Simulating the World & Remodeling Philosophy (Session 5)

Hello and welcome to the fifth session now of Simulating the World and Remodeling Philosophy. I'm going to pass the mic to Brezenegger Stamina. Thank you very much, Theo. Okay, thank you everyone. Again, apologies for being late. Okay, without any more delay, let's hear if you have any questions. questions. If not, I will just delve in into the materials. Anyone? No question?

Nothing? Okay. Meredith or Theo, either of you. I'm actually good to dive in, even though I could probably think of a question or two. Sure, sure, sure. Please do. Joven, that's not news anymore. go on fio sorry don't let us distract you no i will i think something that i was uh

kind of dwelling over this past week too we're just this kind of relationship that models have to theories the reciprocal relationship that they have how how theories are that which kind of couch the possible the framework of possibilities for our models but also i guess i'm curious just moving forward i actually don't want to spend too much time on this how it's possible that models might force us to reinvestigate our theoretical assumptions right uh well i think uh one of the things that you know i i actually thought uh that uh you know just moving to the toy models

today would be a little bit uh too much of a leap uh precisely because uh the answer to your question is precisely in the very question of similarity the vague notion of similarity that we have been talking about, right? In a way, before getting into the details and elaborating the concept of similarity as applied to models, you might think, you know, in terms of Thomas Kuhn idea, theoretical revision, at least the expanded core revision, you know coming back to our previous course and the logical structure of theories, scientific

theories, you would say that revision in the expanded core of a theory, for example moving from Keplerian to some sort of Newtonian framework, requires something more like, I mean, not as a sufficient factor for theory change or theory displacement, but nevertheless you require some sort of instrument, some sort of apparatus in order for you to be capable of, within the context of your current theory, accumulate as much as anomalies you can.

So in a sense, precisely because models and scopes are far more restricted or applicatory in a pragmatic sense than theories, models I would say have a better chance in coming across empirical anomalies the kind of which allow you to over time say that well this theory can no longer a structure or explain these kinds of pattern regularities of data we previously thought it was sufficient for

So that's I would say is like the kind of a naive answer, first naive answer to your question. Well of course I will go and detail different kind of factors, different kind of criteria that are usually understood in terms of how models can actually lead to the revision of models. one of the most important ones, which is of course a humble one, precisely because models are essentially at some point need to be pragmatically understood in terms of their scope of application, is that precisely because their scope of application is quite concrete and far more restricted

than that of a theory within which they work, they have much more chance to come across the kind of anomalies are required to basically lead, lead, again, not a sufficient criteria. We are saying might, might lead to theory dislodgement, moving from one theory to another better explanatory theory. OK. It's like, you see, models and theories, I mean, kind of overextended analogy

is like the chief of staff and foot soldiers. The chief of staff doesn't really understand how the battlefield look like, and it's all messy problems. Whereas the foot soldiers do, because they have to interact with it. Right. Okay, that gives me fodder for thought. I'll continue to dwell on it. You hold on your skepticism. Well, I think that obviously the big question is just like, is just like what are the types of kind of what certainty do theoretical frameworks have

if this kind of like the nuts and bolts traction interaction with the empirical ground is happening at the modeling level i'm not sure if that is even a skeptical question or if it's just right It's not a skeptical question per se, right. The thing is that with regard to the question of certainty of theory, the question of certainty itself, the concept of certainty, when applied to theory, is quite a vague concept. Precisely because we have different levels, different scales, different grades of the concept of certainty.

