I think Kim might be helping if one of you guys want to come in into the Hangout and introduce yourselves to Reza. Basically, the introductions we did last time were like, what's your background? Why are you interested in this course? What do you think you can get out of it? Just so Reza has an idea of where he might go in lectures. Can you hear me okay? Yes. All right. Yes. Hi, my name's Tim. Thank you so much for having me. My background's in media studies. I live in New Zealand. I completed a master's in media studies a couple of years ago, and I'm now sort of in the process of putting together a potential PhD project. I'm hoping to start next year.
and I'm really interested in the fog of war. I'm thinking about doing a study looking at the fog of war as a hyperstitional unit and a lot of the research I've been doing has been into capstone military literature and I found that a lot of contemporary stuff in that field uses complexity theory as a way of generating modern ways of reading Carl von Klappschitz. And so that wasn't something, like, complexity theory wasn't something I'd really encountered much before. And then I saw that this course was on offer and thought, oh, this is very serendipitous. So I thought I'd enroll in . Have you come across Manav Guha's work?
Sorry, what was that? Manav Guha's work. Because, you know, he's basically, is really an exciting kind of military theorist. I think he is a fellow at Bath University of War Studies, and he has written a couple of good books and a number of really interesting, brilliant articles on basically, because there is a, yes, as you say, the whole complexity theory is basically just a standard staple of military studies these days, especially
after the RAND Corporation and all these network-centric warfare and stuff. And Monov is actually, he's still in two kind of continuation of, again, complexity studies within warfare, But his approach is quite non-standard compared to the kind of conventional text that we come across. He's really great. You should definitely check his work. Can you type his name into the chat? Sure. Great, thank you so much.
You're welcome. And yes, he's, by the way, he's actually interested in this kind of hyperstition and and also these kinds, so he has this background of all of these things, so that's why I suggested him. Okay, and then Gregory is having microphone issues, so we'll ask him to do it later when I help him fix it. So we could, so yes, welcome to the second session of Reza's complexity and computation. So we've already kind of got into it, so I'll just pass it to Reza, I just wanted to make sure we were saying what session it was. Okay. So, yeah, Reza, whenever you want to start, we're ready to go.
Okay. First, so last session, I just talked about the basics, where some of the problems are coming, some of the misconceptions around complexity. And this session, I tried to complete this, And so then we can at some point, either end of the session or the next session, we just go to basically the measure of complexity. And then the last session we do a couple of examples. Like probably I picked one example from, for example, climate change and one either from military or from economy. And then we look at them and kind of like see the kind of both problems and the advantages
that these approaches give us in studying, for example, complex systems in either military economy or climate change. But in any case, I'm going to talk about today, again, like the last session, about the kind of metaphysical and epistemic problems that complexity sciences usually have. When I say problem, I mean it both in the sense of problematic interpretation, and also the kind of problems that complexity approach can detect that otherwise the kind of linear,
non-complex kind of canonical approaches can't really detect. So the emphasis mostly today will be on modeling in complex systems and kind of working out the epistemic metaphysical problems. But also I want to, just as a kind of excursion at the end, talk a little bit about the concept of time and kind of time-driven biases. And by word bias, I don't essentially mean it negatively, but nevertheless, you know, bias, basically putting your main emphasis on a certain conception or certain accounts
of temporality and time. So I want to talk about the concept of time and I want to talk about these time-driven biases that undergird somehow at the foundation of basically complex sciences and complexity theory especially when you look at for example you know the account of causation the account of you know emergence account of determinism you see that there is there is a you know a time bias basically there is a concept there's a you know kind of a canonical concept of time that basically these theories are based on. And, well, the kind of, when you have a commitment
to a certain account of time, a certain account of causality, then basically you have committed yourself to a kind of basically a metaphysical assumption. So then I want to talk a little bit about this metaphysical assumption because you know if you are familiar you know with accounts of causation in sciences and also in you know philosophy of science there are various accounts of causation and to the extent that simply accounts of causation are being used as kind of pragmatic tools to basically work in a very context-sensitive domain.
It's not just some sort of canonical account of causation. But nevertheless, when you look at all of these varieties, for example, accounts of causation that complexity theory works with, you see that there is some sort of like a glue, we call it a time glue, that puts all of these concepts together. By virtue of it, these accounts of causation become possible in the first place. And that's basically the account of a time with a direction, basically time with the flow. And I just want to give a background.
Also there are these controversies that whether can you have causation without time, and can you have time without the account of causation that basically complexity sciences are talking about. And when I'm talking about time, and that's what we are going to talk about a little bit, It's not simply a time with a flow, with a direction, but also what physicists call a block view of time, simply a time that has no direction, that has no flow.
Nevertheless, it can be made compatible. There is no incompatibility between the block view of time and R basically as the agential account of time having a flow, having a direction. And this is something that only has recently emerged into kind of literature around philosophy of time and kind of metaphysical foundations of causality and so on so forth one of the people behind this is British physicist and philosopher Hugh Price is a very good brilliant figure and he has done I will type his name
here. Yes, he's basically, his philosophy is very interesting. He's basically coming from a Solars, Brandon, American pragmatism directory. And his view of time comes from basically a Bolsmanian theory of time and entropy. So anyway, before moving to this, we can start with the material for this session. But before that, any of you have any questions, comments, discussions, something that you
you guys want to address or talk about. Hello. Can you hear me? Sure. Yes, please go on. Sorry, no, I interrupted. Don't worry. You start and. I guess I'll just introduce myself now that my microphone is working. Hi everyone, I'm Greg. I'm a friend of Kim's. I have a background in English literature and philosophy. And I also spent a couple of years in the Singapore Army in military.
I combat intelligence. It was part of some compulsory national service. I'm a poet and I also do some art writing. Thank you very much. Interesting. So I think for the first part I'll follow Kim in examining the fog of war. Okay. I gave him a suggestion. Interesting. Okay, I will definitely try to pick up one of the examples in military studies. I try to...
Okay. Questions, discussion, suggestion? Anyone? I was just going to say thank you for the Tretchfield interview. That was really kind of a helpful introduction. Welcome. Actually, we talked in the previous seminar, the one, what was that, like a year ago, we went through Crutchfield, I think, quite a good bit in one of the sessions, talking about the Epsilon machines and his statistical complexity. I will bring it up again in a more detailed fashion in basically the second module, which
would be the computational complexity. Great. I remember that, but the interview style was really approachable. Yes. Yes. It's very good. It's interesting. And actually, James' Ladyman essay that I recommended, it fits nicely with Crutchfield account. Basically, Ladyman's structural account of complexity, the idea of generative entrenchment, you know, kind of basically a state of the system, you know, how they progress.
we have this statistical machine called Epsilon Machine, and how they generate more structure on top of one another is basically based on Crutchfield's account of statistical complexity. Are the other readings that you posted, Reza? I couldn't find those, all the other reference materials? So is that just... Well, I have them. I can... Well, we need to find a solution because apparently New Center can't put, you know, books secured
by sordid means from, you know, pirated websites online. So if you guys have some sort of, I don't know, Dropbox, I can put them in Dropbox and I can share them with you. I have all of those materials. Thank you. That'd be great. Sure. Kim and I actually have a small list of questions, but I think we'll just cherry pick at various points at the seminar as opposed to just pludgy with a whole bunch of questions at once. But I was interested in our, what is a complex system reading, Leiderman. I was interested
in the distinction between a hierarchy in a social system and the hierarchies that you see or that you get in physical systems where he makes a distinction based on the fact that for instance a CEO doesn't actually have Nestor within her employees, for instance at the top of the tree. I guess I'm curious about the fact that perhaps in the folklore, so to speak, of complexity theory or things around it, I was confused in the past by the opposition between kind of horizontality and verticality.
Sure, yes. Yeah, I know, I know, yes. basically when specifically I mean basically all of complexity theory you know theoretical complexity sciences they have an account of hierarchies whereas you know when you look at the kind of pre-gogan or any of these you know, kind of cognitive embodiment, you know, extended cognition, these kinds of stuff. You see, or even, you know, a continental philosophical approach to complexity, you
see basically it's just kind of Deleuzian horizontalism. And so, yeah, they need to be distinguished because the kind of hierarchy that complexity theory talks about, well, it can come in varieties of forms, time-scale hierarchies, length-scale hierarchies. Like, for example, at a level, basically, a scale length is just the idea that, for example, you have a certain magnitude that covers a range of behaviors in a physical system. For example, a crystal, you can have it at a molecular level, a crystal level, atomic level and then basically the hierarchy is simply you know covering these length scales these magnitudes as I'm here yes and or you can have it
simply as an account of how structures and functions are distributed in a system and nevertheless in all of these accounts in when they are talk about talking about hierarchies hierarchies have a specific characteristics one is the nestedness that basically you have components in the system and they are interacting in a nonlinear fashion and the way that this behaviors of models they iterate you know they the kind of behavior of the system and then they
embedded into another form of this interaction this embedding basically this idea of iteration and embedding creates you know this kind of nested hierarchical structures whereby the interactions between components change because of this embedding forms of because embeddings usually you know the whole idea embedding embedding is a form of encoding you know kind of different varieties of behaviors across the system according to the structural component responsible for these behaviors. A good example of how basically, what is basically how they
come up with these hierarchies, and basically why is that it's important to have an account of hierarchy to, for example, talk about a biological structure and how it functions non-classically is a text by Giuseppe Longo it's called I think biological complexity and epistemic organization I can find it and I will put it online but nevertheless yes so the nested nest embedding another thing is basically that they are decentralized. Another, and this decentralization means both in the classical sense that they don't have
a fundamental, basically, law that governs them, but also this idea that, you know, within nested hierarchy, there are different, basically, hierarchies of structural and functional constraints. And these can be either be on top, you know, at the mesoscale or at the microscale. And these can exert their influence on different, basically, structures distributed at a completely different level of the thing. For example, you can have a functional constraint
being exerted on a meso-scale. A meso-scale structure can influence its functional constraints on the lower levels, and so on and so forth. So there is no such a thing as a linearity of this verticality. It basically can the functional constraints and structural constraints can basically exert influences on different levels regardless of the time scale of emergence of the system. So the things that come first don't essentially basically guide the entire behavior of the system.
