Dr. Matt Tompkins: From Magic to Mind Control
Dr. Matt Tompkins, experimental psychologist and magician, explores the intersection of magic and cognitive science. From mastering coin tricks as a child to designing "fake mind control machines" at Lund University, Matt reveals how magicians understood principles like change blindness decades before scientists. He discusses his book "The Spectacle of Illusion," founding the Science of Magic Association, and using mentalism to study public perceptions of AI with deception rates of 97 percent and higher.
Transcript
Coming up in this episode of the Magic Book Podcast …
Matt Tompkins [:We're essentially doing a séance. And one of the things working as a historian, when I talk to the media, a recurring thing is that people will be kind of dismissive. When you talk about trying to talk to dead people at the turn of the century, they'll say, “Oh, but that was so long ago. We're so much more sophisticated as a society now. None of those tricks would ever work on a modern educated audience.” That's why they always work, right?
Adrian Tennant [:You're listening to The Magic Book Podcast, conversations about classic and contemporary books that teach, illuminate and celebrate the art of magic. I'm your host, Adrian Tennant, a lifetime student of magic and mentalism, occasional performer and longtime book collector. Thanks for joining me. Today, my guest is Dr. Matt Tompkins, a postdoctoral researcher at Lund University's Department of Cognitive Science in Sweden, where his role involves designing and implementing what he calls fake mind control technologies. Matt is an experimental psychologist, professional magician and author who conducts groundbreaking research at the intersection of magic and cognitive science. Matt completed his doctoral degree in experimental psychology at Oxford University where his thesis, “Observations on Invisibility,” explored expectation, attention and visual awareness. At Lund University, he adapts techniques from magic to study how people perceive and misperceive emerging technologies related to neuroscience and artificial intelligence.
Adrian Tennant [:Uniquely, Matt is the first member of The Magic Circle to have been admitted to the organization on the basis of a peer-reviewed scientific publication. As a performer, Matt won the Swedish Researchers Grand Prix in 2022 by using magic to make complex psychological research accessible. And he's also a founding member and executive committee member of the Science of Magic Association, an interdisciplinary organization that fosters collaboration between academics and performers. Matt is the author of “The Spectacle of Illusion,” which explores the historical and contemporary relationships between magicians, fraudulent mystics, and scientists, and has co-authored academic papers including “The Phantom Vanish Magic Trick.” Matt, welcome to The Magic Book Podcast.
Matt Tompkins [:Thanks for having me. It's a pleasure to be here.
Adrian Tennant [:Well, let's start at the beginning. Can you tell us about your early experiences with magic?
Matt Tompkins [:So it's a bit of a cliched story for presumably your magic listeners. I grew up in upstate New York. I was maybe 12 or 13 years old and I remember seeing a magician performing at, I think it was a state fair, and he was actually a sword swallower – I looked this up later to confirm – it was a guy named Johnny Fox, who, according to research, he's passed away, unfortunately. But he studied under Slydini and so the sword swallowing was obviously eye-catching. But what really captured my attention as a kid was he had a beautiful routine with silver dollars. And I do have very vivid memories of watching him do a series of coin rolls, vanishes, reproductions. And the thing that really caught my attention was the coin falling up, which many of your listeners might know as a muscle pass. And that really got my attention as a kid.
Matt Tompkins [:And I sort of taught myself to do a coin roll by just sort of like brute-forcing it. And for my birthday that year around 12 or 13, I received actually two books. One that had less of an impact, but was quite interesting, was a classic: "Coin Tricks" by T. Nelson Downs, which was a little bit weird and aspirational. But the one that really stuck with me more was J.B. Bobo's "Modern Coin Magic." And that really got me started on more practical sleight of hand learning. And yeah, from there I started working wWith my younger brother. He was also into it. And we were tiny at the time, but some of our first paid work ever was doing children's birthday parties for smaller children with this very awkward kind of Penn and Teller- inspired double-act.
Adrian Tennant [:I love that. Well, you started performing semi professionally during your university years where initially you were studying medicine. Matt, what led you to transition from medicine to psychology?
Matt Tompkins [:Yeah, so like I said, I'd been doing a little bit of semi-professional work before that, but kept it up. That's how I paid my way through university partially. And when I started out, I went again to upstate New York, a little place called State University at Geneseo or SUNY Geneseo. An American-style liberal arts education, which I still cherish. There's a lot of things I don't necessarily want to translate from the US into Europe, but I think there could be more of that. And so I wanted to be a medical doctor. At the time I was looking at kind of emergency medicine or maybe psychiatry were the things that kind of captured my interest as an 18-year-old. In America you study biology and medical school as like a graduate degree. So I had a focus in biology.
Matt Tompkins [:That was my major. And I did a lot of volunteer work. I used to work on an ambulance, did some medical mission work in South America. But there was a little bit of a branching path for me where I went full into the psychology, which was, I believe it was my second year of university and I was given the choice by my advisor at the time where I could either take one of two paths and I could take another semester of organic chemistry, which didn't delight me. It was a gigantic, several hundred-person lecture. And it is cool little puzzles, but it didn't really engage me in the same way as some other things were at the time.
Matt Tompkins [:Or I could switch my major over to psychology. I could get a job working in a visual cognition lab just as an intern. And I could also get a small teaching fellowship while I would do teaching assistant work in the psychology department. But I couldn't do both. So I'm a psychologist now. But at the time, one of the things that was occurring to me was that there was quite a lot of overlap because it was liberal arts. So I took a lot of courses in addition to biology. So I was already taking Intro to Psychology and different things before that. And I was noticing in a lot of the demonstrations and the lectures that there was quite a bit of overlap between performing magic illusions and the demonstrations of human perception and illusory psychology that people were using. And that sort of sparked my interest. And I did a little bit of very low level writing on that, Discovered some folks that are pioneering the modern era of the science of magic. People like Gustav Kuhn and Richard Wiseman, Stephen Macknik and Susana Martinez-Conde. And that sort of inspired me to sort of continuously mash up the idea of psychology and performing magic as a research trajectory.
Adrian Tennant [:Matt, I understand a pivotal moment for you was seeing Dan Simon's basketball game paradigm, which is a famous selective attention study. For those who don't know it, could you explain what it is and how that experience influenced your thinking about the connections between magic and psychology?