You see, the certainty of theories is not about the certainty of empirical data, but a range, a threshold that within such a logical structure, such and such data and symbols can be regularized, can be explained, can be described efficiently. efficiently again within the parameters of the set theory. And here the models, what you might say that precisely because they zoom in on a specific phenomenon covered by that theory, they have a more chance to in fact revise the threshold, the parameters,

of theory's idea of certainty or truth. But we are not going to talk about truth at this point. Certainty is, I think, is a more modern concept that, you know, the kind of idea of truth that continental philosophers hate. And frankly, it is truly quite a thorny issue and needs to be tackled in its own terms. OK, thanks. I think that answers my question for now. And Adam, would you be able to say what you just stated

on the sidebar, if you could elaborate a little bit? ADAM POSNERAVSKY- Oh, this is me trying to work my way through HES, but not actually working all the way through it, reading that was sort of shared at the stuff of course um but uh she seems to uh lean on the idea of analogies very hard uh when it comes to models and the this the analogous relationship between you know um models is applied to different domains right and and then she I thought it was, whereas Weisberg talked about a lot, rather than analogy.

One of the things she does seem to get from it, even though it seems like a vague concept to me, but yeah, the dialogue piece. then she is able to get this concept of the observed predicates and the sort of predictive predicates out of that. I didn't find it as intuitive, but I thought that idea was kind of cool. But she's getting at it via analogy, which seems like a much bigger concept. Yeah, I think, I mean, surely at some point, at some level, if we are trying to do, you

know, kind of the task of explication in the Carnapian sense, we might say that at some level of vagueness, analogy and similarity can be seen on the same level, basically become synonymous. But the thing is that yes, I generally think that the question of analogy is even more vague than the question of similarity, precisely because similarity is in fact, does in fact have a kind of theory of equivalence relationship. You can explain similarities in in terms of equivalence relationships and certain kinds of parametrizing functions

which can, you know, constrain certain aspects of the model to certain aspects of the world. Whereas analogy is just, or maybe it is not really the fault of Hess, it's precisely because historically the concept of analogy can be said about metaphors, metaphors, it can be said about allegories, all sorts of stuff. And we just don't know what we actually mean by analogy, at what scale what we are talking about it. Right. I mean, it was a bit funny because you also seem to be wanting to talk about models without talking

about models, in a way. I mean, I guess compared to I think Weisberg probably goes to more effort to have to just establish precise definitions maybe where they aren't ones. Right, right, right. Okay, this is what I mean when I mean model. It's exactly this. And, you know, this is the sort of anatomy of a model. Whereas Hess is sort of trying to say, well this is why models operating with analogies are valuable, but sort of is not quite so direct about presenting a definition.

Right. And the thing is that, I mean, given the idea, and by the way, if you guys have noticed that Michael Weisberg is on my Twitter account, so you should ask him all these questions. Pester him. So basically, Michael, I think one of the reasons that he chooses the concept of similarity versus Hess's idea of analogy is precisely because similarity, given the fact that Weisberg is you know this kind of pro the mathematical modeling structure. The concept of similarity, namely isomorphy, can be talked quite coherently in terms of mathematical

structuration. One way to go about it is the concept of equivalence. How two things can be said to be equal which is basically the very question at the fundament of all mathematical innovations mathematical structuration both in uh you know pure mathematics endeavor and also in the mathematical physics so there is a it's a better way of systematizing the concept of similarity than the concept of analogy for which we still don't have that many, what you might call to be, fine-grained mathematical concepts to kind of articulate it without somehow, you know,

lapsing back on the metaphorical, allegorical ideas. Any more questions, anything? Meredith, Joven, Jeff, Artemis, Ian, Sean, Andrea, Alberto, I can call all of you. I mean, I don't want to detract too much from the discussion today on either toy models or whatnot, but I sort of, I feel that the realm of theory and the, you know, so if we

take the realm of theory and using this idea of truth, which we don't talk about, or correctness as the sort of litmus test, and here in sort of the world of models, and we begin with the idea of the error, but really there's a much wider range of sort of parameters that we're trying to judge our model against, you know, like similarity. You know, I mean, we've almost, we're almost like in this, in this theory world where we only have one standard, the error checking mechanism, but really there are a lot more standards that we need to be comparing our models against. For example, for example, would you be able to? So similarity to real world events. So like how isomorphic is my model to the physical world? Like what, you know,

that's like, you know, rather than just looking at the error on the back end, look at the construction on the model on the front end and having a metric for the similarity, how similar is my model to that physical event which I'm trying to model? And there are all different kinds of ways of measuring that similarity. But even, and even to go back, we can talk about the construal and discuss, and I even have problems with sort of like the the four parts of the construal.