something that can come later, can in fact exert a functional constraint on the activity of something that came earlier, basically in the emergence of the system. Yes, I will talk about this a little bit further today. But yeah, I just wanted to talk about that you have embedding, you have nested... Basically nestedness is part of the embedding. You have basically the nestedness, you have this non-centrality, and you have basically varieties, different families of structural and functional constraints, distributed quite
counter-intuitively with regard to the emergence, time scale emergence of the system. So something that can come earlier doesn't essentially mean that it's basically put fundamental functional and structural constraint on something that comes later. Another thing that, I will talk about this a little bit today, something that, you know, is in fact, I think, limited about hierarchical modeling of complex system is a problem known to linguistics,
but also to people who study cognitive science, is the idea of basically the way that all of these models, hierarchical models, namely epistemic hierarchies, as tools of studying different behaviors. All of the things, something that the majority of them have in common is that they use models that basically work with iteration, recursion, and embedding. Now, once you have these kinds of generative mechanisms to create a hierarchical model,
You also basically fall in the trap of creating certain epistemic biases. How does this work, for example? You see, if you have, for example, if you have certain epistemic commitments to a metaphysical, view of your system a for example let's call it how you see this fundamental structure or these component these fundamental components works and how they interact if basically so the whole idea is that you try to epistemically
you know model these and then you try to generate them via recursion iteration and embedding to create more different kinds of interaction between components, different kinds of hierarchy, so on and so forth. What happens is that once you exercise these generative mechanisms to your epistemic model, your rudimentary epistemic model, then it can smuggle basically all your epistemic biases, if you have any, to all the other hierarchies of your epistemic model. Basically, that's the whole point, that hierarchies create long-distance rules.
And these long-distance rules are quite powerful in transmitting biases that might be quite, in fact innocent, kind of like these massive kind of epistemic metaphysical assumptions covering your entire epistemic model. I will talk about this more in detail and how this works. But yes, so there are both kind of limitations with this hierarchical model, also they are kind of you know it allows you to be able to look into the system in a more diversified manner. Brilliant, thank you.
Okay. So let me... So, we talked about, you know, epistemic difficulties in measuring and studying complex systems. And basically two of the epistemic problems that we talked about were, one of them was
this idea of, you know, the mismatch between the model and the system, basically absence of a straightforward mapping. And insofar as there is no such a thing as this straightforward mapping, complex models for a sudden, you know, nonlinear dynamics in actual systems, they become prone to either creating too much complexity or lack of complexity or basically
completely fall off of the trajectory when they are modeling a system. Another one was the idea that we talked about that basically all of the complexity in models involve measurement and measuring, and that measurement and measuring models themselves inject basically disturbances into the model to the point that the model trajectory starts to diverge drastically from the phenomenon that it attempts to study. And another thing that we talked about was that, you know, basically the difficulty of
predicting the behavior of complex systems is not really about that positive global Lyponov exponent, but it's really because of the loss of linear superposition property between the components of the system. And we also talked about the difference between epistemic uncertainty and antics uncertainty. So the case of forecasting individual trajectories of complex systems received the most attention in discussing the implications of chaos and complexity for predictability.
We said that even under the perfect model scenario, no matter how many observations of a system we make, we will always be faced with the problem that there will be a set of trajectories in the model of a status space that are indistinguishable from the actual trajectory of the target system. Even for infinite past observations, we cannot eliminate the uncertainty in the epistemic estates given some unknown antique estate of the target system. One of the important reasons for this difficulty was traced back to faithful model assumption
that I talked about at the end of last session. For example, suppose the nonlinear modelist data space of a weather forecasting model is a fateful representation of the possibilities lying in the physical space of the target system, say, for example, a weather condition over Central America. No matter how fine-grained we make our model state of space, it will still be the case that there are many different states of the target system.
When I talk about target system, I mean it in the sense that then it would be an antique state, that are mappable into the same state of the model state of space, epistemic states. This means that there will always be many more target system states than there are basically model states. Now, the constraints nonlinear systems place on the prediction of individual trajectories, you know, of course, don't lead to impossibility of predictability of systems exhibiting complex behavior. dispel doom for predictability of such systems.
However, and why is that? Because there are other statistical probabilistic forms of prediction that can be effectively applied to such systems. Of course, they have their own problems, and I will talk about their problems. For instance, one can apply various techniques for forming ensembles of initial states surrounding the assumed actual state of the target system, and evolve these ensembles forward in time to forecast the behavior of the target system with specified measures for uncertainties.
Moreover, faithfulness problems, this idea that, you know, your model basically corresponds to your target system, epistemic states to ontic states, for nonlinear mathematical models are problematic for standard approaches to confirming such models. We typically rely on the faithfulness of our mathematical models for confirmation or verification of their efficacy in capturing behavior of target systems. But when the models are nonlinear and the target systems complex, faithfulness turns to be inadequate for these standard confirmation practices.
a target system to be modeled, for example, again, weather over Central America, and invoking the faithful model assumption, there are two basic approaches to model confirmation discussed in philosophical literature and scientific literature on modeling. One is focusing on individual trajectories and the other one is following a strategy known as piecemeal improvement. The one that is about focusing on individual trajectories tries to put the focus on success
Refinements to the accuracy of the initial data fed into the model while keeping the model fixed. Now, the intuition lying behind this approach is that if a model is, in fact, faithful in reproducing the behavior of the target system to a high degree, refining the precision of the initial data fed to the model will lead to its behavior monotonically converging to the target system's behavior. In another word, as the uncertainty in the initial data is reduced, a faithful model's
behavior is expected to converge monotonically to target systems behavior. So invoking the faithful model assumption, if one were to plot trajectory of the target system in an appropriate status space, the model trajectory in the same space would monotonically become more like the system trajectory as the data is refined. So basically the idea boils down to the monotonic convergence between the model and the target system, between the behavior of the model and the behavior of the target system.
And this is when I'm saying monotonic, it's a monotonic convergence of the behavior. It's basically the monotonicity relation is between the model and the target system. It doesn't say anything about the model is actually monotonically built or not. Basically it allows a room for basically a complete non-monotonically built model, but But nevertheless, the assumption is that whether the model, how this model is built, its behavior monotonically should converge upon the behavior of basically the target system.
The second basic approach, and this is due to William Winsat. I suggested his book, his re-engineering philosophy for limited beings, piecemeal approximation of reality. Wimzat is considered to be one of the main figures behind a revolutionary approach to modeling complex systems. So it comes up with this idea of basically a piecemeal approximation. The idea to focus on successive refinements of the model while
keeping the initial data fed into the model fixed. The intuition here is that if the model is faithful in responding to the behavior of the target system to a high degree, then refining the model will produce an even better fit with the target system's behavior given good initial data. That's to say that if a model is faithful, successive model improvements, small approximation will lead to its behavior monotonically converging to the target system's behavior. Now, again, invoking the faithful model assumption, if one were to plot the trajectory of the
target system in an appropriate state space, the model trajectory in the same state space would monotonically become more like the system trajectory as the model is made more realistic. Now what both of these basic approaches have in common is that piecemeal monotonic convergence of model behavior to target system behavior is a means of confirming the model by either
improving the quality of the initial data or improving the quality of the model. The model in question reproduces the target system's behavior monotonically better and yields predictions of the future states of the target system that show monotonically less deviation with respect to the behavior of the target system. In this sense, the idea of monotonic convergence to the behavior of the target system is a key mark for whether the model is a good one or not. If monotonic convergence to the target system behavior is not found by pursuing either of
these basic approaches, then the model is considered to be disconfirmed. Can you guys see the screen? Just to make sure. Yep. Okay. Yes. Now, it all sounds good, you know, in theory, you know, for linear, because for linear models, models, it's easy to see the intuitive appeal of such piecemeal approximations, incremental
strategies. After all, for linear systems of equations, a small change in the magnitude of a variable is guaranteed to yield proportional change in the output of the model. So by By making piecemeal refinements to the initial data or to the linear model, only proportional changes in model output are expected. If the linear model is faithful, then by making small improvements in the right direction, in either the quality of the initial data or the model itself can be tracked by improved model performance.
The qualifier in the right direction drawing upon the faithful model assumption means that the data quality really is increased or that the model really is more realistic. I mean, captures more features of the target system in an increasingly accurate way. is signified by the model's monotonically improved performance with respect to the target system. However, both of these basic approaches to confirming models encounters substantial difficulties when applied to nonlinear models exhibiting sensitivity dependence,
You know, sensitivity to initial conditions and also perturbations or disturbances. In the first instance, successive small refinements in the initial data fed into nonlinear model is not guaranteed to lead to any convergence between model behavior and the target system behavior. As I talked about in the last session, due to the loss of linear superposition, any small refinement, because any small refinement counts in this complex modeling as, in fact, a perturbation.
