https://youtubetranscript.com/?v=feS1zuKz2N8
Okay, when you make progress towards a valued goal, let’s say we inhabit a shared narrative and we’re making progress towards our mutual stated goal. And when we see ourselves making progress, we get a bit of a dopamine hit. Could you say that the fundamental reason for the positively rewarding effect of that movement forward is that as I move forward towards a goal, I decrease the entropy that still remains between me and the goal. And so is even that reward, is even that movement forward, readable as an entropy reduction? I mean, it’s almost written into the mathematical meaning of the word, so if entropy just is uncertainty, and as I get close to resolving that uncertainty, getting my fruit juice, pleasing my wife, or being able to watch the news, if it’s an epistemic reward, it is just expected surprise, just is the uncertainty. And the closer you get, the more, the less uncertain you are, and all the evidence suggests exactly as you say, it’s dopamine. [“Pomp and Circumstance”] Hello, everyone. Thank you for tuning in to Watch and Listen. I have the great privilege today of being able to talk with Dr. Carl Friston as an addition, let’s say, and a signal addition to the recent conversation I had with Andrew Huberman. Dr. Carl Friston is arguably the world’s most renowned neuroscientist, a professor at University College London. He is one of the world’s leading authorities on brain imaging. 90% of the work published in fields employing such imaging relies on methods. He pioneered. Dr. Friston is also well known for his work on many of the topics we will discuss today. Work I find even more exciting, at least conceptually speaking, than his work on brain imaging. We will discuss the ideas that concepts and precepts, categories, that’s another way of thinking about it, bind free energy or entropy. The idea of computation, especially the kind of computation that approximates brain function as hierarchical. The theory of predictive coding and active inference. Welcome, Dr. Friston. It’s very good of you to agree to talk to me on this podcast. I’m really looking forward to it. That’s a great pleasure to be here, thank you. So let me start maybe by helping people understand this idea of hierarchical computation and the binding of entropy. And so if you could walk through that briefly, then I’ll ask some questions if that seems appropriate. Yeah, sure. The binding of free energy and entropy, that sounds delightfully Freudian. And I don’t mean that in a sort of disparaging sense. I think that some of the truisms and the insights of that era have now proved themselves in modern formulations of computation, information processing, sense making in the brain. And one nice link there is to think of free energy as surprise. So one way of looking at the way that we make sense of our world, bringing explanations, concepts, categories, notions to the table that provide the best explanation for the myriad of sensations to which we are exposed is to see that process as a process of minimizing surprise. So binding free energy, I think, can be read very simply as minimizing surprise. But of course, to be surprised, you have to have something you predicted. You have to have a violation of predictions. So immediately, you’re in the game now of predictive processing, predicting, what would I see if the world out there was like this? And then using the ensuing prediction errors to adjust your beliefs and update your beliefs in the service of minimizing those prediction errors or minimizing that surprise or minimizing that free energy. And you artfully introduced the notion of hierarchy in that question, which I think speaks to another fundamental point that in making sense of the world, in making those good predictions, we have to have an internal model, sometimes called a world model, a model that can generate what I would have seen if this was the state of affairs out there. And that notion of a generative model, I think, is quite key and holds the attribute of hierarchy simply in the sense that we live in a deeply structured world, very dynamic world, a world composed largely of creatures like ourselves that has this sort of hierarchical nested structure that has to be part of the models that we bring to the table to understand it. Postmodernism is tearing our world apart. The one thing that may be able to unite us is a mass return to our Judeo-Christian roots. At the individual level, that means developing our prayer life. There’s a ton of literature out there on the benefits of secular New Age mindfulness meditation, but what isn’t talked about nearly enough is the power of a consistent prayer life. That’s where Hallow comes into play. Hallow is the number one Christian prayer app in the US and the number one Catholic app in the world. Hallow features over 5,000 prayers and meditations, including daily prayers to help you build a habit of prayer and gratitude, Bible reflections with Jeff Cavins and Father Mike Schmitz to help you grow in knowledge and understanding, meditations to help you prepare for sleep and rest each day, reflections with Bishop Barron, and much more. Use Hallow as a foundational tool to grow in gratitude and character each and every day. Try it free for three months by going to Hallow.com slash Jordan. That’s Hallow.com slash Jordan. Ground yourself in fundamental truths with Hallow today. Okay, so I encountered these ideas, I would say first, reading a book back in 1982 by a man named Jeffrey Gray, who was influenced by Norbert Weiner and by General Cybernetic Theory, and he regarded the hippocampus at least as part of the central part of the brain involved in contrasting expectation, let’s say, the way you expect the world to lay itself out with the way the world is actually laying itself out, indicating that surprise was a, what was the prodroma to anxiety, and that anxiety in some sense was an indicator of the magnitude between, magnitude of mismatch between expectation and reality. So I wanna ask you a couple of questions about that, because I thought a lot about Gray, and I thought about Gray for like 40 years after reading his book, and also as a consequence of encountering your ideas much later. The cybernetic model of expectation is predicated on the idea that you lay out an expectation on the world, and the expectation is a model, and then you have incoming sensory data, and the sensory data is the world, and that now and then there’s a mismatch between the sensory data and the expectation, and then you have to modify either the representation or the action pattern to reduce that mismatch. But then I thought, well, wait a second, there’s a weird lack in that formulation that’s twofold. Number one, it isn’t obvious to me that it’s expectation. It seems to me that it’s more likely to be desire, and that’s a useful transformation of conceptualization because if you use the word desire rather than the word expectation, you can infer that the models that are being contrasted against the real world are models of motivation rather than cold cognitive computation. And so, for example, if I’m interacting with a woman who I’m romantically interested in, I’m attempting to bring about the realization of my desire, not my expectation, and that’s under the pressure, let’s say, of the demand for sexual reproduction, and so the motivational state grips the desire. The desire manifests itself as a fantasy, and then the motivation is the minimization of the discrepancy between the actual world and the fantasy. And then, so it’s not exactly an expectation because it’s more dynamic and alive. And then one more, I hate to hammer you with all of this right at the beginning, but one more issue on that front is the error. Because you could imagine, and I’m interested in how this matches your hierarchical concept, if I make a mistake such that I don’t minimize the relationship between what’s happening and my fantasy, one of the mistakes I might make is a mistake of perception, not expectation or desire, because I might be seeing the situation wrong. And so it seems to me that it’s more realistic to say that what you’re doing is minimizing the mismatch between a model of the desire and the fantasy. A model of what’s happening, not sensory data itself, but a model of the unfolding present, contrasted with a fantasy, and that the mismatch is what indicates surprise or entropy. Now, I know that’s a lot of questions to answer at the same time, but that whole set of objections makes up a pattern. It does. And despite the fact you introduced a whole bunch of really important concepts there, I think it was nicely introduced in a coherent way. So I’m just trying to remember all the exciting things you brought to the table. So yeah, Geoffrey Gray, a great thinker, in my memory, sort of famed for understanding latent inhibition and sort of how you use these surprise or these mismatches or these prediction errors. And it’s interesting that you speak to his conceptual roots in cybernetics, because there’s a very, very close connection between the things that we’re talking about and the early formulations of cybernetics by people like Ross Ashby. So there’s this good regulator theory, which I think if your listeners don’t know about, I think they’d like, and I think you would like a lot. It’s just the notion. And provable notion that to control an environment, you have to have a model of that environment, to be able to couple with, to engage with the environment. The degrees of freedom of your model have to match those of the kind of environment in which you’re operating. So I think there’s a deep connection there. The other thing, many things you mentioned there, you use the word fantasies a lot, and I think that’s quite important because very often, I sell the brain as a fantastic organ, literally, because it is in the game of generating the right kind of fantasies. And these are the fantasies, these are the motivated expectations that drive our behavior. But in noting that, in noting that, you’ve also made a big move and an important move. I know it’s a move you want to make in terms of our conversation, but just to contextualize