Matt Tompkins [:Yeah, so this is a great example of sort of the parallelism between the concept of misdirection as it applies to psychology research and how it applies to different ways that construct performance illusions. One thing I will say very briefly is for those of you that don't know what we're talking about, I strongly encourage you to pause this and plug “selective attention test” into YouTube. It's a very short video, but watching it for yourself is very different from me explaining it with words.
Matt Tompkins [:But that being said, this is an experiment. They called it “selective attention” at the time. These days we refer to this as “inattentional blindness.” And this was kind of the revolutionary part of this particular paradigm is that they connected this idea of selective attention with the idea, which was relatively new in the 1990s, of inattentional blindness. So this experiment was published in 1999. And the way it works is if you're participating in the experiments, you simply watch a short video clip. There's multiple conditions but the flagship one is you're instructed to watch two teams of people on a screen. So it's just a video recording. One team of people is wearing white T-shirts, the other team is wearing black T-shirts. And you're instructed to watch the people wearing white and not be distracted by the people wearing the black shirts.
Matt Tompkins [:Each of those teams has a basketball to themselves, and they only pass to each other. And your job, your experimental task, or the primary task, is to count the number of times that the players in white pass the basketball. And at the end of it, they get a count. And usually it's people are quite good at this, which is kind of interesting in and of itself. The idea of being able to parse a relatively complex, what we would say a naturalistic, dynamic scene computationally, as people are running into with a lot of the AI and computer vision, stuff like this, is actually a very technically demanding problem. But that's not the interesting part of the experiment.
Matt Tompkins [:The interesting part of the experiment is the twist, as they say. “How many basketball passes did you see?” People answer, and then they ask, “How many gorillas did you see?” And in those conditions we just described, there's ways of dialing it up or dialing it down, depending on particular instructions. But when you're just going in cold, watching the people in white, around 60 percent of people do not notice a young woman in a full-on Halloween gorilla costume that walks through the players, stands in the center of the screen, beats her chest, and then slowly walks out. So very, very visible objectively. But around 60 percent of the people watching under those conditions, they don't think there was a gorilla there. And it is extremely surprising to people when you reveal a lot of times people will initially not believe that it's the same video. I've had that happen where you have to very clearly like scroll back and they have to see you do that. Otherwise they don't think that there was any way that they could have missed it.
Matt Tompkins [:And again, as a magician, the idea that people can fail to detect something that's plainly visible, not necessarily that surprising, although magicians are often very surprised by the gorilla. But this overall concept that you can do things invisibly when people's attention is otherwise engaged is very much parallel to a lot of the ideas of misdirection that magicians have been talking about since, you know, time immemorial, like since we've had written records of any of this. And so that really inspired me to look more closely at these ideas. So my particular specialization actually still is in visual attention. I've gone a little bit weirder in a couple of branches, but depending on who I'm asking for money in grant applications, I work in the field of visual cognition. And metacognition is the other important one, which is really related to that basketball game as well. Metacognition sounds a little bit complicated, but the basic breakdown is it's the way people think about their own thinking and applying it to this. And again, there's great parallels to magic here.
Matt Tompkins [:The fact that people fail to detect the gorilla is a very robust result – around 60 percent of people in this particular case – again, you can do different kinds of conditions, different kinds of stimuli, but this idea of you focus on a particular task and something unexpected happens, even if it's happening very vividly within your visual field, people will often fail to notice that if their attention is otherwise engaged and they're not expecting that surprising event. What's also really interesting from the metacognitive perspective is that sort of, like I mentioned, with people needing to see the video rewound to believe it, this is deeply surprising for almost everybody, more or less like as close as you. And I'm hesitating here because this is something that, like as a scientist, I don't want to say like “universally,” but like pretty close. This idea that it feels like this is impossible and this is like a magic trick, right? This idea of this experience of impossibility.
It's a little bit different because here you're confronting them with the exact situation and revealing a little bit like, because the method is they just walk through. There's nothing secret about it.
Matt Tompkins [:But this idea that this kind of experience that people have, this – “inattentional blindness” is the technical term for it – this happens to folks all the time, presumably in day-to-day life, but you don't necessarily notice it because obviously in day-to-day existence you don't notice the things you don't notice, quite simply. So people don't necessarily update their beliefs around this because it's very unusual for you to be explicitly confronted by this discrepancy. It does happen occasionally, but not in a way that you would necessarily update beliefs and change the way you think about the way that you see. So this is what we would call a metacognitive illusion. And if you're interested in this, Dan Simons and Chris Chabrice were the guys who wrote this original paper. They've done some really beautiful work on this over the last couple of decades. They have a really great popular science book, the spoiler’s in the name: “The Invisible Gorilla and How Our Intuitions Deceive Us.” It is not a magic book, but I think anybody with an interest in magic would definitely get a kick out of it.
Adrian Tennant [:Perfect. Your path to took you to Oxford for graduate studies where your doctorate thesis was titled “Observations on Invisibility.” After completing your Ph.D. you spent several years freelancing, performing, lecturing, doing TV and radio work. Matt, how did that period shape your understanding of how to communicate complex ideas to different audiences?
Matt Tompkins [:Yeah, so I originally wound up at Oxford as part of my undergraduate studies. It was a kind of like a study abroad exchange. I made some contacts with the department and wound up back there for a Master's and then subsequently for doctoral research to do my Ph.D., or as they call it over there, a D.Phil. And the idea here is it was like a progression. So for the master's project I studied very specifically – not gorillas per se, but gorilla-adjacent – it was specifically a project on inattentional blindness, quite traditional kind of psychology, visual cognition methods. But through that time I was again very interested in this interface of magic and research. And I was still performing on again, off again, at the time. And for my doctoral studies I pitched a project where it was going to be a combination of empirical research, but also a historical investigation of the interplay between performing magicians and mostly psychological researchers. But it branched out a little bit as time went on. So that was what my thesis wound up being.
Matt Tompkins [:A couple of little detours, kind of a false start with some neuroscience, but wound up looking at kind of behavioral studies of attention, specifically on how people see things that aren't there, how people fail to see things that are right in front of them, and using sleight of hand very simply to demonstrate that in kind of robust replicable ways, but also looking at the history. So after finishing my Ph.D., I took a little bit of a break from academia, still kept my hand in with some affiliations, kept up a little bit of teaching, but did quite a lot of work with public engagement. So one of the things I developed over those years was I pulled back a little bit from performing sort of straight magic shows, like I said, the occasional close up gig here and there. But what I started to do was integrate the performances with kind of academic speaking. So I would create lectures that would talk about psychology concepts from contemporary research on attention, perception, memory and reasoning. Also a little bit of history of science, a bit of storytelling. But then I would integrate tricks into that.