Any small refinement in initial data can lead to non-proportional changes in model behaviors, rendering the piecemeal approximation or the incremental convergence strategy ineffective as a means for confirming the model. Even a refinement of the quality of the data in the right direction is not guaranteed to lead to the nonlinear model monotonically improving in capturing the target system's behavior. A small refinement in data quality may very well lead to the model behavior drastically
diverging away from the system's behavior. Now, in the second strategy, the WIMSAT strategy, keeping the data fixed but making successive refinements in nonlinear models is also not guaranteed to lead to any convergence between model behavior and the target system behavior. Again, because of the loss of the linear superposition, any small change in the model, for example, by adding additional higher order terms into the equations, can lead to nonproportional
change in model behavior for the systems, for the same basically initial set of data. And in doing so, again, rendering the convergence strategy ineffective as a means of confirming the model. Now, even if a small refinement to the model is made in the right direction, there is, again, no guarantee that the nonlinear model will monotonically improve in capturing the target system's behavior. The small refinement in the model, again, may very well lead to
the model behavior diverging from the system's behavior. So whereas for linear models, piecemeal strategies might be actually expected to lead to better under-confirmed models, presuming the target system exhibits only a stable, again, linear behavior, there is no such an expectation, there is no such a guarantee. In fact, it's not even justified for nonlinear models exhibiting sensitive dependence deployed
in the characterization of nonlinear target systems. I mean, even for a faithful nonlinear model, the smallest changes in either the initial data or the model itself may result in nonproportional changes in model outputs, an output that is not guaranteed to move in the right direction. if small changes are made in the right direction. And this is one of the fundamental difficulties with confirming models of basically designed or built to track behavior of complex systems.
Now, sticking with the individual trajectory approach, one might consider alternatives to these piecemeal confirmation strategies. One possibility is to turn to a Bayesian framework for confirmation. But the thing is that similar problems arise here for nonlinear models exhibiting sensitive dependency. Given that there are no perfect models in the model class to which we would apply a Bayesian scheme, and given the fact that imperfect models will fail to
to reproduce or predict target system behavior over time scales that may be long or short compared to our interests, there again is no guarantee that some kind of systematic improvement can be achieved basically for nonlinear models. Some of these alternative confirmation strategies have been put forward by Leonard Smith. The maintenance, you can see the link, is a very interesting statistician, and basically
his main field of research is on model confirmation, modeling, and especially with a, you know, kind of a main emphasis on Bayesian frameworks for model confirmation. And that paper especially that is kind of a long essay has some, you know, concrete examples, It's kind of rather technical, but they are accessible enough to see basically some of these alternative strategies work. Now another approach is to seek trajectories in addition to the kind of the standard Bayesian
framework. Another approach is to seek trajectories issuing forth from the set of initial conditions in the model space. Presumably the actual state of the target system has been mapped into this set. So it's to seek trajectories issuing forth from the set of the initial conditions in the model space that is consistent with all observations of the target system over the time period of interest. For example, given a faithful model, choose an initial condition consistent with the observational
uncertainty that then yields a model trajectory passing within the observational uncertainty of the desired future observations. Models in this sense can be said to be better or worse depending on the length of their shadowing time, tracking basically the emerged behavior. such trajectories that consistently shadow target system observations for longer and longer times under changes in either the initial data or the models themselves may be quite
difficult. In addition, it's possible to construct models that can shadow any set of observations without those models having any physical correspondence to the target system. So basically the shadowing strategy lacks uniqueness, basically. you know, is basically in terms of correspondence, you can have many arbitrary models that essentially they can shadow the observations of the target systems, but basically the correspondence
between the model and the target system might in fact be quite arbitrary. Now it seems that the probabilistic models utilizing ensembles of trajectories or ensembles of probability distributions would allow for a clearer sense of confirmation. Yet similar problems can crop up basically in this kind of strategies as well.
And symbol forecasting models can give unique but incorrect indications of the target system's future behavior, or such models can give no unique indications of expected future behavior. And even when an ensemble model in fact gives a relatively unique indication that tracks with the outcomes of a target system over a shorter time scale, its indications may diverse significantly from that time point forward. And this is important, and we'll talk about this a little bit in terms of a kind of a
a little bit more intuitive example of the kind of problem that it causes. So again, you know, we face difficulties with formulating a systematic confirmation scheme. Well, you know, these difficulties with determining when a nonlinear model is good causes problems for philosophers' desires to produce a systematic theory of confirmation for all models. Of course, the situation doesn't impede physicists and others from finding ways, basically, to
improve their models and make determinations about how good their imperfect models are. However, it's also the case that these model builders do not follow some universal, basically, a scheme of conformational model building or a scheme for improving their models. But basically, they use varieties of techniques. They use not only varieties of techniques, they use basically different, basically formalism frameworks to be able to frame these time scales of behavior of the system within those
formalistic frameworks. So they have not only different techniques, they have different basically use of different hierarchies of formalisms, and each of these formalisms basically only have a limited range of coverage, for example, a particular aspect or dimension of the system. Not only they have these kinds of techniques and formalisms and conceptual frameworks, but also they have different ways of basically modeling or studying the models they use to study, you know, a physical system.
And I will talk about this a little bit later. I mean, any of you are familiar with the concept of toy model? No one? A toy model? A toy model, yes. Yes, just read that. Who was that who wrote the toy universe? Yes, yes, I sent it, yes. It's a new birth, yes. New birth. So basically, toy models are, I will talk about this, I mean, we shouldn't diverge too much. Toy models are called explicit metatheories.
They are these form of collapsed models that allow you to reinterpret the underlying metatheoretical assumptions behind models and the kind of observational, interpretational, you know, assumptions they have. For example, the best way to talk about this, you know, when we are talking about in a kind of a specific model, whether it's theoretical or it's actually a model designed to track a physical system, a complex system, These models are based on theories, and every theory in a kind of scientific framework is
in fact a meta-theory. It has meta-theoretic assumptions. Now the thing is that with these models, with an actual theoretical model or a model designed to track a target system, all these models have implicit metatheoretic assumptions. And it's like this, that they can't, when you are using the model, when you are either the theoretical or the one that has been built to specifically track a particular complex
system, their observations, their insights, the kind of problems they encounter, they basically can't take, because they have commitments, they have commitments to the kind of that metatheoretic frameworks. They can't step outside of these, you know, metatheoretic commitments and try to solve these observations. In fact, the model, as I said, the model kind of smuggles these metatheoretic assumptions, and because of this, you know, extending, to solve observations simply by having recourse to its metatheoretic assumptions, it encounters
continuous problems in explaining problems encountered, anomalies, observational anomalies, theoretical puzzles that, for example, a certain theoretical model yields. And because of this, you know, you can't really work with, you know, not that you can't. You have to find a way of rendering explicit the implicit metatheoretic assumptions of these models. This way of rendering explicit in quite like a Brandomian fashion, rendering explicit the
implicit metatheoretic assumptions behind the conceptual theoretical resources of a certain model is done by way of a model of models called a toy model. So a toy model is, first of all, a proper explicit metatheory. Basically it tries to basically render explicit all these implicit metatheoretic commitments of theoretical insights and ideas and observations that a regular, a standard, implicit metatheoretic
model utilizes. And by rendering these implicit meta-theoretic accounts explicit, it is capable of reinterpreting these ideas, insights, and observations in a completely different framework than the one offered according to the meta, basically the implicit meta-theoretic framework of the model that yielded, for example, this kind of observation, this kind of idea, this kind of insight. For example, a good way of thinking about this, you see, you know, when we are talking
about how to construct an AGI, you have different ideas and you have different insights and observations. You have different models. For example, the statistical framework of machine learning. You have the syntactic framework of the classical program of AI, which is the syntactic picture of the mind and then you have other companies you know the linguistic computational framework of you know semantic computational semantics another thing is that when these models try sorry you AGI as in
artificial general intelligence right yes yes yes yes go so when when these models try to come up with a basically constructing, or talking about AGI, or talking about intelligence, what counts as an intelligence. They come up with basically the tools of these programs, for example, machine learning, for example, the syntactic program, the computational linguistic program. The tools and the concrete ideas of them are robust. In fact, they are extremely successful. But nevertheless, they have basically underlying metatheoretic assumptions. And
these metatheoretic assumptions, once you try to basically use them in order to construct AGI, for example, by way of their successful tools, these basically metatheoretic assumptions are inflated, basically. The whole idea is that metatheoretic models, implicit metatheoretic models, if they don't have a way of reinterpreting their own ideas within a wider framework, They give rise to inflationary models. So down the line, what you get out of these successful, nevertheless local models are
inflationary accounts of intelligence and the mind, basically. Which you can simply make the entire cognitive edifice simply by way of statistical inference or machine learning, or for example, in the case of classical program of AI, you simply can create semantic complexity of cognition by way of syntactic decomposition, basically. Or you can do the same thing about, for example, some of the stuff about the computational linguistic semantics. So they basically, if you do not keep your meta-theoretic, your underlying meta-theoretic
commitments under check, you are prone to giving rise to inflationary models. And what happens with these inflationary models is that sooner or later they confront observational anomalies. For example, we are talking about in physics, or they confront with fundamental puzzles and setbacks. For example, you see it in AGI and artificial intelligence, and this idea that people get extremely overexcited, that we are really close to making this, but as they move forward, they see that there's
actually fundamental setback that brings the progression of the model to a halt. So basically, toy models, and the same thing about physics, basically toy models were developed in the field of quantum physics, to study and make explicit these implicit metatheoretic assumptions underlying observations, problems, so on and so forth. And then once you make these metatheoretic assumptions explicit, then you enable the
enable, basically you allow for reassessment and reinterpretation of ideas, insights and observations, local ideas, that were otherwise successful and in fact need to be treated as kind of developments and breakthroughs, but nevertheless, once simply extended within their implicit metatheoretical accounts, they led to fundamental setbacks. So basically toy models try to rescue the successful components of models. And there are ways of constructing
toy models. I will share some stuff. They are very interesting and they are extremely school to reinterpret the range of our models, basically. The range of application, their theoretical range, reinterpreting their components, so on and so forth. Okay, so before moving forward, any question, any discussion, anything? Nothing.
With the toy models, it sounds like it's a prototypical sort of version with that toy model tag. Is there an example or something where you can sort of, I don't know, tease that out a bit? Because is it that it's got a smaller, like, data set or...
