The Power of Prediction

Josh Landy
Is the brain a prediction machine?

Ray Briggs
What if our senses are just making things up as they go along?

Josh Landy
Can we predict our way to a happier life?

Ray Briggs
Welcome to Philosophy Talk, the program that questions everything…

Josh Landy
…except your intelligence. I’m Josh Landy.

Ray Briggs
And I’m Ray Briggs. We’re coming to you from the studios of KALW San Francisco Bay Area.

Josh Landy
Continuing conversations that began at Philosophers Corner on the Stanford campus, where Ray teaches philosophy and I direct the philosophy and literature initiative.

Ray Briggs
Today, we’re thinking about the power of prediction.

Josh Landy
You know Ray, our minds are amazing prediction machines, and sometimes they can even make their predictions come true.

Ray Briggs
What are you Josh, some kind of believer in “The Secret?” If you believe something hard enough, ooh, you’re gonna manifest it? It’s just going to drop down out of the heavens and into your hands?

Josh Landy
Oh God no, I’m just talking about the kind of thinking we do all the time. Nothing mystical, nothing magical.

Ray Briggs
If it isn’t magical, how are you going to make your predictions come true?

Josh Landy
Well, I imagine you are a little thirsty. And there’s a nice, you know, glass of lemonade in front of you. Imagine reaching your hand out and grabbing the glass: you predict the way it’s gonna feel when you bring it to your face and take a sip. Then you do it. You make your prediction come true.

Ray Briggs
Yeah, but it didn’t come true because I predicted it. It came true because I did it.

Josh Landy
Fair enough, okay. But let’s take a different example: stereotype threat. So women, for example, often get told they’re no good at mathematics. And unfortunately, some of them internalize that prejudice. And guess what, as a result, they do worse at math. So if you predict you’re going to do poorly, you do worse. If you predict you’re going to do well, you do better.

Ray Briggs
Yeah, okay, that is pretty worrying. It kind of sounds like prediction is a dangerous thing. So maybe we should just ditch all of our faulty predictions and stick with the data.

Josh Landy
But those data are affected by our predictions. Like if you did a study of us mathematicians in 1980, you’d find there were very few women in the field. And that was because people thought women couldn’t be mathematicians. In other words, bad predictions. So the predictions skewed the data.

Ray Briggs
Okay, but there are other ways of gathering more data. Let’s go back to your example of mathematicians in the 1980s. Over in Eastern Europe, there were just tons of female mathematicians, and they were doing good work. If people only bothered to look at all of the evidence, they’d have to adjust their expectations.

Josh Landy
Maybe. But you know, even the evidence of your senses isn’t entirely immune to the power of prediction.

Ray Briggs
Oh come on, surely I can trust what I see with my own eyes. “Seeing is believing,” right?

Josh Landy
No, seeing is predicting. And so is hearing—listen to this clip of somebody saying “green needle.”

Ray Briggs
Okay, I heard the person saying “green needle,” so what?

Josh Landy
Okay, now listen to the same person saying “brainstorm.”

Ray Briggs
Yes, I heard them saying “brainstorm” too—what is your point?

Josh Landy
Well, my point is those were exactly the same clip. But you heard differently each time because you were expecting different things. That shows that whenever you sense something, you’re really just predictng.

Ray Briggs
Wait a minute. Sometimes, sure ,I make mistakes because of my predictions. But those are weird cases. I mean, when’s the last time somebody said “green needle” or “brainstorm” to you in a distorted voice? In a situation where there’s literally no context to guide you?

Josh Landy
Yeah, that’s true. There’s usually more context—but context is exactly what we use to make predictions. Everything’s prediction, Ray!

Ray Briggs
Now that just cannot be right. Predictions are based on observations, and you’re telling me that all of our observations are unreliable. You just have predictions based on predictions based on predictions. How are you ever going to be able to even find your glass of lemonade? You’re going to die of thirst all thanks to your terrible theory.

Josh Landy
Well, I predict that our guest will change your mind. It’s Andy Clark, Professor of cognitive philosophy at the University of Sussex.

Ray Briggs
But first, we sent our Roving Philosophical Reporter, Sarah Lai Stirland, to talk to neuroscientists who have been using the predictive brain in their research on some long standing problems. She files this report.

Sarah Lai Stirland
Every day, we go about our lives, assuming that we perceive a shared reality. But can we really trust that we’re all perceiving our environments in exactly the same way? Take a listen to this. For my part, I heard “virus,” but after a few repetitions, I started hearing “Boris.” yYu might have heard something different. The sound is an audio illusion: no words at all, just a bunch of random sounds. Individual interpretations of the sounds is an example of what neuroscientists call the “predictive mind” in action. It’s the mind’s best guess at what’s happening.

Corianne Rogalsky
The main idea with those auditory illusions is that the sensory information coming into our brains is perceived by us based on what we know about the world.

Sarah Lai Stirland
That’s Corianne Rogalsky, a neuroscientist at Arizona State University. And it makes me think that maybe I heard “virus” and then “Boris” because of what’s been happening in the past three years with the pandemic, and a former British Prime Minister.