Matt Tompkins [:So it would be a combination of like an academic lecture and a performance. So not quite as showmanship-y as a straight magic show, but significantly weirder than a more traditional dry academic lecture. And I spent quite a lot of time doing that. Made some media contacts off the back of some studies that I published, working with particularly folks at the BBC, started writing kind of popular science articles. Again, this idea of storytelling, one of the things that I came to realize was there were a lot of interrelationships between the idea of teaching, “public engagement,” as we call it in the academia scene, which is public lecturing for like non academics, and also even presenting your work for academics. There were all these kind of tools of showmanship.
Matt Tompkins [:And one of the things that I've taken away from that is this idea that when you're presenting a show, all the things that you learn as a magician, this idea of like considering the audience's perspective, considering the goal sets, all of these are toolkits that can be applied for other kinds of engagement. Other people have figured this out in various ways, but the particular flavor of using it for the psychology of illusion demonstrations is slightly more unusual. Few of us in the world are doing it, and for me, especially in the science communication aspect, the idea, again similar to the teaching, is you're trying to give people truer ideas about, particularly in my case, what you would think of as counterintuitive aspects about how our minds perceive and remember the world. It can be really hard to do that in a straight up way if you just describe research verbally. This is why a moment ago I was like, “You need to watch the gorilla video. Don't just listen to me.” It's a lot more powerful if you get those demonstrations.
Matt Tompkins [:And magic gives us a great toolkit for getting those demonstrations. So people have actually this experience of it. And I found that really useful in my work. And again, more broadly, just this idea that all the tools you use in a magic show to construct elaborate deceptions, this is a kind of persuasion. These same kinds of tools can be used in a kind of a teaching and in a nicer sort of environment to guide people towards hopefully more correct ways of thinking. So to my mind, the ideas around deception and misinformation, a lot of these same tools can be used for good science communication. The tool is kind of value- agnostic. But you can use these things for good.
Adrian Tennant [:Well, your book, “The Spectacle of Illusion” examines both the history and science of deception and magical thinking. Matt, how did this book project come about?
Matt Tompkins [:So the book project was part of an exhibition at the time, so it was in some ways independent. But the idea of me getting to write this book was in correspondence with an exhibition that the Wellcome Collection in London – big science organization – put on in 2019, which was a whole big extravaganza of history and contemporary studies of how psychology and magic interrelate. So I knew a lot of the people that were responsible for getting that initially set up.
Matt Tompkins [:One person in particular really got me keyed-in for the book deal, which was a guy named Al Hopwood, or he goes by - A.R. Hopwood is his artist name. He's a really amazing guy. We met actually at the University of Goldsmiths where we had some mutual friends. And he was the artist in residence at the time at the Anomalistic Psychology Research Unit. The guy running it, Chris French, has semi-retired now, he's emeritus. So it's not doing as much booming business as it used to, which is a shame, but the idea behind that was looking at sort of paranormal beliefs, not necessarily parapsychology. And like he was looking at the mechanisms of whether ESP exists or not. Slightly stranger than that, he was looking at how you could cause people, what kind of known and definable psychological experiences could cause people to have these beliefs about paranormal abilities. Everything from ghosts to aliens to psychic powers. Really interesting stuff.
Matt Tompkins [:Anyway, Al was working at the Wellcome at the time. He managed to put this exhibition together which was a huge deal. And he tapped me to help write a companion piece for the exhibition. And it was quite a tight deadline and I was still in the process of finishing up my doctoral thesis. And the way I was able to make that work was that the book is sort of a pop-ified version of a lot of the historical work that I'd already conducted for my thesis. So I took a lot of the research that I'd written up in a much drier, more academic prose, and turned it into more stories. And that was how the book came about. And also the other big factor was I had access to still one of my favorite magic libraries in the world, which is the Harry Price Collection at the Senate House Library. If anyone's ever in London, I can highly recommend ordering up bits and pieces of that. It's beautiful stuff.
Adrian Tennant [:One fascinating aspect you discuss is how magicians understood principles like change blindness long before scientists formally recognized them. How does this historical perspective change our understanding of both magic and psychology?
Matt Tompkins [:So this relates a little bit back to the idea of metacognition that we talked about with reference to the gorilla, and the fact that people can be very surprised about otherwise normal ways that their mind works. Change blindness is a really interesting example of that. So change blindness entered the psychology lexicon in the mid-1990s. There's a kind of an interesting history there. But very simply, and I'm going to describe this, but if you check out my website, I've got demonstrations. One of the fun things is even if I explain how this one works, it's still going be fun, unlike the gorilla. Even if you know what's going to happen, this will still work pretty effectively.
Matt Tompkins [:So the way that the psychologists originally found it, this was a group of people, a team led by Professor Ron Rensink, who was at the University of British Columbia at the time. And the idea was they were interested in how people perceive sort of changes in scenes. There's this interesting sort of eccentricity of the human eye and the way it operates in connection with the brain, where that every time you make an eye movement, just anytime in day-to-day life, for the most part, you actually go blind for a few milliseconds. This is called a “saccadic suppression.” Perfectly standard, very adaptive, but it doesn't feel intuitively correct. And this was Ron's idea of trying to create an easy analog.
Matt Tompkins [:These days there's some interesting applications you can do with like real-time online eye tracking. Like for example, there's like the newer VR that track your eyeballs. They will do fresh renderings and refreshes during eye movements so that they're invisible. So cool modern application. Anyway, with Ron, this was quite a while back. Eye tracking was a little bit less accessible. So they tried to create a system that would create sort of an analog of that without bothering with the actual saccades. It was a little bit like a slideshow. They would have two photographs and a blank screen. And there was actually often very significant differences between the two photographs. So there was almost identical, but with like one major change.
Matt Tompkins [:And then they would show people a looping image where you'd see one picture, you'd see the blank screen, the modified picture, blank screen, the original picture. And it was a reaction time task. You'd hit a button when you detected the changes between the two pictures. And one of the things that they found that was really surprising is that it took people a lot longer to detect these changes than people predicted. So much so, in fact, that when they tried to publish these results, it was actually almost unanimously rejected by peer reviewers in scientific journals.
