"Culture as an additional scale of biological organisation" with Ivan Kroupin and Tian Chen Zeng

This transcript is automatically generated; we strive for accuracy, but errors in wording or speaker identification may occur. Please verify key details when needed.

[00:00] Ivan Kroupin: Just to introduce ourselves again briefly, my name is Ivan Kroupin. This is Tian Chen Zeng, and we gave a talk to your lab YouTube channel recently about "Culture as an additional scale of biological organisation." We'll just recap what the idea was. What we're interested in broadly speaking is developing models of ways of thinking about cultural systems that connect up with biological scales of organization. More principled ways of thinking about culture, what culture is, and how it interfaces with individual human psychology. The first move we made conceptually is to move away from thinking about biological organization as physical entities and towards thinking about biological organization as stable processes. Thinking about an amoeba or bacteria not as the specific molecules involved in its physical structure, but rather the stable processes that can interchangeably bring in other molecules or pieces of physical substrate while persisting across time. Our proposal is that from this perspective, cultural systems, say a post office, can be better interpreted and understood as continuous with other scales of biological organization, because the patterns of behavior within a post office can persist across time even if all the individuals change or the building changes. We can apply this to many cultural systems: rituals, handshakes, the UN. The idea is taking your framework for organizing a hierarchical layering of biological organization at the level of organisms. Our idea is to extend this into cultural systems, which are also layers of self-sustaining processes. You get increasing scales and layers of these processes across time in the same way life has transitioned from prokaryotes to eukaryotes to increasingly complex multicellular organisms. By thinking about cultural systems in this way, we can map this onto changes in the organization of human societies: from small-scale societies where cultural systems are boutique, everything is very interrelated and embedded in a particular landscape, to increasingly modularized, standardized systems that are general across landscapes, such as the difference between learning through cultural apprenticeship within a particular society versus formal schooling, which is uniform across a wide range of environments. The idea is that what changes across historical time, and across biological evolutionary time, is a change in depth of standardization. How many standardized processes are involved in supporting the unstandardized processes at the current top layer? At every scale, the independent unit is doing some non-standard processes. Amoebae are doing non-standard processes in their survival. In small-scale societies, people are doing non-standard processes in their daily activities.

[02:58] Ivan Kroupin: Same thing: highly standardized societies such as ours. There's a lot of non-standard processes. People are writing poetry on computers, but that's only possible because the computer was produced through crazy amounts of standardized layers, modular layered processes that emerge across the course of time. The idea is that this shift from small-scale to large-scale societies, thinking in terms of these layering of modularizing processes, can explain a ton of existing sociological data. Also, if we think about human cognitive systems as adapting to worlds that are increasingly structured through this modularized layered format, we can explain a lot of the cross-cultural differences and historical trajectories in psychological patterns as well. Basically, human psychological systems grow to be efficient at operating within these modularized structures, seeing in the world discrete problems that have definite solutions and rapidly identifying them, rapidly solving them. This is reflected in a bunch of different cognitive domains in performance and IQ tests, relational reasoning tests, formal logic, even structures of memory, et cetera. One thing to note is that this change has nothing to do with progress. The claim here is that objectively something like this has been happening across history. This pattern makes a lot of empirical predictions, which seem to be pretty well supported in the literature historically, and in some of the active empirical work that we're doing at the moment, in everything from solving different cognitive tasks to the visual system adapting to a world with a lot of standardized elements, and that gives us what look like consistent effects in how people interpret visual illusions. It looks like you get these changes all the way down into the visual system. Preliminary results. This is just a quick recap of this first section of what we've been talking about. Cultural systems are self-sustaining processes in behavior space. They organize the behavior of large-scale organisms, primarily humans. Cultural system processes become modularized in layers across historical time. This dynamic is a direct extension of the self-sustaining policies at other layers of biological hierarchy. Thinking in these terms allows us to explain, we think, a lot of historical and cross-cultural psychological data. So that's our part one. I'll turn it over to GC.