Boris Johnson
I say to all the doubters: Dyude!

Corianne Rogalsky
We try and form words out of sounds that may not even physically, energy-wise have those sounds in them. We do that because we are making predictions that auditory stimuli coming into our ears should be grouped together in a way that’s meaningful for us.

Sarah Lai Stirland
Rogalsky studies how our brains perceive language and music. She says two things happen when we experience the world. First, we get the sensation, which is a change in physical energy, such as the sound waves coming into our ears.

Corianne Rogalsky
And then we convert that into neural energy—our neurons in our brain firing. However, after sensation comes perception, but that perception is highly dependent on previous experience, and rules that our brain has built about the way the world works and our conscious and unconscious expectations about how the world is organized and how it works.

Sarah Lai Stirland
This is how our brains perceive our environments with all of our five senses. The predictive brain has launched mind-bending new research. Neuroscientists are exploring everything from how we perceive physical touch, to how we form our beliefs, to how we can help stroke victims to communicate fluently again.

Corianne Rogalsky
To be able to provide them with additional cues that typically aren’t necessary but very explicit cues to indicate that a phrase has ended or it’s their turn to speak, or there’s going to be an important piece of information coming up. So to sort of provide a crutch, to help them with that prediction—do the predicting for them.

Sarah Lai Stirland
We start relying on our predictive brains from the moment we’re born. Our minds must decide what information, like which syllables and sounds, are likely the most important to pay attention to in order to learn.

Celeste Kidd
So for example, if you’re a baby and you’re learning a language, you start out by keeping track of the sounds in your language, you build up over time you track those statistics, and you eventually learn what the rules are for what syllables are.

Sarah Lai Stirland
Celeste Kidd is a professor of psychology at UC Berkeley.

Celeste Kidd
And you just keep building up like that with bigger and bigger components until you are able to speak a language and communicate with others.

Sarah Lai Stirland
Her lab researches learning, language, and belief formation. Audio illusions, like the one you heard at the beginning of the story, illustrate how we can block out irrelevant information if our predictive minds decide that it’s probably not important. Over the years Kidd has realized how our own brain’s filtering mechanism has landed her in the dangerous situation of not noticing pain.

Celeste Kidd
I gave a talk once and then after I was talking with somebody and they’re like, “Wm, you have blood from your ear—” My eardrum ruptured, which can happen apparently, if you have an ear infection, and I didn’t even notice—it was a really good conversation.

Sarah Lai Stirland
One of the most intriguing areas of research that almost anyone can relate to is flavor. It’s well established that flavor results from a combination of our perceptions of smell and taste. But researchers who study smell think that it could be possible to manipulate our senses As to get us to eat more vegetables.

Jess Kanwahl
There is this idea out there can we make broccoli tastes like chocolate or have the flavor of chocolate.

Sarah Lai Stirland
Jess Kanwahl, a postdoc at the California Institute of Technology, has some high hopes for the emerging field of neurogastronomy.

Jess Kanwahl
If we were able to make foods that would normally be unattractive, bitter, but still that are healthy and important for us to eat. If we could alter their flavor so that they do become attractive and exciting to eat

Defending Your Life
Do you like broccoli? Yes,! Do you like it with alot of cheese?

Sarah Lai Stirland
Sadly, this idea involves so many areas of scientific study that nobody has come up with a “sensory hack” yet. And it just goes to show—for now at least—that there are in fact limits to the truism, “mind over matter. Finish your vegetables. He doesn’t like broccoli. For Philosophy Talk. I’m Sarah Lai Stirland.

Josh Landy
Thanks for that great report, Sarah. I’m Josh Landy. With me is my Stanford colleague Ray Briggs, and today we’re thinking about the power of prediction.

Ray Briggs
We’re joined now by Andy Clark. He’s professor of cognitive philosophy at the University of Sussex and author most recently of “The Experience Machine: How Our Minds Predict and Shape Reality.” Andy, welcome to Philosophy Talk.

Andy Clark
It’s great to be here. Thanks for having me.

Josh Landy
So Andy, you’ve been thinking about the power of prediction for some time now. Do you have a story about a prediction that went spectacularly right or horribly wrong?

Andy Clark
Actually, most of my predictions seem to be pretty mundane, mediocre things, but they do make a difference to my experience. So for example, my partner recently started using a little chirpy bird alarm sound in the morning to wake up to this is a pretty subtle sound. And I now find myself often hallucinating that sound in the morning when the alarm isn’t even going off, just because I’ve started to predict it.

Ray Briggs
Oh yeah, I have similar experiences with my own alarm. So Andy, you’ve argued that cases like this show that the brain is a prediction machine? What does that mean?

Andy Clark
Yeah, I think the basic idea here is to flip the traditional picture on its head. So instead of the brain sort of sitting waiting there for stuff to come in, and then processing it step by step, it’s trying to get ahead of the game, by predicting what is most likely to be coming in, and then only using the errors in its own predictions to drive further processing. This turns out to be a very elegant way of getting perception to work.

Ray Briggs
So when I think about my perceptions, it sort of doesn’t feel like that it feels like I opened my eyes and information streams in. Is that not what’s happening?

