How AI can debunk a conspiracy theory in 8 minutes

From recorded future news and PRX, this is Click Here. I want to tell you about two brothers from Saskatchewan. One of them is Gordon Pennycook, and the other is his older brother. He works in the oil fields in Alberta. And like a lot of families during the pandemic, they started having some difficult conversations. Although Gordon's family had a particularly intense version of this. My oldest brother had too much time online, I think. And as time went on, his views got more and more extreme. It kind of fell down the right wing rabbit hole. It became very conspiratorial. Gordon could actually see it happening. Every time he saw his brother, it felt like he'd moved a little further away. A lot of the spare moments that he had were spent watching YouTube videos where the person is just screaming about politics or whatever. Progressively over those visits, his views became more extreme over time. And this wasn't some abstract disagreement about politics. His brother was starting to believe the world itself was a much more dangerous place than Gordon thought it was. For example, at one point, Gordon got a job at Cornell University in New York, which would have been good news, except his brother was alarmed. When I told him that I was moving to the States into New York, you know, his eyes got kind of widened because he was like actively concerned that like, at any day the Civil War is going to start and there's going to be shooting in the streets and stuff like that. And just to be clear, Gordon's brother wasn't talking about the old Civil War, the one with Abraham Lincoln. He thought another one was coming. And he was worried about his little brother finding himself right in the middle of it, which is actually kind of sweet, except for the part where Gordon didn't think there was going to be a Civil War. And there was this other thing that made this whole situation particularly strange. Gordon studies this. He's a psychology professor and his research is on why people believe things that aren't true. He studies misinformation, conspiracy theories, and that's sort of thing. So if there were a person in the family who would know what to do here, it'd be Gordon. But he didn't. A show about how technology is changing everything. I'm Dena Temple Reston. I think most of us have had some version of Gordon's problem. Someone you know believes something that seems completely and obviously wrong to you. And you think, "Okay, I can fix this. I'll show them the thing they haven't seen. I'll send them the perfect article and actually change their mind." Gordon started wondering about that and decided to test it. That's after the break. Stay with us. Support for click here comes from Serval. Your IT team could be building infrastructure, solving real problems. Instead, they're stuck resetting passwords and approving access requests, running the same support playbook over and over again. And that's not a capability problem. That's a work design problem. Serval fixes it. You simply describe what needs automating in plain English, no technical jargon needed and Serval builds the automation for you. No drag and drop builders to wrestle with, no consultant fees, eating your budget, no waiting months for implementation. The result? IT stops being a bottleneck and becomes the operational engine that moves the business forward. And Serval backs this up with a guarantee that half of your IT tickets will be fully automated. Test it for yourself. Learn more or start a free four week pilot at serval.com/click here. That's s-e-r-v-a-l.com/click here. Serval.com/click here. Support for click here comes from Quince. You know those pieces in your closet you reach for again and again, the sweater that goes with everything, the jeans that always fit just right. Those are the kinds of pieces that earn a permanent place in your wardrobe. And that's exactly what Quince does so well. Quince offers elevated essentials and premium materials like Mongolian cashmere, organic cotton and stretch denim designed to be worn again and again. And the price has actually made sense. Those Mongolian cashmere sweaters start at just $50. As the weather cools, Quince has you covered, down jackets, cashmere scarves and gloves, even jewelry for that perfect finishing touch. Quince works directly with ethical factories cutting out the middle men. That means everything is priced 50 to 80% less than similar brands. And I start just clothing. Quince brings that same approach to bedding, bath, and furniture too. All well-made pieces for everyday life. For me, I find excuses to wear my Quince cashmere quarter zip and soft drapes really well and I have it in a bunch of colors. Find the fall pieces you reach for most at Quince. Download the Quince app for app exclusive offers or go to quince.com. Get free shipping on your order and 365-day returns. Now available in Canada and the UK too. And when Quince asks where you heard about them, let them know click here, sent you. It's the best way to support the show. So Gordon Penny Cook had this problem. His brother was disappearing down this rabbit hole of conspiracy theories. And Gordon, a guy who studies why people believe things that aren't true, was standing right there, watching it happen. Which raised an obvious question. So were you able to