Simulating the World & Remodeling Philosophy (Session 13)

Hello everyone and welcome to the first two additional seminars that Reza has offered to the Simulating the World and Remodeling Philosophy course. I'm going to pass the mic to him now. Thanks Reza. Thank you very much. Thank you. So, last session we were talking about, you know, what you might call to be conditions of possibility of toy model as a simple toy model. And we are going to continue with it, with the simple models, its modes of understanding, so on and so forth. And then next session, I will just briefly talk about big toy models and then we make some sort of what you might call to be a conclusion for what we have learned, what we haven't

learned, so on and so forth, what would be next kinds of research. One thing that I absolutely forgot for some reason to actually tell you to read and perhaps you can read it. So I was talking about Henk Direct work. So the spelling is Henk, his last name is D-E space R-E-G-T. Two works you can look at, a contextual approach to scientific understanding, it's an essay in synthesis. The other one, another essay entitled Understanding and Scientific Explanation.

in his edited book called Scientific Understanding, Philosophical Perspectives, and it's from Pittsburgh. So definitely look at his work because it's quite actually good. And I have only recently come across his work. That's why I didn't include it in the reading list. When I was actually talking about this stuff and researching, I came across his work and And I thought that it's actually quite phenomenal. So we can hear

At this point, after the previous session, we can talk about what you might call to be a broadly sketched theory of understanding for toy models. And we can begin such a topic by posing a question. Do toy models yield understanding? understanding in the sense that remember what we were talking about last session in terms of that you know so if toy models are in fact explicitly false explicitly idealized simplified models then why are they good for you know so in that sense this question becomes a glaring one do toy

models yield understanding? And furthermore, is our taxonomy of autonomous and embedded toy models, as defined in the last session, helpful for answering this question? The reason that I mentioned Rekt, but also Dennis Dykes, accounts of scientific understanding of toy models is precisely because their entire project is focused on these two questions and that's why you should look at them carefully.

So I can just go over about what kind of preliminary, what you might call to be data or axioms, do they put forward in order to answer these two questions? In a sense, I'm simply presenting their account in a very rudimentary sense to, one, bring out a number of common

assumptions in several current accounts of understanding models, and two, motivate the account of understanding that we will adopt the refined simple view of a toy model. So according to direct, a phenomenon P can be understood if a theory T of P exists that is intelligible and meets the usual logical, methodological, and empirical requirements. Although he restricts his definition to theories, his approach is intended to be more permissive

since they also refer to models as vehicles of understanding. Both all models, in one way or another, are vehicles of understanding. It's just that their modes of understandings are different. So let's tersoally examine direct necessary conditions for understanding more closely. First, the explanation condition. Second, the intelligibility condition. And third, the usual logical, methodological, and empirical requirements, according to the definition that I just gave you.

First, direct explicitly ties understanding to explanation. And this is in tandem with the philosophy of science. You see, when we are talking about understanding in philosophy of science, we are a little bit more diverging from the Kantian account of understanding. We are making it a little bit more explicit. And this explicitation is that which turns the term understanding in philosophy of science to the term explanation. So in philosophy of science, whenever we use the word understanding,

we most probably are looking at explanation. Okay? In the sense that you have to have that which is to be explained and that which explains. explicandum and explainants. Sorry, explanandum and explainants. So understanding a phenomenon is characterized as having an adequate explanation of the phenomenon, and a phenomenon P is understood scientifically

if a theory of P exists that is intelligible and the explanation of P by T, theory T, meets accepted logical and empirical requirements. Hence, we take it that having an explanation of P is a necessary condition for understanding P according to direct view. So we can simply call this condition as the explanation condition.

Second, Dirich defines a theory T as being intelligible for scientists if they can recognize qualitatively characteristic consequences of theory T without performing exact calculations. That's important. De Richt, for example, argues that physicists consider the kinetic theory of gases to be intelligible if and only if the physicists are able to infer statements from the kinetic theory without performing exact calculation, such as the following statement.

If one adds heat to a gas in a container of constant volume, the average kinetic energy of the moving molecules and thereby the temperature will increase. So you don't really here involve exact calculations. It's just a theoretical statement, so to speak. the intuition motivating the intelligibility requirement that is in contrast to an oracle we want to be able to grasp how the predictions are generated and to develop a feeling for the consequences

the theory has in concrete situations Third, what does direct has in mind when referring to, for example, usual logical, methodological, and empirical requirements? although he does not make this point explicit we can presume safely that he refers to familiar virtues of scientific theories or the criteria of theory choice for example Thomas Cohen's paper Locus Classicals

Thomas Cohen's paper, famous 1977 paper, is actually a locus classical, a canonical work for assessment of the characteristics of a good scientific theory. These five criteria, accuracy, namely corresponding to empirical adequacy, consistency, scope, simplicity, and fruitfulness are all standard criteria for evaluating the adequacy of any scientific theory. For this reason, we henceforth refer to these requirements as Kuhnian criteria for good

scientific theories. explicitly affirms this reading. The Kuhnian criteria determine the goodness of a theory T or a model M on which the explanation of some target phenomenon T is based. Okay? The REC's main motivation for demanding that intelligibility per se is not sufficient for scientific understanding is that, for instance, astrology should not count as providing scientific understanding because despite of being intelligible, astrology fails to be a good theory or model

if judged by the aforementioned Kuhnian criteria. Okay? So, Derrick's account of understanding is one of many possible starting points in the large literature on modes of understandings which models can afford us. However, what matters here is that his account is in several respects a typical account of scientific understanding. To bring out a number of common assumptions in several current accounts of understanding, including the early approaches by Friedman, Kitcher, and many others, we will focus on the explanation condition

and the intelligibility condition. And in that sense, we'll put aside the usual logical, methodological, and empirical requirements, namely the Kuhnian criteria. The common assumptions of these accounts of understanding can be characterized as follows. An individual scientist understands some phenomenon P only if three conditions are satisfied. One, explanation condition. There is a scientific explanation of P. Philosophers concerned with understanding often differ with respect to their preferred theory of explanation. They use different theories of scientific explanation, such as the covering law account, the unification account, the pragmatic accounts, and various causal accounts of explanation, a la mechanistic term, which we have been talking about.

Two, very decolicity condition. In asserting that the understanding of a phenomenon, P involves an explanation of P as a necessary condition accounts. of understanding inherit a feature of theories of explanation that we call the veridicality condition. It is a common view that explanatory assumptions, that is the explainance of an explanation, those factors which explain an explanando, are required to be true or at least approximately true. Consider the following examples. Proponents of the recently dominant causal accounts typically endorse this requirement. That is, the explainants

has to truthfully represent the causes of the explanandum phenomenon. For instance, James Woodward in Making Things Happen holds that the explainants has to be true or approximately so. And Strebens endorses the claim that the explainance is a veridical causal model consisting of true causal laws and true statements about initial conditions of the said target system. Or, for example, furthermore, you can think of Hempel's covering the law, account demands that the explainants consist of true law statements

and true statements about initial conditions. Or for that matter, you can think about Van Fresen's pragmatic accounts, which impose a veridicality constraint on the explainants, but in a kind of a pragmatic sense. Three, epistemic accessibility conditions. If an individual scientist understands phenomenon P, then she has epistemic access to an explanation of P. Direct concept of intelligibility is one possible strategy for making precise epistemic accessibility.

To have epistemic access to a toy model for them just is being able to recognize qualitatively characteristic consequences of that model without performing exact calculations. So you see, the third condition is something what you might call an additional, so far we have been talking about toy model as just simply these kinds of explicitly simplified, idealized models, which can be very much, most probably, certainly, are false.