Andy Clark
That is what’s happening? Exactly, you open your eyes, and let me put it this way raw energies hit your various sensory surfaces, what do you do with those raw energies, the old picture would have been, you process them step by step further, in words, to generate a richer and richer model of the world. On this picture, you try to predict that impingement. And then the errors in that prediction, move further and further inward. So sensory information is kind of replaced by prediction error. But still, it’s real stuff coming in from the world. And it’s, it’s helping you stay in touch.

Josh Landy
And so we talked about, you mentioned your hallucination of an alarm. And we talked about some of these auditory illusions. Are there other pieces of data that are counters evidence for this view that really, what’s going on is that we sort of predicting first, and then slowly correcting our prediction when when errors are flagged?

Andy Clark
Well, there’s an awful lot of evidence of various kinds for the idea that the brain is very much in the business of prediction. Part of that evidence is just the wiring itself. There’s a huge amount of wiring is carrying information from deep in the brain towards the sensory peripheries. So there’s a kind of puzzle in there, you know, what’s all that wiring doing? There’s also a puzzle about just the brain’s own ongoing activity. So hugely expensive piece of metabolic kit that is worrying and grinding and buzzing. And it looks like what that all that expense is mostly about is maintaining the model that is being used to make the predictions, moment by moment, of course, then you can go to all the existing psych experiments involving prime in and admission related responses and so on. And they all now fall neatly into place.

Ray Briggs
Would it be less expensive if I did less predicting?

Andy Clark
Less expensive if you did less predicting is a nice thought. I think the answer is no. But there is a sort of trade off in expense, you’re trading the expense of a nice model to make predictions against the expense of trying to do more and more with the raw sensory input. But I think it’s fairly clear that we can’t really do that much with raw sensory input. You know what? What is it in those impinging energies that specifies cat, you need to add a lot of experience with cats before that works.

Josh Landy
You’re listening to Philosophy Talk. Today we’re thinking about the power of prediction with Andy Clark from the University of Sussex.

Ray Briggs
Have you ever made a self fulfilling prophecy? Did it make you feel good about your ability to control your destiny? Or did it freak you out to know how much we’re controlled by our weird brains?

Josh Landy
Seeing the future and seeing what’s right in front of you—along with your comments and questions when Philosophy Talk continues.

The Sparks
And this song will fade out, and this song will fade out, and this song will fade out—I predict.

Josh Landy
Could the predictive brain help us know when a song will really fade out? I’m Josh Landy. And this is Philosophy Talk the program that questions everything…

Ray Briggs
…except your intelligence. I’m Ray Briggs, and we’re thinking about the power of prediction with Andy Clark, author of “The Experience Machine.”

Josh Landy
We predict that you’ve got questions about today’s topic. So email us at comments@philosophytalk.org, or comments on our website. And while you’re there, you can also become a subscriber and dive into our library more than 500 episodes.

Ray Briggs
So Andy, earlier, you were telling us that the mind is constantly making predictions and you gave us a bunch of examples. But exactly how much of human experience does this theory about prediction actually explain?

Andy Clark
I want to go the whole hog here and say everything, everything about human experience is informed by the brains activity as a prediction machine. So everything we see touch, taste, feel, imagine, fear, all of that even action, even action. So action on these stories, turns out to be a kind of self fulfilling prophecy. It’s the very idea of the action that brings the action about, and this works, because it’s a sort of efficient way of doing motor control.

Ray Briggs
Okay, so So how does this work? Walk me through this. Josh wants a delicious glass of ice cold lemonade, and it’s sitting there on the table in front of him. So what does he need to predict in order to like, reach out and grab the lemonade.

Andy Clark
So what has happened there is over time, the brain has learned a model of the relations between movement commands and perceptions. And what it predicts is the trajectory of sensory motor perceptions that it would arise if that action was being performed. Because the action is not being performed, there are prediction errors that are then generated. And those errors actually act like little motor commands, slowly bringing the body into line. So you predict the shape of the reach. And then you get rid of the errors. Because the reach wasn’t actually happening by making it happen. Hence the self fulfilling prophecy tagline.

Ray Briggs
So okay, so in order to like reach his handout, and have it describe the right trajectory, he needs to like model what that trajectory is supposed to be, and then predict it. And then if he gets off course, he can adjust. But like, what about his decision to reach for the glass of lemonade in the first place? Rather than say, like the glass of water that’s sitting next to it? Like, is that informed by prediction to?

Andy Clark
I want to say yes, but you’re, of course, pushing towards a metaphysical boundaries of our understanding here, I think, because the only thing you can really say about that is that moment by moment, predictions arise because of all of our past experience, and wherever our brains started from. And when you throw that into the pot, you find that you’ve got a system that is making predictions, some of those predictions are driving actions. Some of them are driving perceptions. But if you ask where they came from, there’s not much more to say, than Well, we’re a self organizing system that started somewhere was hit with a lot of stuff, and this is what we became.