[05:58] Tian Chen Zeng: In part 2, we went more mechanistically into how this takes place. We also tried to think about how this way of thinking about things can help to explain certain mysteries in cultural and social theory that have been long-standing. The first thing we talked about is what exactly is happening when all of these different materials are being mobilized by the cultural process to help sustain its continual unfolding. We talked about how these cultural processes need to mobilize different substrates, and among them are mental substrates, such as attention, representation, and motivation, and also physical and social substrates, things that exist outside of the social mind, which include environmental affordances that can be altered so they facilitate other cultural processes to persist. One of the examples we provided was hammer-use and nail-use cultural processes, and toolkits and technological systems and industries in general, which are basically extremely complex networks of cultural processes, all of which help each other to persist. That forms another analogy with processes in biology: when we look at a cell, it's lots of different chemical reactions that are part of an autocatalytic cycle. That's what allows the cell to persist over time. We also saw this in social behaviors. Personalistic norms lead to personalistic networking behavior, and these two kinds of cultural traits are self-reinforcing; they help maintain each other within the population over time. High kinship intensity norms lead to ethnic caste networks. There's lots of game-theoretic or social behaviors that have the same effect as well. After this, we looked at examples of the change in depth of standardization we introduced previously, such as replacing fashioning tools by hand or fashioning built structures by hand, like beds, with fashioning things out of materials that have already been shaped, for example assembling a bed out of IKEA. Or changing from types of social exchange that are incredibly affected by relationship history and community history, by lots of other social details of the situation, versus transactions and social exchanges in large-scale societies, which often take place within a specific domain. Only things within that domain affect the transaction; other aspects of social identity of the people involved in that social exchange don't affect the transaction at all. We talked about how there seems to be a general change in the quality of some of these cultural processes over time. We talked about two things that appear to account for this change, which is that cultural processes take place in a more decontextualized and more compositional way. One example is changes in the way signs are produced, interpreted, and transmitted. We talked about the gradual move from private sign use—one individual using signs for themselves—to informal sign use in storytelling where everyone needs to be present to understand the signs, to semi-formal signs in groups of specialists like shamanic writing, ritual writing, or token systems in trade, to standardized sign systems where only certain geometric or visual features matter and signs don't interfere with each other's interpretation as informal or semi-formal signs can. There's decontextualization and compositionality in both the use and production of signs. We also discussed how you can see this in money. Decontextualization and compositionality appear to emerge in cultural processes when larger cultural processes need to share many tiny units that are constantly recombined to allow the larger process to unfold. One example we talked about already is symbolic communication with written or carved signs. Other examples include money, where you get shared intermediary transactions out of cigarettes or other token systems. We talked about why navigation, the built environment, and navigational processes in a large-scale city can be handled by something like Waymo.

[10:28] Tian Chen Zeng: But in rural Namibia, you cannot do that. Part of why in large cities this has happened is because when you have navigation that involves lots of tiny shared movement steps, it makes a common language for navigation easier to develop. We also talk about construction with shared standardized components. In addressing these questions, it seemed like it answered this other major question in human social and cultural theory, which is in what sense are human cultures superorganisms? This is a very mysterious question, which has confounded people for a long time, because it seems like for human culture and social phenomena, you have both selection and community assembly, and you have learning/development happening on the same time scale. This is very confusing for people. It seems what's happening is you have two kinds of processes: autobiosis, where you have learning and development within a single organism, versus community assembly and adaptation taking place at the population level. These are two ways that you can have adapted biological processes that are fit to a specific situation. In culture, you have a very similar spectrum. On one hand, you have a set of cultural processes that are transmitted together and stay together. That's like the post office — if the post office relocates, you have a single unit that's persisting. But you can also have holobiontic behavior. That's more like the fast food industry persisting in a location instead of a single branch of a KFC. You can have both selection-like processes and learning-like processes happening. With increasing depth of standardization, you have more and more layers of these reassembling processes. This is causing increasing decontextualization and compositionality. You have an increasingly large vocabulary of cultural processes that can constantly recombine. To summarize part 2, we start going deeper into the mechanistic and some of the evolutionary reasons for why increases in modularity, standardization, and hierarchical organization happen. It seems like to go through this, we have to talk about how cultural processes can niche-construct for one another because they frequently mobilize the same substrates, be they mental, physical, or social. Because they niche-construct for each other so intensely, all these cultural processes are constantly affecting one another, and you can have either reassembly of many different cultural processes, or a single set of cultural processes unfolding over time that are adapting to the local situation. This can explain why there is both evolution under natural selection and development, learning/agency happening in culture at the same time. This is an initial attempt to answer and to ground some of the ideas we have learned from your work to see how they play out in the cultural realm. There are still many other questions.

[14:59] Ivan Kroupin: We'll stop sharing.

