think.
Beyond these models and experiments, PED intersected with broader network science through figures like Nicholas Christakis. He was with Harvard and then later Yale, whose controversial work suggested obesity, smoking cessation, and happiness spread through social networks up to three degrees of separation. Those studies spotlighted network effects in public health and sparked debates about causation. PED became a lab linking Harvard scientists with network theorists across the world. Ultimately, some of its research made it into some of the most prominent social networks in our day: Twitter, Facebook. The development on these platforms was not by mistake. It was very much by design from this secret Harvard laboratory.
Now, why do I keep saying it's a secret Harvard laboratory? What we found out from the Epstein files, and one of the things that not too many people are talking about, is that Jeffrey Epstein pledged $30 million to the development of the PED laboratory. Now, you would think that this sort of contribution would mean that there would be a lot of published work coming out of it. And there was some, but not enough to justify that amount of donation. In fact, something very fishy went on here. PED was not located on Harvard's campus, but it was located in an office building, and its space was leased from a private entity. Access was private, control was private, the research was private, and nobody at Harvard really understood what was going on at PED. From 2004 onward, PED was the source of a ton of information about viral technology. Ironically, it started blocks away from where Facebook first began, from where Peter Thiel found many of his initial startups, and where viral social media first got its foothold in the United States. Coincidence? I think not.
Facebook's early growth strategy relied on harnessing social ties, inviting friends, tagging photos, testing the core idea that once you built a web, people would not leave it. That made them cooperators, and it made them easily influenced once you could bombard them with the right signals. Zuckerberg wasn't reading Novak's papers himself, but the people who developed his network obviously were, including his investors. If your friend likes, shares, or comments, you see it, you engage, your engagement becomes the next transmission event, causing your friends to see it. If you like a funny video, you like a meme, you play a game, things go big. And a lot of the games that were put forward on Facebook were really just tests. In fact, that's ultimately how Cambridge Analytica was able to get a foothold.
Tech companies began collaborating with academia in this space. And in 2012, Facebook's data scientists conducted the infamous emotional contagion experiment. They altered the ratio of positive versus negative posts shown to roughly 689,000 users. They found that users exposed to fewer positive posts later used more negative words, and vice versa. Moods were contagious at scale without direct interaction, simply by exposure into the feed. The experiment prompted ethical backlash but underscored the platform's influence and power, echoing contagion theory that Novak had previously discussed. The twist is that in social media, algorithms become vectors, selectively amplifying signals based on what the algorithm wants to amplify. And who's behind the algorithm? Facebook itself.
In parallel, MIT researchers pushed the study of virality into operational territory. Sinan Aral and colleagues used randomized trials to show that product adoption can be turbocharged through viral design, turning contagion theory into viral marketing strategy to sell products. Another MIT-adjacent effort: the Laboratory for Social Machines. In 2018, researchers published a major Science paper on Twitter cascades. False news was 70% more likely to be retweeted than true news and reached 1,500 people roughly six times faster on average. They concluded human behavior, not bots, drove the spread, implicating novelty seeking and outrage sharing. This evidence suggested an uncomfortable equilibrium: absent intervention, online contagion may naturally favor the most emotionally charged or novel content, often false, producing a hyperreal information environment that is built fundamentally on false information.
Cooperation researchers like Novak also noticed these impacts. But this influence had a much darker twist to it. The same principles that spread healthy behaviors or innovation can spread division, manipulation, and manufactured belief. And so, from 2013 to roughly 2019, Jeffrey Epstein's death, academia, tech entrepreneurship, government surveillance, and social manipulation increasingly braided into a single system.
Now, I've talked about Lifelog. Back on February 4th, 2004, DARPA shut down that controversial program. Mark Zuckerberg that same day launched Facebook, and all of the data was at the fingertips of Peter Thiel, Palantir, Mark Zuckerberg, and Jeffrey Epstein. Peter Thiel sits right at the heart of this nexus and is a very important figure. In 2004, he became Facebook's first major investor and joined the board, helping to drive the platform. Remember, he co-founded Palantir, and Palantir needed all of that data from Facebook in order to continue its mission. It's no secret that Palantir was a DARPA, In-Q-Tel, CIA-funded project. Facebook provided the information while Palantir continued to manipulate it and figure out how to exploit it. John Po Dexter, a former national security advisor, led DARPA's Information Awareness Office in 2002. He was one of the first people pushing Total Information Awareness, and he was involved in mining global data footprints to detect networks of bad actors. Ultimately, he was defunded by Congress, but that's when he decided to go underground with In-Q-Tel, linking up with Thiel and Palantir CEO Alex Karp in order to start their mission forward with early adoption of Facebook and other technologies. In fact, it's not even clear that Peter Thiel was using his own money when making these investments in Facebook and other technologies. And given the backdoor that we saw implemented into Facebook, Google, and other technologies, I wouldn't be surprised to find out later on that a lot of that was government funded in the first place.