why that is such an important move, you’re talking about motivation, how our perception and action are all contextualized by motivation and what we desire and what we want and what the preferred outcomes of our acts. That’s, I think, sort of the crux, and will probably be the focus of much of what we’re going to be talking about. But just to say that vanilla predictive processing and vanilla predictive coding does not deal with that. It just deals with sense-making. So it’s usually in the moment. As soon as you bring decisions and choices and actions to the table, you’re immediately in the game now of making inferences about the future, the future that has yet to be observed. So there are no prediction errors. It’s just your expected surprise, your expected entropy or expected free energy or your expected discrepancies between what you fantasize should happen or what you prefer to happen and what you anticipate given your best sense-making at the moment. So I think that inactive aspect, that choice really does profoundly change the game and takes us from simple sense-making into the world of choice and decision-making and motivation. So mathematically, the way that I would articulate that as a physicist would be, okay, if we’re in the game of minimizing surprise and this mismatch, this discrepancy that you were describing in the moment, then presumably I will now choose the things that I do in order to minimize the expected surprise in the future. And the nice thing about articulating it like that is that the expected surprise mathematically is just entropy. So entropy, uncertainty, and I’m using entropy and uncertainty synonymously here, they are the mathematical statement of the surprise I expect when I haven’t actually seen the outcome yet. So in many senses, what you can say is that we are motivated to resolve our uncertainty. And that can be sort of, if you like, carved into two ways of minimizing uncertainty. It can either be through choosing those behaviors that resolve uncertainty in the sort of folk psychological sense, watching the news, looking over there to see whether my fantasies about the cause of that visual flutter in the periphery of my vision was what I thought it was. Was it a bird or a butterfly? But I think there’s another side, well, I know there’s another side to that, technically an expected information gain or an epistemic value. And that’s, of course, I will avoid putting myself in surprising situations, being very cold physiologically or being unloved or being embarrassed or anything that I would find surprising about myself. I will minimize my expected surprise by avoiding those kinds of things. And I think that’s, you know, my reading of what you were saying and my attention to that. And I think that’s what touches exactly on that. A shocking number of stocks are in the red. You may think you’re doing all the things right, but in this volatile economy, many of our portfolios are just not looking good. Why? According to the largest banks and firms in the world, people aren’t diversifying the right way. The average annual return of a classic 60-40 portfolio is down by nearly 20% this year. Goldman Sachs just released a shocking report declaring the 60-40 doomed for the next decade is the allocation from stocks entirely. Goldman will instead be investing in real, tangible assets, specifically fine art. The last time inflation was this high, some segments of the art market appreciated more than real estate and even gold at an astounding 33% per year on average. That’s why this show has partnered with Masterworks. Masterworks results have been exceptional. Six exits for an average net return of 29% to investors, even during COVID and in a bear market. As a result, Masterworks has had to acquire on their platform to meet demand. But you can skip the wait list by going to masterworks.com and using promo code JBP. That’s masterworks.com, promo code JBP, or click the link in the description. See important disclosures at masterworks.com slash CD. I think people hem them, people hem themselves in, in their actions even. In once people, you see this clinically, is once people develop a conceptualization of themselves, which is a box in some sense to put their variability in, they’ll sometimes simplify themselves in terms of their range of action merely so they don’t surprise themselves. And so they artificially constrain the range of their potential behavior and that can become maladaptive if their conception of themselves is too narrow. That would be a good example of a miss, in some sense of a misperception of the self, right? It’s a narrowing of possibility. And so I should maybe try to explain to people a little bit more about this issue of entropy. And so there’s this notion that’s come out of theory of perception that any set of objects or any object for that matter is different from any other object and the same as any other object in a multitude or even a near infinite number of ways. And so you could think, I always think of two goblets standing on a table and you might say, well, they’re identical, but they’re not. And you can tell they’re not if you wanted to paint them in a photorealistic manner, you’d find that the surface of one goblet is very much unlike the surface of another and you’d have to render that at high resolution in order to represent it accurately on a canvas. And then they differ in weight and they differ in the regularity of the surface and they might differ in age. There’s all sorts of differences between them that you might not think are germane or relevant, but that’s really the issue is why do you regard some similarities as crucial and some differences as irrelevant. And so here’s something I’d like to run by you. So would it be reasonable to say, imagine a hierarchy of perception and conception. So the idea is, well, two goblets are identical if I can drink out of either of them using the same pattern of perception and action. So that’s kind of an indication of the entropy constrained by the notion of wishing to drink. Any two things are substitutable as long as they leave me with the ability to drink as an axiom of my current action set. And that’s something like maybe a computational definition of resemblance. Because I can’t understand how you can bind resemblance otherwise if any two things differ in a near infinite number of ways. They differ in terms of the microstates that they might be composed of. That would be a way of thinking about it from the perspective of entropy theory. So if they’re substitutable within a higher hierarchical conception, then they’re identical perceptually and conceptually. So do you know the change blindness experiments that was Dan Simon managed at Harvard? Yeah, well that’s a good example, right? So I’ll just run through one of those experiments for the people listening. So imagine you put a camera on a two story building and it’s looking down over a university yard. And now imagine that you have tourists walking through the yard and an undergraduate research associate comes up to a tourist with a map and asks the tourist to help him find his way. And while that’s happening, some other experimental compatriots walk between the person who’s asking for directions and the tourist carrying a door, two of them. And the person asking the question grabs the door hidden from the tourist and continues walking with the door. And the person who was carrying the door stands there and continues asking the question with the map. And what Simon found was that the largest percentage of people, the majority of people do not notice that the student has been swapped in the middle of the conversation. And the reason for that has to be that the perception of the student by the tourist is something like generic student asking for directions. And any student who’s asking for directions at that point will fulfill that function or act as a player in that story. And so there’s no reason to be surprised by the transformation, even though, quote, it’s a different person. And that seems to add credence to this notion of a theory of identity or resemblance. Does that seem reasonable to you? It seems very reasonable. So mathematically, I think you would describe that exactly in terms of finding those latent states or categories or structures or resemblance that are conserved over different contexts. And that enables you to have a simple but apt for purpose model of your world. So if you can find the causal structure that is as simple as possible, that provides an adequate explanation of your sensorium, then that is the good model. I mean, literally, as a statistician, that would be the model with the greatest evidence, the accuracy minus the complexity. You also brought something interesting into the conversation again, which was what I would read as affordance. Do you like sort of Gibsonian notions of affordance? So you were saying that this ability to extract the latent structure, the essence, the resemblances that are conserved over different contexts, they have, at very low levels of the hierarchy, these microscopic or detailed differences, but at higher levels of the hierarchy, they are the same kind of thing. They have the same affordance. I sit on this, I drink this, I love that, I don’t like that. What you’ve done though, again, in terms of talking about affordance is bring action to the fore. I can perceive lots of inferences in my world. There are categories, and there are exemplars of particular categories, and I can unpack that hierarchically. And many people think that’s how the visual hierarchy, the cortical hierarchy that subtends or subserves visual perception and visual synthesis is organized. But what you’re saying is this also holds true in the way that I plan to act and the outcomes of those actions. And all of these, if I get those inferences out, then they are the right kinds of affordances that speak to my model of the way I’m gonna behave in this situation. Putting center stage, the self model. What kind of thing am I, and what will I do in this situation? And is my model good enough? Is it too rigid? Should it be more flexible? All these are crucial questions