Apple Podcasts | Android | Spotify | Email | RSS | YouTube

Can artificial intelligence transform how we navigate the most challenging dialogues on campus? 

Join us for a thought-provoking episode featuring philosopher and educator Simon Cullen, as he unveils his pioneering work at the intersection of education, technology, and constructive disagreement.

In conversation with John Tomasi, Simon explores how open inquiry is both advanced and imperiled by disagreement, and describes his academic journey from Australia to Princeton and Carnegie Mellon. Central to the discussion is ‘Sway’ an AI-powered platform developed by Simon and his team to foster rigorous, evidence-based dialogue among students on controversial topics. Sway intelligently pairs students with opposing views and acts as a “guide on the side,” scaffolding reasoning, encouraging intellectual humility, and ensuring that exchanges remain constructive and charitable. Simon shares the empirical findings from thousands of Sway-mediated dialogues, where measurable increases in students’ openness, comfort, and analytical reasoning have been observed—even on divisive subjects like gender, immigration, and the Israel-Palestine conflict.

In This Episode:

• The pedagogical importance of constructive disagreement and argument visualization
• The design and implementation of Sway: an AI tool for dialogue across differences
• Empirical research on autonomy, self-censorship, and openness to opposing viewpoints
• Strategies for motivating student engagement in difficult conversations
• Scalable, evidence-based methods for promoting viewpoint diversity in higher education

Want to experience Sway for yourself? You can instantly start a conversation—no account, no signup, no hassle. Just generate a link, share it with a friend (or a friendly opponent), and jump straight into a thoughtful dialogue: https://me.swaybeta.ai/

Read more about research findings and reports: https://www.swaybeta.ai 

Explore Simon’s Argument Mapping Tool: Mindmup.com: https://www.mindmup.com/tutorials/argument-visualization.html?orig=/

About Simon:

Simon Cullen is a faculty member, Dean's Innovation Scholar, and Artificial Intelligence and Education Fellow at Carnegie Mellon University. He developed the award-winning "Dangerous Ideas in Science and Society" course and serves as founding co-chair of CMU's Heterodox Community. 

Simon's innovative research combines philosophy, cognitive science, and educational technology to improve reasoning and communication across moral and political divides. He is the co-developer of "Sway,” a chat platform that uses AI to facilitate constructive dialogue between students with differing viewpoints. His research has been published in Science Advances, Nature Science of Learning, Cognition, and the Review of Philosophy and Psychology. He holds a Ph.D. from Princeton University and specializes in the psychology and pedagogy of reasoning, and evidence-based approaches to promoting open inquiry in higher education. Next year, Simon will be a Faculty Research Fellow at the Segal Center for Academic Pluralism in New York City.

Learn more about Simon https://www.simoncullen.org

Episode Transcript

John Tomasi: Simon Cullen, welcome to Heterodox Out Loud.

Simon Cullen: Thanks so much for having me.

John Tomasi: It's a really special pleasure having you on the show today. You've been one of our HXA leaders whose work we've been following with particular interests. As you know, you won the Heterodox Academy Open Inquiry Award for Best Teacher a year ago, I believe, last year in Chicago. And we've followed your teaching at Carnegie Mellon, which we'll be talking about. The main thing we'll be talking about today that I want our listeners to hear about from you is artificial intelligence and some innovative things you've been doing with your team to model constructive disagreement, teach constructive disagreement using AI. So that'll be, it's a tool called Sway. We'll be coming back to Sway significantly in this conversation. Well, let's just begin by, let's begin with you. So you're a philosopher. You're from Australia. Tell me how you got to here. A short story. You can tell it short or long, whatever you like.

Simon Cullen: It wasn't expected given my childhood. I wasn't very into school. I hated school really. But I always was very interested in big ideas and thinking about big ideas and talking to my friends about big ideas. And in the end, I got to college. I wasn't sure I was going to go. But I did. And I took a course with my first teacher in philosophy, Neil Thomason, who is still a mentor, a friend, and a teacher. All these, nearly 20 years later. But he, you know, I had met a lot of intelligent people by that time. But I, when I.

John Tomasi: Was it your first course in philosophy?

Simon Cullen: It was, actually.

John Tomasi: Oh, interesting.

Simon Cullen: It was a philosophy of science course called Science, Philosophy, History. And in many ways, it has influenced everything I've done since. It was mostly my first teaching gig. Neil managed to somehow get the head of the department to convince the university that before I had an undergraduate degree, I could be a teaching assistant.

John Tomasi: Were you teaching and grading papers as an undergrad?

Simon Cullen: Yeah.

John Tomasi: Interesting.

Simon Cullen: I think I might have been the first at that time only.

John Tomasi: Interesting.

Simon Cullen: Might be quite common now, for all I know. But the thing with Neil was I had met a lot of smart people, but I had never met anyone who really struck me as wise. And he was the first person I met who made me think.

John Tomasi: What made him wise? What made you think he was wise?

Simon Cullen: I remember exactly what happened. Well, not quite all of the details. I wish I remembered some of them. So I must have had a pretty good question from a lecture. And I went to office hours. And I wish I remembered what the question was. It would be a better story if I did. But whatever it was, I went to office hours and I asked my question. And Dr. Thomason sort of looked off into the distance. This was in the old quad at Melbourne University. I don't know if you know. It's very pretty. And he just span around and looked out his window over the quad for what seemed like quite a long time. And then he turned back to me and said the following, I have nothing intelligent to say about that. I'll have to think it over. And I just thought, wow.

John Tomasi: Nice.

Simon Cullen: That is what I'm here. I mean, that impressed me so much. Me, it's a pisher. No one. I'm an undergraduate. And, you know, I was obviously very impressed by him. He was a charming lecturer. And here he was just interested in my question, interested in my idea, interested in what I had to say. And he didn't want to respond quickly. He wanted to think.

John Tomasi: Well, I had a tremendous reaction to that. That may have been the moment that set it all in action quite possibly.

Simon Cullen: But then, you know, he then also helped to support me to go as an undergraduate on exchange to Berkeley. And that's when I sort of first got the taste of what American college is all about.

John Tomasi: Did you take philosophy classes at Berkeley?

Simon Cullen: I did, yes. I'm a fantastic department. So, yeah, I was a wonderful department. I had such great courses. I was also taking music courses. I was playing in the Baroque Orchestra. It was just an amazing time. It was so good. I miss it dearly. But I was so energized by that time in Berkeley and finding my people, my nerds, you know, coalescing around our shared interest and, well, at the time, smoking cigarettes and talking about big ideas. And I came home just knowing this is what I wanted to do. And so I then sort of set my mind to that.

John Tomasi: And you went to Princeton to study philosophy.