But now, with this latest criterion, toy models also get a new definition, in the sense that a toy model can talk about theoretical consequences of certain observations without actually engaging in exact computations or calculations. Just exactly like the example that I mentioned, the kinetic theory of gases in a closed system.

questions questions questions before i move forward are these um explanation criteria vertical verticality and epistemic access are they building on each other no no no they are not building each other but nevertheless there are dark as you say as you are implying there is undercurrent connections between them yes but they are not explicitly building on each other no okay so to talk in fact adequately about

about modes of understanding as pertaining to different models, particularly in our case, toy models, this discussion bottoms up in the differences between many competing accounts of understanding, which consist in alternative ways of spelling out, making explicit each of these three conditions. Explanation condition, verticality condition, epistemic accessibility condition. So, here, with this semblance of an introduction,

about what we do actually mean by the mode of understanding. When we are talking that toy models have a different mode of understanding than other kinds of models, okay? Now we can actually talk more coherently and tersely about toy models, the corresponding mode of understanding and also which also means a refined as I mentioned we are simply at this point talking about small toy models which means that the mode of understanding

we are going to talk about is about a mode of understanding for a refined simple view of models toy models so Michael S. Revens, let me spell his name for you in case that you want to search him out. Michael S. Revin's account of understanding offers a promising strategy for avoiding the challenge from idealized models.

According to his view, his simple view, scientific understanding is defined as follows. An individual has scientific understanding of a phenomenon just in case they grasp a correct scientific explanation of that phenomenon. Now, you should understand, when Estrevens talks about correctness, he explicitly refers to the veridicality condition, the second condition that I mentioned earlier.

The notion of grasping is Stravans's way of articulating the epistemic accessibility condition. Stravans does not provide an informative definition of the concept of grasping. Instead, he takes grasping to be a fundamental relation between mind and world, in virtue of which the mind has whatever familiarity it does with the way the world is. One may be concerned about the fact that grasping is taken as a primitive, but of course we will talk about this shortly. What is central for what we are talking about now is that the simple view offers a strategy

for dealing with the challenge from idealizations, as mentioned in the previous session. The Strebans is fully aware of the fact that the simple view cannot be applied to idealized models straightforwardly. The reason is that toy models are taken to be literally false, as we mentioned in our previous session, while the simple view, implying the veriticality condition, requires them to be true. As he points out, most standard theories of explanation require that the explainance that which explains to be true or veridical.

For instance, his own cheiretic account of explanations in its simplest form requires that the explainance consists of true causal laws and true statements about initial conditions. His account for idealized modelism goes in the following way. Although idealized statements are literal falsehood, his terminology, these statements can be interpreted by using an account of idealizations as being approximately true, which is to say, veridical.

Stravins' specific account of idealizations appeals to an optimizing procedure, one vital component of his chiratic account, whose function is to filter out or to ignore or to forget explanatorily irrelevant information that need not be explicitly stated in the explainants, Making use of this idea of ignoring explanaturally irrelevant information, he develops a minimalist account of idealizations. He argues, the minimalist account implies a veridical reading of idealized assumptions.

Idealized assumptions truthfully, i.e. veridically, report which factors are irrelevant for the explanation at hand. in the sense that coming back also in connection to what we have been talking about in the previous course on the work of Himpel and the concept of scientific explanation, you might say that the simple view of explanation is not too further apart from the idealized view, okay? It's just the simple view, which is explicitly false, as in the case of small toy models,

also explicitly mentions the very irrelevant explanatory factors which are being excluded as opposed to the idealized version. An example of this you can think about it like Impulse pneumological expectancy law. And his example, like, I am sitting behind a desk.

On the top of the desk, there is an ink holder. OK. So I have a knee jerk. And this knee jerk hits the bottom of the table. and the ink holder falls and I get ink all over my clothes. Now, this can be formulated as a causal law, as an explanatory law, from a statistical even perspective. We call this a nomological expectancy. Under such and such conditions, given such and such factors, if P takes place, Q would follow.

OK? So you see, these conditions are idealized, simplified in a kind of a very, what you might call to be, contorted way in the idealized version of explanation, in the sense that we sometimes arbitrarily choose one condition, one factor against another. The idea of the toy model, or the simple view explanation, is that we explicitly say why, in fact, we have not excluded other kinds of factors which could be otherwise

included in our accounts of a nomological expectancy, explanation of a physical phenomenon. Okay? Any question here? So to speak, both idealized version and the simple, refined simple view, view, in fact engage in the same kind of procedures of idealization. It's just that toy models

try in the best capacity, in the best capacity, that is important, in the best capacity to to make such sacrifices, such eliminations of possible variables, factors, and conditions explicit. Why is that in fact we don't care about this? Like, if I have, like, you know, you can have a knee jerk for so many other factors, you know. You can have, you know, your funny nerve got irritated. I don't know, you have some sort of someone tickled your foot. It can be so many conditions. It's just that we don't want to know about this in the Torah model.

Literally, we don't want to know about them. And we are not shy about saying that. OK? Is it the same? Oh, sorry. Go on, go on, go on, please. I was just going to say, explicitly mentioning irrelevant factors is that the same as explicitly stating what you're taking as the relevant factors? I don't think that it's the same. Okay. It is not the same. It's absolutely not the same. This is why if there would be such an equivalence relation between making explicit the irrelevance and making explicit what you actually take relevant, then they wouldn't be false models to begin with.

You know? But they are false models. Precisely because... But they are false models in a helpful way. Essentially, you see, this is the whole idea that people usually talk about. You see, first of all, any person who thinks making... pluralizing the idea of mistake or error can lead us to some sort of enlightenment is just off chart. Science doesn't work like this. Science actually works with carefully curated errors toy models are ways of curating the errors that's important

what if all what if all factors are not modally equivalent sorry can can you repeat it Joven? What if all factors are not modally equivalent? Yes that is true most of the times are not modally equivalent. Then by what means can these nomological expectancies... You see when we... okay nomological expectancy you can there is a top view of neurological expectancy, namely what you might call to be the kind of a canonical causal explanation, okay, in the sense that since Hume it has been advanced,

and now you can see the distillation of it in the mechanistic term, particularly by people like like James Woodward in making things happen. So there's a top view in which basically, yes, now I do agree, but not all of them, but the majority of them really don't see the modal incommensurability between such variables and conditions and factors. Now, but the nomological expectancy, as initially put forward by HIMPO, is actually a statistical account. A statistical account is the foundation of modality. Essentially, it tries to see everything in terms of the convergence of possibilities and modalities,

athletic modalities. So there is this kind of what you might call to be a kind of a very confusing view. If you go for the top-down, you might actually say that, well, yes, as you said, which I completely agree, they don't have the same kind of modal valance, so to speak. But the thing is that the actual nomological expectancy formula, as put forward by Hempel, is not really about the top-down view. It's from a bottom-up view. it tries to capture all such conditions in terms of pure statistical inferences and possibility. Modern notion of probability.

Oh, I had a question. So when, I didn't mean to interrupt, but so when curating these sort of toy models are sort of curating these errors how much of that barrett I think I lost you for one second probably was my end time but you hear me you're back back marriage that I lost you could repeat my apologies oh yeah yeah so so when curating these errors through through toy modeling because you know it's not just any sort of error it's an actual like there's like a scope and there's there's a

tethering to something like how much of that because part of me is like well there's also like this deterministic logic to errors as occurring as sort of like natural phenomena on their own in this ability to kind of analyze those through differences of what we already have without necessarily having to have a model, if that makes any sense. So is there some sort of plane or what makes a curated error explicitly and universally more? I don't know if you were asking this, if you were providing this.