change his mind? I mean, you're literally an expert in this. Yeah, although I don't like, I'm from a small town. Most people don't really know what a professor is or does. And my family is included. And so I make an effort not to be condescending or to come in on my. You're a high educational horse? Yeah, exactly. So his brother wasn't exactly waiting for Gordon to straighten him out. He wasn't a fan of experts or the mainstream media. Instead, he had people he trusted. People like Stephen Crowder. Welcome to part two of this installment of Change My Mind. The premise is a left. It's violent. Crowder is a right wing, you two personality. And a lot of his videos work like this. Here's a big argument. Here's a big grievance. And somebody, not you, dear Crowder follower, is definitely wrong. I mean, I don't sure he's doing now, probably the same thing. But just makes videos where you're screaming about stuff. And Gordon knows he could have told his brother, "Don't listen to this guy." But he knew that wouldn't get him anywhere. So instead, he tried something else. He asked his brother a question. That's why I asked him, like, what is it that constrains Crowder? Like, what stops him from lying? For a journalist, what stops you from lying is that you could lose your job. You know, I mean, if anything, for a YouTuber, like, there's incentive to lie because it'll get you more clicks. That seemed to give his brother pause. And then he thought about it for a second, and then he said, "Yeah, but he doesn't lie, though, so." So not exactly a breakthrough. But that little exchange left Gordon wondering about something bigger. Once someone has really attached themselves to something that isn't true, is there any getting them back? There's some pretty gloomy conventional wisdom about this that once people have made up their minds, the facts don't matter anymore. But Gordon doesn't buy it. I don't believe it. And I can say with some reasonable confidence that it isn't true because of our own studies on the topic. In 2024, Gordon and his team recruited thousands of conspiracy theorists for a study. Participants were shown an AI chatbot called debunkbot, and they were asked to have a conversation with it, playing out why they believed their conspiracy theory was the God's honest truth. The AI would listen patiently. And then only after the conspiracy theorist was finished making his or her case, the chatbot replied, "We told the AI, don't be polite, just give them straight facts, don't do any rapport building, don't use any other techniques, only provide the facts and evidence. And then what the AI does is it provides strong counter-arguments for everything that the person says." Take, for example, a conspiracy theorist who believes the attacks on September 11th were perpetrated by the United States government. The person logs on and provides their argument for a particular theory. Someone might say, "Oh well, I think it's an inside job because the melting point of steel is not sufficient. The fire didn't burn hot enough for the steel to melt." The AI would say, "It might be the case that fire doesn't burn at the right point, but at a certain set of degrees, steel loses half its carrying capacity and that's sufficient for the building to collapse." The chatbot, in other words, is Blunt. It attacks the person's reasoning directly with a counter-argument tailored just for them. After the conversation, the team then measured if for how much people's beliefs had shifted. Gordon was gobsmacked by the results. After, like, eight minutes of conversation, a quarter of the people people who bleed the conspiracy don't believe it anymore. So while we like to assume that conspiracy theorists never changed their minds, 25% of them did in just 8 minutes. When we ran the experiment, I was like, "This is not going to work." Way more people than we thought changed their mind. Not only that, two months later, when the team checked in again, those beliefs hadn't come back. And there was this other thing they hadn't expected, the kind of transfer effect. The research subjects became more skeptical of conspiracy theories more generally. Even ones the bot hadn't addressed. The critical thinking was contagious. Now, we don't want to overstate this. Three out of four conspiracy theorists stuck to their guns. But Gordon thinks that, given more time at the chatbot, those numbers would fall too. His bot, by the way, is available for anyone to use. But it does beg the question, "Why did the AI chatbot work so well?" Gordon thinks the machine has some advantages over its human counterpart. The AI keeps a cool head, and doesn't get distracted. AI doesn't roll its eyes, it doesn't throw up its hands and frustration, it doesn't lose its cool or call the other person stupid. And it doesn't make the kinds of mistakes that put people on the defensive. And my brother is talking about all these different things. There's a thousand different points that I want to make, whether they're the right point to make is like maybe not that well predicted by how important I think they are. The most important part is, does the evidence specifically address the things that the person believe, a plausible alternative explanation with facts that makes sense? And the AI is very good at doing that. All this is to say, facts aren't dead. Evidence