Simon Cullen: That's right. Yeah, a few years later. For a master's and PhD. Yep.

John Tomasi: How many years were you there?

Simon Cullen: Oh, well, I also did two postdocs. So in total, I was there just shy of a decade.

John Tomasi: All right.

Simon Cullen: My actual PhD was six years.

John Tomasi: Were you at the Center for Human Values? Where were you at Princeton?

Simon Cullen: I was associated with the Center for Human Values. Actually, the first seminar I taught was funded by them. But yeah, I was just in the philosophy department. I was mostly working in sort of moral, empirical moral psychology. I was always being interested in a lot of different things. So Princeton was kind of a perfect, Princeton philosophy department was kind of a perfect place. But then after that, they really wanted to support me in this experimental work I'd been doing and teaching analytical reasoning, which I'm sure we'll talk about. And so they created a postdoc to keep me around for a couple of years. And during that time, I just completely devoted myself to trying to figure out how to improve students' reasoning or to support them as they improve their own reasoning.

John Tomasi: They invested in you.

Simon Cullen: Yeah, they did. And at the time, it really meant the world to me. It was a very important vote of confidence. And I finished that up, and I was very interested in neuroscience and cognitive control and computational modeling of cognitive systems. And I was very fortunate to get a one-year postdoc in Princeton Neuroscience Institute in the lab of John Cohen, his cognitive control lab. And so that was another sort of extraordinary year of learning and being in this amazing place.

John Tomasi: I first became aware of you, I believe, when you were nominated for the Open Inquiry Award that I mentioned that you won. And what I was struck by was a course you were teaching at Carnegie Mellon called Dangerous Ideas. Tell us a bit about that course and why you taught it, how you taught it, what you learned there. And what you learned and what they learned, I suppose.

Simon Cullen: Yeah, well, it's a big course with a lot of stuff going on. The idea for it really hatched around 2020.

John Tomasi: Was it up and running when you arrived?

Simon Cullen: Oh, no.

John Tomasi: What happened? Did they assign you to do it? Did they say, hey, we want you to teach a course with a lot of controversies?

Simon Cullen: No. No. Nobody has ever asked me to teach a course.

John Tomasi: I wanted to hear you say no twice.

Simon Cullen: So, I was 20, you know, the idea for it, I first taught it in 2021. The idea for it was in 2020. And the reason I even had this opportunity was I had been teaching intro philosophy. And at CMU, there's a variety of sections. And then, you know, there's no pressure to align the sections. But people just usually are pretty aligned. I mean, maybe you do Descartes. But we cover similar stuff. And my course, you know, the person who was running the undergraduate program actually looked into what I was teaching and decided that it couldn't possibly be introduction to philosophy. But whatever it was, it was pretty popular. And so, we should keep it around. So, intro to philosophy was 8100. And she said, well, why don't you take 8101? What do you want to call it? And the course had become more controversial. I'd been teaching guns and abortion and infanticide and euthanasia and whatever at Princeton. But I'd really been leaning into that more and more. And so, I pretty quickly settled on the idea that this speech culture that we were occupying at CMU was telling us, you know, don't talk about this. Don't cover that. Don't ask those questions. And I thought, well, I'm going to make a course which is only about talking about those things. It's only about asking those questions. And then I'm going to try and make it the most popular course on campus.

John Tomasi: I love it. I love it. I love it.

Simon Cullen: So, that was kind of a defiant origins of this class. But what it actually was is a very complicated beast. It was partly an analytical reasoning course, very much about trying to give students tools that can scaffold their reasoning so as that we don't just throw them into a room and say, well, now chat about abortion, but we actually give them tools.

John Tomasi: Giving them tools that can scaffold their reasoning. Lovely. Lovely. Which, by the way, is a phrase that very much comes to mind when I think about sway. But we'll get to that soon. But go on. So, you're thinking about providing them with, I think, scaffolding is the next word. That's the word you chose, right?

Simon Cullen: Yeah.

John Tomasi: So, say more about scaffolding.

Simon Cullen: So, well, this actually is a specific technique that I started working on at Princeton. It was what I had done this initial two-year postdoc on called argument visualization. What argument visualizations are? Basically, tree diagrams. These are the maps that you work on. Exactly, the maps, yes. Argument mapping is the other common term for it. So, what they do is they try and strip an argument normally presented in prose or speech down to its very bare essentials and then use visual conventions like color, line, and shape to try and make it really easy to parse the structure of the argument. So, we have very simple visual conventions, you know, like a green bracket indicates support or the because relation. And, you know, a red bracket indicates a bracket can contain multiple claims, which are the building blocks of all arguments. And then these things can be recursively built. So, you're supporting something that's playing the role of support. And, of course, these trees get extraordinarily big. But the real teachable moment for him when I introduce it to people is that I'll use a very simple toy argument, 80 words or something like that. You know, soccer is better than baseball, whatever. And, you know, I present it to sophisticated faculty and they're sort of thinking this is really baby remedial stuff. But it's the case that I've not really ever found a student, very bright students at Princeton and CMU, who get this 80 word thing and can point to the conclusion, can tell me how it's supported, can notice that the objection is actually an objection to an implicit claim that wasn't made out loud, which is totally standard in ordinary argumentation. It's just simply not the case that the structure of argument jumps out and slaps people in the face, even very literate people. And, of course, most of the students are literate, but that has involved, most all, that has mostly involved reading, you know, fiction, textbooks, poetry, maybe a couple of Federalist papers, but never really seriously engaging with dense argumentative prose. And so, you know, I would present this little 80 word thing to the philosophers in the room and then I would have to show them the results, the students can't do it. And then, of course, you see them all think, oh, wait, I just assigned the prolegomena to any future metaphysics. I wonder what they're getting out of that. And, but so this was a real problem that I was very motivated to try and solve, to help students develop tools that will allow them to engage in a deep way with argumentative prose.

John Tomasi: Let me understand something. I'm trying to, I'm trying to understand if you're describing a, a way of reasoning, a way of thinking, a way, a set of skills that are specifically relevant to the discipline of philosophy, or are you describing a set of skills and abilities, techniques that are important to people across their lives? Is that a, is that a sensible distinction between two questions or are those the same question for you?

Simon Cullen: I get the distinction. but I think the answer is the skills are ones that have been honed by philosophers and cherished by philosophers, but they are absolutely skills that are important across all, you know, all of the academy, but also just anywhere in life. If, if you're trying to persuade people with reasons, well, you're in the same business as I am and you didn't need to study philosophy to get into it. It's just what we do.

John Tomasi: I ask because it sounds, it can sound arcane.

Simon Cullen: It does.

John Tomasi: On, on academic.

Simon Cullen: Yeah. Yeah.