Like, I'm just having trouble tying it to an example. Yes, you're right. You see, toy models are coming, you know, in physics for example, they come in as small toy models and big toy models. Now the whole small toy model are simply, you know, the couple, what you would call a prototypical or what they call it a collapsed model, simply truncations of varieties of data sets and models that you try to reinterpret. Now the thing about it, and these are, yes, they are fairly rudimentary, but for example in physics, The kind of toy models that they work with usually are big toy models.
And the means of construction, I don't want to get into this, the means of construction of this model is by way of formalism called a CHU space. It's basically a category theoretical reinterpretation of your existing data by regrouping these data and be able to basically study their objects, the objects that these categories, namely these models, have. And you have basically, you can, it's basically, it's completely, I mean, when you look at the best way of thinking about this, how a big toy model made is precisely is, you have a
set of Legos with certain colors and these pieces represent you know objects your observations you know your components of models a smaller data sets so you decompose them and then you try to simply built different models different basically different construct different categories of models from this given set of pieces. Instead of having put them, you know, you have made all of, turned basically all these pieces into making a house in your Lego model, you try to take them apart and basically make different things with them rather than just, you know, one model basically.
So it's just, it's, toy models, you know, offer a number of, you know, advantages over particular model building. One is a theoretical arbitrage over your set of data. Basically, they allow you to not put all of your eggs in one basket. Another one is basically, as I said, once you reconstruct according to different, this kind of category theoretical construction, reconstruct your components and put them into different categories or different constructs, different explicitly meta-theoretic constructs,
then you are able to see these basically components, these Lego pieces, in light of a completely different interpretation. So basically they allow you to play with them, play with your ideas and components, your basically rudimentary sets of data within completely new frameworks. Nevertheless, the whole point of the toy model is exactly like this, that they are falsifiable. They are not supposed to be applied to anything. They have only been designed in order to be broken in the real universe, but not before you have systematically played with them, exactly like a toy. So it sounds like they're deliberately low fidelity
models, right? So it lets you explore the problem space cheaply by showing you can rapidly try out these different configurations of the low fidelity models. Yeah, the models come in low fidelity range, but also the big toy models, the ones that are constructed by Chua space, for example, the big toy models of physics, they try to faithfully accommodate classical, mechanical classics and quantum mechanics faithfully. So they try to faithfully capture the interpretations of that theoretical framework and the other theoretical framework. But as you basically, yes, as you move down to, the more collapsed
your toy model, the more truncated and rudimentary, the basically the lower fidelity of your toy model. But the more, you know, your construction is, you know, vast and you have basically, you put the constraint on how you can construct this toy model, the higher fidelity. Basically, the more constrained the toy model, they are the most useful ones, because they allow you to accurately represent, basically, the underlying theoretical frameworks. But yes, yeah, there is, there is basically the whole point is to create this balance
between low fidelity and high fidelity, but nevertheless allow you, the whole point is to emphasize on constructability and reconstruction. And it sort of brings the act of modeling into view, right? Yes, makes it explicit, yes. Because the whole point with models is that people, when they use models, they bring, as I said, implicit meta-theoretic assumptions into their models without knowing it. Their modeling has basically massive sorts of, you know, these meta-theoretic assumptions put together and what toy models want to do is by this act of reconstruction and
recomposition it tries to again bring that those assumptions to the foreground in this context I'm not sure if this is appropriate but I'm wondering about yeah approaches perhaps like this drawing on your notion of play where the deformation perhaps or the intervention or interference is actually kind of emphasized, foregrounded, maybe so as to surface these metatheoretic conditions. Can you elaborate this a little bit? A little bit. I think I'm just drawing on your notion of play, you know, recomposing, reconstructing
in these different contexts perhaps, and wondering about, you know, so far you talked about the notion of faithfulness. Are there approaches based on unfaithfulness where, does that make any sense even? Well, I mean, the whole point is that you see faithfulness, when we are talking about faithfulness, we are talking about simply a model that was designed to be applied to a target system, to a particular antica state. So you need to have the criteria of faithfulness. Yes, now the toy models is that, as we're just talking about, that they also, they do
not have, not that they don't have it. Basically this constraint is not as important for them as it is for a regular model designed to be applied into a target system. So yes, because they don't have this kind of constraint fully in place, then they allow for you to use basically modal vocabularies inside your modeling system, counterfactuals, hypothetical scenarios. With respect, Adam mentioned the cost savings of using a toy model.
Is that in a formalizable or specifiable way true where you're trading a certain amount of noise or extra information that would come with a full-scale model in exchange for a certain cut in thermodynamic costs? like in computation time or in the amount of noise generated in one place in the modeling or computation process versus another? Hard to say really. I'm sure there should be basically a cost analysis of a toy model construction versus regular model construction. But I have an actually superb question. I have not thought about this. But I assume so, yes.
There is a, you know, the whole idea is this, there is a trade-off here. But the thing is that, you see, as I said, the toy models are not essentially collapsed models. They come in varieties and, you know, the most canonical variations of toy models are the collapsed forms and the big forms. The big forms don't essentially mean that you are really reducing the cost. In fact, the concept construction of the kind of big model that you are using might be extremely higher than a regular model that basically this was originally worked with. Those are completely different kind of models. But another thing
is that these models, as I said, have been designed to reinterpret models. They are not designed to be applied to anything. They are simply reinterpreting tools, reassessment tools. They are basically models for studying models. Right. Like the elastic band in relativity. Like the planet going around the sun. In fact, in toy models, basically in physics, they are completely basically made by way of geometric algebraic approach. You know, basically, and I said heavily use of category theory, topos logics, you know,
bring all of these things together and faithfully, as I said, faithfully represent, for example, classical mechanics, quantum mechanics, they try to faithfully represent. Whereas the small toy model, they want to have rapid reconstruction and effective reconstruction by basically eliminating, almost eliminating, the faithful representation requirement. Whereas in the big toy models, you do not have this elimination of faithful representation. You need to keep that, in fact. But nevertheless, they give you, as I said, they give you a still, because you are basically
connecting different worldviews together, different models together. They give you a constructive edge over your typical, basically, model view, whether it's classical mechanics, quantum mechanics, so on and so forth. Would some of the more brute force AI methods of winning games, and from what I've gathered, the whatever it's called, deep thinker, or whatever that want at Go is one of them, would it be considered one of these large, faithful toy models when instead of like quote-unquote playing the game, you have a gigantic memory of game states
that you're able to search through in a tree structure, and so you're able to pretend to play chess or pretend to play Go because you actually remember almost the entirety of the state space of Go or chess, and you're using that to replicate playing behavior? Is that something that sort of functions like an operative? I know you're not supposed to apply it to anything, but I don't know, are those comparable at all? Search spaces, I think, yeah, they are search spaces, but the thing is that with the toy model, they try to, you know, construct, yes, I think it's a good thing that if you represent different, all moves of your chess, or at least not all moves,
but let's have, like, you know, limit this thing, you know, good, robust, you know, account of the moves that you can make, you know, for example, a small board game, and you call them, you know, basically search spaces. And so, yes, I think toy models can be said to be constructing pathways inside these possible search spaces. They basically, yes, I mean, the whole point is that toy models do not come up with a new, basically, database. They simply try to use what they have, but in a decomposed way, basically different status
basis of every basically things that was, you were working with. Decompose it into completely a searchable form, and then once you have it in a decomposed form, try to reconstruct back into these small local, not global, local basically models. And that's the whole point of toy models. Toy models only work with local models. There is no such a thing as a global toy model. When people were saying that the big toy model, it means the way that basically tries to construct itself from these faithfully represented global models. So the whole idea of a toy model is that it turns the global into the local.
If you have a global model and you use it in a big toy model, it will turn that into a local model. And basically that was the thing that I talked about. They are designed to reduce inflation inside your model system. If you have inflation that was caused by this overextension of underlying metatheoretic assumptions, basically they try to deflate it. and bring it basically to, because the whole point is that once you work with regular canonical model, at some point if you do not keep this method, theoretic assumptions under check,
at some point your model gets inflationary. Basically you try to model behaviors or, you you know, that absolutely have nothing in common with the meta-theoretic assumptions or meta-theoretic formalisms behind your model. So in terms of not being able to have a global model, so would like, you can't, you You couldn't have a toy model of the entire standard model in physics would be an example. You could only have specific toy models of like boson-gluon reactions or all of the particles
but no anti-particles or something along those lines, but you could never have a toy model of the entire thing. I think you can have more of the toy model of the entire thing, but as I said, it depends on basically this low fidelity versus high fidelity toy model. In the collapsed form, toy universes, basically you can have everything you want. You can have everything because there is no constraint of faithfulness here. When I say faithfulness, I do not mean faithfulness to the actual universe. faithfulness to the model that tries to picture the universe, the theoretical framework.
As I said, toy models are not models of anything. They are models of models. So the criteria of faithfulness is simply the model that have those conceptual components integrated within the system. For example, classical mechanics. You have concepts of classical mechanics, observations and insights of classical mechanics. So it needs to be faithful or not faithful at all to the theoretical framework of classical mechanics, not the world that it tries to represent. Right. Okay, any other suggestions, comments before we move forward?