matters. But you have to be unemotional. A more elaborate, personalized evidence-based argument can have a big impact. Except Gordon's experiments turned up something else, something a little more unsettling. Because if this works so well at changing someone's mind in one direction, it begs the question, "What happens if you point it in the other?" So your experiment really is about truth, or is it about persuasion? It's about whether truth is persuasive. So they ran the experiment backwards. This time, instead of asking the AI to debunk a conspiracy theory, they asked it to convince people that one was true. The AI convinces them that it is true, and then we just call that bunging. They're just like convinced some of the conspiracy, and those things work pretty much the same. Which is not just a bit of a bummer, but it also complicates our tidy little tale about truth. Because it suggests that truth may not be the thing doing the persuading. People don't know, they don't have divine access to the truth. What they have access to is whether the argument makes sense, and whether something makes sense depends on whether it seems to be based in facts and evidence. So Gordon's good news, and his bad news, are basically the same news. People are persuadable. You can use that to pull someone away from something that isn't true or toward it. Which brings us to another way of attacking this problem. Instead of getting better at changing someone's mind after the lie gets there, what if you got there first? Stay with us. Support for Click Here comes from Odo. Business software is expensive, and when you buy it from lots of different companies, it also gets confusing and slow and hard to integrate. Odo solves that, because all their software is connected on a single, affordable platform. So you can save money without missing out on the features you need. Odo has no hidden cost, and no limit on features or data. It has over 60 apps available for any needs your business might have. Everything from websites to sales to inventory and accounting, all linked and talking to each other for no additional charge. Check out Odo at odo.com. That's odo.com. That's odo.com. If you sold somebody a loaded gun, who you knew was in a vulnerable state, and they shot themselves, I think it is murder. Just because you're using the internet doesn't mean you get away with murder. I'm Damon Fairless, host of Hunting Morehead. This season, I take you inside the business of suicide, and the places desperate people go when they can't find what they need in the real world. Hunting the suicide salesman, available now wherever you get your podcasts. So Gordon has shown us something pretty surprising. You can change somebody's mind even after they've gone pretty far down the rabbit hole. And another researcher named Sandra Vanderlinden heard that and thought, okay, but wouldn't it be better if they'd never gone down that rabbit hole in the first place? If you come after the fact, people have already integrated the misinformation into their identities into their mental model of how the world works, you know, you can't unring a bell. It's very difficult. It's an uphill battle. I actually want to avoid. Sanders is a social psychologist at the University of Cambridge. And he started wondering if you could get to people earlier, before the bogus headline, before the conspiracy theory, before that thing that isn't true, has already started to feel true. Could you somehow prepare people for the bad information headed their way before they ever saw it? So he started looking at something he calls pre-bunking. Which is all about trying to prevent people from encoding misinformation into their brains in the first place. And we do that. Sanders and trying to tell people what's true. Instead, he wants to give them the tools to recognize manipulation when they see it. Rather than just giving people facts or telling them what's a good source, what's a bad source, what we're actually doing is exposing people to the techniques used to produce misinformation and exposing that playbook. These techniques can be hard to spot. They're baked into the way facts are mixed, matched and presented. Pre-bunking teaches people what to look for. Because people who believe in conspiracy theories don't think what they have are lies. They think they've discovered a truth everyone else missed. Tell them they're wrong and they just dig in. But show them how they're being manipulated and they may start to recognize it for themselves. What are compares it to giving someone a vaccine? When you take a vaccine, you get a weakened dose of the virus that triggers the production of antibodies to help confer resistance against future infection and it turns out you can do the same with the brain. The brain needs lots of microdose examples of what it's like to be deceived in order to mobilize enough psychological immunity to identify, resist and neutralize falsehoods. All that's to say, if you know how disinformation works, then you're more likely to notice it in the wild. And when you can better notice it, you build up what Sander calls "cognitive antibodies." That's why Sander teaches people about manipulation techniques, things like fear-mongering. To explain that technique, he goes all the way back to 1796 when the world's first modern vaccine appeared for smallpox. And