John Tomasi: but you're saying that the arcane is tied to something very familiar.

Simon Cullen: That is that I hope it's familiar. I think you and I would both like it to be maybe more familiar.

John Tomasi: The idea that we interact in a culture where exchanging reasons where up on the wall in front of us, we have the HXA way, which I often refer to on the podcast. Make your case with evidence, be intellectually charitable, be intellectually humble, instructive, be yourself. We want to be in a world, I hope, on campus, especially beyond campus as well, where some kinds of human interactions are progressive in the sense that progress can be made together.

Simon Cullen: Oh yeah.

John Tomasi: Through interaction and all the kind of things we talk about. But, but.

Simon Cullen: And through disagreement.

John Tomasi: And through disagreement. Instructive disagreement.

Simon Cullen: Of course. We can agree on that.

John Tomasi: Tell us about, tell us about MindMup.

Simon Cullen: Absolutely. so MindMup is free open source, nearly completely open source software, originally developed by some wonderful, developers. It's free for, it's been out there for over a decade. I was at the time, using some software. So building these maps, you can do it with post-it notes or on the board, but it quickly becomes quite unwieldy. and so having computers to edit them as it makes it just tremendously easy. It also makes it very easy for students to collaborate. And then it makes it even easy for us to provide feedback as well. but we were using some software that sort of was turning a bit evil. you know, they were starting to make you get a subscription to download your own data and this kind of thing. and so I, you know, I really wanted there to be a free alternative. and so we built on top of mind, the basic mind mapping software, MindMup, an argument visualization tool, which is available and widely used. It's at argument.mindmup.com. And if you go there, you'll see as a three minute video sort of walkthrough. It's designed to deliberately be very simple. Could you share the domain name again? Yeah. MindMup, M-I-N-D-M-U-P. sorry. Argument.mindmup.com. but you can get to it from my website, philosophy mapped, and that's philmaps.com. And there you'll also find sort of handy hints for just introducing the basic ideas in a simple and accessible way. and also some cool arguments maybe that you can try and map yourself.

John Tomasi: so, so, so tell us a bit, but I want, I want to, I want to talk about, I want to get to a sway.

Simon Cullen: Yeah.

John Tomasi: Just bring us, bring us, bring us through.

Simon Cullen: Yeah.

John Tomasi: Talk about a discussion market incentives if you want to.

Simon Cullen: Okay, yeah.

John Tomasi: Or more about MindMap as you like, but tell us more. Well, take us on a path towards sway.

Simon Cullen: Yeah, let's, let's do it. So, this seminar at Princeton was a very intense, seminar. It also was totally viral. it was like half of the incoming course applied. They wrote essays to try and get in and only 15 seats. So, we had like 500, 600 students. And it was totally unprecedented actually in, in living memory at Princeton. it just became a phenomenon. it was called Philosophical Analysis Using Argument Maps.

John Tomasi: Wow. What a world to live in, right? Where that course was, like, stagnant. Only at Princeton, maybe?

Simon Cullen: It was great. You know, I, it got a big, it got a big boost from one of my favorite, and, and earlier students who, just as a joke, he was a real joker. He went on Facebook and posted in the Princeton class of 2000, whatever it was, Facebook group, you know, sorry guys, bad news. Philosophical Analysis Using Argument Maps has been canceled. Don't bother replying. I only learned about this because, of course, I get hundreds of panicked emails from students. They post-show me this Facebook thing and I realized he's a Princeton student who had already applied to the course. Okay, this, this kid's got to get in. He's got to get in. but that, that probably helped really launch it in a, in a big way. And then, you know, we really cared deeply about actually helping students to develop skills that matter to them and, in a meaningful way that, that you can really tell the difference.

John Tomasi: And I think that, may I just, I'm sorry to interrupt, but that your, those numbers, see the numbers again, how big was the class?

Simon Cullen: Only 15 students were ever enrolled at a time. It was in the freshman seminars program. but it was, I think really half of the incoming course in the fall, incoming class in the fall would apply. And then many of the other half would apply in the spring. It was, it was completely viral.

John Tomasi: It goes to the question I was asking before, are these methods specifically for philosophy only or are they of broad interest to a person who wants to learn how to reason?

Simon Cullen: Yeah.

John Tomasi: But go on, take, continue our journey towards how the discovery of sway and experience.