Matt Tompkins [:So these were visual scientists – like I said, this was back in the Nineties – and they looked at Ron's results and they said, like, “Your experiment's broken.” Like, “There's no way people would be that bad at seeing changes this major.” Like, “There's no way it could take that long.” The only way he managed to get this published is he got one of the editors to the journal and managed to sit him down in the lab and make him do the experimental tasks so he could experience for himself how it actually felt like and really worked.
Matt Tompkins [:But back to the magic angle. What's interesting about this is there's really strong parallels, particularly in card magic. So I think one of the easiest ones is another bit of a T. Nelson Downs connection. It was originally published as the Prince's – possessive – Card Trick. And you can see advertisements of that going back around the turn of the century. And you can see it formally written up, most easily accessible in 1909, in T. Nelson Downs’, “The Art of Magic.” And again, the idea here is you show people an array of playing cards.
Matt Tompkins [:You have them mentally select one of the cards, and then you apparently remove one of the cards, and then you show all the cards again, and people see that the card that they were only mentally selecting has vanished. And kind of spoiler alerts … There's a couple of ways of doing this with different gimmick cards, or sleight of hand, but the way it actually works is you change all of the cards.
Matt Tompkins [:And again, similar to this change blindness idea, when people are not expecting this – and again, you have them focus on the one card for the most part – the first time people see this, it seems like you've only just taken the one away. And it feels impossible that you could know that. And you get this very nice, very simple kind of mentalism effect with playing cards. But what's interesting is magicians have understood that principle at least since the early 1900s. You can kind of go a little bit further back with Hoffman and various things, but that's, I think, the cleanest example.
Matt Tompkins [:And so what this tells us is not necessarily the magicians understood how the attention operated, not mechanistically, what was happening on a cognitive level, but they understood that it was possible in a way that, like I said, visual scientists, almost 100 years later, didn't believe that it was physically possible. Like I said, they thought the experiment was broken.
And what this kind of demonstrates – and a good friend of mine, Cyril Thomas, has talked about sort of a parallel of the cryptozoological approach to magic and psychology research – is you can look back at these kind of illusions and you can find these weird examples of things that seem to violate what we understand about psychological principles. And you can use these and then study them to sort of expand our understanding of how the mind works. And this is a really interesting example with change blindness.
Adrian Tennant [:Excellent. If you're enjoying this episode of the Magic Book Podcast, please consider leaving a rating on Spotify or a review on Apple Podcasts. You can also follow The Magic Book Podcast page on Facebook. Thanks. Matt, you're currently in Sweden. Can you tell us about the Choice Blindness Lab and the type of research you're conducting there?
Matt Tompkins [:Yeah, absolutely. This is a great transition from the Princess card trick and change blindness idea. So right now I'm working as a researcher in the Choice Blindness Lab at the University of Lund. This lab is headed up by a couple of guys named Petter Johansson and Lars Hall. And they got their start actually with a variation of change blindness. So choice blindness is a pretty direct evolution of the change blindness idea. And interestingly, they found it again by going back to playing cards and magic trick methods. There's been a lot of variations on this over the years, but the kind of the flagship one was a science publication from 2005.
Matt Tompkins [:And the experiment goes a little like this. So it's face to face with a researcher and a participant, and they sit across from each other at a table, and the researcher holds up images of faces, just a small sort of profile shot, one in each hand – and this is not a coincidence – playing card-sized images. And so one card in the researcher's right hand, one card in the researcher's left hand. The participant, their job is simply to indicate, just by pointing, which face they consider to be the more aesthetically pleasing or the most attractive face of the two images presented to them. And these can be sometimes very different looking individuals.
Matt Tompkins [:So the participant points to an image and ostensibly the experimenter then reaches across the table and hands them the image that they point it to. They then pick up that image and they look at it again. And the experimenter does a very sort of quick interview where they ask, “Can you explain to me why you made that choice, as opposed to choosing the other image? Why did you make this selection?” And people will explain to you sometimes in great detail why they chose this face as opposed to the other one. And they'll give you very specific visual reasons and personal reasons.
Matt Tompkins [: Matt Tompkins [:But where it gets really interesting is in that interview section where people will proceed to justify in extreme detail why they made the choice that they didn't actually make. And this opens up some really interesting … So I'm actually, technically, I'm part of a philosophy department here. This isn't strictly a psychology department.
Matt Tompkins [:And this idea of the way people – you would say they confabulate, or they create a false memory of their own reasoning – as to why they made this. So all of the explanations have to be post hoc and they have to be illusory because it isn't the choice that they made. The choice was made for them without their knowledge. And so when they come up with those reasons, those reasons are purely fabricated. But what's interesting is they don't notice this. And if you look at transcripts of real choices and forced choices, there is no discernible difference. And people have tried a lot of different ways to look at the difference, but there's no difference between them. You can't tell just from the transcript if this was a false choice or a true choice.
Matt Tompkins [:It looks exactly the same, the exact same patterns of reasoning. There's no tells, which is a really interesting way that you can take this kind of perceptual process and expand it into kind of this metacognitive and reasoning domain. And like I said, that was a science publication. That's a very big deal in the academic industry. It helped set the lab up and it was kind of fun for me, because it's this nice evolution of this principle that magicians had been using that psychologists had then discovered. And then you go back to the magic to bring it further and expand upon it in ways that is really surprising to a lot of people and opens up a lot of possibilities for further research to expand what we know about how people think.
Adrian Tennant [:An aspect of your current work is using mentalism techniques to simulate advanced AI and neurotechnology. Matt, can you explain this approach and some of the things you're learning?
Matt Tompkins [:Yeah, sure. So my elevator pitch for the research that I currently do in the lab is that I say – and this is mostly true in a selective sort of way – is that when people ask me what my job is I say, “Oh, I design fake mind control machines for the Swedish government.” All of that is technically true. The fuzzier part of it is the government thing. So I work for a public university, so in a technical sense, I'm a government employee. But that's about as close as that gets. The fake mind control machines, though, is the headline. The idea is we're interested in exploring people's perceptions of emerging tech, particularly in kind of the interface with AI and neurotechnologies. And if anybody's been to the Internet at all in the last five years, you're probably aware of some of the controversies and excitement around the emergence of AI, which has actually been messing with my results a little bit in interesting ways. That's science.