Zuckerberg founded Facebook just down the street. Novak is doing his research. The proximity is telling. The campus incubating advanced network science also incubated the defining social media platform of our age. Facebook grew like contagion, just like Novak had predicted. As Facebook became the global public square, it confronted consequences. Researchers were modeling echo chambers, viral misinformation, coordinated harmful behavior, and Zuckerberg was nowhere to be seen. In fact, it wasn't until the Cambridge Analytica scandal that Zuckerberg even got before the American people to say something was wrong here. And even then, it wasn't the most truthful. Internally, Facebook had already run research confirming its ability to shape sentiment and behavior. It knew it had this power, and it also given certain people access to it. But the American public wouldn't be warned of this harmful impact until it would actually disrupt an American election and cause Mark Zuckerberg to come before Congress.
Cambridge Analytica's strategy was to use big data, especially Facebook data, to perform psychographic microtargeting. That's what they called it. Its origin involved academic crossover. Cambridge researcher Alexander Kogan built a personality quiz app in 2013 and 2014. Many of you probably even took it, and harvested data from about 270,000 users, but also from their millions of friends and their millions of friends' friends. He used Facebook's Graph API, which allowed unfettered access into the data of friends and friends of friends. Cambridge Analytica obtained data on an estimated 50 million users from that original 270 quiz takers, and it did so without consent. With these profiles, it used machine learning to predict personality traits and political leanings, then delivered tailored content designed to change behavior. This was an influence system built on contagion principles created by Novak and hyperreality tactics. The goal was to identify susceptible individuals and infect them with propaganda, seeding narrative cascades in vulnerable communities.
Now, I'm going to continue and talk a bit more about Cambridge Analytica, but so far, if you like the content, I'd invite you to click like, subscribe, and make sure you share, but also smash that bell button so that you can get notified of future episodes. There's going to be a part three coming out soon.
Now, Cambridge Analytica's technique depended on the cooperative nature of like-minded groups sharing content internally, just like Novak predicted. But it also required viral dynamics amplifying this sort of persuasive messaging and the hyperreal effect that ads, memes, and very highly charged information entails. Steve Bannon served on Cambridge Analytica's board. Bannon's philosophy of flooding the information space with misinformation leveraged the sort of outrage that researchers knew would ultimately go viral. And crucially, whistleblower testimony revealed that Palantir employees helped Cambridge Analytica exploit the Facebook data itself. Palantir initially denied formal involvement but later acknowledged that a Palantir employee had engaged with Cambridge Analytica in a personal capacity and obtained the entire data set of 50 million users. Emails showed Palantir staff brainstorming with Wyatt, one of the founders of Cambridge Analytica, about refining data mining, including ideas like building an app to gather more Facebook data. This Palantir, Thiel's firm, Peter Thiel's firm, bridging intelligence-style data thinking into political influence campaigning shaped the way many Americans thought. Government surveillance, DARPA programs, misinformation, the misuse of individuals' data, political manipulation converged into a single system. This was game, set, match for DARPA, Peter Thiel, and the powers that be. And from 2003 to 2019, a Jeffrey Epstein-funded program was providing the information and the research necessary to make this happen.
The outcome of the PED research program is twofold. Intellectually, it produced frameworks and it showed how cooperation and contagion modeling can be used to exploit people. Practically, it foreshadowed some very dangerous events. But privately, it was sold and used to manipulate individuals, to make neighbors and relatives fight, to socially engineer people's thoughts. Novak has framed cooperation as something that's essential. It's the only thing that can redeem mankind. Is that true? Or is Novak and his buddy Epstein attempting to socially engineer people so that they control reality?
Now, we're not done yet. We still have a lot more to talk about when it comes to Peter Thiel and Cambridge Analytica and its impact on elections. We have to dive deeper into this, and I know this was a relatively complicated episode, and it certainly took a lot to put together. But I hope that you can be patient with me throughout this Epstein series because we're going to get into something that is groundbreaking, exclusive, and something that you will only find on Offair Air on the next episode. I also want to give a big shout out to all of the members. I see you all joining up, and it really helps the program and helps us produce high-quality content for you all. Let me know what you think about the episode in the comments section and consider joining up for that membership. Thanks for joining me.