Simon Cullen: Yeah. So this class was, I had very motivated students and, I was also very motivated as pretty much, I'll, I'll, I'll, I'll match you. Whatever you put in, I'll, I'll put in more. And they, worked unbelievably hard. So we published the findings of this, quasi randomized, course, in npj Science of Learning in 2018 in a paper, the title of which I'm struggling to remember exactly, something to do with argument visualization and argumentation. But we, we published our results there. And I think that what we measured there, we were using the LSAT logical reasoning test to try and measure growth in students' analytical topic, neutral analytical reasoning. So we, it was a philosophy course. We didn't talk about the LSAT or the logic or any sort of tests. So this was really a measure of transfer of whether the skills that they learn in one context could be applied in another. And that's really the gold standard of learning. So we got some extraordinary results there. the, the, the people who appreciate effect sizes, it was about 0.7 standard deviation improvement over and above the control group. And to put that into context, an equivalent difference in an IQ test would be about 10 points, on the actual LSAT, because there are two logical reasoning sections and they improved their scores by about 2.5 points. So on average, that's the difference between roughly a top 10 law school and a top 40 law school. So we're talking about a real improvement in one semester. I think these students hold, hold possibly a world record, but they worked extraordinarily hard, which is of course what you have to do if you're trying to budge skills that are so intimately connected to something like basically intelligence, very, very hard to, to boost it, but should be, maybe, I mean, if you can't boost it, we got, I don't know what the heterodox academy is at the universities, but, it's, it's hard to do. And it takes a lot of work, a lot of training, a lot of practice, like anything that you, you have to learn how to do reasoning is something which you can't really get by osmosis. And that's the standard thing. You know, we have extraordinary reasoners, philosophy professors or professors in whatever discipline, and they stand up in front of the class and deliver 20 or 30 lectures. And then, it's not very surprising at the end of that course that you can't measure any improvement. It's sort of very similar to trying to teach people to play the violin through lecture. It doesn't work. It's not how it works. You have to try. You have to actually get your hands messy. So these students were doing it and achieving extraordinary things. Around that time, my postdoc finished and I went to Carnegie Mellon. and there I started teaching and I thought, you know, I'll try and scale this thing up. I did at Princeton. It was, you know, a great, great course. it was part of the reason I was hired. We just published this cool paper. but it was very, very different at Carnegie Mellon. First, first of all, students were not fighting to get into my class in Carnegie Mellon. At CMU, they're effectively coerced to take my class because it's a gen ed requirement. I mean, not my specific class, but a class in the vicinity. So this is completely changing the approach that students have to the course. And I remember after one, lecture, a student, came up to me and he sort of looked me dead in the eyes. And he'd been talking to me about how much homework there was and how long it took to do these maps and how, you know, students were crying and whatever. and he looked at me and said, you do realize you teach at a STEM school. well, I was obviously very offended initially, but I thought about what he said and I realized he actually had a pretty good point. he was taking a ton of very demanding courses. And here I was a philosophy professor trying to get him to do this strange visualization technique. And he had no appreciation of why that was valuable or worth doing. and so that sort of set me on this path of trying to think very deeply about motivation and autonomy. so the basic sort of, insight that, that drove this was the students at Princeton were there because they wanted to be there. They chose to be there. Actually, they didn't even get grades in my class. They had to agree to have their grades suppressed. I mean, I wouldn't hide them from you. If you came to office hours, I'll tell you how you're doing. But I would strongly encourage you, please, let's just reflect on, on, on the narrative feedback and working together because most of the students are in A's, but that's not as good as you can be. That's not even close. That's a low standard. so they were there for very different reasons to the Princeton, to the CMU students. And I, that sort of set me on this path where I started to think, well, what I should try and do is find ways to challenge the students, so as that they take this on voluntarily. and so that was, so it started me on a new experimental path, where I began to make the homework, what I call optional mandatory. so you choose if you want homework to be mandatory. And, that, that work, I, I then continued with my wonderful colleague at CMU and SDS, Danny Oppenheimer, who is also a sort of autonomy, not the first thing we ever bonded over was a, autonomy. And second thing was heterodox Academy. but, you know, Danny was also very interested in attendance and, and sort of tried this, this general strategy of optional mandatory and applied it as broadly as I could to my class. so this is, you know, at the beginning of the semester, I tell you, well, you know, I hope you want to come to this course. I hope the lectures are going to be really meaningful to you. I'm going to put a lot of work into them. but you don't have to come if you don't want to do the readings, two essays, pass this course, easy B. but I think you can get a lot more out and I should give them the experimental evidence that it's possible to have these sorts of improvements. And then I would give them a period of time to experiment with it. So they didn't have to worry about their GPAs. And then at the mid, you know, just before the midterms, they would, the students would make a decision. Am I going to stick in this grueling argument visualization based problem set course, which they have to do, you know, pretty hard problem sets every week. or am I going to just write two short essays, you know, five pages each. And I actually surveyed my colleagues to see what they predicted would happen. And they, the most optimistic prediction I got, was from one colleague who said, initially the students will start out wanting to do the hard thing, but then their STEM courses will fire up and they'll triage into the easy option. That'd be my bet as well. That was the most optimistic. the actual, the actual answer was, 90% of them initially set out to do the hard thing and only five or 10%, triaged into the easy thing. And as I continued to increase the amount of autonomy the students were receiving, I found the same, same thing over and over again. and so Danny and I, you know, studied this in a, in a randomized control trial, and a field study as well. And published that, last summer in Science Advances. It's, the title of that one. I do remember it's called the, importance of student autonomy in higher education. Oh no, sorry. I missed the best part of it. That's the subtitle. It's called choosing to learn the importance of autonomy in higher education.

John Tomasi: it reminds me of a course I taught at Brown, where I made a deal with the students that we would never grade them on how well the answers they wrote matched up with the questions that I asked. Instead, we evaluate them on the quality of the essay that they wrote, the origin, the originality, the rigor and so on. So that meant that when I would give them an exam with questions, they were always free to change the questions, scribble them out, adjust them, turn them into some different kind of questions. It was inspired by an experience I had myself as a sophomore, as undergrad, taking an economics class and finding a question that I thought could be, allow me to say something more interesting if I changed the question. I went down to the tech teacher as a 19 year old kid and I said, I just make it like this instead so I can talk about this other connection. And he said, sure. If you let me change the question, which shocked me at the time, but it stuck with me. And I did that systematically with large classes of Brown and it's sort of a way of doing autonomy. It's sort of saying, well, here's the question. Here's what I imagine will be a really good answer. You've got the material mastered that I wanted, but you're a human being with a mind of your own. You may see the frontier above and beyond this question available to you. I can't see that, but you might see it. And if you do, let's go there, right? This is somewhere.

Simon Cullen: Well, I mean, for a student who's never experienced anything like that, such a thing can be really very meaningful. I mean, for one thing, you trust the student's judgment in some way. I mean, you're there. You're going to hopefully look over the questions they come up with and help them formulate the questions. But there's a certain kind of trust there, a respect for them as independent intellectuals, who have something worth contributing. And then by having them write their own questions, you're also giving them ownership, which is profoundly motivating. And if they choose, and sometimes what happened, they would change, that's actually happened with some students, would change the questions in ways that did not produce a wonderful essay. And we just knew. Unapologetically, so that's not what we're talking about here, a few.

John Tomasi: So, but, but, I want to, let's, let's go to the main course.

Simon Cullen: Okay, let's get there. Well, I was getting, I was just getting to, Dangerous Ideas.

John Tomasi: So, Dangerous Ideas was very- This is the amuse-bouche. Go ahead.

Simon Cullen: Yes. So, this was a course that, was really just about, talking about things that were challenging to talk about, but it was not a political course or a class in politics. It was a class about arguments. so this course was, as, as it was recognized by HXA, was, it was a really profound, of course, and, and, it, I didn't ever expect that it would become that, that important to me or to the students. but at the end of it, I, you know, the last time I taught it, we got the biggest hole we could get and filled it up, 200 students. And then there was another 300 students or whatever it was on the wait list. And they, they have this metric, maybe they have it elsewhere as well, but it's, they don't, you know, how many students are on the wait list is great. But what they really want to know is by the end of the wait list period, when they all get kicked off, how many of them stayed clinging on for dear life until the very end? I think we topped CMU on that metric. but at the same time to do this class with just 200 students, I had 15 papers with me, 16 people, handpicked and trained teaching assistants and it absorbed my entire life. And I just realized, it sort of was in the same, same problem I had had with argument visualization. It's intense. It's very labor intensive for the instructors. You're not just standing up there lecturing, you're coaching them. And so I, I wanted to make this possible for faculty to do without devoting their entire lives to it. And without having to find people that they could trust to run these discussions. But the basic motivation was students came out of this class saying all the things you hope they'll say about college, you know, saying, you know, I, how difficult it was for them to realize that they had to separate being right or having the right beliefs from who they are. And they had to learn how to just try and interrogate an argument with someone else in the shared pursuit of truth or wisdom or whatever. That's a super profound thing. And I wanted to see how we could make that possible at a very big scale. And so that was around the time, that chat GPT, GPT-4 especially, was really dazzling. and it was immediately, I was, with my, at the time, my postdoc, Nicholas DeBella, my co-founder and colleague, we were talking about that. We were reading this incredible feedback from, from students and we just started to think, wait, this AI could be an incredible discussion mediator. If only it wasn't just one person talking to it at a time, right? Cause it's, we've conceptualized it as a chat bot for productivity more or less, but it doesn't need to work that way. And so we, we built a prototype that, you know, two people would jump on and send messages back and forth and we'd be processing the messages in the background and prompting the model to try and see if it could be intelligent about when it intervenes, about whether it could prevent discussions from going off course.