[15:05] Michael Levin: I usually think about it in the opposite direction because I don't know much about the cultural side of things, but I always wonder what would the developmental version of that look like? What would it look like in a cellular system if that were happening? What would we be looking for? What are the signatures? How much of this do you see in things like ant colonies, beehives, the Internet of Things, and swarm robotics? Is that something you guys do? Do you ever look for these kinds of things?

[15:46] Ivan Kroupin: One thing that we've talked a lot about is that the kind of scale of organization that we're talking about in terms of culture, it's actually a bit of a misnomer to say it's cultural. It's not technically tied to us or to culture. It's just that humans are, to date, uniquely good substrates evolutionarily for supporting these kinds of scalar processes. So you get protocultural systems in great apes. Colonies have this. But by virtue of the fact that we're so adept at social learning and have symbolic communication, we can allow these processes. We're a great substrate for these processes. So the processes can start to maintain themselves, pass themselves along, and allow themselves to adapt and change across time when they're implemented on the human substrate. Much better than in other substrates so far.

[16:44] Tian Chen Zeng: Part of the reason why the human substrate, and if we generalize this to other less human-looking things like corporations or, in the future, humans plus AI or AI themselves, is that we as a substrate are so useful for cultural processes is the agential material aspect of things. The substrates, when we look at how a culture has to reproduce, how a cultural trait has to reproduce itself, there's going to be some level of goal-oriented behavior; it takes up some of the mental space in human beings. And that means that you can have highly abstract cultural processes like science. So the concrete details of what the cultural process should look like don't matter as much. Take the Roman Empire: the number of different goals and processes that have to be unleashed for such a highly abstract cultural entity to persist over time is gigantic. In some sense, it doesn't fully matter what the exact details are.

[18:10] Ivan Kroupin: Every layer below that abstract level has its own agentic.

[18:12] Tian Chen Zeng: Has its own agentic.

[18:13] Ivan Kroupin: The sub-parts of the empire, and then that goes all the way down to humans' world to have drives, and that grounds out in the organism side of things.

[18:21] Tian Chen Zeng: One question that would be very interesting for us to know more about is, because from what I understand, in the biological side of things, you can have microbiome mats, you can have biofilms. They, when the environmental conditions change, the community composition also changes, but there's also some level of coordination within the mat to be able to develop. So what is a signature of development? What is the signature of collective agency in biology? We can say this may have to do with organisms sacrificing their fitness, but any kind of signature, whether on the theoretical or empirical level, that tells us that collective agency has emerged and development is happening? Because if we have that, then the question for culture becomes an easier question to answer.

[19:28] Michael Levin: That's a great question. I think fundamentally what you're asking is how do we recognize an agent and the borders between an agent and their world? Because as far as I understand, all intelligence is collective intelligence. All of us are made of parts. There are no indivisible agents anywhere. I think what you're asking is how do we recognize agents? Here's my approach to it. I'm sure, as you said, there are many other approaches, including an evolutionary benefit approach. My approach is that agents are held together by goals. In other words, if your parts are aligned towards some specific goal, then you can draw a boundary around what are all of the parts that are aligned towards this goal, and that's your agent. The idea comes from asking what an embryo is, because when you look at an embryo, what you're seeing early on is a blastoderm of a couple of hundred thousand cells. What are you counting when you say there's one embryo? What there is one of is alignment. It's a story that all of the cells are committed to. They're all going to go into a particular location in anatomical space. There's a goal that they have that has a certain size. It might be a limb or a whole embryo or whatever it is. What you're counting is the fact that all of them agree on this. If you cut them in half, you'll have two agents or more because they'll develop independently. That requires answering the question of who gets to determine what a goal is. I think this is very much an empirical process because you can hypothesize a goal, but then you have to do the hard work of actually proving that the agent cares about it and that it has some degree of intelligence to get there. At that point you're doing perturbative experiments to block them from their goal. If I think the goal of this set of cells is to build something, I'll put a barrier in their way and see what they do. If they have some degree of resilience so that they can come back to it, and it's not just for damage but for all kinds of changes, that's informative. In the examples I talk about, you can make fairly radical changes to the hardware of the agent itself. You can change the parts, not just the environment but the agent's own parts. Living things, things we call alive, are really good at getting their goals met despite all kinds of things going wrong. They can also shift goals. If I take cells and make it completely impossible for them to make a frog embryo or a human embryo, they will make a Xenobot or an Anthrobot. If you push them far enough, they'll go to a different goal and make something else. It'll still be a coherent thing that they do, but it'll be something different. You have to do these perturbative experiments. Different observers can see different goals and they can both be correct. I don't think there's a privileged single position. There could be multiple perspectives. You also have to deal with the fact that the cognitive light cone is flexible; it can shift. The same system can pursue larger goals or smaller goals, or it can fragment because pieces of it decide to go off and work on something else. This is what we study in cancer: when the border between the self and the world shrinks for some components. That's what I think you should be tracking: some degree of goal-directedness.