But my whole feeling is that we sort of have, if we look at theory as having this intellectual history coming from like Aristotle and logic and the syllogism and how do we talk logically and models as, well, how do we represent the world, sort of like in all of the particularities of the world, you know, you know, how, you know, what are the, what are, what are those things that we can only, that we can use models for and how can we think about models? Like how can we lose the legacy of the theory? Like, I don't think, I don't think we can just say models are a new way to talk about theory. Like models are a new way to talk about the world. No, no, absolutely. They are not, they are not, they are absolutely not ways of,

of course, there are some people like Van Fresen or, and some other people who do think actually theory is essentially what you might call to be assemblage of models. But that is a fundamentally different kind of question and what they mean by model is not essentially the kind of models that we have been talking about or what Weisberg is talking about. However, I disagree with this idea that models do start from what you might call to be the error constraint. Essentially, the idea of similarity, as we were going to talk about, established from a kind of a presumption of mathematical physics in a modern sense,

in the sense that geometric structure, geometric structure, should be able to give us a semblance of isomorphy, namely translation of atomic facts to axioms of geometrical system, whatever that the geometrical system is. For example, I see a red dot time t1 space 1, location 1. Now, and I see red dot t2 location s2.

Now, the idea of similarity here at the most basic minimum level kind of is similar to Vienna Circle, to Carnapian idea, of for us to be capable of establishing a similarity or a semblance of isomorphism, we should be capable of translating such elementary experiences, observational protocols into geometrical statements within a system of mathematics. By real numbers, geometrical relations, so on and so forth. That's all about it.

At that level of basic similarity, we don't even talk about the error tolerance, so on and so forth. It is simply the idea of similarity, how to establish a kind of rudimentary isomorphism, which always, of course, entails a massive amount of idealization about the location and time and the color of the so-called dot. in the visual field, into a kind of geometrical system. So essentially, from that perspective, Weisberg is kind of right.

Essentially, the entire whole idea of modeling begins somehow as a presupposition. But of course, from a philosophical standpoint, it should be explained why is that we can establish such an isomorphism to begin with. But I'm not going to talk about this right now. It's just that everything that we are going to establish as a rudimentary level of isomorphism should be, in fact, in one way or another, be capable of establishing a translation point, a bridge between observational protocols, namely elementary experiences of an observer. And that observer does not need to be human.

It can be just a tape that is sensitive to light. And the location, the spatiotemporal location of such observational protocols or elementary experiences within a geometrical system. That's all isomorphism at that very, very beginning means. It doesn't require any sort of fidelity criteria, error tolerance, truth confirmation, certainty or anything. It's just like a very purely descriptive.

Any question on this? Or shall I begin finally? Okay. And by the way, one thing that I was going to tell, you know, once we get to the toy models. Actually, the two of the models that we have already been talking about, Volterra's mathematical model of predatory-prey dynamics and Thomas Schelling's segregation computational

model, from the standpoint of modeling are considered to be in fact toy models. And we'll argue why they are toy models in fact. They are usually categorized under a category which is called the small toy models. A small in the sense that they involve extreme idealization and simplification. And we will see their task. Their task of modeling is quite very different from the kind of more lavish modelings, like model of, I don't know, climate change, chaotic behavior, so on and so forth. We'll talk about this.

But OK, let me begin. So before even starting, the question of similarity and how we can formulate it. Allow me to continue something that was left off in the previous session with regard to the different stages of modeling. If you remember, the whole point was that by elaborating the structure, the controls and the distribution of model, we could finally formulate rather more accurately accurately different stages involved in the enterprise of modeling. Now let me just go a

little bit in that direction and then I will get back to the question of similarity. So, you know, if you remember, Volterra began his investigation of Adriatic fish, not by looking directly at these fish or even the statistics gathered from the fish markets, but by constructing a model. This is characteristics of the first stage of modeling. Volterra imagined a population of predators and a population of prey, each with only two properties. Setting this idea to paper, he wrote down equations specifying the model that he had imagined.