Let me reformulate. in the sense that, you see, okay, what actually guarantees that if we curate errors, we are not ending up in a situation where we hadn't curated those errors, a la idealizer. Yeah, no, I get that, yeah. Yes. So let me tell you, and this comes back to Jovan's idea. So actually, you see, not every toy modeler is actually understanding this. But here is a point. And what I mentioned, and this is something that only I've recently thought about it quite carefully.

And that's why I think that the concept of statistics and modern notion of probability is so fundamental to how science works. Essentially, yes, you can do the top-down view of error curation, but then you are actually possibly making even more errors, making more irrelevancies, implicit irrelevancies. Yeah, that was one of the things I was about. Yes, but from a statistical perspective, these things take a fundamentally different shape these these problems it's a question of

probability distribution so on and so forth and in that vein we can no longer talk about you know error mitigation or error explicitation the way that we used to talk about it when we were on the top-down view. Now, another thing that you said, deterministic. What do you mean by deterministic? Everything is deterministic. Oh, yeah. No, I know that. But I'm just saying at what level is this kind of – like do these things – because I understand there's a necessary function in creating this, as you say, like a toy model as like a place of differentiation so that we can actually curate error and cure...

Right. ...analyze things. The thing that I'm wondering is like, on what plane does this exist with natural occurring phenomena as these things can be like explicated as an error, they can be seen as... That's exactly where you need determinism. Oh, of course. In the sense that determinism is all about causal association, about causal laws. So as I mentioned to you... But where does the... Just one more thing. Go on. Where does the role lie in the person who's kind of... Instead of creating this sort of statistical analysis versus the person or subjects or

collective of the people who are actively trying to well okay yes yes i understand this this is actually a far more difficult question in the sense that uh when we are talking about a a community of scientists essentially a collective we are not just simply talking about statistics or probability but we are also talking about perception of probability. Yeah. Perception of probability and how it should be applied. Yeah. And that's when you get, you know, a kind of what you might call to be a communal framework, which is a way of science. But it's not...

That's why I didn't mean the same one person's role. Like, I was like... Yes. Yeah, I know that. Yes. Never one person. Yes, but also with regard to the question of causal determinism, which I think that any person who thinks, I'm not talking about you, I know how you think, but I'm talking about so many people think that indeterminism is some sort of like a heroification, getting out of the Laplace demon, so on and so forth. No. If you don't have determinism, you really don't have any belief. Yeah. If you don't have any belief, then you have no judgment. Essentially, causal determinism is as simple as this.

I'm making a very, very badly reductive example. To believe that, for example, when I see blood, I might actually faint, a counterfactual, you know, is actually the result of a certain causal determination. in the sense that there is a causal reliability between me seeing red liquid and me being fainting. So this determination is absolutely the fundamental feature of every, not only scientific procedure,

but every even ordinary talk that we make about the world in our own ordinary language. Without causal determination, literally nothing can be held together. Oh, yeah. But determination doesn't mean compulsion. You see, people think about determination as some sort of impulsion. Impulsion, maybe not compulsion, it's not a good word, because compulsion means you can be compelled by a rule, which is not exactly some sort of a natural pre-established law, but you can be impelled, impulsion, impelled by a law.

So many people actually confuse the concept of causal determination with the concept of impulsion. that, you know, if there is some sort of causal state or some causal processing that is going on in my goddamn nervous system, then I only simply react to these. I never proact. I only react to these. Essentially, what I ever do is going to be the effects of such impulsions. But that is just totally pretty scientific, to be honest with you. It really doesn't have, sure it has.

When we actually take off some other stuff, and in certain cases when basically we are talking about certain kinds of systems, like a thermostat, where they are bare, they are minimal. They don't have the other auxiliary options like theoretical, practical reasoning, so on and so forth. But natural impulsion hardly ever happens among those people who actually believe in theories.

or basically another way to say it is that natural impulsion is hardly an issue when it comes to rule-based compulsion. And causal determination is actually not about either of these. It's not either about rule-based compulsion, like a Brazarian rule-based compulsion, if you have read about that, or simply natural compulsion. It's just the tissue between the two, the bridge between the two.

It is not either and or. It is the commensuration of causal factors as reliability factors of our certain kinds of beliefs. should we have a should we have a should we have a a a break and then we start and i know that you and some other people who are uh playing with their beard let's not say who those people are they know who they are they maybe we can have a question All right, that sounds good. How about five minutes?

Sure. All right. All right. Thank you.

Thank you. Thank you.

Thank you. Thank you.

Thank you. Can everyone hear me? Yeah, I can hear you. Yes, also, essentially, this is not something that should concern us right now, but the

The more I have been thinking about it, and I think I mentioned a little bit of this in the previous session, is that a lot of this stuff that is going into the philosophy of science with regard to the question of a causal law, neurological expectancy, modeling, so on and so forth, has something to do with the modern concept of probability and a statistical inference and its status within the scientific canon. And to the extent that we are not really familiar that much with these things, many of these discussions look,

even though we can say, OK, it's plausible enough. But when we look at it really carefully, we say that, oh, how is this going to actually work? And what do we exactly mean when we say that at a statistical level, it is objective? What actually makes a statistical inference objective? Right? Particularly after all the riddles of induction, all the problems of induction. Well, that's why I'm saying that I think that one of the things that is definitely going to haunt philosophy of science, but also philosophy in general,

most probably not consensual philosophy, is going to be the concept of probability, as defined in the modern canon, after Dauphiné onwards. Can you elaborate more on the statistics and probability thing? That was my question. And also, can you give some references? Yes. So essentially, in the canon of philosophy of science, there are at least three concepts of philosophy. So Carnot talks about two concepts of probability. But these are both modern concepts.

Now, let me tell you that there is a naive interpretation of Hume, that if you have certain kinds of observations within such and such frequency, you say that this is the case. OK, let me, Ian Hacking's example of witch burning is a good example, actually. It's a precritical notion of probability. So we have observed the Council of Evangelical Christians at Fairfield County, USA, that these number of women

hold the broom. These number of women do stuff with brooms. And then these number of women also engage in such and such subversive activities. The correlation between these observations and the frequency of observations lead us to believe that these numbers of women who hold brooms are actually witches. We shall burn them at a stake. So this is how a frequency, a naive, a critical frequency idea of observation works and probability. And I really suggest to both of you,

it is essentially not that great of a book. But if you really want to actually understand the solid ground of what we mean by probability today, I really do suggest reading Ian Hacking's Taming of Chance. It's actually a very, very enjoyable historical account, kind of like Julian Barber's account of the history of dynamics. So this is like the ancient pre-critical notion of probability. Now, there are two concepts of probability here, modern concepts. One, the frequentist, and the other one, logical.

The frequentist account of probability is not exactly like the Humeian or precritical account of probability. It's not exactly about how many times you have seen such and such observations of such and such associations. It is about the statistical distribution of such and such observations and frequencies within a probability space. It is constrained. Number one. second one which is what you might call to be the modern concept of probability, truly modern is purely logical there is observations

don't play any main factor in your inductive inferences precisely because such observations are indexed by logical references so it is only through these logical references all couched in terms of the logic of a statistical inference or Bayesianism or whatever you think about it actually lead us to believe that under and such conditions within such brackets of probability namely the probability of space we might arrive at such conclusions.

That's a very, very brief idea of what probability is, really, from pre-critical to post-critical. So in terms of references, I actually do suggest you reading Daufinetti's classical work, The Subjectivist Account of Probability, which is the modern account of probability. So probability is not objective actually, in terms of your wager. Now here is a confusing thing. The probabilistic description is objective, the probabilistic wager is not. The probabilistic wager about a coin toss

is actually subjective not objective according to Daufinetti. So Daufinetti's classical works on foundations of probability, Carnap's work logical foundations of probability, then Ian Hacking's statistical influence and and taming of chance. I think this should give you some sort of a groundwork about what we actually do mean by probability, and what is the role that contingency plays here, or chance for that matter.