[23:16] Ivan Kroupin: That resonates very well with our conception of things. There's a lot of work to be done in formalizing stuff, but you think about ritual practices as an example. They persist across time and consistently will reproduce even with large perturbations. The community can move somewhere, there can be a drought, somebody could — the head person can die — and you still get a really stable configuration of behaviors. You can scale all the way up to corporate systems that are going to survive whatever series of changes and persist across time. They can be quite hard to kill. Similarly, with the bound processes and becoming more local, there's a lot of work on scales of cooperation, and one can imagine nepotism. You go down a level and suddenly you're cooperating at this level, and that's pursuing your goal of advancing family success at the cost of larger scale cooperation. In that sense, on the process level, there's certainly a lot of problems.

[24:13] Tian Chen Zeng: It's fascinating that the way we are talking about this, because we have recently been thinking about morality and ethics and the different moral foundations and what they're actually doing. You might have heard about this: there are the hedonic foundations and there are the binding foundations, and the hedonic foundations are care versus harm and equality/proportionality. But the binding foundations are loyalty, authority, and some people call it purity, but really what it is, sanctity or divinity. It's the capacity to say that some goals are just worth it. Their value is not changed by practical concerns, like comparing it in the context of other goals. For example, if you're married, you're not going to say, "Oh, if someone’s going to pay me this amount of money, I'm going to marry another." That type of capacity to show that kind of commitment. We've just been talking about this because from the game theoretic understanding of morality and coordination, it's assumed that everything is known. It's assumed that morality and cooperation exist independently of learning. The contract specifies everything and there is nothing unknown. Cooperation has nothing to adapt to. But really, once you start thinking about cooperation as something that also has to maintain itself in the face of unexpected challenges, then the binding foundations suddenly become extremely, extremely important.

[25:55] Ivan Kroupin: You're saying, developing the same thought, one of the things that we thought a lot about is that, taking a very anthropocentric perspective, we have very strong intuitions, which we think is one of the reasons it's very hard to see cultural processes, to your point about observer dependence, understanding of what goals are. If you think about only humans and our individual goals, it's hard to see that a value system can have goals and can have adaptive features to one degree or another. One of the adaptive features of a value system can be that it rules out questioning the value system. And that makes it a better survivor within what we've been calling a behavior space. It's going to survive better, it's going to maintain itself better. And so from that perspective, another set of ideas within social science, we think. We get more on that.

[26:45] Michael Levin: I've been told that one of the things you're not allowed to do in court is to question the validity of eyewitness testimony in general. What you can't do is show up and say witnesses screw up all the time, and that's a way of preserving the system as a whole. You can talk about specific cases within the system, but you can't come back and say the whole witness testimony thing is misguided. You're not allowed to do that. Assuming that's true, I think that's an example of the kind of thing you're talking about. It's self-protective.

[27:19] Ivan Kroupin: We think that there's lots of things here. We've talked a lot about how these systems would have immune responses, where you have mechanisms where if people start violating it, undermining things, you get reactions in various ways. Cultural conservatism has some of these properties.

[27:42] Tian Chen Zeng: One question about this. Do you think, speaking from an evolutionary point of view, what is the relationship between game-theoretic fitness, where fitness is extended between lots of different individuals — inclusive fitness — and the agency that you were talking about? Do you think, in your way of thinking, there is an agent to the extent that there is a group, a collective that has inclusiveness?