Josh Landy
Yeah, that seems fair. I am very curious about the places where you go all in, you know, we ran I’ve read your book, it’s super interesting. And, and there’s, there’s tons of it that I’m very excited by and persuaded by but the action things as you say, you’re kind of sort of gambling on a rather strong hypothesis that we’re, you put it as we are pulled along by our our highly predicted future states and, you know, you might think that you’re in control. You know, you want to become a better surfer, for example, it’s an example you give and, and so you seek out the actions that will help you achieve that, but in reality, you’re being drawn along by your own predictions. And the worry I have there is, well, you know, what if I’m in the presence of a potential danger, so for example, there’s something hot, you know, like, the stove is on, right? And, and I and my brain predicts that if I touch the stove, it will hurt. When that situation, I’m not pulled by my prediction to touch the stove. On the contrary, my prediction of future harm helps me avoid touching the stove. So isn’t something like volition, willpower decision? wanting something? Isn’t that a really big part of this picture? That’s not it’s totally separate from prediction.

Andy Clark
Yes, I can see the challenge there. I mean, the right thing to say is what you’ve already said, the right thing to say is that you constantly predict not damaging yourself not harming yourself in those ways. And that’s one of the forces that gets thrown into the pot, it also gets balanced against the local prediction that this is the trajectory that you currently have, and it is going to hit the hot stove, or whatever it is. And of course, we vary these balances, or our brains vary these balances all the time. There’s a sort of quagmire here, that would become a discussion of freewill, where I think we would probably have to, we’d have to use a different set of conceptual tools to approach it. But I don’t think predictive coding or this sort of model of the predictive brain is any worse off than any other picture that I’ve heard. So if you have the traditional picture that says, you know, you’ve got your separate beliefs and desires, and they interact, and then you ask, Well, why have I got this desire? Now? Why is this desire winning out over this one, you’ve got all the same problems. So I don’t think it’s a special vulnerability of the predictive mind story. But it’s certainly a tricky topic that needs very careful handling.

Ray Briggs
So I want to pick apart something that happened in the in that like exchange there. So Josh said, When I reached toward the hot stove, I predict that if I touch it, it will burn me. And then Andy said, Well, I predict that I won’t burn myself. And those, those both seem like they’re compatible. But I’m also a little bit confused about like, what why Josh’s thing is a prediction in the first place. So like, I picture myself reaching out to touch the hot stove, I picture getting burned. I’m like, Okay, that would be bad. So it won’t happen. Like, is it a prediction thing, something I think is going to happen, and I don’t think that the burning myself is going to happen. Like isn’t that an imagining? That’s not a prediction.

Andy Clark
So a lot of the predictions that the brain is trading in are kind of counterfactual predictions, their predictions about what would or would not happen if such and such. And that, and that’s how these accounts merge into accounts of planning, for example. So predictive brain with temporal horizons, does an awful lot of this kind of counterfactual prediction, which can sound sort of slightly odd when it’s just translated into the way we ordinarily use these terms. I do think that, that you’re right to lean on that a bit.

Ray Briggs
Yeah. So I’m really curious about like, what a counterfactual prediction is so like, I want my predictions to be like, correct, rather than incorrect as a counterfactual prediction. Correct? If like, if I were to do the thing, the result I am predicting would have happened, even if like, I don’t do it, and the result doesn’t happen. Is that Is that how it works?

Andy Clark
Exactly. Yeah. So you know, a good counterfactual prediction is one that correctly anticipates the causal structure of the world if you like, so that you know that if I poked and prodded it in this way, this is the sensory impressions I get, if I probably did it this other way. This is the set of sensory impressions that I get.

Josh Landy
You’re listening to Philosophy Talk today, we’re thinking about the power of prediction with Andy Clark from the University of Sussex. And Andy, you know, there’s a question from a listener from John, via email, which I think connects up to what we were just talking about. John says, I say we have a selective brain. I agree that it does predict, but I think select is a better way to describe it, as our senses perceive the raw information, the brain selects and interpret according to what it’s used to. So what about that, you know, reframing, should we should we really say we’re predicting or is selecting better?

Andy Clark
Yeah, I don’t think that prediction and selection are exactly exclusive options here. It’s just that selection under these accounts needs to be reconceived as something like a competition between different predictions. So you have many different predictions operating at many different levels of processing, and they compete with each other in order to gain control of action and that select

Josh Landy
Okay, so I have a really flat-footed question for you, which for which I’m going to apologize in advance but if I understand the picture, correct clay, you start out with a prediction. So for example, you predict that you’re going to hear the chirping alarm sound. And then an error correction mechanism kicks in if in fact, there are errors, and your brain realizes, Oh, actually, there isn’t an alarm, there’s no sound, I’m just imagining it or something like that. But if the brain is capable of error correction, doesn’t that mean it’s capable of getting it right about the world? And if that’s true, why did you need the prediction in the first place? If your aim was to detect what’s out there? Why not just go straight for whatever the whatever is in the circuitry of the error correction mechanism, and use that to tell you what’s going on.