Thank you. Is the subjectivist account of probability, is that a full book length? Let me give you the exact title. One second.

One second. It's just simply called theory of probability, a critical introductory treatment. Thanks.

And he made, surely, the logical probability has gone a long way since the time of Dufinetti. But essentially, Dufinetti gave a final blow, in the sense that for a long time, the Bayesianists, the probabilistic people, thought that a wager is actually an objective affair.

Daufinetti completely showed that it is thoroughly a subjective affair. but nevertheless commensurable with the objective probabilistic or a statistical description of a set phenomenon. And here, there is actually that every single, not every single, almost all modern Bayesians are followers of the finetti, the subjectivist account of the wager. Yeah, the taming of chance is superb, but also, one second. One second. Let me give you the other title.

Sorry. I can't find it. Well, let me just... Is it the emergency probability? Is that the one? Yeah, I think that's the one. Does it have the thingies, the dyes as the cover? No, that one is the logic of statistical inference.

Yes, yes, that is the one, yes. Logic of a statistical inference, Jan Hacking, yes. So these books should give you very good. And also, there is another thing, even though as a digression from the main topic of our class, You see, well, I'm sure that you will get a little bit more about these kinds of topics from Inigo Wilkins' forthcoming book for Rubanomic. The thing is that, so here we have some sort of kind of like a cognitive designance of some sort,

in the sense that we think that if the universe is purely contingent, then there is no possibility for what you might call to be rationality or reason. Yan Hacking actually tries to show that the irreducible, a stochastic configuration of the universe is, in fact, purely commensurable, but also responsible for the kind of thought that we have and why we can actually change nature. So Jan Heuking is a kind of a rationalist reading, gives a rationalist reading of the thesis

of absolute contingency. Whereas, for example, Mea Su gives, ultimately not gives, arrives at an irrationalist reading of pure contingency. So, very quickly, just to briefly talk about, introduce or define the simple view.

An individual scientist S understands phenomenon P via model M if an only model M explains P and S grasps M. Now, let's refine this working definition in four ways. One, naturalism about grasping. If qualified properly, we are willing to follow a strivence in assuming that grasping is a fundamental relation between mind and world.

For present purposes, of course, we are prepared to accept that the notion of grasping is philosophically primitive, but not scientifically primitive. What does it mean to take the notion of grasping as philosophically but not scientifically primitive? Well, understanding has a subjective component in addition to the publicly accessible component represented by explanation, in the sense that understanding takes place in an individual mind. Following this, we adopt a naturalistic approach to the subjective component of understanding. That is, what grasping turns out to be is a scientific matter, not in fact a philosophical matter.

The subjective component of understanding can be studied by cognitive sciences. For example, cognitive science tells us that grasping toy models sometimes consists in being able to visualize the behavior of the target system of a scientific toy model or to have a mental model of the toy model and its solutions. Second, the contextual character of understanding. Understanding a phenomenon is always contextual. Some model in, say, population ecology may generate understanding for an expert in this field, but not for an expert in statistical physics or a layperson.

so we can agree we direct in assuming that the individuals who gain scientific understanding are experts regarding the kind of phenomenon that is understood we express this thought by saying that an individual scientist s understands a phenomenon p via model m in a context c where context c is a scientific discipline and s has expert knowledge of that discipline verified by the objective criteria of its community ala kun's paradigm

third different modalities of explanation and understanding the kind of explanatory information scientists receive from toy models is not always the same. Or so we'll argue that it is in fact useful to distinguish two different modalities of explanation. And that's how we are going to shift understanding, sorry, maybe treating toy models in terms of the mode of understanding that is usually associated with idealized explanation, move from this toward a mode of explanation that is more commensurable with the simple view of explanation,

a la toy models that means moving from how actually understand mode of understanding to the mode of how possibly understanding and we are going to talk about them what is the difference between these two how actually explanations possess an explainance that which explains satisfying the veridicality condition i.e. consisting of actually true or approximately true statements

whereas the explanation of how possibly explanations refer to merely possible explanatory factors for instance to possible causes and mechanisms bringing about or realizing the explanandum phenomenon the target system, how it behaves, so on and so forth if the explanation is causal So, the distinction between how possibly and how actually explanations can be accounted for by and integrated into many standard accounts of explanations. For instance, with the covering law account and various causal accounts of explanations

such as Woodward's seminal theory of causal explanation or as I mentioned earlier, S. Treven's cheiretic accounts. So how possibly explanation has a very, in fact I would say from a historical scientific perspective, has a far deeper history than how actually an understanding. You can even think about Archimedean mechanical method. Does anyone know what Archimedean mechanical method is? I think I have talked about it in our class.

So it goes like something like this, very briefly. You have a certain geometrical problem that you don't know how to solve. OK? You don't know the explanation behind this geometrical phenomenon. So what do you do? You make a mechanical model of this geometrical problem. An example of this would be solving the problem of the ratio between a cone and a sphere inside of it.

What would be the ratio of the cone to the sphere? Well, that's actually from Archimedes time, it's a totally, you know, so to speak, difficult problem. He has a novel solution for it. He creates a mechanical model. What would be this mechanical model? It's essentially a lever with an adjustable prop. On one side he puts the sphere, on one side made of full solid, and on the other side of the lever he puts a sphere made of the same homogeneous solid matter.

And then the adjustable prop actually shows the exact ratio, approximately so to speak, between the sphere and the cone when they get balanced in weight. Okay? It's totally genius. And the thing is that, so here is a mechanical method, essentially. The mechanical method, so you have first a geometrical problem and you want to find an explanation for this geometrical phenomenon, whatever sort it is. I'm saying here, a geometrical phenomenon. I'm not even talking about physical phenomenon.

I'm even going to more deeper abstract phenomenon, definition of phenomenon. So you have a geometrical phenomenon and you want to find a solution. You don't have a solution of it in geometry. So what? You make a mechanical model. Now, this mechanical model is made by virtue of a certain equivalence or equivalence between certain families of relationships which hold between certain sets of geometrical concepts and certain sets of mechanical concepts. We call this an equivalence map. So it's not just an arbitrary model making.

The equivalence relationship is absolutely necessary. So once you, by virtue of this equivalence relationship, you make the mechanical model, then you study and observe the behavior of your mechanical system or your machine behavior. The observed mechanical phenomenon can, by virtue of the aforementioned equivalence map, can be translated into a geometrical solution for your geometrical problem.

But it is not yet an instance of how actually the geometrical problem works or how the geometrical phenomenon is explained. It is how possibly it can be explained. this is an Archimedean toy model how possibly understanding a certain phenomenon questions, questions yes, Adam, yes, absolutely, Pythagoras definitely, definitely you get a lot of this stuff in the age of analytic geometry.

But the method itself is attributed, either urban legend or not, nevertheless, we are going along with this story, to Archimedes. The Archimedes was the first person who actually tried to create a model of a system. by actually dissociating from the actual contextual problems of the system, creating a very simplified machine. This machine is not going to yield you an understanding of how the actual geometrical problem should be solved

or how the geometrical phenomenon behaves. It is nevertheless going to give you a key. They should refine it. And this is how possibly it can be solved. How possibly does it behave? Questions, questions, heckling stuff, whatever. No?

Okay. So, that was the third. Fourth aspect is neutrality with respect to different theories of explanation. And it is about assuming that, one, a toy model have explanatory power of either how actually or how possibly kind in the sense that I mentioned with regard to our Archimedean

method example. And that there is a philosophical account of explanation that applies to toy models, instance by identifying merely possible or actual causes if a causal account is adequate and so on. Now with that said we can formulate a refined simple view of of explanation for toy models. An individual scientist understand the phenomenon P via model M in context C if and only if one of the following conditions holds.