[28:18] Michael Levin: I don't spend a ton of time thinking about the fitness aspects. You could talk to Richard Watson at the University of Southampton. He's the best person to talk about that aspect of it. But I do think that groups will often acquire collective goals that are not necessarily known to the members of the group. Humans in particular have some ability to know what their goals might be, but they're completely blind to a huge number of them, even at their own scale; that's psychoanalysis trying to uncover it. They're part of collectives and they're made up of subsystems that have all kinds of goals that we have no idea about. I do think that those kinds of things can potentially arise at every scale. I think figuring out what those are is an existentially important aspect for humans going forward: how to predict, detect, and discover what the goals of composite systems are. I also think it's very analogous to the network interpretation problem in AI. In machine learning, we have a number of ways to train networks to be good at specific things. Being able to come back and say, how did you do that, and what is it that allows you to do that, is extremely difficult. So you end up with something that's very good at doing things. Some people claim that that's actually a huge part of human cognition too: you basically end up being good at doing things. Later, the top layers of your mind paper over a bunch of explanations that are good for social interactions about what really happened. The idea is that this is added later. You actually have no idea what happened any more than you know what your liver is doing. You can tell a story you've read in physiology textbooks, but you don't actually know what it's doing. The other thing to bring up is this notion of where the agents are. Up until now we've been talking about agents that are basically physical systems—cells, embryos, humans, or populations. They are the agents. They join into collectives that have certain kinds of memories or goals. What about flipping it and looking at it from the perspective of the patterns themselves? The patterns are the agents. That's been done, for example, in memetics, where people say it's the information patterns that actually propagate. To my knowledge, most of that stuff assumes that the patterns themselves are dumb, low-agency things, passive data that ends up being propagated by physical systems. The idea is you can talk about what's in the interest of the genes, but it's assumed that's metaphorical because the data is taken to be passive and what's really happening is that it's just being copied or propagated differentially. So I want to stand that on its head. I've written a little bit about this and I've been working more on the idea that the actual patterns might be agents in the sense that they might be high-agency things that make decisions and have memories. What makes that plausible is the idea that from a particular perspective, we are also temporary metabolic patterns. If you have a large enough perspective, consider my story about creatures from the center of the Earth. It's a science fiction story that I've changed around to suit my purpose here. The idea is that these creatures come out from the center of the Earth; they live in the core, and they are incredibly dense because that's where they're from, and they have gamma-ray vision. They come out of the core; they come up to the surface. What do they see? Well, as far as they're concerned, everything that we see here is like a thin plasma to them. It's not a solid object. They move; they walk through it the way you would walk in a garden through smells and structures of odors from the flowers. It's not physical; you just walk through it. They walk around and it's a thin gas that surrounds the planet. All of our physical objects are like this thin gas. One of them, a scientist, is watching patterns in this gas. He says to the others it's almost like some of these patterns are agential.

[32:39] Michael Levin: It almost looks like they're doing things. They hold together for a little bit of time — hurricanes, magma flows. They also look like they're doing things. They almost look like they have goals and purpose. Others say, well, that's crazy because those are just patterns in the gas. We're real physical beings. We can be agents, but patterns in the gas can't be agents, right? It seems like they're doing something. How long do these things hold together? They keep it together for about 100 years. Okay, well, that's nothing. Nothing interesting can happen in 100 years. I think that tells us that the distinction between who's the agent and what's a pattern within an agent is relative. It's in the eye of an observer. It gets back to the standard Turing paradigm: you have the active machine that operates on passive data. But you could also flip it and say everything the machine does is just dictated by patterns in the data. The drivers of the machine are just a stigmergic scratchpad in the world. It's the patterns that are driving the machine; everything else is just a scratchpad that follows along. And that has all kinds of implications in our biomedical work: whether you're addressing the patterns as the agents or the cells themselves as the agents. How is this different from mimetics is specifically that it's not that the patterns look like they're doing something, but they're actually dumb. It's the exact opposite. They're the ones much like us. I think William James saw that when he said, "Thoughts are thinkers too." This idea that the patterns within a cognitive system can spawn off their own patterns and so on. The reason I bring this up is because you were talking about fitness. My bias is that the way we should think about this is not that some patterns are evolutionarily more fit and therefore will get copied as passive data; I think we should look at it not as a selectionist process, but as a variational process. The patterns are active agents and they will do things and they will propagate or not. I think that's different. You're not talking about the fitness of these things necessarily. You can ask that too, but I'm more interested in looking at it as a cognitive system and saying, what can patterns know? What goals can they have? There was a paper by Randy Beer called "The Cognitive Domain of the Glider in the Game of Life." What's super cool about it is that there are not actually—if you want to be a reductionist about it—gliders in the Game of Life. There are just pixels that turn on and off. The gliders are an imaginary construct observed by our visual system as we try to make sense of what the thing is doing. And yet it has a cognitive domain. How is it that a pattern has a cognitive domain and what does that mean? I think it's an extremely prescient paper. That's what I think we can also think about in the New York case.