in terms of updating not just your beliefs about what you’re doing and the context in which you’re operating, but the kind of thing you are and whether you’ve got the right simple explanations, whether they should be more complex or simpler. Right, and so the optimal simple explanation is the least complex affordance necessary for the operation at hand. And so then we could talk about affordances a little bit. So people who are watching and listening might think that you see the object and infer the meaning, but the theory of affordances would suggest that you see the meaning and infer the object or something like that, it reverses it. And that what you see first is something like functional utility. But I think you can bring both those positions together if you think that what we see in the world are patterns that have functional utility. Now there’s all sorts of patterns that don’t have functional utility or that only have potential functional utility. And that might be, you might consider patterns that don’t have specified functional utility as something like the infinite domain of potential empirical facts. If a fact is something like the accurate identification of a pattern that exists stably across contexts. But then we can’t see all the set of all possible facts because it’s way too large. And so we wanna reduce that further to that set of patterns that offer us a grip on the world. And Gibson who wrote an ecological approach to visual perception talked a lot about affordances, tools and obstacles in particular, if I remember correctly. But I’m wondering as well, is the domain of affordance, here’s a way of conceptualizing the domain of affordance. It’s tool, those things that you can get a grip on the world with and move forward towards a desired goal. Obstacles, so things that get in your way of different size. And then pathways. And the pathways in some, so I would say the pathways are littered with tools and obstacles. And the description of a pathway that’s littered with tools and obstacles, that’s a story. And so these, it looks to me like those hierarchical, conceptual boxes are best conceptualized as something like stories. If they’re an amalgam of affordance and pattern, then they’re not either just an object or just the subject. They’re the bringing of the two together. And the story is where the pattern and the purpose meet. That’s a way of conceptualizing. That’s a way of thinking about it. And I would think maybe even when you perceive a tool, it seems to me that what you’re perceiving is something like a micro story. You know, like if I perceive this pen, say, well, that’s obviously, it’s an objective fact that this pen exists. And it is insofar as the pen is a stable pattern in time and space, but it’s the function of the pen that makes it a pen and gives the unity that it gives the unity to the perception because you might ask, well, why is the pen, a pen this way and a pen this way and also a pen this way? And speak well, because it’s objective characteristics transform tremendously when it’s moved or when it’s illuminated differently, but its function remains constant. And so I think even an affordance is a micro story. And so a tool is a positive, it’s a comedy. That’s a way of thinking about it. A tool is a comedy and an obstacle is a tragedy. And we lay out the world like that, even at the levels of our fundamental perceptions. Is there any, what do you think of the ideas of a concept or a percept as a story? Because you see, the reason I like that, the reason I think this is so important is that there’s an entire literature on narrative, of course, the entire literature on literature is about narrative. And if it is the case that our fundamental concepts are narrative in their origin and their nature, then that allows us to lay the narrative world on top of the objective world. And I think that’s a, well, it would be lovely if that was a possibility. It would alleviate the terrible tension between them. Then of course, the question comes up with what’s the best narrative, but that’s something we can also talk about in terms of hierarchical processing. Yeah, I think there’s some deep truths there. A number of ways I could sort of paraphrase what you’ve just said. So this notion of micro stories and narratives, I think that is exactly the plans into the future that we were talking about before, that, you know, choosing the right ones. And mathematically, you can think of that exactly as you articulated as a path. One perspective on that is that of Richard Feynman in terms of the pathological formulation of quantum electrodynamics, where he was playing exactly the same game, finding the path of least action, the path with the least obstacles, the path that isn’t the most fluent, the most ego-syntonic, leading to exactly what we talked about before, either the information gain or avoiding surprising outcomes, which I would read as the obstacles, I don’t go there, that’s not the kind of thing that I’m in. So if you read our active engagement with the world, our active inference, just as a process of committing to the right paths, the right plans, the right narratives, the right micro stories, I think that is the essence of, well, one could argue, sort of sentient behavior and existence. So I would put myself very much, or I would certainly find it very easy to commit to that kind of formulation. I think a lot of other people would as well in different fields. So what you’re talking about, you mentioned before about all the facts that are knowable, and yet we only register and recognize those that matter, essential variables, for example. Change blindness is a nice example of that. You don’t see student of a particular identity, personal identity change, it’s just a student. You only see and you only model and rehearse and sample the world using the simple level of explanation that is apt for getting those right paths forward. Yes, well, and it would be lovely if your hierarchical conceptualization consisted of micro stories, let’s say, that were flexible enough so that they would apply across a very wide range of microstate transformations, because then you can use the same simple model in all sorts of different situations. So you might say, well, if you could extract out a universal ethic for dealing with people, then you could apply that ethic to everyone that you met. And then, and that idea might help us triangulate in on what might constitute a universal higher order ethical narrative. And if you imagine that interactions with people are generally constrained by the necessity of iterating the interactions across time, and then you might say, by the additional constraint of iterating the interactions across time so that they increase in their utility. And so then you might say, well, how do you have to treat someone, or everyone for that matter, so that that reality makes itself manifest as you move forward into the future? And I would say, well, that looks like something like genuine altruistic reciprocity. Now, Franz de Waal, I had a chance to talk to Franz de Waal about his work with chimpanzees. You know, when chimpanzee alpha males are often parodied as dominant in a sort of Marxist sense, power-driven. And it’s the most dominant male chimps, so the one with the most physical prowess, the biggest tyrant in some sense, who gets to dominate all the other chimps, and who in consequence has preferential reproductive access. And so it’s a theory of power, and social structure and reproduction all tangled up into one. But the problem with that is it’s not true. So de Waal has shown very, very clearly that, first of all, sometimes the alpha chimps, so to speak, can be the smallest male in the troop. Frequently, he’s allied with a powerful female, and he is generally the most reciprocal individual in the troop, very concerned with the long-term maintenance of social relations, and very good at making peace, not war. Now, that doesn’t mean he doesn’t have, let’s call it power, at his disposal, especially in coalitions, if necessary. But de Waal has shown very clearly that the alpha chimps who rely on power and force are very likely to rule over an unstable polity, and to meet an extremely violent end in the relatively short term. And so if your fantasy about the future, let’s say, is motivated by an underlying motivational state, it could be hunger, it could be thirst, it could be sexual need, it could be rage, it could be the desire to make anxiety decrease. Then you could imagine that there are ways of interacting in the world that satisfy multiple motivational states simultaneously, and then you could imagine that those modes of being satisfy multiple states of motivation simultaneously in a social context, so also for other people, and then you could say, and that make that occur as it iterates forward into the future, and then you could say, well, you want to extract out a representation that allows you all those advantages simultaneously, and it looks to me like something, maybe that’s marked, here’s some hypotheses, it’s marked by the sense of active engagement that you might have in a good conversation. It’s marked by the sense of the emergence of the spirit of play, and Joach Panksepp has detailed out the psychophysiological structures underlying the play circuit. It would underlie something like maximal, no, optimized stress. So you talked about minimizing predictive error, but here’s a variant, what if you optimize predictive error so that you lay out a fantasy on the future and then work so that there’s just enough predictive error so that you encounter something you don’t expect at a micro level, small enough that you can manage it, but large enough so that it expands the confines of your hierarchical presuppositions. And maybe you do that, see, I was thinking about that relationship to play, because if you’re on a team and you’re playing against a well-matched opponent, the opponent pushes you right to the limit of your skill, not past it, right? So it’s not too stressful, but it isn’t exactly in that situation that surprises minimize, it’s more like a little entropy is allowed to enter the system at just