One, a scientist S has how actually understanding understanding of phenomenon P via model M in context C if and only if model M provides how actually explanation of P and S grasps M. And that's usually the case with, you know, what you might call to be models which are not toy-like. Two, a scientist S has how possibly understanding of phenomenon P via model M in context C if and only if model M provides a how possibly explanation of P and S grasps M.

being able to distinguish between how actually understanding and how possibly understanding will prove to be central in our discussion of understanding in the context of autonomous toy models, which we were talking about last session. So if we accept the refined simple view, then we can ask, do scientists obtain understanding with toy models, with understanding that toy models are patently false? Now here we can provide an answer in the form of the following three claims.

One, an embedded toy model, M, yields how actually explanations, if three conditions hold. 1. The well-confirmed embedding framework theory T permits an interpretation and justification of the idealization of the model M. And 2. This interpretation and justification is compatible with the decality condition. If one grasps the how-actually explanation provided an embedded toy model satisfying the condition A and B, then one has the paradigm of how-actually understanding.

Two, some autonomous toy models do not provide how actually understanding because major interpretations of idealizations do not support an interpretation and justification of the relevant idealization of these toy models that is compatible with the very decollity condition. In other words, major interpretations of relevant idealizations do not support the claim that all autonomous toy models provide how actually understanding. Three, there are central examples of autonomous models that are best interpreted as proving how possibly explanations and respectively how possibly understanding this sort of understanding is valuable.

because it has what we can call important modal, heuristic, and pedagogical functions in scientific research and scientific education. And coming back to what Jovan said about modalities, the question of modalities, Well, you see, one of the greatest problems in modeling, so remember we talked about this, that when you are talking about a certain kind of thing, whatever it is, the universe, as big as the universe, and you try to model it.

Or you can try to make a systematic theory out of it. this systematic theorization always and all the time relies on certain kinds of counterfactual scenarios fictions so to speak okay modal concepts but then how can we go actually make such counterfactual scenarios such modal vocabularies Right? Just simply saying that there are certain kinds of analytic modalities or possibilities of how we could approach and talk about a certain kind of phenomenon

doesn't really give us anything. It's just like when you are invited to a party And you're supposed to eat the most lavish dinner, and some people give you raw vegetables at the end of the night. No, this is not how it works. Just talking about plentitude of modal scenarios or counterfactual scenarios doesn't get us anywhere. Yes, you can talk about them logically, But that is a different kind of things. You have to make models. And what are these kinds of model models,

counterfactual models? Well, they are toy models. Toy models which are not giving us how actually a phenomenon can be explained or how actually a phenomenon can be understood. They instead give us how possibly a phenomenon can be explained and how possibly a phenomenon can be understood. And the labor of the refined simple view, the kinds of idealizations that I have talked about

with regard to calculation theorizations, you can think about them as requirements, conditions of possibilities, of arriving at a modal view of a target phenomenon, whether it's a physical phenomenon or it's another model. Nevertheless, without that, if you don't have that kind of toy model in which you are in the realm of how possible understanding and how possible explanation, then literally you cannot talk about counterfactual scenarios, modalities, fictions from which science buddhist raps itself. Fiction is not in a high-prestitial sense. Don't get so overexcited. Fictions,

we are talking about decent fictions, scientific fictions. Questions? What is the requirement again for the condition of possibility of arriving at the modal view or target phenomenon? Well, as I mentioned, it's simply the understanding of how toy models can be constructed. the refined simple view of explanation remember those three conditions radicality condition explanation condition and empirical accessibility condition these these three these three

when you actually commit to them you are already in the realm of how possibly so explanation and understanding plus the auxiliary condition that I mentioned. These are the kinds of models in which you do not need the exact calculations, the exact calculations to arrive at conclusions of what might actually happen. think of it about the Archimedean machine the lever do you really need to know anything about geometry

no you don't need anything about the axioms of analytic geometry allow Euclid you don't even need to know basically the axioms certain axiom about a sphere or the cone all you just need to know is that how you can actually make a sphere and a cone put them on two extremes of a lever a machine a machine here is a mechanical machine of course it has geometrical properties but that is not your concern That is not your concern. Those exact calculations are not your concern. Then all you need to do is to make a balance by adjusting the prop such that you see the lever on a balanced position between the sphere and the cone.

And then using this mechanistic solution, mechanical solution, in conjunction with the set of the equivalence relations between mechanical problems and geometrical problems, then you can translate it simply back into a geometrical solution, but, and that's a counterfactual. there's a counterfactual you couldn't in fact make such a model counterfactual model if you were simply abiding in the axiomatic realm of analytic geometry you had to make a model

you had to make a model but also to make a model you had to not only sacrifice so many details al-idealization of all models, but actually make a patently false model. False in the sense that this model is not geometrical. It's just a mechanical machine. What's the this isn't supposed to be a philosophical clarity a little bit but what's the significance of describing a toy model as possible understanding as opposed to actual understanding

You see, possible understanding is essentially works exactly like a modal vocabulary or a counterfactual. In the sense that it is, its realm is a fuzzy explanation, so in a way, without getting too much about the idea of the vagueness, fuzzy, logically fuzzy. In the sense that a possibly so understanding works like a hypothetical. There can be many other solutions to a hypothetical. You see, it is in fact a broader range that's how actually

understanding from a logical radicality condition. In the sense that how actually is just a special case, certain kinds of causal relations that we try to get, extract, in accordance to the theory that we use and the context of the phenomenon that we have observed. But the hypothetical gives us much more in the sense that, let me tell you this, you you can actually solve this problem in a fundamentally different way. In the sense that you can go to some sort of Riemannian geometry, where the sphere and

cone are fundamentally have different definitions in a technical sense. And in that sense, still you can make this happen, but through a different method. So the hypothetical, the modal, is kind of like Peirce would have called possibilities. So all of these possibilities can be actualized if they are veridical. It's just that your method always requires to be exact, always needs to be contextual.

So you only are going to move through one of these possibilities, not the whole range of possibilities. But a hypothetical actually gives us what you might call to be a very mystical allegorical, I would say, example of this would be the Oracle of Delphi. So she tells you that you are going to die tomorrow. OK, now, well, you can, of course, die in so many sorts of ways. It's just that it is the truth.

It is literally what you might call to be the sum of all existing or available possibilities within the current capacity of theorization and modelization. at sort of paraphrasing that. And so it seems to me like the toy models, okay, they're constraining the problem, they're constraining the solution space without fully providing a solution, right? Because they're showing certain properties. Yes. You see, right? Yes. So there's that aspect where they constrain the solution space, but the other aspect is

they're like training the intuition of the scientists or the community of scientists, the sort of, so that they can comprehend what's going on in either big toy models or, you know, a whole model of the system. Is that a fair way to describe it or is that, Yes, yes. But one of the things is that I think that, Just like hypotheticals, how they function, toy models can also backfire inside a community of scientists. in the sense that when, I mean, if we are really believing the social criteria of Thomas Kuhn

set forward for science, we can say that not all scientists are aware of what has actually gone to the construction of a hypothetical, and hence they cannot wield it correctly. only those who construct it can wield it correctly and those who are familiar with the construction can wield it correctly but not everyone else, otherwise we were in a utopia but yes, also absolutely in the sense that it is plotting without drawing the graph it is outlining without making a drawing that's what toy models are

outlining without the drawing the figure. That's it. They show us the knowing, the theoretically knowing boundaries of behaviors, conditions, and constraints. That's good enough for me. I mean, that's not raw vegetable. That, I would say, is like a kebab at the end of the night. GERALD REIMER- Sure. I guess this is also why it's trained by experts as well,

right, that we're talking about this process. It's not all. GERALD REIMER- Absolutely. And one of the things that, for example, If you have read the second chapter of Intelligence and Spirit, when I actually mentioned toy models, particularly big toy models. So here, let me make a kind of a very rudimentary example of this, the shitty work Intelligence and Spirit is, but let's not talk about it in public, even though it is going on public. The idea is so simple. You see, so I remember we actually talked about this in the course on computation and complexity. So here, there is a problem.