Theorists do not often record the details of this process, so we do not know how satisfied Volterra was with the initial model. Perhaps it did not match the model he had imagined, and so he refined the model. Or perhaps he had correctly specified the model he was imagining and was able to proceed to analyze the model. In either case, the first stage of Volterra's analysis, involved in constructing something. He did not start by looking for patterns in data, what you usually do in the earliest stage of theory construction.

He began by constructing a model. Once Voltero had constructed his mathematical model, he could go on then to perform his analysis. And this is the second stage of modeling. Now, Voltero's analysis involved studying the effect of a perturbation on the ratio of the average abundance of predators and prey. The results of Voltero's analysis proved useful for determining the behavior of the Adriatic population of fish. But it is important to see how these first two stages of description and analysis are autonomous from any application to the real population.

So this coming back to what I was just trying to relate to Meredith. this. Although guided by what he knew was happening in the Adriatic, the second stage of modeling involved finding out some very general properties of predator-prey models, ones that apply far more widely than the particular case he had in mind. For example, Volterra's discovery about the effect of fishing on the predatory prey balance was merely an instance of a more general principle he discovered in the analysis of his model he discovered what is now called the volterra principle general pesticides whatever that pesticides might be you

you know real pesticide war climate change what whatever you might call it essentially some intervention that kills both predator and try increase the relative proportion of the pray this principle was first discovered using the model described in his original predator prey model but Volterra and subsequent theories showed that it emerges from many other predator-prey models as well. This discovery is an example of how analyzing a model can be autonomous from any particular system being studied.

So this is a moment where even though models are quite narrow in their scope and assignment, you see that at the level of analyzing a model, you might in fact arrive at certain kinds of patterns, regularities, or first order principles, which are no longer within the purview of your particular model or its particular scope of application. There are more on the side of the theoretical mobilization of the model.

See, this whole idea of Walter's principle is more of a theory rather than just a specific model applied to a specific aspect of a real phenomenon. Questions? my apologies so of course there are other types of cases that are even more striking examples of the autonomy

of the first and the second stages of modeling description and analysis some modelers construct models simply to explore or illuminate a hypothesis and there are quite a great range of such models in the history of science and engineering at the very least starting with Archimedes for example population biologists often examine very simple models of sexual and asexual reproduction in order to better understand the evolution of sex these

These models are not, however, intended to describe any actual organism. They are far too simple for that. Their importance, however, lies in helping us to understand very general facts about the differences between sexual and asexual reproductive systems. In this type of case, the model itself is clearly the object of the study. remember what I said in the previous session, that the whole idea of scientific model should not, in fact, be restricted so narrowly to simply its application to a real-world phenomenon. In the majority of cases, in fact, in the history of science, we use models to study

the imports, the significance, and implications of models, other models. Also, it is possible to study a model of a phenomenon that is known not to exist. For example, Arthur Eddington, a British physicist, once wrote, we need scarcely add that the templation in natural science of a wider domain than the actual leads far better understanding of the actual.

For example, following this Eddington diktat, Fisher explained that the only way to understand why there are always two sexes involved in sexual reproduction is to construct a model of a three-sex sexually reproducing population of organisms. Constructing a model of such a phenomenon is the only way to study it because, by a stipulation, the phenomenon does not exist. Modelers often interested in phenomena, such as three-sex biology, you know, perpetual motion machines, or non-aromatic, you know, cycloristians,

because insofar as we can understand why these phenomena do not exist, the idea of the counterfactual, we will have gained a better understanding of phenomena that do exist. Again, in this case, it is clear that the model and only the model is the object of a study. So, you see, not all models are about an actual phenomenon in the world. Models, for a great part, can be counterfactual models. just the way we use modal vocabularies to decide the context, the regularities of certain kind of patterns.