enough rate and intensity so that you can push your development in a manner that doesn’t stress you too badly physiologically. Yeah, so you’ve brought in about sort of four really important themes here. The two key things that you’ve brought to the table there were sort of putting sentient creatures together. So you’re talking about social interactions now and social hierarchism and sense-making when the other, the thing I’m making sense about is also trying to make sense of me. I think that’s a really important sort of and challenging sort of move there. You’ve also brought to this sort of highlighted this paradox that we might be in the game or we might be seen as in the game of trying to minimize our surprise, minimize our prediction errors, and yet we seek out novelty. So I think there’s a fundamental paradox there that needs resolving. I think you’ve, in your setting up of the issue, I think you’ve implicitly resolved it. There is, I think, a very simple way of resolving that and it comes back to this sort of isomorphism between expected surprise and uncertainty. And I notice you also use the word angst and anxiety. To my mind, uncertainty just is a state of, or recognized as a state of angst or anxiety. So that sort of imperative to minimize expected surprise just is choosing or can be complied with by choosing those plans that minimize uncertainty. And what would that look like? It would basically look like responding to epistemic affordances that resolve that uncertainty. So I think that’s the kind of surprise that we aspire to. It’s the novelty that affords the opportunity to resolve uncertainty and thereby resolve angst. And if that’s true, then taking it to your context, how would I do that if I was in a social hierarchy of chimpanzees or I was in any social setting? In one sense, the simplest way to resolve my surprise and make the world as predictable as possible would initially be to resolve my uncertainty about you by asking you the right kinds of questions that allows me to sort of put you in a particular category in one of my narratives, my pro-social narratives about the kinds of people that I can talk about. But also, ultimately, I’m gonna try and make you like me or make me like you because the closer we are, if we can share the same narratives and the same language, then together we’re mutually predictable. So that mathematically would be like a generalized synchrony from a social neuroscience perspective on Dalek interactions. It’s basically aligning ourselves so that we come to know each other and that we can dance and synchronize and exchange. And after a while, I don’t need to ask you any more questions and you don’t need to ask me any more questions. We are now on the same page, singing from the same hymn sheet, the same generative model, the same world model, the same kinds of narratives. Having said that, of course, there’s also in the background the putative or potential novelty of finding out what somebody’s not like me like. I think asking questions about the right kinds of narrative that resolve uncertainty, responding to epistemic affordance, novelty-seeking, information-seeking, whilst at the same time still avoiding those surprising states of loss or physiological extremists, put that into a social context. And I think you’ve got some really interesting questions and possibly a structure and a framework to understand social organization and sort of information exchange and self-organization, not at the level of just the individual negotiating with his or her body, but negotiating with another individual with a very similar kind of body. I spent a lot of time studying Jean Piaget and his description of how cognitive structures organize themselves across time. And you might imagine a two-year-old as a collection of micro-narratives, each of which are driven by a somewhat independent motivational state. And so two-year-olds will cycle between being too hot and too cold and too tired and too playful and too enthusiastic and too anxious without any real overarching integration. And they start to manifest that ability to integrate those already integrated motivational states, which are probably in large part hypothalamically controlled. They start to be able to integrate those into a continuous narrative through time and in different locales apt to. And so then they start to be able to play by themselves. And so, but it’s at three or thereabouts when they can adopt a shared narrative with someone else. So a little boy and a little girl might get together and the little boy will offer to play, or the little girl, and say, would you like to play house? And so that specifies the goal. And then the little girl has to say yes, it has to be voluntary. And then the little boy might say, well, I’ll be the dad and you be the mom. And then they can integrate their two identities within that overarching concept. And so they can develop mutual understanding that way. And it’s partly because you said earlier that the best way to make someone else predictable and you to them, which is also equally useful, in some sense is to be inhabited by the same conceptual structure. And that’s sort of what we do when we decide to play a game together. It’s like, here’s the game, here are the principles by which we’re going to operate. Those are the rules of the game. And if you’re operating by those principles, and I’m operating by those principles simultaneously, then we’re gonna share perceptual reality because that’s instantiated in relationship to the game. And we’re gonna share emotional response. And so I can now predict you if you’re playing the same game, merely by reading off of me. Because the same thing, and I think this is what we do when we go to a movie and we watch the hero, we adopt his hierarchy of attentional prioritization. And then we can feel the same emotions because we’re in the same state. And we really are, like neurophysiologically, we’re in an inelegate state. And then the understanding comes not from me deriving inferences about the character’s motivations because of his actions, but by me adopting his goal or his story, and then reading off the emotions I have, which are now isomorphic with his. And I think children are doing the same thing when they play a game. And I would say that you and I, in so far as we’re playing the same game in this conversation, are very likely to remain predictable to one another and also to occupy the same emotional states simultaneously, similar at least, similar enough so we’re not jarring and off-putting to each other. Yes, I wanted to say that. So the shared narrative is, I think that’s a perfect example of a shared narrative. And you crucially point out that you’re talking about theory of mind, and one of the easiest ways to get to theory of mind is just to commit to the hypothesis that you are very much like me. So what I would feel if I sensed what you sensed is going to be a very good proxy for my inference about your feelings. And of course, making inferences about you just is theory of mind. I also wonder whether you were gonna develop that argument sort of almost pre-stages of Piagetian development just to how a newborn infant starts to make sense of its world and the very emergence of selfhood that self is distinct from mom or the rest of the world, getting into notions of motor babbling, babies that will rattle their toys to say, yes, I caused that as opposed to mom causing that. So I think that what you’re talking about, the denouement being the ability to actually recognize, oh, you are another that is like me, and if we can share the same narratives and there is some, not only a deep connection and a communication, but also very sophisticated theory of mind, that would be the denouement. But en route, you’ve actually described a mechanism as say a structure learning, learning the structure of this hierarchical world model or self model that entails the emergence of selfhood in and of itself. In order for us to take turns, we both have to have this fantasy that I am me and you are you. And I have to recognize I am me and you are you, and we have to take turns. So now it’s your turn. Right, so those are the axiomatic preconditions of that game and the axiomatic preconditions. Some of them are you exist as a separate entity and I exist as a separate entity, but we can be joined together in a shared vision. And one of the, just for those of you who are watching and listening to understand is, well, if this is true and we establish a shared, two people establish a shared narrative, and that shared narrative simplifies the world and that simplification constrains entropy, then that shared narrative constrains terror. And so you might say, well, why is it possible to be calm in the presence of someone else given their infinite complexity and also their almost infinite capacity for mayhem? And the answer is, well, in so far as the two of you establish and inhabit a shared narrative, then all that in tropic complexity is constrained by the desire, let’s say, to keep the shared narrative intact. And then you might also point out that if you and I, our shared narrative might be, it would be worthwhile to have an interesting conversation because both of us might learn something and we might have the opportunity to bring a bunch of other people along for the ride, which seems like a good additional bonus. And so we both come to this conversation with that story in mind and then we can play the game as a consequence. And even though we don’t know each other because we assume goodwill on each other’s part, not only are we not anxious in each other’s presence, apart from whatever additional relevant features might have to do with being on a podcast, but we can also take some pleasure in joint movement towards the shared goal because the dopamine system, to me, seems to indicate progress towards a shared, to the instantiation of a vision. It’s something like that. And then if you’re acting in a manner that makes the vision appear to me making itself manifest, then that’s rewarding. And that reward has an existential element, a phenomenological element you can feel. It feels good, but then