John Tomasi: May I pause?

Simon Cullen: Yeah, absolutely.

John Tomasi: Because this is to me, what's really profound, interesting about what you're doing. And let me just chart a course from there's, you said before that learning how to do argumentation is something that one needs to experience oneself. And so here's a, here's a progression at Oxford. It's a form of teaching that you probably know about this, the tutorial method, which I went through as a grad student there where I'd walk into my professor's office with a short paper paper I'd written, I'd read it out loud, then we'd spend an hour or two hours just doing this mind melting analysis of that ideas. The next idea, super labor intensive, one student, one grad student in a room with a famous Bernard Williams for two hours per hour, you know, whatever it would be. And that was considered the old Oxford Rolls-Royce model of teaching was unsustainable financially. And then they started moving to what they described as the American model, which is you put the famous professor, the sage on the stage. Of course. And people go along and you, and you watch a mind at work and you learn from that. That can be very fun. And it's wonderful. And now you're thinking, okay, you're, so that's phase one is that intensive thing with you and the professor, but you're the student and he's the professor. And then the difference model is the sage on the stage where you're learning that way, but you're not really engaging as deeply as you might. You're engaging. You're following the argument. You're reading, but you're not bullying. Maybe things are going well. But what's really striking to me about the Sway idea is it's the third thing, the third way, because it uses this tool, this guide to bring together two students, both learning. So go on, please. Yeah. So first of all, I just want to, I'm sorry, I just want to really, that's a little piece because I'm sorry, I didn't make the point I wanted to make fully that. I said, destructive disagreement involves interacting with a person, let's say in real time, person who may be struggling like you, who may be an equally imperfect learner as you are, not a person like my Oxford Don's who knew way more than I do about technical philosophy questions or helping me do my best to catch up to level one, two, and three before I get near number seven, but rather actually interacting in real time with a fellow imperfect learner trying to get better together. That's like a certain dynamic cultural inquiry that we're always talking about around HXA. Yeah. Now, please, I'm sorry. Yeah. Please take it away.

Simon Cullen: Okay. Well, you just said a thousand very rich things. I just wanted to say about the Oxford tutorial style. Yeah, it is old school, but it is still also the gold standard and doesn't need to be at Oxford, but one-on-one tutoring or very small group tutoring is absolutely the gold standard. It has not been matched. There's a famous paper, the two sigma problem by Bloom, and two sigma refers to two standard deviations. And that was a finding that had been made in this sort of context with very close one-on-one personal tutoring, but you could improve people by two standard deviations. So that means taking someone who's average and pushing them past the 97th percentile. This is the only method of teaching which reliably can achieve anything like that.

John Tomasi: I didn't know that. That's fascinating.

Simon Cullen: It's extraordinary. Unfortunately, like you said, it's very, very expensive. And well, one of the very first things I thought about when I was getting familiar with these new models was the two sigma problem. Other people had the same idea, but it is very much connected. And the connection you saw there is one I'm very glad you saw. So on Sway, like you said, it isn't just the don and the student. It's the guide. And rather than the sage on the stage, we think of it as the guide on the side. And it is there.

John Tomasi: I haven't heard that expression. That's great.

Simon Cullen: I didn't come up with it. I wish I had. I wish I could take credit.

John Tomasi: Is that Nicholas?

Simon Cullen: No, I don't remember which of us said it, but one of us said it and thought it was great. And then we found out someone else said it.

John Tomasi: The guide on the side.

Simon Cullen: Yeah, not the sage on the stage. But this was the thing that we had seen in Dangerous Ideas was students having these remarkable discussions, and we wanted to see if we could make that happen, but in a sort of small and scalable way. And by small, I don't mean small not reaching a lot of students. We want to reach millions of students. I mean small as in small groups, ideally actually just one-on-one. And so Sway is actually designed, it's not only for controversial topics, but it certainly has a lot of design built into it to make it possible for students to have discussions, especially about controversial subjects. And that is one of the main things instructors are using it for at the moment. And so in those situations, we do get two students who, before they have a discussion, I'll just go back two steps. Sway is basically a chat platform. So think iMessage or WhatsApp, very similar to that, except you can only talk with people you disagree with, and there's a sort of bunch of AI stuff going on to try and help you express yourself more effectively, to think more clearly, to identify the implicit assumptions that are required to make your arguments fly, to make sure you're actually listening to and understanding your partner, to make sure you're not just ships sailing by in the night, but that you're connecting to the actual points being made. These are the sorts of things that the model is sitting in the background trying to judge and trying to figure out how can I promote that. And we really modeled it on our own intuitions about teaching and on what had worked really well in Dangerous Ideas. So that, it ends up being a very small discussion, just two students, ideally, if there's an odd number of students, we have to make one group of three. And that's also for an important reason connected with concerns that are very important to HXA. And that's changing your mind and thinking something through is much harder in public. And if, yeah, if it's just you and me, it's less of a contest. And there's a wonderful quote. I can't remember who Newton was corresponding with. It was someone. But he said in, you know, when we have these disputes in public, it's a contest. We're trying to win. We're trying to show who's right. But when it's just one-on-one, Newton says it's like more of a consultation between friends. And so I hope it will be between you and me. And we found that in these small groups of students, they very frequently comment on how much easier it is to talk. The other great thing about it is that in a regular, anyone who has taught or participated in a regular philosophy precept or whatever discipline will know that many of them function kind of like a ping pong game. There's one student, not necessarily a student who's actually read anything, but just a student with a sufficient chutzpah to talk about it, whether they've read it or not. We'll say some, make some comment to the lector, the preceptor or the tutor or whatever. And the tutor will make some comment back to that student. Then some other student will make an unrelated comment and ping pong it back to the tutor and the tutor will respond to that. And it goes like that with maybe three or four students dominating the discussion and other people hoping they won't be called on. This is a very familiar, very familiar thing. Every teacher. So there's none of that. On Sway, both students have by design to participate to progress the discussion. So you can make as many messages as you want, but until I respond, this will not count as a turn.