Matt Tompkins [:But the idea is there's a lot of hype, there's a lot of excitement, there's a lot of trepidation about what the impacts of these new systems are going to be. How is this going to impact people? How is this going to change the way people interact with each other, interact with the world? And there's a big push, for example, to look at how do we create regulations to anticipate some of the problems that might happen here? And there are a number of problems with that kind of complications and challenges.
Matt Tompkins [:I can't solve all of them, but I do have a solution for a significant chunk, or we do in the lab. And so the idea is, when you're trying to design guidelines, regulations, even laws around tech before it exists. So, again, there's a lot of really amazing things that this stuff can do, and there's a lot of misperceptions about what it can do. But if you think about the future, and you want to get ahead of this stuff with some sort of regulation, the challenge is that you're trying to design a regulation or a policy that's based on imaginary behaviors of future people who are reacting to imaginary technologies that don't exist yet, that creates a lot of complications.
Matt Tompkins [:For starters, we've already talked a lot about the metacognition. This applies pretty fundamentally to the way people predict their own behavior in future circumstances. So we know from loads and loads of research that if you just ask someone, like in an interview, how they're going to respond to a scenario, even mundane scenarios, what they tell you, even if they're being honest with you – which is not a guarantee, and that's a separate issue – but even assuming that someone's being honest and they're trying to explain what they're going to do, the accuracy on that is very unreliable. People are not very good ...
Matt Tompkins [:We all are. Yeah, you, me, everybody can be quite unreliable in predicting their own behavioral responses to future events. And it gets even weirder when those events get weirder. So I can't solve the technology part. Like that's not my department. I'm agnostic about the developments of AI tech, but I can solve the behavioral part of it. So what we do is we create simulations of future technology. So this uses illusion and deception. So when people sign up to our. experiments we say, “Oh, we have this new piece of technology.” So some things that we've said it can do that we're currently writing up - hopefully I'll have results in the next couple of months that are publishable –
Matt Tompkins [:We say, “Oh, we have this machine that it's a high-fidelity prediction device. Like with a little bit of calibration it's going to know all of the actions that you're going to take in the situation before you know them.” Or we say, “This is a mind-reading device. Like it can look at your face and pick up brain activity and determine exactly what you were thinking of in like an online real time setting.”
Matt Tompkins [:Or we say, “This is a mind control machine.” Effectively, like “This is going to cause you to enact certain behaviors basically under the instructions of the AI system and it's not going to feel like you're being influenced. But we'll show you how that all of your actions were determined by this system.” All of these are lies. But like I said, there's a lot of cool stuff that AI can do.
Matt Tompkins [:There's a lot of really interesting ways of interfacing like the large language models with neuroscientific FMRI to interpret brain activity in really cool ways. But we're definitely not on the level of like sort of Star Trek kind of mind-reading that is sometimes portrayed in the media. But using magic, we can create these experiences. Like I said, my job is to design convincing simulations. So when people engage with our experiments, my job is to create false evidence so we don't tell people, like I said, that they're going to be seeing a magic show, but they're going to be seeing a magic show! And the idea is to create these experiences. We do it undercover of various things, but we say, “Oh, this is part of the calibration of the system.” Again, the system doesn't actually exist. We don't have a system - purely fictional. But the idea is we use that as a convincer where we show them what are magic tricks to prove the efficacy of the system. In relation to whatever claims that we're making about it.
Matt Tompkins [:And the idea behind this is when people really believe that the system is working, they've experienced for themselves firsthand pseudo empirical evidence and they actually believe that this thing exists, that it's real, that they're interacting with it, we can then ask them questions about it. We can conduct different kinds of tests. And this is a lot more powerful, potentially than doing this more honestly as an imagination game. We could ask, “Hey, can you imagine a future where a machine could predict your actions with high fidelity? How would you feel about that?
What would that be like for you? How would you behave in that situation?”
Matt Tompkins [:And people will be able to answer quite fluently, but as I mentioned previously, those answers, not very reliable. With the deception and the simulation, we can cut out a lot of that unreliability because for them, it's not a role-playing exercise. Like, they actually believe that they've just experienced this. So we know how they're going to react because they think they're in that situation. I should also note that for ethical purposes, we always, we have very careful protocols in place to make sure we, as you say in the business, de-hoax participants. So nobody walks away from these studies thinking this tech exists. We make it very clear to people - like, we don't fully reveal the exact methods. For those listeners in the know, one of the examples we use is Corinda's first step.
Matt Tompkins [:Yep, very effective. And that's a recurring theme. Part of this is for me. But we draw on stuff that is often very old kind of mentalism methods. I like to pull stuff from fraudulent mediums. And effectively what we're doing is a séance in a lot of traditional ways. Except instead of saying that we're talking to dead people, I'm saying we're talking to a fictional algorithm. And it's very effective in this day and age when people have a lot of ideas about how this tech could or ought to work.
Adrian Tennant [:Now, when we were preparing for this podcast interview, you mentioned to me that you can achieve a 97 to 98 percent deception rate even with – or maybe especially with – computer scientists. Matt, what does this tell us about human psychology and our relationship with technology?
Matt Tompkins [:So that 97 percent deception rate is combining a couple things from the conversation. The 97 perception rate refers to that prediction machine paradigm that hopefully will be published in the next couple of months. We're very careful actually, about how we measure deception because this is a tricky thing. Like I said for the simulation, it needs to be deceptive in order for us to get good data. But at the same time we can't just ask people like, “Hy, were you deceived?” It's a bad question and it's not going to be like a very methodologically effective.
Matt Tompkins [:So what we do is very carefully, we have what's called a manipulation check, which is where we probe to see if the deception was bought by the participant. The way we do that, that is we ask people at the very end, we say – roughly, not exactly the quote here – but the question goes something like, “This is a question we ask all participants towards the end of this study when we first tell people about the system. A lot of times people can be very skeptical. Sometimes they don't believe that the system is possible, but you have just experienced it for yourself, so you've interacted with it, you've seen what it can do. So our question for you” – this is to the participant – “is, was there any point throughout the procedure where you yourself were skeptical or you didn't believe that this could, could really be happening?”