[37:00] Tian Chen Zeng: Can you describe what features of a variational process make it different from an evolutionary selection process?

[37:13] Michael Levin: This is something else that Richard Watson emphasizes. A selectional process means you generate a set of variants of something, usually at random. Most of them disappear and one of them propagates forward. You do that again and again. That's a selection: you're generating a whole bunch of variants, and then one gets to move, or some number of them get to move forward, and the others don't. A variational process is more like learning where when you learn something, you didn't generate a million copies of yourself at random and pick out the one whose brain looks like it has that information. That's not what happened. You had exactly one cognitive unit and you manipulated it. You transformed it somehow so that it's closer to what it needs to be. So those mechanisms tend to be — there's not as much randomness there because you can't lean on randomness for that one. You have to explain what was done. So you need a mechanism for it: what was done to change that brain or whatever learning system it is, and how it has retained and then deploys this information. You can't just assume the randomness thing. Development, learning, all of those things are processes of transformation fundamentally.

[38:53] Ivan Kroupin: Closer to autobiosis in the description than the technology that we've been using. I think this whole way of talking is extremely close to our... to our hearts. Because we also think that it's exactly like this, first of all, we need to be really comfortable taking really massive perspective shifts and moving away from this anthropocentric assumption that humans are the agents and anything above humans is something dominant, some mimetic events. We'll actually, eventually, if one thinks through these layers of processes, humans end up occupying a really strange position. On one hand, we're a collection of biological processes. On the other hand, we serve as substrate for many different patterns at the behavior level concurrently, so we participate in post office and handshakes and chess and corporations all at the same time at different levels of abstraction, and we have this interesting dynamic and you get competition between layers where you've got cultural systems pushing us in one direction, and then the organismal level poses limits. One example we've thrown around is the Industrial Revolution, where you get these cultural systems which are, the factory is the most efficient thing that's going to persist the most, going to spread the most. It has all these properties. But at some point, you're putting kids in factories for 16 hours a day, and there's something organismal level that's saying, this is not sustainable. You can't do this. And so you get changes that have to adapt to that. There are interesting questions about culturally, historically, how that plays out. To your point about thinking about variation, to think about the cultural level would be specific patterns and how they maintain across time and whether what conditions are required and whether they can adapt or do adapt, et cetera, which presumably would be going into various historical data, or even, you could do that experimentally, generating some pattern of how people interact, and then, as you say, trying to perturb it. That's always going to be more brittle and artificial than real-world stuff, but that seems tractable. Or at least potentially.

[41:13] Tian Chen Zeng: It's really, really interesting thinking about it as a pattern, quad pattern, because we've also been talking about some of these processes and how they affect mental health. Nowadays everyone's talking about the attentional economy. It's not just about the amount of time that you spend on devices. It's about the motivational states, the kinds of rumination that they produce. We see all kinds of phenomena; for example, there's an epidemic of people who, after watching a large number of videos of Tourette's, end up believing that they also have it. There are many of these cyclical phenomena, and it's very timely to investigate, because we are in a stage where more and more things are competing ever more intensely to use us as a substrate in ways that are very, very damaging to agency at the level of the individual and at the level of smaller collectives, for example the family, as opposed to Facebook. So it's very, very fascinating.

[42:32] Ivan Kroupin: We've been thinking about this inter-level competition a lot, because that's where a lot of these human-level real issues are playing out. One of the things we have been trying to sketch out is what the nature of these entry-level competitions are and how exactly, what are the mechanisms by which one level is regulating the other, that's setting bounds for the other, what the dynamics are here, because there does seem to be this constant process within cultural systems that are trying to take up. One way we've been talking about this is behavior space. So you just take the total amount of behaviors or time in a day that a person has, and you take that across all people, and that's the total space in which these processes occupy. This is a simplification, because then there's details to it. But more and more of that is being taken up by these modularized separate structures, and I think people don't feel comfortable in that. We like being in these environments where things are not standardized and sitting and chatting. What we'd be super interested in is trying to think through some way of mapping mechanisms and understandings of the nature of competition between levels from your work, which is really well spelled out, and seeing how we can think about it at this level, if there's a generalized way of thinking about it.