I don't think so. Okay. Let me share this screen again. I see that Sean is saying toy models should look like a subset of the target model. Yes, I think it's a good definition. Yes, not subset of the target system that the model tries to track, but subset of the target model. And they are not usually, the whole point is that they are not simply, they don't use
just one model. They are basically, you know, use a model pluralism framework. They use different models. Basically the whole thing that they need to get components from different models and put them together. Okay, can you see it? No? Sorry.
modeling and not only modeling but also confirming models and basically whether they are actually faithful or not well we can think of some of these problems in terms of you know their implications for the use of nonlinear models in the development and assessment of public policy you know you know policy formation and assessment often utilizes model forecasts. And if the models and systems lying at the core of our policy deliberations are nonlinear, then policy assessment will be affected by the same lack of guarantee as model confirmation
due to the loss of linear superposition. Suppose for example, government officials are using a nonlinear model in the formulation of economic policies designed to keep GDP ever increasing while minimizing unemployment among achieving other socioeconomic goals. While it's true that there will be some uncertainty generated by running the model several times over a slightly different data set and parameter settings assume policy taking these uncertainties into account to some degree can be fashioned. Once in place, the policies need assessment
regarding their effectiveness and potential adverse effects. But such assessment will not be merely a function of looking at monthly or quarterly reports on GDP and employment and unemployment data to see if targets are being met. The nonlinear economic model driving the policy decisions will need to be rerun to check if trends are indeed moving in the right direction and are on the right course with respect to the earlier forecasts. But of course the data fed into the model have now changed and there
is no guarantee that the model will produce a forecast with this new data that fits well with the old forecasts used to craft the original policies. The question then, how then our how then are policy makers to make reliable assessments of policies? You know, the same problem that the small changes in data or model in nonlinear context are not guaranteed to yield proportionate model outputs or monotonically improved model performance, you know, also basically plague policy assessment using nonlinear models.
So you know, I think it would be, as I put the assignment for this whole module on the classroom page, it would be great actually, you know, talk about some of these problems in concrete examples. For example, coming up with a kind of a model of, you know, a concrete model, whether it's economy and policy making, in military, and see basically the way that basically these models are presented, they are not basically free of serious problems of interpretation
and epistemic basically quandaries, conundrums, dilemmas that in fact the people who are trying to present these models are either completely oblivious to them or basically they are not simply interested to talk about these. But all of these, as I said, all of these have ramifications not just for how you make the model, but also how this model behaves and how you are actually tracking the behavior of a target system.
So we talked about modeling and epistemic, ontic uncertainty. Now we can talk a little bit about determinism in complex systems. Intuitively, one might think that if a system is deterministic, then it surely must be predictable. But the relationship between determinism and predictability is much more subtle to support this intuition.
Predictability of systems has much to do with epistemic estates, while determinism has to do with ontic estates. And while the characteristics of ontic estates should have some implications for the character and behavior of epistemic estates, it is difficult at best to draw any conclusions about the ontic estates of a system based on our access to epistemic estates. This is a kind of a fundamental reason why the often-discussed unpredictability of chaotic and complex system by itself does not undermine the determinism of the underlying ontic estates
of nonlinear complex systems in classical mechanics. So arguments, you know, like Karl Popper's, you know, to the effect that a breakdown in the predictability implies a breakdown in determinism, trade on this conflation of epistemic conclusions and antique conclusions, basically epistemic states and antique states. The trend since the early scientific revolution has been to support the belief in metaphysical determinism by appealing to the determinism of theories and models from physics.
So this strategy is not without subtleties and surprises. A standard way of characterizing mathematical model as deterministic is through the property of unique evolution. So what is unique evolution? A given state of a model is always followed, namely preceded, by the same history of state transitions. The basic idea is that every time one returns to the mathematical model to the same initial estate or any estate in the history of estate transition, it will undergo basically the same history of transitions from estate to estate and likewise for the target system
if the faithful model assumption holds. Now, in other words, the evolution will be unique given a specification of initial and boundary conditions. For example, the equation of motion, you know, of a frictionless pendulum will produce the same solution of the motion as long as the same initial velocity and initial position are chosen. Now, is it not uncommon to find arguments in scientific literature that claims
to show that chaos and complexity tell against determinism. For example, John Polkinghorne has pushed this claim that the kind of sensitive dependence exhibited by complex dynamical systems should lead us to view even the deterministically rigid world of classical mechanics as ontologically indeterministic. Now, this is a quote from him. The apparently deterministic proves to be intrinsically unpredictable.
It is suggested that the natural interpretation of this exquisite sensitivity is to treat not merely as an epistemological barrier, but as an indication of the ontological openness of the world of complex dynamical systems. Now, he attempts to make this line of thought plausible through demanding a close link between epistemology and ontology under a critical, realist reading of the two. Now if we remain at the level of dynamical system, i.e. mathematics, mathematics of dynamical systems, then clearly there is a serious problem with this kind of reasoning. Namely, the mathematical
equations giving rise to exquisite sensitivity and attendant predictability problems are deterministic in exactly the sense of unique evolution described above. So our ontic description in terms of these equations push in precisely the opposite direction that Polkinghorne, you know, pursues. It's true that apparent indeterminism can be generated if the status-based one uses to analyze chaotic behavior is coarse-grained. This produces
only an epistemic form of indeterminism. The underlying equations are still fully deterministic. Instead, to bring these questions about determinism of real-world systems, one has to pursue the nature of these complex models and their implications, as well as, you know, kind of investigate the presumed connection with the target system via the fateful model assumption. Now, as I talked about in the first session, the mathematical modeling of actual world
systems requires us to make distinctions between variables and parameters, as well as between systems and their boundaries. These distinctions become problematic in the context of complex systems, where linear superposition fades, and such systems can be exquisitely sensitive to the smallest of influences. Such features raise these questions about our epistemic access to systems and models in the investigation of complex systems. But they also raise questions about making
sense of the supposed determinism of target systems. You know, an example again, considering applying a deterministic mathematical model to forecasting the weather over Central America, where the identity and individuation of that system is questionable. Then the question would be, what all do we have to do to include in this model to be able to make some reasonable pronouncement about whether Central America's weather is deterministic or not. Do we need only to include a particular fluid mass over the particular continent or over the Earth's surface, or that plus, you know, that's that, stratospheric,
magnetospheric, and so on, you know, boundary conditions, so on and so forth. So there is a further problem in our application of deterministic models to actual world-complex systems and our belief that those systems are deterministic. Although the faithful model assumption appears fairly unproblematic in some simple context, if the system in question is nonlinear, the The faithful model assumption raises serious difficulties for inferring the determinism of the target system from the deterministic character of the model.
For example, there is the problem that there will always be many more target system estates then there are model states as described in terms of the correspondence and scope of the model to the actual system. More fundamentally, there is this problem of, as I just said, the mapping between the model and the target system itself you know even a faithful model we don't have a guarantee that the mapping between the model and target system is one to one
you know as we customarily you know usually assume the mapping may actually you know be you know many to one or for example different nonlinear faithful models of the same target system, as is the case with competing weather forecasting and climate prediction models. Or the correspondence can be many-to-many relationship. For so many of the classical mechanics problems, namely where linear models or force functions are used in Newton's second law, the mapping within the model and the target system appears to
be a straightforwardly one-to-one with plausible empirical support. By contrast, in nonlinear context, where one might be constructing a model from a data set generated by observing a system, they are potentially maybe as empirically adequate to the system's behavior—sorry, there are potentially many nonlinear models that can be constructed. And each of these models can in fact be empirically as adequate to the behavior as any other. For the inference from the deterministic character of our mathematical
model to the deterministic character of the target system to hold appears to require either a one-to-one relationship between a deterministic model and target system, or that the entire model class in a many-to-one relation to be deterministic. Now there is a different approach attempting, you know, to bring attention to the ontological determinism of the macroscopic world into question via complexity is basically the research on far from equilibrium systems put forward by Ilya Prigogin in his Brussels Austin group.
Physical physics usually describes physical systems using particle trajectories as a fundamental explanatory element of its models. If a system of particles is distributed uniformly in position in a region of space, the system is said to be in thermodynamic equilibrium. In contrast, a system is a forefront equilibrium, namely non-equilibrium, if the particles are arranged so that highly ordered structures appear, basically like a good example of a cube of ice floating in T. This means that the behavior of a model is derivable from
trajectories of particles composing the model. The equations governing the motion of these particles are reversible with respect to time. They can be run backwards and forwards like a film. When there are too many particles involved to make these types of calculations feasible, coarse-grained averaging procedures are used to develop a statistical picture of how the system behaves rather than focusing on the behavior of the individual particles. In contrast to Prigogin, Brossel, Austin's approach, views these systems in terms of models whose fundamental explanatory elements are distributions.
That is to say, the arrangements of the particles are the fundamental explanatory elements and not the individual particles and trajectories. The equations governing the behavior of these distributions are generally irreversible with respect to time. In addition, focusing exclusively on distribution functions opens the possibility that macros, non-equilibrium models are irreducibly indeterministic and indeterminism that has nothing to do with epistemic access to the system. Now if true then this would mean that probabilities are as much an ontologically fundamental
element of the macroscopic world as they are of microscopic. One important insight of Prigogin's approach is that it shifts the way from trajectories to distributions as fundamental elements in... basically as fundamental elements is that explanation also shift from a local context. So the whole idea that, you know, the basically, the shift
away from trajectories to distributions as fundamental elements is that explanation also shifts from a local context set of particle trajectories to a global context, distribution of, for example, the entire set of particles. A system acting as a whole may produce collective effects that are not reducible to estimation of the trajectories and sub-elements composing the system. Now, the thing is that, I mean, we can't get much into detail, but there are serious open-ended questions about this kind of view. You know, for example, what could be the physical source
of such indeterminism, and what is the appropriate interpretation of the probabilistic distribution? Now once you look into basically that there isn't no really explanation of what these things are, what these sources are, and what is the appropriate interpretation of the probabilistic distribution in Brussels-Austin's approach. And to the extent that there is no robust explanation or interpretation on these critical accounts, the whole Brussels-Austin's approach view on indeterminism and this collapse of the ontological epistemic distinction collapses
is, remains, you know, kind of highly speculative. Now let's get to the, you know, causation and either I'm going to talk about it or I I just maybe, in fact, maybe just put it on the classroom page, you know, so you guys can read it and have some thoughts about this and, you know, share your ideas. So I'm going to talk a little bit about causation in complex systems and the material that I'm
going to put in the classroom so we can talk about the beginning of the first session without that interfering with our progression is the kind of metaphysical, time-driven biases behind in five fundamental theoretical components of complexity sciences. Now, the analysis of determinism in complex systems is complicated by the fact that there are additional forms of causation arising in such systems that must be taken into account. Now, understanding whether a process is deterministic or not depends on understanding the underlying
causal mechanisms. There is no consensus account of what causes are. There are rather a set of accounts of counterfactual, logical, probabilistic, process-based regularity accounts of causation, structural. And each of these have their own strengths and weaknesses. And exactly like the definitions of complexity, they have different scopes of applicability for different purposes.