Like for example, you know, that idea of if this would have been the case then what would have happened? Right? This is a kind of a modal inference, a very rudimentary one. Now you remember that I mentioned I think was it, Theo, was it in this course or was it in the previous one with regard to the match? That was last question. Okay, so just let me just quite briefly. So this kind of counterfactual line of reasoning allows us to actually talk about some sort

of empirical regularities and their context without simply settling for some sort of primitive fundamental regularity. It is to the extent that we would say, so whenever, so here in the atmosphere I use a safety match, I strike it against a frictional surface and it's being lit, okay, that's fine, all good. But this is, however, it is not the way that science works okay science actually uses quite heavily counterfactual scenarios exactly like this counterfactual

modus by for example coming with a counterfactual and model scenario where we say that if we would have a struck this safety match against a frictional surface in the vacuum then what would have happened? Well you see it allows you to move from one context that is quite mundane and empirically available to a kind of a counterfactual context. Well if you when you do that then you notice that two things

can happen. Either we are talking about an old match, old match heads didn't have an oxidizing agent, it was just like a combination of sulfur. So basically it wouldn't have been lit if you had struck it against a frictional surface in the vacuum. So that's old match. But a new safety match, in the match head, you not only have sulfur combination, but you also have an oxidizing agent, which means that even if you go to vacuum and you repeat this experiment that you had just done

in the Earth biosphere, the match head is being lit because of the oxidizing agent. However, the matchstick doesn't burn. So you are seeing a range of counterfactual robustness for what it means to light a match by way of these counterfactual scenarios, these counterfactual models. Questions here? Jovan? Counterfactual models are not empirical then.

No, they are not empirical per se, but they can, doesn't mean that they cannot be empirically applied. Yeah, so they're non-empirical but... Yes, yes, yes. And this is in fact one of the main assaults of what you might call to be conceptual rationalists. rationalists against you know the kind of you know classical empiricists precisely to cite evidence in the history of science where the empirical observations don't give you an actual regularity it is only when you try to

elaborate what you have seen by taking them to the realm of hypotheses and model counterfactuals that you would be able to articulate something called a natural law or irregularity how exactly do you specify or articulate a counterfactual scenario anything that involves with defeasibility. Essentially, within this structure of a common sense language, you can see the structure of such defeasibilities, defeasible statements,

in the grammatical order of this would have been the case, what would have been happened. You see, here, would is a defeasible modality of the alethic kind in the sense that we just yet don't know what would happen there. And the whole point of science is to discover defeasers for a set of statements such that it is capable of articulating counterfactual robustness. A defensor in a linguistic sense is essentially exactly like the idea that if I take this match, this modern safety match, to vacuum and strike it against a frictional system, what would have happened?

You see, the idea that it is being lit is a defesa. Because from the perspective of the old match combination, the biosphere, oxygen reaction pattern regularity, we only would be able to stipulate or postulate that only requires oxygen, and it has nothing to do with basically the exact combination of match. But here we show a diffuser that all these regularities require

bracketing of certain kinds of context sensitivity. in order to be considered, you know, a pattern, a regularity, a piece of invariance or law. And here you can see that, for example, the same thing happens in, you know, in early experiments of, you know, quantum phenomena where for example you say that you know you you know you have some sort of you know kind of activate an electrode inside a polarizing

chamber and then you see its phenomena well it all looks good but then you would say that what would have happened if I had changed the structure of the polarizing chamber. And then you get different kinds of stuff. And through these kinds of concatenations of counterfactual scenarios, you say that under such and such constraints, namely contexts, this would be an invariant of how basically a quantum phenomena would behave so i'm just going to use kind of iterative language but it seems like you don't go about constructing like a positive counterfactual the counterfactual

is something like a negative defusing absolutely they're always diffusory diffusory they're always non-monotonic. They always need to be, do something nasty to a previous observational empirical statement. Yes, that's exactly what they do. And the whole thing is that, you know, the examples of modeling are quite similar to that precisely because they allow us to, in fact be capable of refining empirical statements and elevate them to the order of law a natural law or a principle

or an invariance or a regularity because irregularities are not given by empirical data, they require counterfactuals they require modal vocabularies defizors No, I'm not going to shame you. There is no way to shame a skeptic. They don't know any shame to begin with. Okay, let me continue.