Andy Clark
So I think the thing to notice there about the chirping bird alarm case is that if my prediction of the chirping bird alarm was strong enough, then it would actually overwhelm the fate sensory information. And I wouldn’t correct, it would just be the experience of hearing the chirping bird alarm, it would, in that case, be a hallucination. And it’s interesting that the ability to do that kind of thing is so important to so many of the things that we do, because we all the time have to deal with noisy and degraded data. And that means allowing the prediction to do more of the work. And so I think that that’s why you can’t just say, let’s have a brain that is always given maximum weight to the incoming sensory information, you have to go to vary that weight.

Ray Briggs
So we have a question from Lidia, who’s in Sofia, Bulgaria. And Lydia asks us to connect the question of prediction to the Forgotten meaning of wisdom. So she quotes Plato’s the Titas, which says that man is wise who causes a change and makes good things appear and to be to Him? So Lydia says prediction is not a hallucination, but the work of creating a new reality. What do you think about that, Andy?

Andy Clark
I think there’s something very interesting in the work of creating new reality is what we’re doing all the time. And it’s important, I think, not to be overwhelmed, at least by the notion of hallucination there. Because that can be a very misleading notion in these contexts. And that’s what I get from that question. Very often, I don’t like the tagline that perception is controlled hallucination. I rather prefer the idea that hallucination is uncontrolled perception. And I think that kind of puts a boot on the right foot as it were. So normally, we’re in good contact with the world. And it’s only when these things go wrong. But we suffer from hallucination.

Ray Briggs
Yeah. So I think this brings me to kind of a big question I have about the predictive brain, which is, how do I get it to stay in touch with reality? So if I make predictions, and then my perceptions are partly based on those predictions, and I feed those perceptions back into new predictions that I use for my new perceptions? Like what’s to prevent me from kind of going off the rails?

Andy Clark
Yeah, well, the sad thing is, of course, we humans do very often go off the rails, we can enter into delusional states that feed themselves by helping generate the perceptual experiences that the higher level delusions predict. And that then seems like evidence for the higher level delusions, and we get locked into these vicious cycles. They’re also these vicious cycles of low self expectation that Josh was talking about at the beginning as well. If the question there really is what can we do to try to sort of avoid those cases, I’m afraid that it’s mostly down to what your brain is doing. And so there’s not very much that we as individuals can do that. The good news is that most brains most of the time, are rather good at striking a balance between predictions and the incoming sensory information.

Josh Landy
I love that’s one of the things I love about this position the most is this idea of a balance, right? That it’s it’s not all prediction, and it’s not all error correction. But it’s a balance, and each has its own contribution to make. So paying attention to the details to sensation that’s really important because it prevents us from getting locked in, in prejudices and stereotypes and fantasies, hallucinations, but also, as you were saying a moment ago, don’t discount the power of prediction. You know, it’s a little bit hard to make out what’s what you’re hearing on the radio. Well, your predictive brain can can help you figure out what’s in there. So I’m sort of curious, you know, what are the gains and losses from each balance? Can we can we tweak the ballots? I think, you know, there’s that lovely boreholes story foodista memorias about a character who’s cursed with being unable to forget things he notices everything remembers everything. And he can’t see the forest for the trees. Right? So it’s clear you have too much just raw sensation would be overwhelming. So what’s Yeah, how do we strike the right balance? Can we vary the balance?

Andy Clark
Yeah, we can certainly vary the balance. The right balance, of course, is just a balance that is useful for doing the kinds of things that you’re currently trying to do. And I think it’s going to vary moment by moment and case by case, and probably person by person, because different human individuals probably have different native tendencies to balance incoming sensory information against prediction. So you know, some of the accounts of autism spectrum condition, invoke a difference in that balance whereby sensory information is enhanced, and enhanced sensory information can sometimes be an advantage, and sometimes a disadvantage. So it’s all about the situation you’re in, and the task that you’re trying to perform. Even something that cases where there’s no degraded information, like here in what I’m saying, Now, you’re hearing gaps between my words, but those gaps are being inserted by your brain where it predicts the word boundaries to be. So that’s not really about degraded information, but it’s a useful tweak.

Ray Briggs
We have another kind of skeptical email from Mark in Aptos, California. So Mark asks, What if much of human intelligence is also just a sophisticated kind of autocomplete, like GPT-4? What impact would this have on some of the more renowned philosophical stances on human intelligence, consciousness, sentience sapiens and thinking generally? So little question?

Andy Clark
Yeah, that’s a lovely question. The notion of an autocomplete here might be the right place to begin, because an autocomplete is a very sort of low level way of proceeding, it’s just taking the most likely next successor element from a very simple kind of array. Now suppose that what you’re doing is a kind of intelligent, autocomplete operation, where you’ve got a rich world model that is trying to predict level by level all the way down to something basic you’ve been given, maybe it’s a word or a letter or something, then you’ve got something that looks much more like understanding or intelligence. And so I think it’s not, it’s not right to think that this is a sort of auto complete model of human cognition. But it is right to think that it’s a multi level predictive model.

Josh Landy
You’re listening to Philosophy Talk. Today, we’re thinking about the predictive brain with Andy Clark, author of “The Experience Machine.”

Ray Briggs
How do we harness the power of prediction to our advantage? Could we use it to correct mistakes instead of just generating them? Can we predict our way to a better life?