But nevertheless, even though there are different accounts of causation, but once you really look into basically all of them, almost all of them, including the statistical probabilistic ones, there are assumptions about time and the direction of time. especially using basically the low entropy point as the initial and then basically the progression of time as it progresses, then you have basically this whole low entropy point becomes for you as a kind of observational cue.
Basically, that's the whole point of causation, studying causation, is that it has, it basically it's based on this model of time and its correspondence with the patterns of observations. Patterns of observation, for example, at the fundamental level, the entropy levels in a system and then you try to then map these observations onto some prior accounts of time, a prior model of time. And this is something that we'll talk about this, we'll write a little bit on this for the classroom page. One person who really, you know, get into
this, the kind of metaphysical time-driven biases behind all of these and how you basically map observations and patterns, deduce patterns out of them to the intermediation of a model of time was Boltzmann, really. And this is like his basically Copernican view of time, the way that he discusses that time as having a flow is only a secondary quality, is basically only a secondary account of time. And that's not really, in fact, we should be able to come with...
And the whole point is that with the secondary quality account of time, the time with an arrow, you see that once you get into the nitty-gritty conceptual assumptions behind these models, flow of time, becoming, basically time with an arrow, you see that they are not false, but they are incoherent, in fact. And so that's why he argues for a black view of time, for basically a time that does not have an error. And that can be in fact made compatible with our secondary quality accounts of time, namely the flow picture of time,
with an arrow. And it gets kind of fairly complex, the discussions, but there are nevertheless, I think there are fundamental importance, especially people want to look into basically deep into the philosophy of science and the kind of metaphysical assumptions behind causation and how we talk about complex systems and so on and so forth is that if you do not really keep your metaphysical assumptions about time under check, then you are really prone to come up with all sorts of, as I said, inflationary models of causation, inflationary accounts
of basically determinism, function, structure, so on and so forth. A good example of this is this whole idea that if basically low entropy as an index of information and the way that basically you make an equivalent correspondence from that to basically a flowing account of time allows you to come up with this idea that something that comes earlier should in fact cause something that comes later. So basically
you have these kinds of accounts of causation that are simply, you know, basically based on these implicit irregularities in the flow of time. Something comes earlier, basically the cause should be sought, should be found in something that is coming earlier rather than something that comes later. So it does not allow for basically causal retroactivity. Now the thing is that once you, and that's why I suggest that Hugh Price, once you get
rid of the slow picture of time and adopt the block picture of time, then you in fact can come up with a robust account of causal retroactivity. And so many of these paradoxes that are in fact puzzling inside the kind of canonical arrow-based picture of time, like the time-traveling paradoxes, they can be solved within the block view of time. And Hugh Price covers a lot of, you know, discusses a lot of these, basically, these
puzzles and conundrums. So anyway, so back to, you know, the account of causation in complex systems, before, you we end our session, is that these accounts of causation required rethinking in the face of the richness of nonlinear dynamics. Chaos, complexity, self-organizations are behaviors where complex wholes play important roles in constraining their parts.
interlevel causation more generally has received little philosophical attention relative to bottom-up efficient model of causation. As I said, it's basically the whole idea that a bottom-up model of causation allows, it basically comes from a completely canonical, Absolutely, in the kind of traditional philosophical science text, unexamined model of time. Basically, everyone takes it to be simply true, because it looks like it basically matches
our observations. But the whole point is that the kind of the observation, that's really, I think, Boltzmann fundamental revolutionary view that the whole way, the reason that we do make these patterns of observations about how things progress is precisely because of a, is because of that secondary time as a secondary quality. But secondary quality time doesn't tell us anything about time as a primary quality drawing inference from the former basically doesn't yield any you know good robust metaphysically and epistemologically resolved insights about
the latter, namely the primary qualities. Time as a primary quality rather than as a secondary quality, which is basically the arrowed picture of time, time with a flow. Now, Immanuel Kant was one of the first people to recognize the peculiarities of what we Immanuel Kant was one of the first people to recognize the peculiarities of what we now call self-organization in living systems. He classifies such phenomena as intrinsic physical ends because they are in some sense both cause and effect of themselves. For instance, according to Kant, a tree in the genus as effect, now as cause, continually
generated from itself and likewise generating itself, preserves itself generically. This is from Kant. An entity is an intrinsic physical end if its parts, both as to their existence and form, are only possible by their relation to the whole. Self-organizing systems, particularly organisms, are produced by and in turn produce the whole. Each part, such as they are distinguishable, exists in virtue of the activity of other parts and the whole. Furthermore, each part exists for the sake of the other parts as
well as the whole. Again, this from Kant, only under those conditions and upon those terms can such a product be organized and self-organized being and as such be called a physical end. Now in Kant's view of self-organization, self-organization basically in Kant's approach to this kind of part-whole relationship, self-organization has nothing analogous to any causality now and basically to us. And causality now and basically to us. And that's basically, you know, Kant wants to come up, basically imply a new form of causation, a new model of causation. And because the dominant,
so he said, because the dominant concrete conception of causation available, and that's The reason that he says that, the reason that he says that, you know, this scheme of self-organization has nothing basically in common with the kind of causality known to us is because all Kant possess, you know, in terms of this conceptual toolbox, is basically, you know, conception of causation available to him derived from basically modern Newtonian mechanics so all he has is this kind of
you know understanding of you know causation and how you basically can account for causation in a system based on Newtonian mechanics now given his recognition that self-organizing systems require some kind of time-irreversible processes, and that Newtonian dynamics was fully time-reversible, he relegated our lack of understanding how self-organization comes about to a limitation of reason, namely counterfactual, logical, regularity analysis of causation, basically. He doesn't, basically, for him, the reason
that we cannot talk about these is basically not that there is an actual form of irreversibility in systems, but that our reasons are doomed to basically only be able to describe reversible processes, because the only thing that was available to him was that kind of physical model, physical system model. Now, modal and regularity analysis of causation fare no better at penetrating this lack of
understanding, this difference between causation irreversible and reversible processes. While process and structural accounts each appear to have some pieces of the puzzle for understanding self-organization, process theories lack an adequate account of the structural constraints of wholes and parts, while structural theories lack an adequate account of processes." This is very interesting that when you look into, for example, some of the text about mechanistic account of causation, generative mechanisms are these, what are mechanisms?
Mechanisms are these specific patches of causal fabric which are basically playing a generative role in basically generating, occasioning structures. Now, you see these kinds of, you know, limitations, especially in, for example, mechanistic philosophy. Mechanistic philosophy has a strong structural constraint of basically these part-whole relationships, but it does not have whatsoever a robust account of processes. You can read William Bechtel, for example, or Carl Craver.
They are extremely interesting. For example, they apply the new mechanistic picture of causation to, for example, models in neuroscience and biology. And you see there is no account of processes. And the same thing about those people who work with account of processes. For example, a good point of entry to process-based picture of causation, you can see it in Ruth Milliken or Johanna Zeit works on you know process ontology nonlinear interaction picture of causation as you know under the product the so-called
process ontologies of of causation and then you see that all you have are processes and interactions but without any structural constraints without any mechanistic constraints. So I think a fruitful way of approaching these accounts of causation is to see how much you can integrate these two accounts together, and kind of analyzing the kind of constraint that processes impose on structures, and the kind of constraint that the structures put on processes, basically.
You know, causation really, I think, has been the least studied concept in, basically, complexity theories and complexity sciences. in complex system has been given very little sustained analysis in the philosophy literature relative to causation, for example, in general, in terms of those canonical models of causation. But nevertheless, as I said, even in the classical models, not only that we don't have a robust
account of causation in complex systems. But even when we have a robust account of causation, according to that kind of classical models and physics, we do not have an account, we really do not know where it is coming from. And basically there is, as I said, there is this kind of underlying unexamined assumptions, basically how you map your model of time to your model of observations. And that becomes a pattern of causation for you, basically causal activity. And that hasn't been explored that much. So,
I will talk about, I mean, I will try to put, you know, write some comments on the class page room about causation and time and, you know, some of the stuff that usually crop up about the role of, you know, these counterintuitive pictures of causation in complex systems. I try to kind of come up with a few helpful comments. But also this also brings back the stuff that we talked about during the questions and answers about downward causation. And I kind of expressed a little bit of hesitancy when I was talking about Jake Won Kim's critique
of downward causation against the strong emergentic approach. The thing is that causation in complex systems is both against the strong emergentic approach and also is against the traditional causal accounts in classical views. One this, and the other point was that I think the kind of classical views of causation that are straightforwardly this bottom-up approach,
basically, they themselves, I think, they need to justify, basically, their position without rather taking it as basically something that doesn't need to be examined precisely because everyone has been using it. But why is that everyone has been using it? First you need to answer that, and also you need to come up with what exactly justify you to not going with a different model of time and accordingly completely a different model of causation. Just because something produces apparently positive observations doesn't mean that basically it is a robust theoretical model.