So different research groups want to develop a strong AI, AGI. But nevertheless, they are entrenched in their own theories and methods informed by those theories. And hence, they make idealized models. You can think about neural network, Bayesianism, symbolic AI, so on and so forth. And a lot more yet to come. But the thing is that all of these, when you look into the historical progression, the historical development of AGI research, you see that these are what we call bottleneck models.

they start great with great ambitions they create justifiably successful results but they never meet their ambitions namely the creation of AGI why is that precisely because the whole point about AGI making sapient intelligence as Dennett would have said it is to create, it's a design problem, is to create as vast as possible of a space, as vast as possible of a space, a design problem. How can you go in the realm of theory and modeling

and commit to the vast design space? Well, one way to go for it is by actually using toy models, particularly big toy models. They are false, as we talked about them. But nevertheless, they, in fact, have a better opportunity to outline the constraints of what we can actually do, what the resources are, what the methods are, what are the conflicts between different theories and models in the current moment.

And that's the first step to make a strong AGI. The design of space problem is exactly the task of a toy model. Instead of focusing on an answer, trying to look into the constraints or the frame of the problem as broad as possible within which different kinds of answers can be plucked. That's, yeah, no, no, I like that a lot actually.

That's quite cool. I was trying to think about it before the class, and it seems like the sort of toy model, sort of the understanding aspect of toy models is kind of anti-accelerationist in a way, in that a lot of the accelerationists thought it's all about the world was too big to comprehend, you can't possibly catch up with it. Yes, yes. No, no, I know, I know. I know. I have already got comments on this that how did I end up to be a decelerationist? I mean it seems to me it's all about this feeling of incoherence or coherence,

making things increasingly more coherent rather than sort of being lost in this inevitable interference. Absolutely, absolutely, absolutely. Not simply coherence in two sense, in the sense of what when logical coherences between the theories and methods involved but also integration of all existing methods and theories that We have other . Any more?

By the way, Theo, how much time do we have? We have about 40 minutes still. Good, good, good, good. So we can still hear a few questions before I go on. Some people have been so silent. Yan. Artemis. Mikey. Sean, Alan. Maybe I would just like to ask,

I don't know if I lost this distinction or we haven't arrived to it yet, but the distinction between a big toy model and a small toy model, have we already reached that point? No, no, no, no, not yet. We are simply, we are still talking about, what you might call to be refined small toy models. Next session, we will talk about toy models, big toy models, and we will make a distinction between the two. The distinction between the two is you can think about it like this. So, think about it metaphorically for this point, at this point,

And I'm going to talk about it later in the next session. But with regard to what I was just talking to Adam. So, like the question of HGI, it's a design space problem in a Danetian sense. That you have, the idea of the intelligence is vast. So, you have a lot of assumptions, a lot of variables. And you cannot simply cover them by one single theory or one single set of methods. You have to have many, many plentitude of models and theories

without succumbing to bad pluralistic accounts in the sense that, well, these are all equal. No. They have priorities. They have rankings. It's just that we need to incorporate all of them, but at different contextual levels, when they are actually relevant, when they are addressing relevant problems. okay so in that sense you can think of a small toy model and I here I mentioned a level so you have different levels of AGI you know different levels of problems that should be answered as exactly like Kant's idea of conditions of possibility

of being an agent of being a perceptive agent So you have sensation, intuition, imagination, understanding and reason. And sensation can also have two levels, outer sense and inner sense. roughly speaking the simple neurological setup and two neurological setup that has an inner computational so to speak way of processing information so with that said the toy model

Well, you can think about it. If you are going to design this space, this kind of Kantian hierarchy of conditions of possibility for a thinking cell, or for that matter, which is exactly what the question of AGI, design of a strong AI is. The design of the entire hierarchy is a task of a big toy model. Whereas the refined, simple view of a toy model, a small toy model, is a toy model that can only address specific questions

of certain conditions of possibilities residing at a very specific level. So imagine this slices, this is a stack. The whole picture of the stack is your big toy model. Essentially your big toy model should increase the scope of the stack, whereas each stack requires small toy models. models. Toy models don't go vertically, don't deepen and broaden the scope of the stack, but simply they broaden the scope of each layer, level, or stack, each conditional possibility.

And that's, I would say, is a relatively decent metaphor for the distinction between a big toy model and a smart toy model in terms of their application and their scope thank you that's really helpful welcome thank you very much so I'm probably jumping ahead to what will be tomorrow's conversation, but do the simple

view and the idealized view of toy models... Tomorrow's... We don't have a class. I'm sorry, not tomorrow. Yeah, yeah. Next time. Next time. You are giving me a heart attack. Yes, next session. Yes, go on, go on please. Sorry. Are the simple view and the idealized view, did these only apply to the small toy models or is this a way that we should be thinking also of big toy models? I think certain components of refined, simple view should be also extended to the big toy models. It's just that the big toy models have an additional, what you might call to be injunction, variable, or factor,

and that factor is the fact that toy models, of course you can think about this with regard to everything that goes into what we have been talking about, that the small toy models target a specific phenomenon or model, Whereas big toy models are actually trying to see one phenomenon by way of the perspectives afforded by different theoretical systematizations

or understandings of different kinds of phenomena. So here, like imagine like this, that you have a small toy model of biology, right? Biology is a great field. I'm just reductively speaking here. Let's say that biology was just this small thing. Certain kinds of ranges of stuff that are happening in the physical domain. So you have a small toy model of biology, right? The big toy model is actually, the big toy model of biology is when the model tries to approach biology,

not only from the perspective of the biological concerns, the theory of biology and all of its constraints, But in a very simplified, carefully curated manner or idealization from the perspective of quantum mechanics, chemistry, so on and so forth. That's when you get a big toy model. so there are you see here big toy models are not that innocent they are actually more sinister than small toy models essentially you are dealing with much more

possibly what you might call to be pernicious assumptions, theoretical assumptions nevertheless it's what you might call to be the courage of the fool. Sometimes the fool actually has a better chance to solve a problem. But that only happens if you actually carefully start to see biology from the perspective of these other kinds of fields. or for example, you have biology 1 theory,

and then you try to see it from the perspective of biology 2 theory, biology 3 theory, biology 4 theory, so on and so forth. That's again a big toy model. It is, big toy model is by no means, so to speak, a less biased version of a small toy model. It's just that it is in fact what you might call to be that you are putting, in fact, all your metatheoretical cards on the table. So when you are dealing with certain kinds

of biology, obviously you have certain kinds of metatheoretical assumptions about other biological theories. Or if you are a biologist, in the general possible sense, you in fact have certain meta-assumptions about how molecules work, what chemical theories are supposed to behave, so on so forth and that's when if you actually put them on the table that's when you are in the realm of the big toy model but that doesn't simply mean that you have got away from the biases

or those thorny issues which plague a small toy model it's just a it's just like a different configuration and is the movement from theory to meta theories in the plural sense I think this is something we spoke about previously I can't remember which session but is there some way to model the relation or use a big toy model to to try to get at modeling big toy models in relation to each other? Or is that already implicit within doing big toy models?