Josh Landy
The future of predicting the future—plus commentary from Ian Shoales the Sixty-Second Philosopher, when Philosophy Talk continues.

Kate Bush
I just know that something good is gonna happen, I don’t know when. But just saying it could even make it happen.

Josh Landy
If the brain is a prediction machine, saying something make it happen. I’m Josh Landy. And this is Philosophy Talk, the program that questions everything…

Ray Briggs
…except your intelligence. I’m Ray Briggs. Our guest is Andy Clark from the University of Sussex, and we’re thinking about the power of prediction.

Josh Landy
And we have an email from Tom. Tom says, perhaps the predictive brain plays a role in the placebo effect and alternative medicine. What do you think, Andy?

Andy Clark
Yes, I think it really does. I think that the placebo effect is one of the effects, which draws attention somehow to something that’s really important about the predictive brain, which is the expectations that engage relief in the placebo effect, or multi level expectations, many of them are unconscious. And so you can even give people what they called on this placebos, you tell them that this is a placebo, it’s an inactive substance, they believe that they understand it. But if it’s presented in the right packaging, by people wearing the right kind of clothes, very often, you’ll get the placebo effect there too. And I think that’s drawing attention to the fact that our conscious predictions and expectations are just the tip of the iceberg here.

Ray Briggs
So that’s how it’s kind of convenient, like the placebo effect is a way to predict something good. And then like actually make it happen because your predictions that you’re going to get better from your illness, because you’ve taken this pill or whatever. Is that how we should think of the predictive brain is something that could be advantageous?

Andy Clark
I think definitely there are many, many ways to leverage the predictive brain model to our advantage. I’m even in the ballpark of placebo effects. So very interesting thing is that you can train much more specific reactions. So there is an anti Parkinson’s drug called apomorphine. And if you give people a few doses of that, and then a placebo, they can get the apomorphine type response. But if they haven’t been trained on the real drug, their body can’t do it. So you’re sort of training the body to be able to make rather more specific predictions that can then bring about the right effects.

Josh Landy
And that’s extraordinary. So the body is actually producing the chemicals itself. It’s synthesizing the chemicals that were being given to it by the drug.

Andy Clark
Exactly in that particular case. I mean, there’s no claim here that that could be the case for every possible drug, but for some reason, for that one, that seems to work.

Ray Briggs
So I’m also thinking about, like, ways that people use environmental cues to sort of remind them how to regulate their bodies, even something like studying for an exam and same kind of environment that you’re gonna take the examine. Seems like it might be explained by the predictive brain. Do you think that’s right?

Andy Clark
Yes, I think that is right. And that has a sort of good side and a bad side. Because a lot of our experiences with things like chronic pain can come about as a result of starting to make predictions that in this sort of circumstance, you’re going to feel worse or be unable to perform well. And some of those predictions can become self fulfilling. So likewise, in the good sense, predictions about your performance can become self fulfilling, you think I’m wearing the right clothes, I don’t know I’m in the right room, I’m in the right situation. One of the lessons of these models, I think, is that we have to be very careful about what we slowly train ourselves to predict because that’s going to play a big role in bringing about results and structuring our own experience.

Ray Briggs
So another thing that is like kind of puzzling to me is so sometimes you can just make the prediction come true. Like if you predict that you’re going to like experience an alleviation of your headache, or you’re going to experience chronic pain, but like this doesn’t always work. So if I predict that, that taking some random drug will prevent me from getting COVID 19. And there’s no medical evidence to back this up. That sounds like a random example, just a random example. Like I can’t predict my way to whatever state of my body I want to all the time. How do I balance trying to make good things happen with trying to stay in touch with reality?

Andy Clark
Yes, I think a good way to think about that is that we have to make realistic but optimistic predictions. That’s the way to stay in touch with reality. If you make an unrealistic prediction, not only will it not bring about its effect, it can even have exactly the opposite effect, because you find yourself falling so short, so quickly, that things just go from bad to worse. So learning to make optimistic but realistic predictions, I think is what goes on. When you become expert at anything. If you become expert at driving the car, you’re making optimistic but realistic predictions about whether or not you can get through a certain tiny gap in the traffic.

Josh Landy
That sounds like dangerous advice. For me anyway. I love your range of exams, because obviously we’ve talked about placebos. And we’ve talked about getting yourself in the right sort of frame of mind for an exam, for example. And the the body’s potential to synthesize certain chemicals. What about self affirmation? What about you know, the old Stuart Smalley thing looking in the mirror telling yourself you’re good enough? You’re smart enough? And Doggone it people like yours. It does that work, too?

Andy Clark
Yeah, there’s good evidence for self affirmation as long as it’s practiced in this realistic, yet optimistic kind of way. So you know, it’s no good practice in self affirmation that says, I know, I’m the greatest poet of the century. You never even written a poem, but self affirmation that says, you know, I’m perfectly capable of doing these things. People like me do these things successfully, I can do this successfully. There’s very good evidence that systematic applications of training and self affirmation can kind of change the performance of minority groups, for example, in standardized tests. So I think self affirmation is something that we should learn to practice. And again, we need to practice it wisely. So like with all of these hacks, I think there’s a there’s a very delicate balance between the power of prediction to pull you along. And the need to have the predictions be grounded in a model, which is actually a good model, therefore a realistic one.