[43:56] Michael Levin: Unfortunately, we're just at the beginning of creating tools to help with this. So typically, all the modeling and simulation tools that we have require you to state up front who's the agent, where are the patterns, what do you want to track? You have to know ahead of time where that is. We don't have tools where it will tell you where the agency is. We're just starting to develop some of those in our lab. We have examples of these things. For example, we've made a line of flatworms, planaria, and it's not a genetic line, there's nothing different about the genetics, but what is happening is that it's like those perceptual bistability illusions, like the Necker cube or the rabbit-duck. It can't quite decide if it should interpret that as one head or two heads. When you cut the thing into pieces, sometimes you get one head and sometimes you get two. Pieces from the same animal don't agree. There's a ratio that they'll pick. One thing that we're working on is to model that as an attention mechanism: if you're a two-headed pattern, what can you do to grab the attention of the cell so that they implement you instead of that competing one-headed pattern? Another example is the memories from caterpillar to butterfly. If you're a memory in the brain and body of a caterpillar, you are never going to persist if you don't change, because the actual memories of a caterpillar are useless to a butterfly. The butterfly lives in a different world; it wants different things; it moves in a different way. Specific memories are useless. If you are either able to change or at least are changeable—how much is the pattern and how much is the brain of the butterfly? We don't know that yet. But if you're at least amenable to change and you can resonate with the brain of the butterfly, you could be remapped, and in fact that's been shown: some of these kinds of things get remapped into the thing that the butterfly actually cares about. Again: as a pattern, what can you do so that you grab the attention of the system that you're going into? When we inject a specific kind of ion channel into the tail of a tadpole, that sets up a little voltage pattern that says, "build an eye here. You should be an eye." If we do that, the cells that we inject start telling their neighbors, "you should help us build an eye. Let's build an eye." We know that because if you take a section through those eyes, you can see that there's a few cells that we injected, but there are other cells participating that we never actually touched. At the same time, there's a cancer suppression mechanism where the other cells are seeing this bioelectric pattern and they're saying, "No, you're crazy. You should stay skin or you should stay gut or whatever you're going to be." What you have there then is really a battle of worldviews. You have a battle of models about what should we be? Should we be an eye? Should we be skin? I don't know which of these is better yet. We will find out, or maybe they're both useful—whether we should be looking at this as a competition between cells, I really don't think so. I really think this is a competition between ideas or visions that cells can maintain. One says, "Our future is being an eye," and the other one says, "No, we're going to be skinned." Those patterns compete because in some cases you get a nice eye and the eye wins out, and in some cases you don't. They start making an eye and it goes away, and it goes back to being skinned; the other side wins. You can watch; you can see the patterns. There you have a competition for the attention of the physical, of the machine. Maybe it's some kind of a new formalism where the data are not passive, but the data compete for attention. Again, it all sounds like memetics if you assume that the data are stupid and that it's just selection forces. I don't think you can assume that. I think that, quote-unquote, all of us are data. We are, to some extent, somewhere on that spectrum of agency. They may be putting forth some amount of effort to make sure that they get noticed, to make sure that they get paid attention to.

[48:55] Ivan Kroupin: I think that's right. One thing aligning with your perspective is a phrase we thought about: "the old structure is process a step down." So everything that looks like structures, if you take a step down, it's some sort of process. And the process is often agentic in some way if it's stable enough to, in spite of environmental perturbation, persist. We completely agree with that and that would apply to the sort of cultural systems that we're interested in here as well.

[49:27] Tian Chen Zeng: I've heard you talk about this idea quite briefly in some podcasts and in some YouTube thumbnails. About this platonic idea, was that where the cave idea came from as well?