And I think that's why I was hesitant to kind of buy or kind of endorse Kim's critique of a strong emergence and downward causation. I mean, for me, they all have all of these accounts, from a strong emergence to complexity theory to, you know, kind of classical views by virtue of which someone like Kim, you know, challenges downward causation. They are all basically, I think, you know, have problems because of, you know, the kind canonical views of causation and time that underlies their conceptual frameworks.
Okay, we can go to questions and answers and then we are done for this session. Questions? You are screen sharing right now, Reza, just so you know. Oops. How should I get out of the screen share?
Just hit the screen share button, the green button on the left, the icon. Yep. Okay. Gregory, you had a question? Oh, was that block view of time, like B-E-L-O-C-K? Yes, block view of time, yes, yes. I mean, the thing is that, you know, it's not something new. I mean, philosophers talked about this, you know, basically ancient philosophers talked about this. Quite actually, that was a controversy between the Heraclitian picture of time and the Parmenidian picture of time. Parmenides was the one that came up with, you know, the kind of idea that in fact time is a blind,
and there is no such a thing as a flow. this flow is simply a pragmatic point of view from the point of an ideal observer. And you can, now the thing is that the great thing about the, you know, people like Hugh Price is that they try to render compatible the secondary quality of time, namely the flow and the block view of time, the primary quality. I mean, it's really interesting that the stuff that it gives so much, leads to so much confusion and so much, you know, assumptions, this whole idea of time and causation and how it is used to model the pattern of observation but also observing.
To the extent that some people say that we can in fact get rid of causation, I mean get rid of time and preserve causation. We have for example statistical frameworks of basically simply causation. But the whole point is that they also fall into some incoherencies. And not only that, but also the kind of that they, some of these people who also want, you know, to have causation, but not time, because they know that basically having a canonical model of time, in fact, puts you in a kind of a slippery road. So they want to get rid of the concept of time.
But the whole point is that I think as Hugh Price suggests that we need to come up with an integrated framework that allows for us to not have time as a flow, but also have time as flow, namely a secondary quality relevant to our observational, basically, capacities. Questions? Talks? I have another question, but this relates more to maybe some of the stuff in the last
but I might wait for someone else to... Okay, I'll go. This was around memory in a complex system. Am I right in understanding that memory has to do with the ability of a system to persist in a global state in spite of local disturbances or fluctuations? When you say memory, what do you mean exactly by memory? That was my, I think this was the term used in what is the complex system. You mean history of the system, that basically draws moving forward based on history, basically
the paths are history sensitive, trajectory sensitive. This was in the context, I think, of robustness in the Ladyman piece. And yeah, I understood to mean some kind of persistence of order despite... Yeah, but I'm not sure. I wonder if someone else can... Yes, no, when Ladyman talks about memory here specifically, it means it's a history sensitivity, yes. The idea that... Winsart calls it generative entrenchment.
Now, generative entrenchment means that you have, again, that kind of a bottom-up or the kind of hierarchical picture. It's the idea that, you know, in order for you to, you know, for example, let's think about this whole idea in terms of evolution. Once you have certain evolutionary organs in place, then the more structures come on top of one another, the more entrenched they get, they play a more significant constraining role for basically diversification of functions.
So, is that as the complexity grows in a hierarchical system, because of this entrenchment of structures that are already in place, and they basically, they have, they exert, you know, basically constraining influences. Because of this, the range of complexity of function, the range, namely diversity of functions that can be added to this basically evolved structure are significantly reduced. That's why, basically, hierarchical systems of this sort, of this kind of generatively entrenched, yield robustness, basically.
Yield robustness. Because they increase complexity and they reduce diversity of functions, or diversity of a structure. That's why, for example, a structurally stabilized organism like us cannot evolve into that kind of massive, drastic, you know, for example, morphogenetic variations. Like we can't give up certain functions that are completely diverging from what is already structurally and functionally in place. And because of that, these kinds of organisms like us, generatively entrenched hierarchical
systems, they do not evolve to diversity, but they evolve by basically designing a specific levels or a specific complexity levels that have been designed to do, basically to function as a platform rather than having a specific function. You see, when you have diversity of organs, you have basically these diversification in the evolutionary sense, basically address
you know, certain specific functions and functional criteria that needs to be solved. For example, a hand is basically for, you know, from a evolutionary perspective, does a range of functions, specific functions, and basically these functions respond to specific requirements and constraints. Now, the thing is that it's due a specific functional range, whereas the kind of complexity that generatively entrenched hierarchies give rise to are not designed to do a specific function. They are designed to function as a platform, namely supporting a massive amount
of functional diversification. So they can give rise to diversification of complexity, but they can give rise to this robustness of structures that can function as basically platforms, exactly like the concept of platform. They can support various functions being installed on top of them. A good example of this is that the kind of the platform function that you see in evolution of human is language. Language, in fact, reduces diversity, precisely the point that, you know, the diversification is no longer a tenable solution for the increasing cognitive capacities.
You can't add just simply different organs and think that you can increase the cognitive capacity of this complex system. So what instead is being used is a structure, a platform structure that supports additional functions that couldn't be added if you had simply diversified the structure. And that's precisely language. You can plug different new cognitive capacities into a platform called language. And this is something that we will talk about for the third module, that there is a reason,
I mean, not only that in terms of the content form of language and the content of our linguistic you know, activities are important to generate, you know, cognitive semantic complexity. Well, I will talk about this a little bit in terms of the actual, you know, evolutionary story behind the language. And this is the idea that, you know, language is simply a platform structure. And platform structures are basically emerged in complex systems when, because of the entrenchment, because of the
constraints that are already in place, you can no longer diversify the structure. So you need to have to come up with a unitary platform structure that can support new functions without basically diversifying a structure and by virtue of that endangering the integrity of the complex system. Because if you add one new organ to humans or any kind of organism that has massive generative entrenchment in place, you simply endanger that, you know, the integrity of the structure because of these layers and upon layers of entrenched constraints, the structural functional
constraints. So what you need instead is a platform structure. Cool. Awesome. Questions? Discussions? So the platform idea seems really rich, but I think I will have to chew on it a bit first, rather than wait into the middle of it. But that's really intriguing. I guess the other loose connection that I'm making with the focus on the history of the systems and that's sort of expressed in one form through the initial conditions but also
because the generative aspect. So there must be a theoretical link to like historicism and Hegel there as well? Yeah, I mean, well, I think if you look at it in terms of comparison, you see that the stuff that Hegel talks about is actually the same stuff that these people are talking about. But I really haven't really, and this is something, I think it's more of a comparative point of view rather than actual substance substantial link between the two between historicism of you know hegelian accounts of the history and the kind of
basically the the the picture of you know generativity and this past presence you know a structural link in you know complex theory I think it's It's more of a, I think it's more of a kind of similarity of the accounts. But I haven't really thought about this, you know, in a core year, if there is actually a substantial link between them. But yes, I mean, because you see the whole point in Hegel, history for Hegel is a conceptual Basically, thing.
Only people, only creatures who are endowed with concept can have history. This is a very different thing. You already enter into a different kind of a notion of history than the one that basically in complexity sciences are talking about. What complexity science and the history sensitivity that they are talking about is what Hegel called simply nature. And he distinguished a creature endowed with a history against a creature that has merely a nature. And the medium that allows for something to have a history rather
than merely a nature, and basically a sequence of transition, what he calls nature, is basically the power of the concept, the power of semantic content of cognition. Okay. Yeah, no, that's the, yep. Going back to your sense of extension via platforms as opposed to diversification of structure, I've been interested in thinking about systems like, you know, not just natural language but like money. Do you think that corresponds to a system kind of explicitly that facilitates exchange?
Yes, definitely. Well, I'm probably the worst person to talking about money and finance to talking about money and finance and stuff. But yes, I actually think that there's a good, you know, in fact money itself is a platform. You know, a platform that probably at some point needs to be upgraded. But nevertheless, yes, I mean, when you look into the history of money and the kind of, you know, it is absolutely a, Basically, you see these kinds of, these are special kinds of platforms. And the same thing about, well, natural language is, you know, of course, it's evolved.
You can have, you can provide a naturalistic account of the evolution of the natural language. And the thing is that it seems that all of these platforms have something to do with increase in capacitation for abstraction, basically. And basically, which is, you know, the idea of the encoding processes, basically creation of autonomous space for abstraction. And I will talk about this, I mean not money, but I will talk about this in terms of language. Yes, I mean money is a form of abstraction
and it increases capacity for social abstraction. Questions? Comments? by creating abstraction are you saying that it actually induces more hierarchies and creates more more distance like you were talking about before yes it does yes it creates I mean not essentially
you know negative account of hierarchy or but yes basically the whole point is that it creates a difference basically a structure that that the structure can also be hierarchical and allows for you know again different kinds of relations different kinds of scalings, different kinds of structural, functional distribution, so on and so forth. Yes, I think so, yeah. In language it's more complex though, I would say, rather than money, you can easily intuit the idea of hierarchy as social relation hierarchy.
But in language, I think it's more difficult, this understanding of hierarchies. But nevertheless, You can talk about this, I think, at hierarchies at the level of syntax, but also hierarchies at the level of semantics, you know, namely different basically functional role of concepts that generate different semantic content. This is something that I will talk about. So yeah, I think there is this question of hierarchy and platform, yeah, there is a different a connection between them. Actually, can you give a little bit more detail when you say platform? Are you talking about more a field that sort of enables
certain structures to be induced, or is it sort of foundational? Yeah. It is not foundational. You see, it's usually platforms emerge or design, emerge, and let's talk about just the evolved kind of platforms. Emerge usually because of two things. One is the canalization of processes. Basically, positive functions are canalized into one specific domain. So, the canalization process. And also, increase in a structural constraint existing in the system. So basically they are in response. So basically you want
to have, you want to be able to single out all the positive functional abilities, put them together as basically a core while responding also to the existing structural constraint of the system. So usually platforms are reply to these. That's why I think people, I will think about this, but I think we need to be careful not to intuitively connect this account of platform as in evolutionary hierarchical complex systems with the idea of platform that we intuitively imagine in software industry, in technology, in urban design, so on and
so forth. But nevertheless, the whole point of them is that basically the kind of platforms that they talk about is simply canalization of positive functions, of enabling functions. So it's almost like a centrifugal force of functions that collects a bunch of things and kind of binds them together where the constraints... Yes, canalization, assimilization, and also with a view in response to massive constraints that are already in place. So it's not something...