John Tomasi: May I just pause? I was going to go back one step if I may to be sure we got something. So this would be a tool used in a class, of course, and the computer would sort all the class in pairs of two. Are they sorted according to criteria? And what are those criteria?

Simon Cullen: Yeah, so suppose you're a professor teaching, I don't know, a course on gender or immigration or whatever it is, dangerous ideas. And you want to give your students opportunities to engage, not just in sort of discussions where different viewpoints are aired, but to really deeply engage with the arguments of an opposing view. So Sway is something that we hope you'll be really interested in. What it will do is it's very simple for instructors to set it up. We've really tried to be very mindful of the time and to try and save instructors as much time as we can. But you'll go and put some statements in there. Perhaps, you know, I just saw one that an instructor put up this morning about trans women in sports. Easy topic to discuss. So you want your students to be able to have this discussion. We require you to provide three topic statements. You just upload them, then the system will survey all of your students, and it will then compare every student with every other student to find sort of optimal mappings where we maximize viewpoint diversity.

John Tomasi: So ends you maximize diversity within each pairing.

Simon Cullen: Exactly, yeah. And across the course. So we're looking for the distance between your opinion and trying to...

John Tomasi: Trying to maximize that.

Simon Cullen: It's not something any human instructor could do with just 200 students in the class, whatever. It's 200 factorial and 198 factorial. It's like 20,000 or something. It's impossible. It's totally impossible. But this thing takes care of it for you. And so we, 95% of the chats are between two students who disagree with each other. We've had instructors from all over the place. Our very first instructor was, we launched with a Heterodox Academy member workshop. And our very first instructor is a wonderful professor at UCLA, Professor Abigail Saguy, who teaches gender studies in a very heterodox way. And she was the first professor to deeply integrate AI-mediated student-to-student discussion into a large course. So she actually put her 200-student sociology of gender class on sway. You can imagine what it was just an extraordinary thing to see the students having these discussions and frequently commenting on how refreshing it was to be able to have, to be able to think this stuff through. There's this problem of pluralistic ignorance where it's assumed that everyone shares the same belief. But in fact, there's considerable diversity. Or maybe the thing you think is the majority opinion might even be the minority opinion. But they never find out. And then they get on to sway. And of course they do. So we measure a lot of things as part of a bunch of empirical studies. And I chat about that, but wherever you want to take it.

John Tomasi: And so the students now are interacting. And from what I've seen at experimental modules, the ones I've seen on your website, a student might type in some answer to some question. A student might make a statement. It could be quite a statement, say, about a conflict in Gaza, for example. And the guide would be listening in, let's say, and would decide to intervene if they thought the statement was made in a way that was not likely to bring progress towards a pointed piece of disagreement. Is that what you would say? Say it better than me, though. When does the guide intervene? So I'm chatting away. I'm running down some thoughts off the top of my head. I'm serious about that. I'm passionate, let's say. And the guide is going to say, well, pause on that one. Do you want to say it that way or do you want to say it this way? Or can you find a specific fact, for example? Could you make a case with evidence? Can you be more charitable than looking up on the wall again? This is sort of the instructions for guide. It embodies heterodox values. That's how it strikes us. It strikes us as highly values along with HXA, plus being brilliant in a whole other variety of ways. So the students now are trying to capture for our listeners how Sway delivers that intellectual experience that we were talking about that can be so effective on our human brain that's so hard to create. And it's going to be doing it now in the specialized context of a disagreement between two equally, let's say, imperfect undergraduate learners. So here they are learning from each other to do this reasoning thing with a computer.

Simon Cullen: Yes.

John Tomasi: That's the setup, right? James Bond. James Bond. This is your mission.

Simon Cullen: It's not James Bond. This is your mission to choose to accept it. So this is the setup. We really don't want the AI part of it to be the draw. So people are by now pretty familiar with the chatbot and what it's like, and it's amazing that this thing is already ordinary. But connecting with someone you disagree with on a topic about which you feel quite passionate, that can be a profound experience that can really move students and bring them back for a lot more than a session with ChatGPT. So it's really all very focused on the human connection, and the AI is there in the background. So imagine you send a needlessly hostile message, perhaps in the context of a discussion about Israel, Palestine. A lot of students have tremendous moral passion around the subject. So the first thing we do is before we transmit that message, we read it, and we have a model that we've fine-tuned to generate charitable rephrasing. So our aim with this model is, you know, you type some angry statement, and then it sends you back a thing that says, hey, would you, maybe you want to send this one instead? And our aim is that you will get that message and go, oh my God, this thing understands me better than my wife of 20 years. We want it to deeply get you. So you can be, you know, it's going to retain the core thing you were trying to say, the point that is actually worth making, but without whatever invective or, you know, needless...

John Tomasi: Or judgment or implied judgment. It's simply a dispassionate analysis of what it is you're trying to say.

Simon Cullen: Yes. And we find people overwhelmingly accept the suggestions. And that, moreover, over the course of a discussion, they trigger this thing less and less and less. I will say that I was just then reflecting on empirical studies with sort of regular American partisans. I will say one of the most important things to come out of our studies with students, and, you know, we've now got a fairly large sample, thousands of students who have engaged in very challenging discussions. They are absolutely nothing like what they might be like on social media. So, you know, there's an archive of tens of thousands of messages on the most challenging topics. And it's, I think, 0.3% that are flagged as potentially unconstructive. So students are nothing like the caricature you would get from social media. So that first level of the AI doesn't really actually play that bigger role for the students, although it does in empirical studies. And it does, it's funny, you know, with high school students, they kid around a lot more and they trigger it a lot more. But the thing that's really there for college students is the guide. So it's basically got two sort of functions. It's in the background and it's agentic. So it will be trying to figure out if it's a good time to interject. And if so, you know, what's the way it ought to interject? But it's already filtered out, you know, that filtering system that I mentioned, the rephraser, that actually you only get to if you've gone through one earlier check. So Sway is not a free speech platform. It's a free inquiry platform. And that means that there's a lot of speech which just doesn't work on Sway. Simple example, slurs. Whatever you think of slurs and whether they ought to be banned universally or whatever, it doesn't matter. We can all agree that slurs are going to undermine the possibility of having a constructive disagreement across, you know, a significant moral, political, social difference. So we just block them. You just get a message from the system that says, we're not sending that. Try another message. So we filtered all that.

John Tomasi: I noticed, I saw that and I was quite startled when I first saw it because it seemed like a heavy hand was coming in. But on reflection, I liked it a lot because there is this thing we talk a lot about about HXA. We're trying to do a certain kind of dialogue, a certain kind of discourse, something like open inquiry. And again, I just want to emphasize for our listeners, you're constructing this in such a way where the interlocutor is of a very different viewpoint to now directly, well, through a mediator, interacting with someone who is not in your group by definition.