Okay, so as these cases make clear, modeling can terminate after the second stage, namely the analysis stage, and involve no assessment of the model-world relationship, namely the the concrete application of a model to a real-world phenomenon. When however it does terminate here, any analytical result gained in the analysis apply only to the model, not to the world. But even when modeling does not terminate after the second stage, the first two stages involve the study of a model as an autonomous object in itself.

This is true even when these steps are carried out with a particular real-world phenomenon in mind, such as, for example, Volterra's analysis of the predator-prey model. Now the third stage of modeling. We can say that modelers sometimes proceed no further than the second stage of modeling analysis. much of the time theorists construct and analyze models in order to study real-world phenomena. This requires a third stage in which the theorist attempts to coordinate a model with a real-world phenomenon. The model-world relationship is still the subject of much debate however in literature

and detailed analysis of the third stage of modeling will depend crucially on how this debate is resolved. So I'm going to only describe the third stage in outline form, you know, following Weisberg who believes that this to be more or less compatible with all of the major accounts of the model world relation currently under discussion, you know, without getting into the debates of model world relationship. No, comparing a model to a real world phenomena involves first preparing the real world phenomenon using what Supes calls a model of the data and what Weisberg calls a parametrized target system.

Essentially, the theorist has to subject the phenomenon in the real world to a process of abstraction, deciding which aspects of the phenomenon will actually be considered. The result of this process of abstraction has variously been called a physical phenomenon or a target system. The next step is for the modeler to represent the target system mathematically assigning variables and parameters to the properties of the target system. system how to articulate the next step of the process depends on one's account of the nature of model world relationship in the status space version of the

semantic view as you know defended by van frozen the next step involves comparing the states of the target system to trajectories associated with the model. In Van Fressen's version, the curve which best fits the data must be isomorphic to one of the model's trajectories. Given the unrealistic demands of this requirement, several authors have weakened the requirement considerably. For example, Costa and French. now the relationship should be made even more flexible from this kind of narrow constraint

the assessment of the adequacy of a model can only be made with respect to the fidelity criteria which we described in the previous session a theorist has chosen sometimes this will involve something as strong as isomorphism like a map and territory, but other times only qualitative agreement between measurements and the model will be required. This is settled by the pure pragmatic interests of the theorist or modeler, not universal criteria. For example, let's consider another example of modeling, paying particular attention to

the third stage. Molecular models are often used in the course of explaining three-dimensional molecular structure. Say, a chemist is interested in explaining the major conformational change caused by a tertiary butyl substitution in cyclohexen. She might employ a classical molecular model in this analysis. In such a model, bonds are treated as rotating harmonic oscillators and atoms as the masses being oscillated. The model further builds information about the spatial position of the atoms

and the potential energy associated with these atoms at different distance from one another. For a very simple model such as this one, the tertiary-butyl substitution would be simply be treated as a change in the mass and in the amount of space filled by the substituent. Now, of course, using these techniques, collectively referred as molecular mechanics, the theorist can calculate the minimum energy conformation of substituted and non-substituted cyclohexane models.