the dopamine, what it does neurochemically, seems to be to track the neural systems that were activated just prior to the manifestation of the success and make them grow. And so once you establish a shared narrative, if you negotiate it successfully, you also increase the probability that the neural architecture manifesting that vision, making that vision manifest in skill and apprehension is more likely to dominate in the future, more likely to take priority in the future. Yeah. No, no, go ahead. Well, no, I’m just noting you’ve now sort of brought in dopamine as an important neurochemical part of the anatomy of sense-making and exchange. So I just wanted to acknowledge that because in my world, the way that one might describe that and interpret all of the empirical evidence is very much along the line of this notion of pursuing a narrative, and in this instance, a dyadic narrative or a narrative about how I should behave in a dyadic context. And if I resolve my uncertainty, and just to open brackets, just to come back to your nice observation about we might learn something from this exchange, I think that speaks to that novelty bonus, you were talking about a bonus, and certainly in machine learning, this information gain, this epistemic affordance that is part of the good narratives and the good paths into the future would be seen exactly as this sort of novelty bonus. So that is part of the reward. So if in this exchange, we are both realizing, you use the word instantiating, instantiating or realizing through committing to the right narratives, literally a narrative of verbal exchange or linguistic exchange here, and that secures a resolution of uncertainty about what you think or indeed about what I think, then that will be rewarding in the sense that it’s minimizing uncertainty, minimizing expected surprise, minimizing, if you like, entropy in the sense that you originally introduced it. So I think that’s an important, if you like, generic thing, concept that you bring to the table here, that all these ways of looking at good narratives and good engagement with the world, I think are all very internally consistent, especially when placed in terms of interpersonal interactions, and they also actually have biophysical correlates in our brains, and you’ve identified a really important one, which is dopamine, and we could talk about, and I suspect you would want to talk about, what’s special about dopamine relative to all the other chemicals that are responsible for message passing and belief propagation and sort of getting the hierarchical fantasies aligned in order to explain what I’m sensing, what I’m hearing, and indeed what I’m saying. So do you want to elaborate on the role of dopamine? I’d be more than happy to hear what you have to say about that. So, because we talked about surprise minimization, but this is reward and reinforcement, and the propagation of growth, and that can happen artificially. I mean, if you dose yourself with cocaine, you can produce a cocaine-seeking narrative that’s instantiated in your brain, and that’s actually what constitutes the addiction in some sense. It’s a cocaine-seeking personality that’s a unidimensional monster that now comes to dominate your neurophysiology in conditions of deprivation. That’s a very bad idea. You’ve generated an internal parasite that’s fed on this externally applied chemical. But so if you could elaborate on the role of dopamine, I’d be more than happy to hear what you have to say about that. Yeah, no, absolutely. Just picking up on that nice notion of a cocaine addiction being sort of parasitic. I think that’s absolutely right. It’s sort of almost as if there’s been a short circuit, a hijacking of the normal mechanisms that we would, our brains would certainly bring to the fore to actually choose and register the choice of the right paths forward. So for me, reward just is that minimization or realization of expected surprise or uncertainty. So it is intrinsically rewarding to resolve uncertainty and to secure and seek out those novel things or avoid those unfamiliar, uncharacteristic obstacle light states, states that do not characterize me. So dopamine, I think, red light, that just is the fact you have resolved uncertainty. So if I get a cue in the world, say a conditioned stimulus that tells me, oh, I now know exactly what I’m going to do next. I’m going to, if I’m a little monkey in an experimental paradigm, I’m going to receive a drop of juice and I’m going to drink that. If I am somebody engaging in a social conversation, then I know exactly where this conversation is going. That’s great, I know exactly what I want to say. So I think that’s when you get the dopamine blush, that resolution of uncertainty. Suddenly you see the path forward clearly, and it is exactly, and I’m using path in your sense of the micro story, that the micro story that’s responding to the affordances. What’s special about dopamine though? Well, it’s a neuromodulator. So it plays the role in the brain as not of sending information from this neuronal structure to this neural structure, or this set of neurons to this set of neurons, but greasing the pathway by setting the excitability or the game by being the chemical mechanism by which you will switch on this set of messages or that set of messages. Another way of saying that is it sensitizes, for example, let’s come back to your sort of hierarchical structure, that the micro stories are informing or perhaps the mismatch at the lowest level, the prediction has at the lowest level, are inducing belief updates at a higher level to get to these simpler, more abstracted inferences you were talking about before. But how much does the high level listen to the low level, and how much does the low level inherit or respond to top-down constraints afforded by your simple high level abstractions, which could of course be the narrative. So chemicals like dopamine, and I wonder whether you also want to talk about things like serotonin in relation to things such as depression and learn helplessness. All of these neurochemicals have one thing in common. Their role in the brain is just to sensitize one set of neuronal representations to messages from another set of neural representations. When placed in your hierarchical context, that can have a profound effect on the balance between how much you’re attending to what’s going on out there. So the microstructure, the low level sensory constructions or categorizations that say you might think are being played out in the early visual cortex or the primary auditory cortex relative to your coherent, deeply structured narratives about me in a particular world. So I would imagine that a lot, your way you might want to go with this is just thinking, well, how might that go wrong? And what would that look like if I had an abnormality of these neuromodulatory transmitter systems in the brain? And of course you’ve highlighted one of the key or a key abnormality that which is induced by drugs of abuse or misuse nowadays, such as cocaine. So that’s, I think drug addiction is a really good example of what tends to happen if you mess with these really important systems. Yeah, well, cocaine addiction prioritizes the micro behaviors associated with cocaine self-administration, prioritizes those over all other potential behavioral microstates over all other stories. And it does that neurologically. On the serotonin front, so here’s a pattern of depressive cognition and you can think about it as the collapse of a hierarchy. So let’s say you have a tiff with your wife and if you’re operating, let’s say normally, in terms of your neurological hierarchy, you might say, well, you know, I’m just having an off day or I’m having an off hour and it’s only one little upset. It’s only one little anomaly. It’s only one little surprise. I can safely ignore it. But that isn’t what a depressive person will think. A depressive person will think, oh my God, I just had another fight with my wife. I’m doing nothing but fighting with my wife lately. My marriage isn’t going very well. I’ve always fought with my wife too much in the past and I’m fighting a lot with my friends. I’m not a really good person to get along with. I mess up everything I do. I’ve always messed up everything I do. I’m gonna keep messing up everything I do in the present because that’s what I’m like and there’s no hope at all for me to change in the future. And you can see that an error that could have been bounded at a low level, which is, well, maybe I didn’t have enough to eat in the last two hours and so I’m a little irritable, has cascaded through the entire hierarchy of self-conceptualization. And so imagine that each level of the hierarchy has to be protected against the propagation of error messages from a lower level. And then imagine each level of the hierarchy has a resistance level that’s set by something like the tonic level of serotonin. So the higher the serotonin level, corresponding to higher social status, by the way, the more error has to accrue at a given level of analysis before a message will propagate up the hierarchical system. And so one of the things my wife and I have worked out in terms of modulating our reactivity to each other and to other people is, well, when should you respond to a disruption in social communication? When should you call someone on it? And our answer has been something like the rule of three that’s fairly typical of narrative descriptions of such things. If it happens once, you can ignore it. It’s just random fluctuation. If it happens twice, you could mark it, but still discount it. But if it happens three times, it establishes a pattern and then something has to be called into question. So I might say if I’m interacting with my wife and it doesn’t go well three times in a row, I might say to her, I tried to be friendly three times in a row and I’ve been rebuffed. What that indicates to me is something else is going on here. That’s like a Freudian slip in some sense. It’s like, I think this is what’s happening. I want this to be happening, but it’s not happening. Here’s the evidence, three instances. Thus, we have to reconfigure the