this is where things get a little weird, because people were told the vaccine might turn them into cows. We give people graphic propaganda from the 1800s that if you took the cowpox vaccine, you were going to turn into a human cow hybrid and that picture showing like others sprouting out of people's mouths and ears and they were trying to scare people, don't take the cowpox vaccine, you're going to turn into a crazy cow. Sander often deliberately uses propaganda from a long, long time ago to get people to open up to the idea. Remember, the smallpox vaccine was created in 1796. So, with 230 years of separation, it's easy for everyone to see how ridiculous this argument is. Then people got threatened, right, by what happened 200 years ago, but they can connect to dots and say, "Huh, well, that is kind of suspicious that, you know, the same technique is being recycled." It might not fix their vaccine hesitancy, but it is empowering people to spot manipulation and make them reconsider their stance a little bit. And there's evidence this actually works. In 2025, Sander's research team invited Instagram users to complete a poll. It showed people news headlines and then tested their ability to identify emotional manipulation within them. Next-line discernment of emotional manipulation among young people to Instagram in our sample was 38%. That's low. Only 38% of people can identify very basic emotional manipulation. I mean, that's not a good number for democracy. Then Sander tried an intervention on an Instagram itself. He made a 19 second video about pre-bunking and slipped it into thousands of users' feeds. There's no voice telling you what to think, just music. And a question appears on the screen. Can you spot the trick it says? Then a pops a headline, yoga linked to terrifying full-body cancer. Which sounds pretty alarming, and that's the point. A moment later. the video reveals the trick, fear-mongering. The word "terrifying" is what Sander calls a "scareword", language designed to get an emotional reaction before you've had much time to think. Then, people who saw that video were invited to take the same poll. Before the video, 38% could spot that kind of emotional manipulation. After watching it, 60%. And 19 second video on Instagram. And this all sounds like progress. Except, think about what we've spent this hour asking you to do. So, don't necessarily trust this video. Check the evidence, check who's showing it to you. Ask whether someone is trying to manipulate you. All useful advice, but also a fairly exhausting way to move through the world. Because spawning manipulation doesn't come naturally to us. Just ask Gordon Pennycook, the researcher from Cornell, we spoke to earlier in the show. We generally just intuitively expect people to tell the truth. There's a truth bias in communication. It would be a kind of shock if someone that you knew or even someone that you didn't know just came up to you and completely lied to you about something. And mostly, that instinct serves us pretty well. We couldn't get through the day if we had to fact check every person we talked to. The problem is social media plays by different rules. On social media, or just on YouTube, when engagement is the thing that brings success, you can see why the truth doesn't come first. And we're much better at getting people's attention than we are at getting to the truth. Which sounds pretty bleak, except the truth hasn't disappeared. That's one of the things Gordon's research tells us. People can get something wrong, really wrong. They can believe it for years and still change their minds. The trouble is, often the lie gets a head start. It can be outrageous or frightening or just irresistible enough that you really want to send it to someone. That's the gap we're all living in now, the space between seeing something and knowing what actually happened. So maybe the trick is learning to live in that gap for a minute, to leave a little room for doubt and give the truth some time to arrive. I'm Dena Temple Reston and this is Click Here. And that's the show. Click Here is written and produced by Megan Dietree, Erica Guida, Zach Hirsch, Maya Fawaz and me, Dena Temple Reston. Karen Duffen and Sarah Kaveto are our editors. Darren Ancrum is our fact checker and our staff writer is Lucas Riley. Our illustrator is Megan Goff. Original music is by Ben Levingston with additional music from Blue Dot Sessions. Jake Cook and Jesse Nyzwanger are our sound designers and engineers. Click Here is a production of Recorded Future News and PRX. I'm Dena Temple Reston and thanks for listening. Support for this program comes from Recorded Future. In cybersecurity, the biggest risk isn't what can be seen, it's what gets missed. Recorded Future analyzes billions of signals to help organizations stay ahead of threats. Recorded Future, know what matters, act first. If you're looking for a daily guide to cybersecurity news and policy, sign up for the Cyber Daily from Recorded Future News. It serves up the day's most interesting and important cyber stories from our sister publication The Record and then aggregates all of the big cyber stories you might have missed from news outlets around the world. Just go to the record.media and click on Cyber Daily to get all you need to know about the world of cybersecurity right in your inbox.