I think it's already implicit. And of course, it's implicit, by no means explicit. It's implicit there, yes. And the reason that I'm saying that it's not explicit precisely because, remember what I just said that when a scientist uses a toy model, doesn't mean that he made this personally, but also it doesn't mean that every single member of the scientific community can understand actually the configuration of the model or what it's supposed to do, so on and so forth. It's assumptions. The same thing about the big toy model.

But here, with the big toy model, we are not simply talking about variables within a theory. We are not talking about, you know, different ways of explaining a certain phenomenon within established theory. we are actually trying to look at a specific mega theory from the perspective of its corresponding meta theories. But again, in that case, also, we are dealing with the same problems that I would say that we were confronted with when we were dealing with small toy models.

So... Let me talk a little bit about how possibly interpretation for autonomous toy models. Those of you who have forgotten the distinction between autonomous and embedded toy models, please do watch the last session video where I actually made this distinction clear.

Now, let us suppose that there is a considerably large class of autonomous toy models that cannot be interpreted as providing how actually understanding. We hold that, for example, Schelling's model of segregation is a member of this class. if some autonomous toy model fails to provide how actually understanding, what kind of understanding then does it provide, if any. The proposal is to take these autonomous toy models

to yield how possibly understanding, applying the refined simple view, as we have discussed it, A scientist S has how possibly understanding of phenomenon P by using an autonomous toy model M in context of C if and only if M provides how possibly explanation of P and S grasps M. For instance, we take the Schelling model to explain how it is possible that racial segregation occurs. And we take this to explain how, or for example, the DY model, to explain how it is possible that income distributions with the specific qualitative features emerge.

Both of these models only provide a potential explanation of a general pattern that is segregation and a certain kind of income distribution. And this pattern happens to be actually instantiated. For instance, the pattern of segregation is actually instantiated in Detroit, in Chicago. So certain kinds of income distribution is contingently instantiated in the United States. So both models and the evidence we have do not tell us whether they have correctly identified the actual relevant explanatory factors.

The question then arises why scientists are interested in how possibly understand it. as one appears to gain considerably less, appears, boldface, appears to gain considerably less from how possibly than from how actually explanations and their corresponding modes of understanding. direct for example is very quick in dismissing how possibly understanding as mere intelligibility which is clearly intended as a you know basically negative term and in taking how possibly

understanding to be necessarily on par with pseudoscience such as for example astrology In fact, how possibly understanding plays a central and legitimate role in research and in science education. More precisely, we hold that there are at least three central epistemic functions of how possibly understanding can actually help a community of scientists and modelers. Okay. One, the modal function, Jogan's question. Two, the heuristic function.

And three, the pedagogical function. So to conclude our session, I'm going to briefly talk about these three central epistemic functions of how possible is so understanding. one modal function. How possibly explanations are valuable if the phenomenon to be understood is a modal phenomenon, that is, if scientists want to understand whether and why some phenomenon is possibly or necessarily the case. One of the most famous illustrations

of the modal function of toy models is in fact Schilling's model of segregation. Schilling's model is concerned with the question whether it is possible to understand the emergence of segregated neighborhoods without assigning explicitly racist attitudes to agents. Remember that. Essentially all agents are just coin toss, heads and tails. This is over a 30 by 30 grid board game. Equipped nevertheless with a utility function, 30%. Shaling's model shows that in contrast to the view that segregation is necessarily a result of racism,

it is possible for segregation to arise in a population of Asians following the 30% rule. Even if the Asians would actually prefer to live in non-segregated cities, If the goal is to explain a modal phenomenon, then how possibly understanding an explanation is an appropriate tool for achieving this goal. Two, heuristic function. How possibly understanding via autonomous toy models is not always an end in itself. How possibly understanding often plays a heuristic role in the process of constructing less idealized and often, but not necessarily, also more complex models.

that latch onto the target system more accurately than the original toy model. For example, the DY model has inspired the construction of CCM model. or, you know, certain kinds of toy models in biology have allowed us to create far more refined models about, let's say, hormonal pathways. And if you really want to look at this, you know, example of biology,

then look at the work of Giuseppe Lungo, his recent book, which covers a lot of stuff about modeling in biology. And three, pedagogical function. The how-possibly character of autonomous toy models is often also used for primarily illustrative purposes in science education.

These models enable students and researchers to quickly grasp the idea behind the solution to a problem or the description of a phenomenon. Generally speaking, the pedagogical function of toy models is to enable students to learn how to calculate with and how to use a particular model or theory. Once the students have learned how to calculate by practicing with a toy model, the training is put to different uses in the case of embedded toy models and of autonomous toy models.

Regarding embedded toy models, science students learn how to calculate with the embedding framework theory by practicing with toy models initially in order to prepare the students to mathematically handle less idealized, i.e. with more details and sometimes also far more complex models of the embedding framework theory later on. Regarding autonomous toy models, the goal of practicing with the toy model is different. The acquired ability to handle an autonomous toy model mathematically enables the students to make use of the toy model in a modal or a heuristic function.

You know, more like the design space problem that I was talking about here. And sometimes it is not good to learn the rules of the game, but just play with the blocks of Lego that you have at your disposal, rather than simply trying a certain kind of, based on the instruction, a certain kind of a Lego model. So this is more on the side of the play. And play always precedes ontologically the game. The rules of plays are never set. In fact, play doesn't really have rules.

The only rule that it has is the interaction. how possibly such and such Lego blocks can fit onto such and such other Lego blocks. That's it. You do not need to know, understand how a robot looks like, how a Death Star looks like to make a Lego model of them, namely the game. So to sum up, we have argued that some central examples of autonomous story models yield how possibly understanding as opposed to how actually understanding. Moreover, we have claimed that scientists value how possibly understanding

because it has a modal and a heuristic function which is indispensable in scientific research. but also it has a pedagogical function which is absolutely necessary in scientific education okay Dear friends, say something.

if someone doesn't speak then i will but i'm going to wait a second for someone else to say something okay anyone any person Apparently not.

I kind of, well part of my question wants to come back to the earlier explanation of scientific understanding and try to make some more explicit connections to that and the what you call the epistemic functions of the how-possibly descriptions, but maybe they're also the epistemic value yes everything when I say that systemic functions you should understand that so I give I gave three you know and I think the educational one is really the epistemic value rather than epsilon function but

But nevertheless a value that plays a function. Right absolutely. What is the, why do you I guess disavow the actual in favor of the plausible? No I am not absolutely disavow actual. Essentially, you see. I know it's not a method called either or, but like. Yes, yes, absolutely. It's not an either or. No, no, no, absolutely. So let me tell you this. Essentially, so the either and or scenario

can get you into the hells of naive empiricism or the hells of modal realism, a la Lewis. Now the thing is that the idea of actual is quite vague. So as possible, they're co-constituted. Essentially, as I mentioned to you, If a scientist actually thinks that observation means something that a frequency of what a tape recorder records or an ideal observer observes, it is absolutely pre-critical, pre-scientific.

It's just human bullshit, so to speak. The idea of observation is not either or. It has sensory data, perceptual processing, a la Helmholtz. It has rules, a la Wittgenstein, but it has something far more... even important as the tissue that can actually, in a Kantian sense, bridge the domain of rules, the domain of the sensory observations. And what are they? Poincaré called them hypotheticals.

Today, we call them toy models. so absolutely I'm not against that it's just that I think that these ideas of actuality reality possibility if not explicated adequately they are just metaphysical fairy tales they don't have anything any role to play in our understanding of how science works. Actuality definitely has a role, but actuality, you can, well, and with actuality, you can go two ways. Actuality in the sense of something being substantiated

in a kind of empirical sense, at least I'm saying this. I'm sure there are so many ways that you can talk about actuality, substantiated in an empirical sense, namely in terms of observational data, or two, in terms of Hegel. That which is actualized is that which is true. And that which is true is only dialectical and speculative. So even the idea of actuality doesn't have, there is no single term for it. There is no agreed upon idea of actuality. So then why do we need to actually, part of the pun, talk about such terms without explicating them?