narrative that we’re using to structure the space and we have to say, well, what actually is happening here? What needs to be resolved? And you don’t say, well, we were rude to each other three times today, therefore our marriage is over and we’re both terrible people because that would be leaping too far up in the hierarchy. You might say, well, is there something else going on in the background that’s disturbing you so that you’re more irritable in relationship to me that’s part of a different conceptual structure? And maybe the other person will say, well, I didn’t have a very good day at work. I was arguing with my boss. He’s a bit tyrannical. Then you can go off on that narrative and try to resolve it. But you can see depression as the collapse of that resistance of the hierarchy to the propagation of errors upward. And so when you give people serotonergic reuptake inhibitors what they seem to do arguably is make each level of the hierarchy more resistant to the propagation of upward error. And the reason I tied that into social status is because we know that animals that have higher social status and therefore occupy a more secure position in the social and environmental hierarchy are more resistant to anomaly partly because they can rest comfortable in the supposition that their superordinate status actually means that they’re globally safer. They have better social relations. They have better access to necessary environmental resources. The world isn’t as dangerous a place. And so can imagine that your brain computes how likely an error message is to propagate upward partly by looking at your social status which would be the value that other people have attributed to you by their distributed computation. And it does that with trait neuroticism which is your own genetically mediated, mostly partially at least, genetically mediated initial propensity for those error messages to propagate up the hierarchy. You might say they’re more likely to propagate up the hierarchy, the internal representational hierarchy of women. And women are more sensitive to negative emotion than men. And I think the reason for that is is because they have to take care of infants. It makes sense for them to be more sensitive to smaller errors of prediction because the consequences for someone who’s truly vulnerable, an infant, can be cataclysmic. And so anyways, there’s a lot in that, but that’s a theory of the relationship between hierarchical processing of entropy and the proclivity for depression. Yeah, there is a lot there, but it all makes perfect sense from the point of view of hierarchical inference in the brain and particularly hierarchical predictive coding. So if you’ll indulge me, I’m just gonna say exactly what you said, but using slightly different words because I think that that notion that you’ve just described and its implications for things like depression has a lot of construct validity in relation to sort of more machine learning, artificial intelligence formulations of this hierarchical processing. So one way of articulating that insight is to think of the message passing in a hierarchy that literally is our brain under the rubric of predictive coding. So in this sort of framework or scheme, the idea is that each level of the hierarchy, it receives information from below and it tries to explain away the information based upon top-down predictions. And that which cannot be explained is ensuing prediction error. So these are the mismatches you were referring to before. And then these prediction errors are used to revise beliefs or representations, sub-personal beliefs or representations at the higher level, the more abstract level, until the top-down predictions are more apt to explain away what’s going on below. Now, the key thing about this architecture, well, there are a number of key things and we’ve spoken in depth about a number of them, but framing like this, then the game of minimizing surprise, the game of is just the game of minimizing prediction errors, how are you doing that? Well, you’re explaining away what’s going on down there based upon higher level hypotheses or belief structures or expectations or representations in your hierarchy. The key aspect of this sort of message passing scheme, this predictive coding scheme, is that it really matters how much weight you afford to the prediction errors that are passed and ascend the hierarchy. You use the word sort of cascading up the hierarchy. And this is exactly the image that an engineer would have when building a predictive coding machine. And the degree to which they cascade up is exactly proportional to the gain or the sensitivity that is set by the neuromodulators, in this instance, serotonin. So the resistance now is set by having entrenched, if you like, beliefs about this level of the hierarchy relative to this level of the hierarchy that is mediated by decreasing your sensitivity to the ascending prediction errors. So if I read your, well, let’s take depression. And I also like to talk about psychedelics because they act upon exactly the same neurotransmitter systems. But let’s just take depression, which is a particularly pernicious, I think, sort of set of narratives to find yourself in because I’ll just cut to the new model, this argument. In many senses, this predictive coding formulation, when put in the context of me discovering and learning and optimizing my models of the world, is all about accumulating evidence for my explanations that are updated in a way that minimizes the prediction errors. But in accumulating evidence, I have to expose myself, I have to choose to expose myself to the world. I have to actively sense and go out there. Depression is pernicious because, of course, a lot of the symptoms of depression prevent you going and getting evidence that you’re not this kind of person or that you could have coped with this particular scenario. So depression, I think, a little bit like the cocaine using that, it has a slightly self-subverting or self-maintaining aspect that sort of hijacks the normal ways that we get out of it. But just to come back to this sort of, I think, quite fundamental notion of sort of inducing, either through physiological setting of that particular resistance or that what we call precision. So we call it the precision. It’s the inverse uncertainty. It’s the reliability that you can afford these ascending prediction errors that tell you you’ve got to change your mind. You’ve got to find a new way of coping, either in a marital relationship or just in terms of where you’re actually looking from many, many different levels. So this notion of precision translates into exactly what the neuromodulators control, which is the excitability of the neuronal cells that are broadcasting the prediction errors to the next level. And as such, that looks very much like attention. We were talking before about sort of change blindness and we just don’t seem to be able to attend to things that are irrelevant. To the extent we don’t even notice changes when they occur, if we don’t assign them the right kind of informative or salience or precision that is necessary to explain the narrative that’s unfolding before us. So one way of reading this sort of this state, this syriacnergic state or continuum where these high level, hierarchically higher level beliefs are recalcitrant or insensitive to the lower level information is the remarkable and important capacity to ignore stuff that is irrelevant. And what you’re saying is when it happens three times, perhaps I shouldn’t be ignoring it anymore. And I now have to redeploy the precision, the neuro, refocus my neuromodulator systems, perhaps away from serotonin at the top and perhaps more to say acetylcholine at the bottom, just given the anatomy of these neurotransmitter systems that would indeed render me in a state where I’m now much more attentive to what’s actually going on out there. And what is actually going on out there will engender prediction errors that will change my mind or indeed change my generative model that entails this hierarchical structure. So I think everything you’ve said makes perfect sense from the perspective of the mechanics of belief updating and structural learning in the brain seen through the lens of an engineer who thinks about the brain as a predictive processing or coding machine. So what does acetylcholine do? What does acetylcholine do? You contrasted acetylcholine with serotonin and you associated acetylcholine with increased precision of attention focused outward and that’s at the lower levels of the hierarchy? Yeah, so this is a vast simplification, but I think it’s sort of a useful mnemonic. So if you think about the hierarchy that you were describing before and you now want to discriminate between a situation where all my high level beliefs are insensitive to changes at the lower level. So this would be say the dominant alpha male, very, very self-confident, has very precise, what we sometimes in a Bayesian reading of this predictive coding scheme called prior beliefs. Prior beliefs, this is the way I behave, this is the way you behave and I am going to realize instantiate those fantasies by behaving in this way and indeed that’s what normally happens. So I’m very confident and that translates into a high degree of prior precision which could be mediated by things like serotonin. The equivalent neurotransmitter at the lower level is often just looking at the neuroanatomy and the neurochemistry and physiological experiments. A similar role might be played by acetylcholine. So you can think of, if you like, too much prior precision as being mediated by serotonergic neurotransmission. And of course, you know as well as I do that it’s a very complicated game with different receptor subtypes and sort of inverted you behavior. So I’m not saying it’s more or less, but certainly rests upon the way that you deploy your serotonergic firing that will have a profound influence on the higher level prior beliefs. Exactly the same kind of role may be ascribed for cholinergic neurotransmission from the nucleus, basalus, and mine heart which is another neuromodulator. So you’ve got dopamine, you’ve