Ray Briggs
So I want to ask about something This kind of chin tangentially related, which is the value of surprise. So sometimes it’s really good for me to run into information that violates my predictions. And it’s sort of how I learn. If I’m always trying to make predictions with my predictive brain, why should I seek out surprise? Like, is it a good thing for me to seek out surprise? And why?

Andy Clark
Yeah, so that actually leads to one of the, to my mind, at least most interesting dimensions of the predictive brain, which is that predictive brains are automatically information foraging brains, they kind of love to drive the whole organism to get more information, so that it can reduce errors in prediction better in the future. And so one of the reasons I think the that we engage in a lot of the exploratory behavior we do is, because we kind of estimate that that’s a good way of minimizing future prediction error, as they say.

Josh Landy
That’s super interesting. It actually brings me I hope you’ll forgive me, I’m in literature person. So I can’t help asking a question about literature in all of this, because works of sophisticated ambitious, interesting literature very often violate our expectations in different kinds of ways. So for example, the the character you thought was great, turns out to be monstrous, or, you know, that this thing that looked like a comedy is actually a tragedy. And these are very powerful effects. And we very often praise them, I praise works of, of literature for for producing these effects on us. Is that doing some kind of work for our brains, these predictive machines? Or are we getting fine tuned or something like that? Does literature have any role to play here?

Andy Clark
Yeah, it’s an interesting question about the possible functional role of literature. I mean, one, one thing to say at the outset, I think is that works of literature. Or what Karen could Conan is quite nicely called probability designs, they deliberately lead us through building up expectations, and then generating surprises, building up new expectations, just like works of music as well. Or indeed, going on a rollercoaster ride, a roller coaster is a kind of engineered set of building up expectations and surprises. And the funny thing is, even when you know the surprises is coming, you still respond to them. So I think that literature is training us in something, it’s very hard to say exactly what it is, it could be improving the models that we use to generate predictions by kind of showing us what it might have been like to live in a certain country in a certain historical moment. It’s also just giving us training in responding to surprise, and perhaps training in the sorts of different trajectories that our own expectations can take.

Ray Briggs
That’s really cool the value of like training and surprised through literature. I’m also curious about this related question of like, what we make predictions about. So it seems like one way that I can get a surprise is to learn that something I thought wasn’t true. But another way is to learn that there’s like this whole way of categorizing the world that hadn’t occurred to me before. And this is what happens. I think, when I learn a new language, I’m like, Oh, I have just like a new set of concepts. Does the predictive brain hypothesis have anything to say about which categories to predict in?

Andy Clark
I don’t think that there’s anything very systematic there. I mean, it’s clearly the case that sometimes you’re encountering surprises that are quite small, they’re not changing the larger structure of the model that you’re using to make the predictions. It’s just that that’s not quite the way that the model thought that this particular bit of the world would pan out. So you know, I spot something unexpected on my desktop, there’s a different kind of surprise I could get if my desk started behaving in ways I’ve never seen a desk behave before it starts levitating. Or every time I move my hand towards it, it behaves oddly as if in some kind of virtual reality there, I’d have to start to revise the higher level bits of the model that I’m bringing to bear. And I think that’s a very important thing to do. I think we use art and literature and science to challenge our own existing models, so as to generate those kinds of particularly beneficial surprise. But I don’t think that there’s a well worked out account yet that shows how it is that that we’re actually able to deliberately construct these environments that challenge us in these very interesting ways. It’s actually something I’m working on at the moment.

Josh Landy
Andy, we’re almost at the end of our show is there one small piece of advice you’d have for our listeners, you know how to make the rest of their day, a better one harnessing the power of their predictive brain.

Andy Clark
Yeah, have optimistic but realistic expectations about the shape of the rest of your day. Don’t expect to get everything done that you would like to have got done today, but just maybe a little bit more than you might otherwise fall into.

Josh Landy
That’s great advice. And it’s been a great conversation. And I predict we’ll be thinking about it for a good long time. Thank you so much for joining us today.

Andy Clark
Thank you. I really enjoyed the conversation. And thanks to everyone that sent things in too.

Josh Landy
Our guests is being Andy Clark, Professor of cognitive philosophy at the University of Sussex, and author of “The Experience Machine: How Our Minds Predict and Shape Reality.” So Ray, what are you thinking now?

Ray Briggs
Well, I am ready to be optimistic yet realistic. I think I’m a little bit unsure how to figure out what’s realistic and whether I should be optimistic when I’m figuring that out. But once I’ve got that, I’m going to go go forth and apply Andy’s advice.

Josh Landy
It’s brilliant advice. I mean, you think about the power of placebos, even placebos that are labeled placebo, I mean, we really do have more control, or at least our brains, parts of our brains, our unconscious brains have more control than, than we know that we have known. But we’ll put links to everything we mentioned today on our website, philosophytalk.org, where you can also become a subscriber and gain access to our library of more than 500 episodes.