We have to. It is our business as philosophers to explicate all of these terms because these terms are as vague as a fog at... 5 o'clock in the hottest day of summer. When you get both the headache, you get the fog, you can't get back to home, so why do we need to go for these kinds of stuff? We need to explicate them. These are not, yes, they are philosophical terms, but philosophical terms have no, so to speak, value

theoretical or practical unless and unto we give them the proper context which they deserve until then we could go all sorts of basically ways we could defend actuality as an empirical thesis, we could defend actuality as a totally rationalist, stigmatic thesis. And at the end of the day, we see ourselves

at the Pyronic bridge. If all of these things are so vague, And if all of those things which are vague can be equally justified, then everything can be justified. Then what are we doing here? This is not how science works, nor should philosophy, I would say. Absolutely, we have to take the idea of actuality in whatever context we mean seriously. But we cannot take it seriously unless and until we explicate what we do actually mean. And then in that sense, actuality, you see that even in the empiricist or the dogmatic rationalist sense has something to do with possibility.

Just like the concept of observation, sorry, the concept of observed data, has something to do with the concept of observation in the modern scientific sense. And the concept of modern observation has something to do with that which is not observed. Hypotheticals, fictional entities. And I do really believe that Brando makes a really fantastic, I mean, even though I'm growing wary of the Brandomian pragmatism, but I think that he actually makes a fantastic point in his essay on modal vocabularies and empiricism. You just can't have empirical actuality without modality.

I always thought that your defense of actuality was in some way consistently rationalist until right now, and this has been really helpful. no no I think that it was it was necessary for me to be aggression on the on the on the offensive side precisely you see even though that I know that you are going to read all of this stuff but nevertheless let's say look like when an educator should be like a diplomat So you never actually tell anyone what they should read. You just egg them on. You know, that's the whole point of corruption.

So the point is that the idea is that I have to first eliminate the threat of those kinds of naive understandings of actuality such that we can actually do agree on what we are talking about with regard to the concept of actuality. And from there, we are going to make the Tower of Babel, the philosophical Tower of Babel. Yeah, I mean, I completely agree with your understanding of actuality right now. And I think this changes my perspective on what you've been. I mean, the question of, you know, to what extent do we understand this as a speculative enterprise?

And how do we define speculation in this way? Yes, yes, that's a good question. That's a damn difficult question. And I don't think that I have a good answer to you at this point. But literally, take this so seriously. That's okay. When people talk about actuality, usually they mean some sort of empirical smuggling, right? Sure. But imagine that Hegel considers himself, and also Hegelia considers him, as a philosopher of actuality. How can it be? How can Hegel be the philosopher of actuality? Well, because his understanding of actuality is fundamentally idealistic

and even teleological. Yeah. You see, this is one of the things that I think is both a curse and also a bliss for those of us who are working in this field, namely philosophy. Essentially, in the history of philosophy, like any kind of history, but more so in the history of philosophy, we are dealing with multitudes of concepts, connotations, context, terms, so on and so forth. and literally history cannot be renewed in this sense the history of philosophy unless and until we can actually converge on some agreed upon perhaps

even compromised idea of actuality and then from that we venture and branch out and that's how history of philosophy will be reborn, will continue, as it has always done. But until then, essentially, we should think of the history of philosophy, its terms and its connotations, as traps. These things are way too vague to, for example, us talking about the idea of the real, you know? Right. We don't know what we actually mean by it. I really don't know what I mean by real, to be honest with you.

Yeah, I think like what, like the question about speculation, at least from what I'm understanding of your work right now is that it, the domain dimension, yes. The domain that we're thinking about is through the scientific exploration of what is, but then, you know, what is smuggled and what is not, etc. And on that domain, can we speculate? And that speculate, yeah, sorry, go ahead. Yes, yes, that is one. There is another. So speculation, when I'm using the word speculation, I mean it's in completely Hegelian sense.

So a speculation is what you might call to be the movement of Aufheben, suspension, sublation. Suspend that which is immediate. Show that which was deemed as immediate was all along mediated. And then rinse and wash. Do the same thing. So this is, for Hegel, is a movement of speculation. And so, for example, when we are talking about speculating about the concept of the real, the first thing that we need

to do is the Carnapian tactic. We need to agree upon what exactly context we are talking about. And that cannot happen unless and until we a conversation which we are but of course this conversation can you know even more intensified and then from there we start to identify certain kinds of problems of immediacy which are illusory we are going to suspend them we are going to ascend to a different level where we have a better more refined concepts of the real, real too. OK? It's a new video game.

And then from there, we are going to do the same thing that we have been doing at level one, so on and so forth. And of course, there is the movement of speculation to suspend that which is immediate for the sake of that which is beyond. People think of this as some sort of esoteric mystical recipe, but there is nothing mystical about it. It's purely a logical, rationalistic, and scientific enterprise. Go on, Joven. No, no, no. I think this is, I think, yeah, this might be like the only last thing I have to say,

which is that I was really thinking about, like curious about the relationship between the target phenomena and because the question is, you know, even the target phenomena cannot be remained at the level of actuality that we think that it was. No, no. The target phenomena, and this is how I'm understanding the real in the sense that the target phenomena is indeterminable in some sense, not as in a how do I put this there's no distinction to the target phenomena has to be able to be open enough to encompass

the myriad capacities of integration and allowing all these modal relations to do their work or whatever. Yes. Yes. So essentially, I know what you are saying. It's essentially, okay, actually this idea is far more, I would say, is in tandem with Poincare's idea of chaos. Okay? Chaotic systems. in the sense that a target phenomenon... Okay, first, let's not talk about that real as independent. Poincaré doesn't really actually think that real is independent, okay?

In the sense that we have a transcendental co-constitution between the epistemic norms and certain kinds of physical behaviors. And this coordination is important, and it should always be kept balanced and disbalanced such that we can make better theories of the behaviors of a real system, of a target phenomenon. But now here, sorry, Poncarette says something like that. that, okay, with a dynamic system, that's all, sorry, with a static system, this is all great and good, this coordination.

But with a dynamic system, you are in a different kind of trajectory in the sense that that so-called real might actually, over time, behave fundamentally differently than its initial conditions. And then if you actually try to frame the system, this target system, from the perspective of initial and boundary conditions, then how can you actually keep the pace of tracking the evolution of this system in chaotic trajectories? That is the idea of the real. it's not that it is above episteme

or it's some sort of ontological ineffable it's just that it is real in quite an objective epistemic sense even from the perspective of our rudimentary modest epistemological resources This is something I completely agree with. I guess this was in some way my attempt to stumble into that. Definitely, definitely, I really do suggest to read two works. One, the classical works

on Poincaré mostly are translated to English to, do you know Jean Petiteau? Which should I read of Petiteau? Well Jean Petiteau, I would say that read some of his essays like the one that he wrote on perception and of Searle and he tries to look at the phenomenological eidetic exploration in terms of basically this vagueness or what you might call to be this unstability of a dynamic system in the sense that the dynamic system can have chaotic branchings

trajectories. Let me just one second, one second. Let me give you the title. It's called Cognitive Morphodynamics. Thank you. Okay, dear friends, any question, anything, or shall we just stop at this point? Perhaps now is a good time to end, unless someone has an urgent question, just because

40 right now. All right. I'm going to take that as a we'll conclude next week. Silence. Okay. Love you everyone. Have a great weekend and we will see each other next week. Sounds good. Bye everybody. Thank you.