got serotonin, you’ve got adrenaline or norepinephrine, you have oxytocin, there are lots of them. And they may all have particular roles in setting the precision or the recalcitrance or the sensitivity at different hierarchical levels. So it’s generally, in my world, if you think of serotonin doing one thing then the complement of that is the astral currents doing it in the inverse way. So it’s sort of like a yin yang. So when I talk about attention at the periphery it’s likely that that does, sorry, the low level of the hierarchy, it’s likely that that does rely upon intact cholinergic neurotransmission with possibly a barren serotonergic neurotransmission that may be due to psychopathology or it could be due to taking drugs that affect say 5H22A receptors like all the psychedelics, psilocybin for example. Okay, so I want to go in, I’ve got three directions to go in now. The first question that’s been lurking in the back of my mind for a while is, okay, when you make progress towards a valued goal, let’s say we inhabit a shared narrative and we’re making progress towards our mutual stated goal and when we see ourselves making progress we get a bit of a dopamine hit. Could you say that the fundamental reason for the positively rewarding effect of that movement forward is that as I move forward towards a goal I decrease the entropy that still remains between me and the goal. And so is even that reward, is even that movement forward readable as an entropy reduction? Yeah, absolutely. Because I’m closer now, so there’s… Okay, okay, so… I mean… Right, see, I didn’t know, I didn’t understand that before, okay, okay. I mean, it’s almost written into the mathematical meaning of the word. So if entropy just is uncertainty and as I get close to resolving that uncertainty, getting my fruit juice, pleasing my wife, or being able to watch the news, if it’s an epistemic reward, it is just expected surprise, just is the uncertainty and the closer you get, the more and the less uncertain you are and all the evidence suggests exactly as it says. Partly because the… Well, the closer you get, the fewer things you have to compute in order to get there. So that’s a good working definition of entropy. It’s like I have to do less, I have to handle less doubt between me and my eventual destination. So that, okay, so that’s cool. So that reduces dopaminergic reward to a subset of entropy reduction. Right, and we should point out, living creatures are always fighting entropy. They’re trying to violate the laws of thermodynamics in some, not fundamentally, but in some local sense, by insisting upon the maintenance of order in the face of this proclivity for things to go every which way at once. And so, all right, so movement towards, movement forward towards a shared goal. That’s also gonna reduce the entropy between us, because it means if I can rely on you to accompany me as I move forward, that means I can predict you better. It also means that both of us are now in a situation that’s less entropic because there’s less variability between you and me and the joint us and that shared goal. And so I’ve fortified my belief in your reliability, and I’ve reduced my apprehension of your entropy. Yep, a physicist would love that because of course the nice thing about entropy and free energy, which we’re here sort of reading as surprise and prediction errors, is an extensive quantity. So your free energy and my free energy or your entropy and my entropy, we just have to add them together because they’re extensive. Then our free energy is exactly the sum of our free energy. So if we can both render our mutual worlds more predictable and less surprising, then our joint free energy will fall. This is, if you like, it could be read as a statement of imperatives, but you could also read it another way in a much more deflationary way that stuff societies through to cells that exist are just these free energy minimizing systems. So these conversations are just the kinds of conversations that can only be there simply because they are free energy minimizing. And if you like, what is left when you wanted to say, resisting the second law of thermodynamics, the very fact that we are here, having this low free energy conversation and this exchange rendering everything mutually predictable and resolving uncertainty about ourselves means that this little dyadic exchange is in itself a free energy minimizing system. Free energy here again is just being used as a proxy for uncertainty, for unpredictability, disorder. So by minimizing free energy, we are implicitly going to be minimizing the disorder and the entropy and the expected surprise. So it’s all very consistent with the physics of self-organization. What you’re doing though, I think, is thinking about what would these things look like in a social context, in a dyadic context. Just to say also that one way of reading what you were saying about, well, if we can both shorten the path to a state of orderly predictiveness and uncertainty resolution, then if one thinks about that in terms of interactions either between people or between people and their environment, then you’ve got now a nice model for niche construction and cultural niche construction. Everything I do is in the service of making everything more predictable. And if that involves evolving to have a language and to teach my children language, then even that aspect of very, very high level niche construction in culture, acting upon the world to make it in an encultured way, more learnable and more predictable is all in the service of minimizing this entropy, anxiety, or free energy. Uncertainty, sorry, I slipped you an anxiety there. I shouldn’t do that. Anxiety, I think, would be our remarkable capacity to recognize that we haven’t resolved our uncertainty in the way that we would normally expect to. And that would be the situation where the dopamine just goes away. So how do you view the rule of GABA? So that’s one question. Another question is, I’ve spent a fair bit of time thinking about hemispheric specialization. And I’m wondering if, is there any reason to make the assumption that does the left hemisphere specialize in some sense for precision? Or does it specialize for instantiating certainty at the lower levels of the hierarchy, whereas the right hemisphere is involved in play at the higher levels? Is there anything to any of those concepts that you know of? It’s a very interesting question. And now I’m sort of speaking as a sort of imaging neuroscientist about the function anatomy. First of all, if we just go back to what we were talking about before, which is the cybernetic view and the good regulator theorem, and the notion that we are, or we entail, good models of our lived world, or at least our sensed world, then having two hemispheres tells me immediately that there is some lateral symmetry in my lived world. And of course that tells me, of course that is true in the sense I have two arms and two legs. If my world, certainly as a newborn, is basically 99% my body, I think the sort of having two hemispheres tells you something quite fundamental about the universe into which you are, as a brain at least, introduced. Just to generalize that, what that means is if you gave me the brain of a Martian, I should be able to tell you a lot about its lived world and its embodied, and its body, and the kind of world that it lives in, just by looking at the structure, the anatomy of the brain. So I think that there’s an important aspect to that sort of lateralization issue. I think a sort of more scholarly and more specific answer to your question is that there is certainly in neuropsychology asymmetry in the way we deploy attention. So if you now read the deployment of certain neuromodulators, such as serotonin, or acetylcholine, or indeed adrenaline, as instantiating endogenous attention, then its deficits will correspond to certain kinds of neglect, the pathological inability to attend to, i.e. you’re always going to ignore or just not be aware of this. And of course, there’s really interesting work in terms of hemineglect systems and bilateral asymmetries between the right and the left parietal cortex in these syndromes. So I don’t know very much beyond that other than to be able to say that for reasons that must have a principal explanation in terms of the high order causal structure of the worlds in which we operate, there certainly is some asymmetry in the way that we attend to things, or there’s some benefit in terms of having that factorization that allows certain things to attend to, that this just set the sensitivity or the flexibility or inflexibility of a hierarchical construction. It must be the case that there are certain domains and certain attributes that do show this lateralization. And just to point out also, of course, that the lateralization issue was quite hot in the days of Gray at the Morsley as a possible cause of things like schizophrenia. Yeah, well, I mean, Goldberg, who was a student of Lurius, suggested that the right hemisphere was specialized for processing in the domain of novelty, and the left hemisphere was specialized for processing in the domain of relative certainty. And so it might be something like the more novel it is, the more likely the right is to attend to it. And that sort of maps on to Gilchrist’s, McGilchrist’s conceptualization of hemispheric specialization with regards to both predation and predator detection. So the right hemisphere seems to be specialized for contextual evaluation and the spotting of predators and the left for focused attention in the service of predation. So a bird, for example, will attend preferentially with the prey detection system while eating, but the right hemisphere and the other eye are scanning the environment for signs of context-dependent signs of predation on the bird. Okay, I didn’t know that. That’s very interesting. So I was just thinking, of course, the obvious example of lateralization is language. And if you look at language as really predation for information. So if you think of language as the way of asking questions, that is the tool that we use to predate for information. So that