Ray Briggs
And if you have a question that wasn’t addressed in today’s show, we’d love to hear from you. Send it to us at comments@philosophytalk.org, and we may feature it on the blog.

Josh Landy
Now, a man so fast you don’t know where he’ll be next—it’s Ian Shoales the Sixty-Second Philosopher.

Ian Shoales
Ian Shaoles… Used to be, we’d make maps in our head, and then we’d just remember the way to Grandma’s house, or we’d give the ponies their head, and they’d wake us when we got there. We used to love the maps though. The glove box was crammed with them. But carshare drivers don’t read maps. The smartphones do that. Ancient maps used to show where dragons were. Avoid that! Then mathematics made for cartography on steroids. We could calculate by the light of the moon where we were in the world. Thanks to the speed of light, we learned that many stars have been dead for thousands of years. We learned too much. So why leave the house? It was supposed to get easier. Because along the way, we got eyeglasses and motorcars and ships and planes and rocket ships. We got telescopes and microscopes, concrete methods of telling us just how long it would take to never arrive at the end of the universe. If our behavior was predicated on predictions, leaving the house became safer all the time. Predictions became data analysis, giving rise to the insurance business. If there was a travel-related death, your loved ones cashed in. Giving rise to true crime programs and insurance conglomerates. As insurance got bigger, so did the world, at both ends of the spectrum. Planets proliferated, and quanta got smaller. Until you realized if you look at it, it’s not what you’re looking at, and if it is what you’re seeing you don’t know where it is. That’s the Heisenberg Uncertainty Principle. Don’t blame yourself. It’s the Quanta. It’s Chinatown Jake. Just walk away. That’s how the 20th Century began. And the old empires began falling apart, making the maps useless. What map to follow in the new world we’re destroying? And we had brand new concepts to predict the way, marked by Darwin, Freud, Marx, Hitler, Einstein, Henry Ford. Existentialism, Hollywood, Internet, and beyond! All this may indicate that the old hippie shout, “What is reality?” remains unanswered for a reason. Our vision of reality is based on experience, and true crime shows tell us what eyewitness testimony is worth. Something as simple as a car, for instance, entered into the database of our inner self, can give us anything from a Buick to a Tesla to a boulder and log steamroller convertible, with Fred Flintstone driving. Not to mention bumper cars which can trigger undue amounts of nostalgia for mild whiplash and carnivals, which are also dead as distant suns. How does all that help you make that left turn at Albuquerque? We’ve also recently pretty much stopped going out to movies. They were the realest thing on earth for a hundred years, even though they were just a bunch of photographs strung together. Video, DVDs, streaming, we may still be transported, but we’re not going anywhere. AI, 3D environments, fake news. No reason to leave the house, much less cross the street. This year a spy balloon was shot from the sky, and a spy drone knocked down. Proof that the 21st Century has new ways to map the world, and help us make our way, but also new takedowns, making sure that maps will always be controlled by those in power. Perception is a mixed bag. Sometimes you learn things, sometimes things are not meant to be seen. But how do you know it’s forbidden until you do it? It all goes back to Adam and Eve. The forbidden fruit gave knowledge, but not about the world East of Eden. Adam and Eve had to draw their own maps. Which we are still doing. Now we are living in a map as big as the world, and yet we’re lost. We’ve forgotten how to read the signs, and nobody’s left to teach us. We flip coins for directions. We buy bitcoin and try to flip it. We are dug so deep into our world we don’t know what we’re seeing, and we don’t know what is doing the seeing. So, as I understand it, and of course understanding is really anybody’s guess, we don’t actually know what we’re doing or where we’re going if we leave the house, it’s all guess work, playing hunches, placing bets, hoping our legs work the way we remember them. On the other hand, global warming might kill us because we won’t recognize the signs, but we still know a dragon if we see one. If we ever find ourselves in dragon country. I gotta go.

Josh Landy
Philosophy Talk is a presentation of KALW San Francisco Bay area and the trustees of Leland Stanford Junior University, copyright 2023.

Ray Briggs
Our Executive Producer is Ben Trefny. The Senior Producer is Devon Strolovitch. Laura Maguire is our Director of Research

Josh Landy
Thanks also to Jamie Lee, Elizabeth Zhu, Emily Huang, Merle Kessler and Angela Johnston.

Ray Briggs
Support for Philosophy Talk comes from various groups at Stanford University, and from the Partners at our online Community of Thinkers.

Josh Landy
And from the members of KALW. Local Public Radio San Francisco, where our program originates.

Ray Briggs
The views expressed or (mis-expressed) on this program do not necessarily represent the opinions of Stanford University or of our other funders.

Josh Landy
Not even when they’re true and reasonable. The conversation continues on our website, philosophytalk.org, where you can become a subscriber and gain access to our library of more than 500 episodes. I’m Josh Landy.

Ray Briggs
And I’m Ray Briggs. Thank you for listening.

Josh Landy
And thank you for thinking.

Firesign Theatre
We think that is a fair and a wise guy, uh, rule to be guided by. What is reality? And we’re not afraid of it, are we. Eat it! You bet.