Matt Tompkins [:So one of the issues in psychology experiments is something that they call a demand characteristic, which actually they learned about from stage hypnotism, which is another interesting story, which is that people who volunteer for psychology studies, they really want to help, which can be very useful, but a potential pitfall with deception research. Traditionally, psychologists aren't exactly trained in deception in the way magicians are. So if the deception is not necessarily effective, people might realize that there is a trick involved. And if that happens, they might, in the interest of being helpful to the researcher, just play along and pretend that they were fooled because they think that would be useful to the scientists, but that would be devastating for our results. So we need to make sure that doesn't happen.
Matt Tompkins [:So what that statement does is it's effectively, it's inverting the demand characteristics. So now they know that if they had suspicions about this being a trick, that it would be very helpful for us, the researchers, to know that. And as you mentioned, when we do this, at this point with these current paradigms, it's around 97 to 98 percent effective in terms of people just fully buying into the fact that what they experienced was real. And they'll explain things like, “At first I was very suspicious, but then I experienced it for myself.”
Matt Tompkins [:And we have a follow up question with this latest round which is quite fun, where we then after they've said that, then we say, “Well, what would you say to somebody who hasn't seen this in action before but doesn't believe that it's possible?” And the response is fairly stereotyped across a few hundred people now, which is they say, “Well, if someone doesn't believe it, then I think they would need to do what I did. And if they saw it for themselves, anyone who experienced what I did would have a similar kind of belief adjustment,” which is quite fun. And they'll sometimes explain to you exactly what part of the paradigm caused them to shift from suspicion to belief in the system.
Matt Tompkins [:Which is an interesting measure because it doesn't necessarily mean that's what changed their belief, but it means that's what they think changed their belief. Which is something that's going to be quite interesting to investigate more. What does this tell us about technology? So I think it tells us to be quite careful about some of the more extraordinary claims emerging in the AI and neuroscience space. I think it tells us to be very careful about what qualifies as evidence, especially when it comes to personal experience. These are all lessons that we've learned in a lot of different ways for various things. Particularly, you can draw a lot of parallels with like paranormal research.
Matt Tompkins [:And I think this is really important. This idea of kind of a cognitive humility is something that I really think is viable. Cognitive humility just refers to this fact – it's related to the metacognition – but very similar to the James Randi proselytization of this idea that you can be deceived, even people that study deception, even people that are experts in the very cutting edge of aspects of this technology, under the right circumstances, you can design things that can potentially deceive these people.
Matt Tompkins [:And that's not necessarily a problem unless you think that you can't be deceived. So again, I don't necessarily think that there's any way to create some sort of like perfect inoculation where people are not going to be fooled by this stuff. But I think teaching people that it is possible to be fooled by it, and that itself is not an impossibility, can really be quite helpful in the way that people think and process these things. It's also good fun for me, like I said, because we're essentially doing a séance. And one of the things working as a historian, when I talk to the media, a recurring thing is that people will be dismissive when you talk about, like, you know, scientists trying to talk to dead people in like the turn of the century, in the 1880s, 1890s, they'll say, “Oh, but that was, you know, so long ago.” Like, “We're so much more sophisticated as a society. Now, none of those tricks would ever work on a modern educated audience.” That's why they always work right? Like (laughs) and it's quite fun.
Matt Tompkins [:Again, I try not to use methods like … Mentalism's had a lot of really fun developments over the intervening century, but for simplicity's sake, and also because I find it aesthetically pleasing, I try not to use stuff that's like post-1880 as the methods for our cutting-edge AI neuroscientific systems. And part of that is the rhetorical and methodological move to demonstrate that these things are still very effective at deceiving people under the right circumstances, just as they were back in the day.
Adrian Tennant [:Just a reminder that you can be notified when new episodes of this podcast are published by subscribing to the email alerts. You'll find all the details on the podcast website at TheMagicBookPodcast.com, which is where you can also find find transcripts plus accompanying blog posts with summaries, timestamps, and links to resources mentioned in each episode. Matt, let's talk about the Science of Magic Association, which you helped found. What was the motivation behind creating the organization?
Matt Tompkins [:Yeah, so the Science of Magic Organization, or SOMA, if you acronym it, is an interdisciplinary organization that's been designed to bring together not only people that are interested in magic within the academic context. Everything from psychologists like myself to philosophers to historians, but also for people outside of the academic community. So people that are in the performance community or magic historians that don't necessarily associate themselves with formal academia.
Matt Tompkins [:And the idea is to bring these folks together and create like dialogue and collaborations to how we can move the field, which sort of been dubbed “the science of magic,” which is a slight misnomer in that it's not necessarily completely focused on scientific aspects, it's more of kind of research aspects. So again, we welcome philosophers, historians, folks from the humanities disciplines just as much.
Matt Tompkins [:So we started conducting conferences. So back in 2015 and 2016, I believe, were the first ones. And this was originally the brainchild of a good friend of mine named Gustav Kuhn. Anyone who's interested in this field, I highly recommend seeking out his work. There's two very excellent books that are great gateways into this field for anybody. There's his book, “Experiencing the Impossible,” which was through MIT Press. For magicians in particular, he, along with another excellent researcher named Alice Pailhes, published a book which was out with Vanishing, Inc., “[The Psychology of Magic:] From Lab to Stage,” which is very much written for magicians to look at kind of applications of psychological researchers’ research to performers.
Matt Tompkins [:And again, the idea behind SOMA is very much related to that. It's how can we create a dialogue and create kind of mutual beneficial relationships? How can we move science forward with these ideas – in ways that we've been talking about – a little bit? And also how can we enhance the appreciation and maybe move things forward design-wise a little bit with people in the performance sphere?
Adrian Tennant [:SOMA has held several international conferences. The most recent one was in Las Vegas. Matt, how have these events evolved and what role do they now play in advancing the field?
Matt Tompkins [:Yeah, so the last SOMA conference was hosted in Las Vegas. We were at the Golden Nugget Casino. One of the great things about this, obviously Vegas is a mecca for the performing magic community. A particular highlight was we had Jeff McBride's Mystery School partnered up with us to do some really great events within the conference. We also had Raymond Teller of Penn and Teller as our kind of guest of honor speaker, who was excellent needless to say – kind of like I mentioned a childhood dream situation for me – also just a really lovely guy and very, as many people know, like has a highly intellectual approach to performance that mixes very nicely with the crowd that we try and bring together.