She can then analyze the model, examining the factors which predict and explain structural changes, such as torsional strains, van der Waals strains, or esteric hydrants. for this particular substitution, the conformational analysis of molecular models suggests that the drastic change in the conformation is a result of a steric interaction between the tertiary butyl group and the axial hydrogen, which is sin to it. So far, the chemist is still at the second stage of her analysis, having accounted for the properties of the model system, but not the real system. However, if model is similar to the real molecule,

but with respect to aspects of its structure and the effect that substitution has on the model's conformation, then the chemist can move to the third stage of her analysis. If the model molecule is similar to the real molecule, then she can indirectly account for the conformational change as being due to the interaction between T-butyl group and the syn-axial hydrogen in the real molecule. The chemist has indirectly represented and analyzed structure and properties of the real molecule by describing and analyzing a model. model the chemist description of her model can serve as a description of the real phenomenon

because of the phenomenon's similarity to the model given the construals and fidelity criteria had this theorist engaged in a in abstract direct representation model adr approach she would have attempted to describe the actual phenomenon without the mediation of the model. Therein lies the key difference between modeling and abstract direct representation. And another example which we discussed was Mendelius' periodic table.

Shall we have a break before you all have meltdown? Sure, let's take a five minute break. Okay, excellent. Cure the meltdown. Thank you.

Thank you. Thank you.

Thank you. Thank you.

Thank you. Thank you.

A question for everyone. So, you know, trying to tie this back into, you know, the canonical rivalries and philosophy. with all so far with all the stuff that we have been discussing with regards to the question of models, empirical data, isomorphism, similarities, on so forth, how do you see the question of models can be

re-staged reposed and mobilized within the competition between empiricism and rationalism Any answer? Jean-Pierre? You feel so comfortable. I'm not going to let you to be so comfortable here. Actually, I wasn't because I was struggling. No, I'm just kidding. My internet connection

for like 40 minutes at the beginning of the session. So I missed most of the debates. one i'm not sure i'm equipped just to jump in right now post brazil election yeah you know yeah okay anyone can you restate the question at least so so the question is that uh you know obviously you know we are not just trying to do uh we're not going to just talk about this kind of you know purely technical details without in one way or another bringing back the question

of scientific modeling into the pre-neal concerns of philosophy one of which is the rivalry between rationalism and empiricism. And in that case, how can we, in fact, after knowing a little bit about the procedures of scientific modeling, their import, their significance, their specific characteristics and constraints, how can we look at the question of modeling at the intersection of the rivalry, the competition between empiricism

and rationalism? Yeah, that's a great question. We don't have an answer for that. It's a good question, nevertheless. Anyone? Joven, Mari, Justin, Adam, anyone? I'm quiet because I don't know. When I know, I will speak. I have a question about that. I would wonder if ADR would be more on the side of the empiricist tradition, whereas modeling would be some attempt at achieving a synthetic view between empiricism and rationalism. I wonder because there seems to be like a temporal, like a temporal, a temporality to modeling that anticipates its own capacity to be mediated by theory.

And so in that sense, I don't think it's, it would be like purely like within the realm of like rationalism. But I don't know. That's, that's a super. That's really good. Okay. That's really good. Well, to be honest with you, I myself have absolutely zero, you know, at this point, clear cut position on this question, even though, you know, I consider myself broadly as a rationalist. But the thing is that I wouldn't actually say that, you know, direct ADR, you know, abstract direct representation is essentially empiricist. It's just that it is direct within the realm of theory, you know, and how theory somehow basically finds its bridge with the observational data.

Whereas for models, there is a lot of more details with regard to the criteria of fidelity, analysis of the model, so on and so forth. But I do agree with your statement that, yeah, from a very perspective of a very quite technical analysis of scientific modeling, it appears, I might be wrong, that models are more on the side of rationalism precisely because not only just they anticipate the kind of theoretical change or transformation, but also, as I was telling Theo earlier, they also anticipate accumulation of anomalies,

which basically give the indication that a theory, even though it is settled, might at any point be dislodged by a fundamentally new theory. Adam, elaborate. I think Adam was abducted by a rationalist UFO.