makes entire sense. I didn’t know that about the comparative ethology and anatomy of predators, but if it’s a little- Yeah, well, if McGilchrist’s new work, details that out at some length, the relationship between attentional breadth and focus and hemispheric specialization. It’s quite nice and it maps very nicely onto the concepts that we’ve been discussing today. So shall we take a brief foray into psychedelics? And then we’ll have to close this part of the conversation, unfortunately, although there’s about 50 other things I’d like to discuss with you. But while that’s, I guess, we’ve covered a fair bit of territory for one day, but I’d be, I’m very interested in your conception of the relationship between, say, psychedelic experience and its antithesis in some sense, if I’ve got this right, with the action of antidepressants. Yes, well, yeah, I’m not sure it’s an antithesis because there are, in the past few years and indeed months, an increasing number of papers looking at sort of 5-H2A agonists and partial agonists and drugs that act upon the social-nerdic system, namely psychedelics and their ability to remediate certain conditions that would have usually pronounced effective states. So the game is very complicated, but what we certainly know at the moment is that the actions of psychedelics, from the point of view of their definitive effects on the brain, namely the abnormal perception and the characteristic way that you can’t attend from the microstructure of your sensations, we do now know that that is probably best explained when I say we know, we conjecture that it is nicely explained by exactly the same kind of mechanism that you were talking about before, which is a sort of changing the balance of recalcitrance or precision or sensitivity away from these high-level constructs deep in the hierarchy, at the top of the hierarchy, and reinvesting that kind of precision or sensitivity under predictive coding models to prediction errors much lower in the hierarchy. So this would look basically like I’m now going to ignore my prior beliefs about the narrative that I’m currently committed to in terms of this interaction. And I’m just going to focus in on what can I sense. So you were talking before about a mother’s predisposition to be very sensitive to cues that could engage or could represent really important affordances like responding to a baby crying. So this would be one way of viewing the effects of psychedelics that you are forced to by reducing or relaxing the precision of the high-level beliefs in relation to the lower-level evidence or belief updating or evidence accumulation. You are putting yourself in an attentional set where everything is interesting, you can’t attend away from it at a very elemental level. And it just struck me that this is very, very similar to what you were talking about before in terms of the, a neuroticism, and I presume this is sort of an Einsink-like construct, where you would have some people who are just jolly confident that their prior beliefs are the most apt explanation, and they will ignore lots of evidence to the contrary simply by suppressing the precision or the importance or turning down the gain on that kind of thing. But you can’t do that if you get more than three errors that you can’t explain. So there is an adaptability built into our brains that will actually say, well, no, actually let’s just attend to these lower-level ones, and at that point, you’re gonna have to relax the high-level and become more flexible and more adaptive. Well, okay, so that ability to relax at that level seems to be indexed by the personality trait openness. And open people are more creative, and so creative people have more play in the higher-order conceptualizations. And one of the solid empirical findings emerging out of the research on psilocybin is that a single mystical experience induced by psilocybin produces something approximating a one-standard deviation increase in trait openness that’s permanent. It never goes away. Yeah, yeah, and so it doesn’t look exactly like it’s a reduction in, or say an increase in neuroticism so that error messages can propagate upward. It’s something akin to that because if you’re more open, then there’s more play in the system, but it doesn’t seem to be tied exactly to error per se and negative emotion. Like, I can’t puzzle out the distinction exactly, right? Because if you’re high in neuroticism, you’re gonna propagate error messages, but then things are gonna collapse. If you’re high in openness, the error messages propagate, but you generate alternative theorems at a very rapid rate in order to recontextualize the anomaly. I didn’t know that. That is very interesting. So I think then it’s the openness that I was talking to, and certainly that’s the aspiration or the motivation behind the use of these chemicals in, say, end-of-life care, or indeed in terms of psilocybin-assisted psychotherapy. It’s really to open you up to new possibilities. Well, maybe with creative people, what you see, so imagine category rigidity, and category rigidity might be something like the probability that activation of one category will activate adjacent categories. So imagine that constraint is the constraint of openness. The more open you are, the more flexible those boundaries, the more when you activate one category, you’re gonna co-activate a network of associated categories. So then imagine you dump psilocybin into the system, and what happens is the barriers between adjacent categories become more permeable. And so then as information propagates up, there’s more play in the systems because the category boundaries have become wider, and that would increase your probability of a false certainty, right, which is an idea derived from insight that’s wrong, but it would also increase the probability that you’d get some true positives out of the deal, which is really what creative people are doing all the time. A lot of creative ideas just aren’t functional, but some are crucial. And so it’s a high-risk, high-return cognitive strategy in some sense to generate, to have looser categories or more co-activation of categories at the higher levels. And certainly that is akin to what people report in psychedelic experiences, that ideas flood in on them, and they see how things are connected in ways they couldn’t perceive before. And so that’s different than the flexibility that high neuroticism in some sense produces, because that’s more like the probability that a conceptual system will collapse rather than it will expand. But your use of the word barriers, I think, is very nice. And I’m just wondering, and certainly, I think you’d enjoy speaking to Robin Kaha Harris, who has described, I think effectively what you’ve just described, but instead of casting it in terms of jumping through barriers, he would describe it really as a reduction in the height of a barrier. So if you can imagine, and it’s in my world, it would literally be a free energy landscape. And our ideas, our prior beliefs are basically sitting at the minima, at the bottom of a well. And sometimes we can get stuck in a rut. I see, yeah, yeah. For example, if you say I was depressed, or I had the hypothesis, I am going to die, and this is how things that are going to die behave, and this is how I’m going to behave. And that may not be the most functional way of that kind of end-of-life self-modeling. Then by making the barriers more permeable, simply by reducing the height, you now enable a jumping from one minima to another minima to explore more options, exactly in the spirit that you meant in terms of creativity. But it could be creativity about other ways of being me in this situation. And that flattening of the landscape is just one way mathematically of writing down the reduction in the precision or the rigidity of these high-level beliefs, prior beliefs, relative to the lower ones. So I think there’s a beautiful conciliance there. Okay, so that would imply that, that would imply as those walls come down, let’s say, that it would require less novelty propagating up the system to produce a phase change. Exactly, yeah, yeah. Yeah, okay, okay. Okay, well, I’m afraid we have to bring this part of this conversation to a close, even though I don’t want to. There’s other topics I would love to discuss with you. I would very much appreciate it if you would consider putting me in touch with, is it Carhartt Harris? Indeed, yes, I will do that. Yes, yes, because I know some of the papers that you’ve written jointly. And I would like to discuss those further. Maybe I close with an observation, if you don’t mind, is that one of the most functional narratives, as far as I can tell, is predicated on the idea that you should conduct yourself in a manner that leaves you open to exposing yourself to information that will allow you to update your narratives. Right, so it’s a weird loop. It’s like, well, narrative itself is dependent on exploration, and so the best narrative in the most fundamental sense is one that leaves the option of exploration continually open. And that’s something like a voluntary confrontation with the anomalies that characterize existence as the central pattern of adaptive being. It’s something like that, and that’s existence on that border between chaos and order in some fundamental sense. So anyways, I appreciated the conversation very much. It would be fun to meet in person sometime. I think we could probably talk for about 36 hours. And I’ll be in London again in January, and so maybe we could meet then, if you’d be amenable to that. And in the meantime, I would like to let everybody watching and listening know that I’m gonna continue my conversation with Dr. Friston for half an hour on the DW Plus site. I like to go behind the scenes with people and to investigate the process by which their narrative unfolded, the process by which they made their path through life, their successful path through life, because I think it’s very useful for people to be provided with models of how that occurs. Hello, everyone. I would encourage you to continue listening to my conversation with my guests on dailywireplus.com. Thank you.