[49:46] Michael Levin: It's not the cave part, it's the other part. My vision of it is not really close to what Plato himself said, but I was trying to use words that give an idea of things people are used to, because mathematicians talk about it that way. My point was that there are patterns that are functionally important, certainly for biology, certainly for physics, possibly for psychology. There are patterns that are not specified by anything in the physical world. In other words, there are truths of mathematics. If you start looking for the things biology likes as causes, so that we like history, meaning evolution — it is this way because it has a history of selection — and physics, it is this way because the laws of physics make it so. The most conventional aspects of it are aspects of mathematics. This is what a lot of mathematicians believe: these are not set by either of those things. There is no history of evolution for these mathematics things. There's nothing you can change in the physical world that will make EB a different number than it is now, the natural logarithm and Feigenbaum's constant and things like that. There's nothing you can do that will change that. So my point is simply this. I make the claim that there is this, and what the mathematicians offer, at least some of them, is that what we are doing is exploring this pre-existing Platonic space of patterns. We do not invent them; we discover them. They exist. It's structured space that you can investigate, and that's what we're doing. So my point is this: the low agency parts of those are what we look at, what we recognize as mathematical facts, but there are also other patterns that are much more active and much higher agency that we would recognize as kinds of minds. When we build interfaces into that space, and that could be anything from a dumb triangle to a cell, a robot, a biobot, an embryo, when you make these things, they are making our interfaces into pointers into that space. That's why I talk about the Platonic space, because I think many of the important patterns that we need to deal with, certainly in biology, don't come from this space at all. To say that they're emergent, which is the conventional view, is to say you don't need a whole other non-physical space. People don't like that kind of stuff nowadays. They'll just say they're emergent. To me, that's a very pessimistic view, because it suggests you got surprised. You didn't see that coming. You'll write it down in your catalog of emergent things, and that'll be that. I don't like it at all. I think the way we should look at it is how the mathematicians do, which is that it's a structured space of patterns. We have tools to study it. We can systematically investigate the contents of that space. We can't just be surprised when this comes up. We need to map. What I want to understand is the mapping between the pointers and the thing they pull down. Tell me why, when I make a particular kind of collective system, when I make an embryo, why do certain patterns show up and how can I not be surprised when they show up?

[53:00] Ivan Kroupin: That's the opposite. 100%. And I think that that's something we'd be interested in thinking through on the cultural level, because it just has all the same signatures. So presumably there's properties of the substrate for which attention is going to be competed. Cultural evolution has clearly developed the dopamine loop. That's a successful thing for interacting with the substrate. It's going to have downsides, and there's going to be some equilibrium. But there are all sorts of these levers that are interacting with the substrate, et cetera. We wanted to touch on potentials for some joint project—something we're very interested in developing and thinking through: a scale-free set of ideas that we could organize and apply across organismal levels and intercultural levels, even, first pass, just a general description. If you think about these as processes, all this stuff we've talked about in terms of structure actually being better understood as patterns that are agentic to some degree and interacting with substrate. If we step away from anthropocentric intuitions and this attachment to individual bodies that we need to point to, the actual fundamental thing is continuous across into culture and these larger scales of organization. We would love to work with you on a project that would describe this general view, because I think that would be certainly interesting for our domains, since we even have a pretty good sense of how we would plug in where psychology comes in. All the culturally changing parts of psychology are these systems building affordances for themselves to persist more effectively.

[54:50] Tian Chen Zeng: And to start, because this could go as deep as we want or it could be bounded somewhere. Because it is true that eventually we can hit, purely at the conceptual level, this kind of universe of structures with agency. We can go all the way until we hit that. One good place to start is to notice that there's so much material about the emergence of collective agency coming out through variational processes already in biology. Thinking about what is happening at the cultural levels, using the same set of tools, is already incredible. I suspect that having an initial step of laying all of this out in a very systematic way, in a way that people in cultural evolution and the social sciences will understand better, would already contribute hugely to the conversation.

[55:54] Michael Levin: I think that would be very interesting. I think the connections that could be made here to psychology and psychotherapy, and also economics, because there have been a couple of people like Ben Lyons and a couple of other people that we've been talking about. You probably saw that paper that we did on the price system and so on. Potentially there are lots of good parallels here to

[56:24] Ivan Kroupin: Just to sketch out the general logic of it and then even point in the direction of various things. Because there's so much stuff within the social sciences, social psychological sciences, that could potentially be reinterpreted from this perspective. All of the structural sociology stuff and economics.

[56:41] Tian Chen Zeng: Structure and function exist at the same time as evolution. How does that happen?

"Culture as an additional scale of biological organisation" with Ivan Kroupin and Tian Chen Zeng

Watch Episode Here

Listen to Episode Here

Show Notes

Transcript

Related episodes

“The Unreasonable Effectiveness of the Behavioral Sciences in Developmental Biology and Biomedicine”

Discussion #1 with Katrina Schleisman and David Burke

Conversation with David Resnik 1

"Culture as an additional scale of biological organisation" with Ivan Kroupin and Tian Chen Zeng

Watch Episode Here

Listen to Episode Here

Show Notes

Transcript

Related episodes

“The Unreasonable Effectiveness of the Behavioral Sciences in Developmental Biology and Biomedicine”

Discussion #1 with Katrina Schleisman and David Burke

Conversation with David Resnik 1

Thoughtforms Life Podcast