Simon Cullen: Yeah, when we're on these polarized issues, we've got a lot of tribalism.

John Tomasi: Yeah, you're connecting with someone from the other tribe.

Simon Cullen: So what will happen there is Guide will be there in the background and sort of contemplating whether to interject. It doesn't have to deal with the nonsense that we've already filtered out. So it will be asking you to provide evidence for your claims. It will be asking you to respond if you haven't responded to what your partner had earlier said, some claim that they had made. It will basically be a demanding professor who wants you to reason as well as you can, wants your discussions to be grounded in evidence, and is otherwise friendly and supportive, but very demanding. So it will interject in this sort of way, but then also the students can call it in manually. So sometimes the students just, it's going so well that we just don't interject at all. We're just like, these kids are having an amazing discussion. Let's just let that unfold. If we see a spot where we think we can deepen it, then we'll intervene. But sometimes they're just doing fantastically. In that case, the Guide is still available for them as a tool which they can call. So they can just start a message, hey, Guide. And we see this a lot. So we can give factual information when they want factual information. We can reconstruct arguments charitably when that's what they want. And it's pretty amazing, especially in the empirical studies where there's so much misinformation in people's heads that when they get corrected by an AI, you sort of never know what's going to happen. But we found an extraordinary deference, actually, to the model. Because we've tried to develop it so it has a very high degree of epistemic humility. It doesn't take a strong stand on topics which are sort of reasonably contested, whatever exactly that means.

John Tomasi: I want to invite you to say something about some of your findings. Because I've seen some of the data showing the degree to which something like, I'm going to call it open-mindedness. I can't remember how you said it, but I think open-mindedness, there's a metric of improvements, a delta on open-mindedness. Can you just say something about what you find for Sway, how effective it is, or what the numbers look like?

Simon Cullen: So what you're thinking of is the openness to opposing viewpoints.

John Tomasi: That's it. That's it.

Simon Cullen: Yeah. So that's a scale from our empirical studies. And what we were doing there is studying easy-to-discuss topics like January 6th and the 2020 election, between connecting partisans on that. Israel-Palestine, we studied those discussions as well. And issues around gender and transgender inclusion.

John Tomasi: And I'm just going to say it again, because I don't want to lose it, because it's so striking about this. You put it so calmly that it seems it's easy almost to miss it. These are conversations between college students who had deeply different views on these topics. And now they're interacting to talk with one another. So go on.

Simon Cullen: Yeah, and even on our empirical studies, when we just recruit people, regular folks off the internet, you know, they're the very same people who go on Twitter and are completely unconstructive. But instead, on Sway, we saw people having amazing discussions about these topics. And coming out at the end, I think the delta you were talking about is the move on this scale. And what we find is something like two-thirds of the participants, after just a 35-minute discussion with a partisan of the opposing viewpoint, will show about a 0.3 standard deviation improvement. Or on this scale, it's about a 20-30% reduction. Now, reduction in polarization or increase in openness to opposing viewpoints. And that difference is a meaningful difference, but it's just after one discussion. The real question is, what happens when this becomes a sort of regular part of your life, when constructive disagreement is normalized and, you know, heterodox academy doesn't have to exist for that particular reason? That is, I think, where we're going to see lasting, deep changes of the sort that we see with students in dangerous ideas. So I'll just tell you a little bit about how the empirical results have compared from the classroom onto Sway. Please, please. So in the classroom, we find at the beginning of the semester all the sorts of things that FIRE and Heterodox Academy have been rightly concerned about, very high degrees of self-censorship. Students with unpopular opinions are not sharing them. Students have, I would say, broadly dysfunctional ideas about encountering opposing viewpoints that they may find offensive. At the end of that class, where we previously found the sort of standard 60% who were at least somewhat uncomfortable, we cut it in half. Nearly all students, I think it actually might have been 100% of students, said they were not at all concerned that they would be graded, that they would be punished for agreeing or disagreeing with us, whatever we happened to think. If they could figure it out. We found self-censorship just absolutely drops. We found students were much more open to and could understand the importance of engaging with opposing views, even when we find them offensive. And so now transposing this, we have really great aspirations for these online discussions. And we've seen the same thing. It's over 90% of students say they felt very comfortable sharing their sincere views. Again, on those sorts of topics, that's an extraordinary contrast to what you get in FIRE data or really anywhere else. But I partly attribute it to various sort of deliberate elements of the design and partly just to the pair one-on-one format.

John Tomasi: Can I just ask you about that for a moment?

Simon Cullen: Yeah.

John Tomasi: Tribute to the pairing format. Are you familiar with Vernon Smith's work in economics?

Simon Cullen: Not in a deep way.

John Tomasi: Do you know the basic, do you know the basic, do you know, so the basic concept he's working, he was, he won the Nobel Prize for this work. And he found, surprisingly, that when you bring in two, well, it's the simplest case, a set of two subjects, in this case they're undergraduates, who play a certain kind of game with each other, a prisoner's dilemma, where there's an incentive set up so you can affect and do better, or you can cooperate and both do better. He found that in ways that are economically irrational, people tend to cooperate. So if you take the, turn down a bigger word for yourself and pass the game on to me, knowing that I could, we could both make more if I, if I agree to cooperate, but also knowing that I could defect and walk away, you decide to pass it on to me, even though rationally you shouldn't. And I, receiving the game, in a statistically significant degree, tend to play out the cooperative role that I shouldn't be playing, it's not rational. So that, that, that, that's the simplified model of a, an idea you work into a Nobel Prize winner. So important. And, and, and it sounds like something similar maybe is going on now with these people, who I'm going to say it again, students who have fundamentally different views on these issues and who are 19 years old, 20 years old, and they're doing this thing. It's just, it's remarkable.

Simon Cullen: Yeah. I have my students play the ultimatum game, to demonstrate when we, we talk about spite and, and venge, vengefulness and, and the result is extraordinary. Every time, huge effects, there's all of the rational pressure pushing you to defect or, but people find their way in these small interactions into cooperating. that is very relevant to what we're discussing in the context of sway. I'll just give a little anecdote. So when we first started doing empirical studies, not with students, just with regular folks, we would survey them before and discover what they believed. And then we would put them into a discussion and then they would immediately lie about what they had said moments earlier to, to no longer disagree because people are so extraordinarily agreeable and they want to get along. I mean, you see this wonderful moral psychology experiments. This emerges in very early in childhood and maybe in, even in infancy. this is, this basic agreeableness, is something which we can really capitalize on, for these sorts of things. But we, what we had to do in that case was now the guide actually starts off by announcing, John disagrees with the statement. Simon agrees with the statement. Sometimes we still occasionally see them say, oh, I clicked the wrong button, but, you know, for the most part now, they at least begin by acknowledging that they just, that they actually disagree.