Matt Tompkins [:So in terms of the development of Science of Magic Association and SOMA, [it is] hard to pin causality to it, but we like to think that we've helped things along a lot, particularly in the academic sphere. So prior to say like 2005, there were less than 20 articles that were published in academic journals where you had empirical work investigating magic illusions. In the last 25 years or so that has gone up exponentially. So well over 200 articles at this point. And this is again just specifically looking at kind of empirical studies. This is setting aside some of the kind of commentaries and theoretical work, which is also excellent. But one of the key things for particularly psychology is people are very keen on looking at kind of empirical investigations. That's something that's seen as a hallmark of moving a field forward. So we like to think that we've helped a bit with that by encouraging people and raising awareness about the possibilities of using these ideas.
Matt Tompkins [:One of the kind of underlying goals is the fact that there's a few of us: so like myself; Gustav; a guy named Tony Barnhart, who's also been really influential in founding this; Cyril Thomas I mentioned. There's a few of us in the world that have backgrounds as performing magicians, often from childhood, and then we went on to become psychology researchers. So we have these kind of parallel tracks but what we sort of figured out is that's not super practical, like very effective. But you can't go to somebody who's got a Ph.D. at a university and be like, “Well now hop in a time machine and get your Bobo's ‘Modern Manual’ at the age of 12!”
Matt Tompkins [:The solution to this is to foster collaboration, so creating effective dialogue. So rather than kind of getting a magician to get a Ph.D. or getting a psychological researcher to spend like the next 20 years performing gigs, you can bring these people together and combine the strengths and minimize the weaknesses. And so a big part of what we do is trying to create good dialogues. And historically there's been a bit of friction between performing community and the academy community. Various reasons. There can be some dismissiveness on the part of the academics. I think that you can see going back through the years, where I think that's shown some marked improvement. You still see it cropping up occasionally, but there's now very much more of a culture of mutual respect.
Matt Tompkins [:And part of it is also explaining to magicians about the kind of strengths and limitations of taking a more academic method to exploring these kinds of ideas. Like a good example is if you think about the way you design an experiment. So researchers are often very keen on this idea of like reductionism, how you kind of eliminate down things to create, where you can look at how one variable has a causal relationship.This is similar but also different from the way that you might design like let’s say you're a close up worker. Some people have said that there's this good analogy with the way that you perform things in the real world and the way a science experiment works and that you're empirically testing out stuff: you find things that work and you discard the things that don't work. This analogy can be a little bit tricky. It's not useless, but it can lead down some sort of blind alleys.
Matt Tompkins [:Like a key difference within a performing context is if you're, say you're a close up worker. This is something I've got a little bit of experience with and I think this is nice because you're doing lots and lots of mini performances, so you're iterating very quickly over the course of a couple of gigs. The idea is when you're designing a show as a performer, as a professional, you're getting paid to entertain. So there's a little bit of leeway where you can experiment, but you also don't want to reduce things down to the point where the trick stops working. That's not a very useful or practical thing to do economically or you know, aesthetically. As a consequence, and you see this all the time, you get performers that kind of layer on methods on top of particular effects. And so this is very effective from a performance standpoint, maybe sometimes a little inefficient, but still very functional.
Matt Tompkins [:But the issue is, when you start layering different kind of methods, you can obscure the causality. You don't know exactly what method is doing what, or how they're interacting. This can be very hard to determine in kind of an informal context. In an experiment, we're in a kind of a privileged condition where instead of getting paid to perform, we are paying people to sit there and watch. And this allows us to effectively – and this is where some magicians get a little bit apprehensive – is that if you look at some of the paradigms that are studied in academic contexts, the magic trick from a performance perspective looks like garbage because it is deliberately reduced down and sometimes to the point where failure is a real possibility, because we're trying to unpick, like, very specific elements of a method and what the causality of it is. So you would never necessarily do this in a performance context, but understanding where the causality happens can potentially help downstream.
Matt Tompkins [:So that's one example. Another example is, and this is interesting, given the context of this podcast, the accessibility of literature. I mean, there's a lot of annoying things about gatekeeping and paywalls with academic journals, but it is a breeze relative to finding some obscure magic literature. Sometimes, unless you have access to somebody with a good private collection or a really excellent library, it can sometimes be really hard to track down very specific volumes. And this can be quite challenging for researchers from the academic side because there's a temptation when you start reading magic books to sort of treat them as you would more formal academic literature – because a lot of times they structurally, they feel similarly – but there's a lot of differences. Again, a lot of this stuff is not formally tested in that kind of reductionist sort of way that we talked about.
Matt Tompkins [:These are a lot more opinion-based, which is still valid, but it should be seen more as presenting testable hypotheses. This is in no way to diminish the experience of these people, but it's a different kind of experience than a more formal academic investigation, and it has different connotations. So a big part of what we do is try to clear up those sort of potential miscommunications when people are talking about things like developing methods, talking about literature. There's a lot of ways that people can come together. It's just sometimes the language can obscure and lead to potential frictions.
Adrian Tennant [:This is the Magic Book Podcast. So, Matt, what magic book or books do you most cherish in your own collection and why?
Matt Tompkins [:It's part of the Proceedings from the Journal of the Society of Psychical Research. So this was actually a science experiment. It is, I think, for my money, probably the first formal study of magic illusions. And a kind of – they didn't call themselves psychologists – but it's effectively a psychology experiment. And this was a study that was published in 1887.Title's a bit of a mouthful - it is “The Possibilities of Mal-Observation and Lapse of Memory from a Practical Point of View.” And this is a really beautiful study. And it is effectively a magic book. I managed to track down a first edition, so I have the original printing of this, which is delightful, a little bit of magical thinking there. I don't know what it is about first editions. I also enjoy facsimiles, but there is something kind of irrational and special about the first edition of things that I also tap into, happily!
Matt Tompkins [:But the idea was that this was back in 1887, this was in the heyday of what we now call modern spiritualism. This idea – and there's a lot of ways to unpack this – but the key features for present purposes is this idea that they presented it as an empirical religion. So rather than having to have faith in miracles, they would show you miracles. You wanted to know about the afterlife? They'd put you in touch with a dead person.