Platforms are not ever revolutionary. Nevertheless, they can create basically massive enabling functions on top of that can become basically for a whole new structure, for the evolution of a whole new structure altogether. Cool. Thank you. Welcome. I mean, it's kind of interesting to see that some of this idea of what platform is coming up in libertarian, new anarcho-libertarian, you know, literature that, you know, let's
not have revolution and let's have just platforms and these kinds of stuff. You know, there are these kinds of connections. I think it would be interesting to talk about these and see that how much they are just like kind of intuitive idea of platforms and how much they are actually, there is some sort of like a really robust conceptual analysis behind these things. But yeah, I think definitely it's the whole idea that when there are generatively entrenched structures, when there are massive amounts of constraints in place and abilities like democracy, like for example, like individual freedom, these kinds of stuff, they are basically
the results of these massive existing constraints and structures built on top of one another. So if you take revolution as simply the kind of classical view of revolution as break from the totality of the system, then that basically doesn't really add to increase in abilities. You are simply depriving yourself of all sorts of things that gave you these abilities to make this revolution in the first place which also this brings again a kind of so you have so you have this is I think it's not all good stuff talking about politics and use them as different arsenal against different political
views so you have against the classical break from the totality of Marxism but But also you can use, in fact, this kind of platform jargon as the subject of your critique because it's used, for example, people usually talk about platforms and not platform necessarily, but generative entrenchment in terms of tradition. And this is part of, for example, you know, part of the reactionary, for example, you know, take on the stuff of this constraint. They say that, okay, you know, you have massive structure in place, you have generative entrenchment,
and so, and they equate this generatively entrenched structure that is already in place with tradition. Then they use this as a, basically, as a conceptual springboard to justify that we simply should resurrect the classical concept of tradition, in terms of conservation of tradition, in our political way of moving forward. And if you look into them, you see that there are people who are not really traditionalists in that traditional sense, but they are traditionalists by virtue of having these kinds of structural
metaphysics behind their traditionalism. This idea that generatively entrenched a structure, you can't deprive yourself of them, and contra classical Marxism, because that would just deprive you of all sorts of abilities. Then basically, then this kind of view also pushes you to conclude some sort of really a speculative wacky stuff, like let's talk about that we need to justify, is it worth to change a tradition tradition or not according to, for example, the cost of changing these underlying structures,
namely the computational cost of replacing one structure for one another, as if tradition only was about a structure. But that's the whole point, that tradition is not just about a structure. Tradition also about a structure of beliefs, epistemic beliefs, of normative normative beliefs and they do not have completely, so there is this massive illusion of different conflation of different things together in this idea that natural structures are wedded to basically cognitive epistemic structures and they all become part of one structure, one tradition. But if you really, that's the whole point of, I think, Marx, that if you want to understand which structures are positive and which are not, when you want to come up
with a revolutionary critique of tradition, is that you need to distinguish between the natural, between the social, and the cognitive, basically, structures. Because tradition is a space that accommodates all of these structures. And just because they are structured doesn't mean that they are all on the same spectrum. They are yielding the same enabling constraints that you should be careful not to get rid of them. And a person who says that you are allowed to change the tradition after you have come up with a good computational cost
analysis is basically a person who is also prone to say that we should not say that we should get rid of slavery because simply it can actually yield, you can do much better with it rather than getting rid of these kinds of basically normative beliefs. But the whole thing is that slavery is not a natural structure and you can't treat it with the same criteria of distinction that you use it to draw a structural constraint. Slavery is a normative belief, epistemic, basically, dogma.
And these kinds of basically, you know, these kinds of traditions are purely normative, purely, you know, artificial constructs of ourselves, you know, and then these are according to Marx, these are the ones that they need to be, you know, they need to be undermined and they need to be overthrown. So this is, yeah, I think these are all interesting stuff and, you know, all these various stuff that comes up about generative entrenchment, idea of a structure, platform. We shouldn't deprive ourselves of new abilities by getting rid of structures that are already there and
have enabled us. But also at the same time, all these new kind of conflationary attitudes towards various accounts of structure, various different forms of constraints, so on and so forth. Yes, okay. I think a good example of this neuroreactionary reading of computational cost and generative entrenchment is Nick Jabot.
Jabot is, he doesn't explicitly claim to be a neuroreactionary, nevertheless, neuroreactionary theorists explicitly use him as a star as basically their theoretical elements. And he's quite anti-Marxist and quite superficially anti-Marxist. I mean, I consider myself to be anti-Marxist from a certain perspective, because I absolutely do not buy into this idea of breaking away from the totality of the system. Nevertheless, I completely believe in this Marxian critique of distinguishing different components of a structure from one
another. And the thing is that Javo is like, you know, comes up with this absolutely horrendous traditionalist view by bunching all sorts of structures and all sorts of structures that are already existing into one thing, one mess of basically beliefs and structures called tradition. And then he uses that as a justification that we need to be traditionalists, we should avoid revolution, we should basically come up with abilities that are simply boosting
our traditions. And that's the whole thing. What do you mean by tradition? Do you mean cultural tradition? Is tradition reflecting a series of belief norms, or is it simply an institutional tradition, the Hegelian tradition? Or are you talking about the structures of basically a society infrastructure of how they are hierarchically evolved. And if you do not have basically a robust way of distinguishing these things from one another, you can either become radical, anti-traditionalist revolutionary in the kind of Pol Pot fashion,
of the only solution is basically resetting the system to year zero or you become a pro super traditionalist of the form that no reactionaries are. Resurrecting monarchy because that is a fantastic tradition. So I think you need to have these kinds of critical tools to be able to criticize both of these camps. It's probably not quite this simple, but it seems to me to try to condense this, just sort of confusing marginal utility for any kind of robust concept of reason here, right?
Yes, yes, absolutely. Yes. I think that's why that all of these reactionaries or these kinds of libertarian, you know, traditionalists are basically the ones that in fact they refuse rationality or they reduce it to basically rational choice theory. that reason for them is simply rational choice theory. And basically, yes, this is absolutely, I think it's true. It's the utilitarian idea of reason and rationality, which is, we will talk about this, I mean, when we talked about this
in the basically previous series, that really there is no, you know, there is no common ground between them. And in fact, once you really resurrect the inferential core of rationality, which reason is really about, then that becomes a weapon against these kinds of utilitarian picture of reason, the rational choice theory and the kind of game theoretical accounts of rationality, so on and so forth, pay-off functions, trade-offs, defection. and basically the whole idea of exit, reactionary exit, is simply a defection, what game theory you have called a defection, defecting the game, because it doesn't give you enough fair
function. And yes, I try to talk about some of this stuff as we move forward. Yeah, maybe it's a good place to leave it. The big issue here seems to be abstraction, or like different forms of abstraction, where you're abstracting, I guess, out any sense of practical reason, or I guess in the same way where money can turn a house into something where it doesn't matter if anyone's living in it or not.
tradition, same way gets sort of hollowed out or any kind of social structure gets hollowed out by abstraction into this sense of just... Yes. Would there be a view for like developing different senses of abstraction here? I understand what you are saying and I think Hegel is fundamentally important in understanding what really tradition is And also I really think that Brandon's, you know, basically work on Hegel is important. Even though, you know, it's kind of a still, you know, Brandon is a progressive liberal, but he is a very smart progressive liberal when he tries to politically approach Hegel.
And I think this is the whole idea of Hegel, that tradition is a lived historical experience. And that's why pragmatism, basically Brandon's pragmatism, is useful to basically capture and robustly track the various aspects of this lived experience in history. from the institutions to basically judgments of predecessors, basically our forerunners. They, basically, the reasons that we have arrived at this historical juncture is the result of their judgments
and the institutions that are built according to judgments, so on and so forth. So institutions, judgments, basically beliefs, basically commitments that we have made ourselves to. We have committed basically all of our commitments. That's why the whole idea of, you know, belief is a commitment, and this commitment has implications. It implies other commitments, but also it presupposes commitments. The commitments that you made at some point that you arrived at this commitment. And a necessary historical analysis is the one to be capable of to triangulate your present
commitments, render them explicit, not only in terms of what your current commitments, historical commitments imply, but also what implied them in the past. What implied them? Where are they coming from, basically? And this is, I think, that once you do that, then you are essentially capable of, by, you know, this ramification of commitments and the critique, you are capable of dissociating various aspects of this lived experience in history, and hence the idea of tradition. You can bring, basically, tradition under the, you know, the power of judgment. Yeah, thank you.
Yeah, Sean, yeah, I agree. We should conclude. I should go also. But yeah, this is Brandon's idea of a reading of common law, right? Yes. Yeah. Right, well, yeah, thank you. Yeah, that's all. Okay. Yeah, I think we're... If you guys... Yeah. Yes, if you guys don't have any questions, we can conclude this session. And I will, as soon as I come back, I will post this stuff online. And also feel free to email me if you guys have any questions. I'm glad to answer. Is that your email at the news center?
You kind of, your voice is getting glitchy on this end. Can you repeat? Is that the email address where it's like reza at the new center.org? Is that the one? Yes, yes, yes, yes, absolutely. And you can, I mean, you can email me or just put any question, comments, suggestion, discussion in the classroom and then we can, yes. Feel free to open news threads on the classroom if you have discussions or anything. Thank you. Thank you. OK, guys. Thank you. Question?