John Tomasi: I'm just, I'm just thinking about, a line from Vernon Smith that always really struck me. He says somewhere that, anything on The Economist I was mentioning, he says somewhere that humans are the types of creature that relentlessly seek positive, some solutions. So you're set up in a world where defecting makes a lot of sense, but we're the kind of creatures for some reason. That relentlessly are seeking some sort of positive solution, which is kind of a more positive take on our human condition than is often the case. I want to, as we, as we, as we, as we reach, as we reach the end of this, this conversation, I want to talk with you a little bit at the end, if I may, about the, the deeper why of what you're doing. So there's, there's, there's, there's a, there's a reason to, to develop these tools in a way that are, that make these tools effective. There's a reason to develop these tools in a way that makes them scalable. and I think you're doing, you and you and your colleague are doing that brilliantly. I use that word seriously, brilliantly. but I just want to, I just want to pause on the why. So let's say that we, let's say you're able at very low cost to create this tool, which you seem to have done that allows people to get better at disagreeing, at being more open-minded. And they're being more open-minded now on topics that they begin with a set of convictions that are important to them, probably important to their own identity. And now if they go through this process of swaying successfully, they're now, you're calling it, you'll be calling it successful. They're now, they report things that I've seen in some of your data. They report that, I'm now more likely to see the good reasons that my opponent has for that view that I really dislike over there. I'm more likely to be open to hearing them present a reason for that, that I wouldn't have wanted to look at in a serious way before. And I'm just wondering about what the value of that is. It actually, say we talk about open and crew, we talk about the value of being this way in the world. Is there some, is there a downside? Is there something about having a conviction and not losing it that that's important? Is there anything, is there any kind of uprooting? Is there a danger of people being uprooted from having convictions and having beliefs of their own if they go through this process? Do you see a downside here or is there, what do you, what's it for?

Simon Cullen: I guess I have, I'm probably the least likely person to feel like having strong convictions is a good thing. But I even, I sort of get what you're saying. There's something to that. But generally speaking, I think people hold their beliefs and their convictions far too closely to their identities. And they should be more concerned with truth and reasonableness. And less concerned with the, you know, particular group that they align with.

John Tomasi: I'm sorry, does reasonableness mean moral relativism?

Simon Cullen: No, I actually don't really have a relativist position about reasonableness. I think there are some basic norms of rationality that we can know a priori for the philosophers just by thinking about it. I also think there are very strong empirical reasons to prize these kinds of interactions. The first thing you mentioned there was people reporting that they found that their partner, they were surprised that their partner had better reasons than they expected. That's a really important step to take. I'm not even thinking about changing your mind. I just want you to see that someone could hold a different view without being motivated by stupidity or ill will. That first step, just getting people to take that, is a huge achievement. Why is it so valuable? It's the engine of human reason. There's, you know, for anyone who's familiar with the recent psychology of heuristics and biases, you know, think of Danny Kahneman's book, Thinking Fast and Slow. The basic picture, we have two cognitive modes of engaging with the world. One of them is fast and frugal, effort minimizing, and pretty good in a whole range of circumstances. But human beings also have the capacity to step back and to reflect and to deliberately exercise our reason, our capacity for reason. We want to try and get people doing that second thing a lot more. One of the things that makes people fall onto the, as it's called in psychology system one, the fast and frugal effort minimizing thing, is the task being overwhelming, the task being too difficult. And that goes back to what we started off talking about, the difficulty of understanding argumentation presented in speech or prose. Look, if I make a sort of nuanced political argument to you just right now, the chances are, even though, well, you know, if only you weren't John Tomasi, but a regular person, even a very smart one, will not immediately understand everything I had to say. Probably systematically will misunderstand important parts of it. But nevertheless, there's a person in front of you, so you have to say something. You have to respond. And so what you do is fall back on this fast, frugal thing. And what we want to do is try and provide scaffolding to provide techniques that can help you to reduce the difficulty of these tasks, to reduce the difficulty of understanding the argument, of keeping track of what the hell we've said in this 35 minutes. And by making that easier, hopefully making it easier for you to engage your deliberate reasoning, to engage your rational capacities in a deliberate and effortful way. Why is that so important? So anyone who's familiar with the heuristics and biases literature, well, you know, a lot of people come away with it. It's not the message that Kahneman and Tversky wanted to share. A lot of people come away with it thinking, wow, boy, gee, humans are pretty stupid, aren't they? You have to be a real dope to think the bank teller and the book. But actually, that isn't the message. I don't know whose quote it is. But someone said, well, you know, if we're so dumb, how are we so smart? How is it that we can understand the things that we can, limited though they are? How is it that we can achieve extraordinary things like building these systems, like, you know, whatever, rattle off examples? The answer is through disagreement, through institutions that allow us to greatly amplify our rational capacities. That is a huge promise and a massive peril if it isn't used effectively. And that's really what we're trying to promote through Sway, through argument visualization. And by the way, those two things are going to be coming together soon. So, guide will be able to make argument visualizations.

John Tomasi: This is lovely. I have a dream that Sway will be widely adopted across universities, across the country, and maybe around the world. That we'll have students and professors assigning Sway experiences to students. We'll see this in a variety of courses and a variety of disciplines. And we'll see constructive disagreement and a change in the cultural open inquiry because of the work that you've been doing. Next year, you're going to be a fellow here in the Segal Center for Academic Pluralism. We're so looking forward to having you with us for a whole year next year. You'll be working on Sway while you're here, and we'll be helping any way we possibly can to advance your platform. And Simon Cullen, thank you so much for being on Heterodox Out Loud.

Simon Cullen: Thank you so much for having me. And I'm sure this is just the first of many discussions. Can I add one thing?

John Tomasi: Please.

Simon Cullen: Just for the listeners who are teaching classes who might be interested in Sway, it is funded generously through research grants. It's a research project. We are giving this away. Our website is swaybeta, S-W-A-Y-B-E-T-A dot A-I. And if you go there, you can go and check out the data from real students. The feedback is extraordinary. Although I haven't actually taught in a classroom for a long time, I still get so much of the reward from all of these hundreds of other people's students. Go check it out. We want to support as many Heterodox Academy members as we can. If you have a Heterodox community on your campus, we want to talk to you. We want to help you. We want to help to fill the interstitial spaces between the meetings and just get in touch.

John Tomasi: You are an entrepreneur in the world of ideas. It's fabulous. Thank you, Simon.

Simon Cullen: Thank you.