Matt Tompkins [:Very powerful stuff. And it also piqued the interest of a lot of luminaries in the scientific community at the time. But there was obviously some controversy, which is that people, particularly some magicians … You’ve got folks like Maskelyne or Robert-Houdin, who were presenting a more skeptical take on this, where they said, “Well, you've got all these accounts of people experiencing these seeming miraculous things, but you can also get those accounts from someone watching a stage magic show. So how do we know that when people are presenting these things in a sort of alternative religious context, that they're not using the same methods that magicians use?”
Matt Tompkins [:And at the same time, you've got people saying that “We have all of these things from honest, healthy, sober observers, even if there's some people that are lying or using active deception, like, can you really rule out this incredible body of evidence?”
Matt Tompkins [:And they tried a couple of things to sort of crack this. This was something called the Society for Psychical Research, which was founded in the UK to sort of investigate these ideas from a scientific perspective. And one of the things they tried to do first was they tried to get magicians to attend the séances for themselves so they could watch in a way that is going to be different from a layperson would watch. Actually, they got Professor Hoffman of Hoffman's “Modern Magic.” He was the first guy who was kind of working with them to try and infiltrate the séances.
Matt Tompkins [:But an issue with bringing magicians in is when the mediums knew that a magician was there, they would often get the blank séance where nothing would happen. And so this created a little bit of an impasse. So from people on the less skeptical side, they would say, “Well, actually this is proof” – like Uri Geller on Johnny Carson-style – “this is proof because if it were a trick, they would be able to replicate it and reproduce it on-demand. And by the fact that they can't just do it all the time, if anything, this proves even more the legitimacy of these powers that we can't really understand yet.” And then obviously on the skeptical side, they say, “Well, they knew someone was going to catch them, so they just didn't do it.”
Matt Tompkins [:But it created this kind of problem. Anyway, what these guys did – I love it because it's very similar to what I do now – well, the solution was they created their own fake séances. It was a guy named Richard Hodgson and S.J. Davey. Davey was the magician and they would invite people over to different houses and Davey would present as a medium. So again, they wouldn't tell people it was a magic show. They wouldn't say it was definitely going to be a religious experiment. They say, “Come and see.” Well, it was a magic show – they just didn't tell them immediately. So a lot of the greatest hits of fraudulent mediumship from the 1880s, particularly around slate writing, so it was a scripted performance, relatively well choreographed, and they didn't charge people.
Matt Tompkins [:But instead of charging people, kind of the cost of admission was everybody who saw it had to write a letter detailing everything that they could remember happening in as much detail as possible. And so then they had their scripted, choreographed performance and they had then multiple eyewitness accounts of what people remembered from what was again, effectively a show. And they could do a little compare and contrast.
Matt Tompkins [:So you get really interesting results from this. So the article is a compilation of all of the letters along with commentary from the researchers. What's really interesting is you get all of these things that would not be really accepted into psychology for sometimes more than 100 years. So you get gorilla-style results, right? Like where people would fail to detect pretty significant elements of the method and they would either not notice it entirely or fail to remember it. A little bit tricky to tell with the late stage interview. You also got situations where – we haven't talked about this so much – but another corollary of this is a reconstructive memory. The way we can create false memories of events. Again this is something that under different circumstances can happen all the time. There are ways of inducing it. Quite a lot of magic is scripted in such a way as to maximize the odds of this happening.
Matt Tompkins [:So you'd have these witness statements that really deviated from the reality. So not only where you get multiple people saying the same thing and telling different stories about it, but it was also different from what they knew was the actual event, because they'd choreographed it. Including the fact that people would remember things that couldn't possibly happen. And performers probably have had this experience at least a couple of times where someone describes a trick that you did, and it's not a trick that you can do, or not actually what happened.
Matt Tompkins [:And again they demonstrated this really vividly, in a quite controlled way for the time. And it was revolutionary and relatively ironically underappreciated. So psychologists didn't notice. Nobody really remembered this study but at the time it was really intense. You've got these great fights. Alfred Russel Wallace, one of the co-authors of the “Theory of Evolution [by Natural Selection]” was particularly angry at this because he was a brilliant guy, but also a pretty devoted spiritualist. He thought that just like evolution was really controversial, was ultimately correct, he thought the same thing about spiritualism. He thought that that with time the scientific community would come to embrace it in the same way they would come to embrace evolution. Well, he was half right.
Matt Tompkins [:But anyway, His take was that there is no way, and this is the metacognition again that any kind of healthy sober observer who's being honest would make the level of consistent errors that these guys were documenting. So his solution was that Davey actually did have supernatural powers and that the mal-observation report was actually like a cover-up, to try and prevent people from knowing the truth about the one true religion. It's a delightful letter, but yeah, so that's my personal favorite thing on the shelf at the moment.
Adrian Tennant [:Matt, if listeners would like to learn more about your work, your research, or the Science of Magic Association, what's the best way to do so?
Matt Tompkins [:My personal website has contact details on it and we also have a website for SOMA which is the science of magic assoc – just ends in a ‘c’ – abbreviated, dot org. There you can find some descriptions and some videos of past events that we've had recordings. You can also sign up for our newsletter if you want to be kept up to date. We're going to be having some virtual events in the next year that people can tune into from anywhere. And I can't say too much about this because it hasn't been formalized yet, but the plan is next year sometime to have our next big international conference, will be held in Paris. So we try and bounce back-and-forth between Europe and the US, so the next one will be a European-side one and the current top pick is Paris. But if you sign up to the newsletter, you can get updates on exactly when that becomes finalized.
Adrian Tennant [:Perfect. Matt, thank you so much for being my guest on The Magic Book Podcast!
Matt Tompkins [:Pleasure to be here. Thanks so much for having me.
Adrian Tennant [:You've been listening to The Magic Book Podcast. In this episode, Dr. Matt Tompkins shared his fascinating journey from performer to researcher, exploring how magic techniques can illuminate fundamental questions about human cognition and our relationship with technology. We learned about his groundbreaking work simulating AI and neurotechnology through mentalism, his role in founding the Science of Magic Association, and how the field of magic psychology research continues to evolve. You'll find the transcript accompanying this episode on the website at TheMagicBookPodcast.com plus a blog post with a summary, timestamps, and links to the books Matt mentioned. If you have a question or would like to suggest a topic for a future episode, please contact me: adrian@themagicbookpodcast.com. Thanks for listening to The Magic Book Podcast. I've been your host, Adrian Tennant. Until next time, goodbye.