Wait, wait, still on mute. Okay. So, my comments in the sidebar were back to the sort of counterfactual discussion a little bit earlier, rather than the empirical rationalist thread. I was just trying to think of examples where you deliberately miss in real life, like as a matter of technique, right? And there was, like, I'm not sure if people are familiar with artillery bracketing, but No, no, no, tell us, tell us something about it. So this is a technique, at least in World War I, where you're trying to work out the

range where you can hit your target with your artillery. Okay, yeah, yeah, okay, yes, yes, I know what you're saying, okay. So you sort of make an estimate, you don't know for sure, right? And then instead of using the estimate directly, you deliberately target shorter than the estimate. So you know it's going to fall short. And... Right, right. That... You fire that and almost certainly misses. So it might still be how mortar... It might still be a technique used in modern warfare or not enough of a military... Oh, yeah, absolutely. I mean, this is how modern warfare works. Sure. This is how mortars work still today, so they shoot. Yes, absolutely. I mean, it is essentially the way that's, I mean,

there is actually a fantastic, everyone should read this. It's like kind of a very quite similar to what Adam says. You remember, I think back in the 90s, there was this kind of brief conflict between Pakistan and India. So it was a mountainous warfare. You see, how it serves or artillery fundamentally work differently in the mountainous train. And Indians actually precisely use this kind of estimate threshold in order to effectively bombard the Pakistani positions. It's actually a superb story with regard to kind of engineering side of estimate.

The same that happened in Libya. They were deliberately actually a misfire to chase people away from the position because for a long time the army didn't want to kill anybody. I see. So they would try to adjust the mortars, but then misjudge by to push people back. Right, right. Okay, so that's another level, actually. So that's interesting because you have a model of how the person is going to react in the opposing trench or the imposing position. It's like just trying to find a range. It's essentially post-World War I. It's essentially you are not, in fact, targeting for the precise location of the trench

or your target system, the enemy, like World War I, where basically you are, and it always goes badly. It's just that you actually find the range. The range is important. right so once you find that range you can fire much more accurately you know it has to be within that range and there's a less violent example from when you know um i was a a kid being taught how to hike with a map and compass when you're um hiking sometimes in hiking you can also think about the passion across some high shooters passion team okay yeah yeah I mean the same kind of stuff target on one no they're on one side yeah yeah um so

there's sort of like very rough counterfactuals in a way you you sort of use the fact that you're going to miss any instrument within the um technique that you're using and I don't know if that was just a way for people to grab Right, right. So you are basically saying that counterfactuals, I mean, please do correct me if I'm completely off point here. Essentially that, you know, the idea of the counterfactualism is not always involved with a kind of, you know, use of modal vocabularies in a logical sense, but it actually, in real world phenomena, particularly in the kind of, you know, concrete

modeling, can involve techniques, techniques of inaccuracy. Yeah, yeah. So it's a way of instrumenting inaccuracy, just like these real world examples of it. And that's what sort of Fisher is trying to do, I think, with the three gender examples. or is it might have been Haldane. Right, right. Deliberately missing. I see, I see. What do we learn from what we miss, from how it misses. Right. It's kind of like a metaheuristic device. Yes. Okay. That sounds, yeah, sure, yeah, okay. It was actually a fantastic topic.

But it's actually, it's a very known investigative journalism book where you deliberately state in order to assume a particular arrogant person being interviewed to correct you and then by this confirm the fact of the knowledge or what you actually try and trust when you ask someone directly but when you misstate or deliberately misquote the facts and this is actually a very important conversational technique Right, right. Interrogation technique. It definitely works a lot to correct you. Right. In response to your counterfactual. Uh-huh, uh-huh. Okay, that's good. Yeah, sure. Yeah, well, this is one of those things that I

haven't even thought that, you know, there is a realm of metaheuristics. And yeah, surely, there are very great real world examples about it, which they try to do the same kind of things that what you might call dubiological modalities, modal vocabularies do. Yeah, okay, let me think about this. Is there any kind of, Adam, do you know if any kind of actual serious research has been done on this idea? I haven't seen it mentioned in research actually, that's an interesting question. I'd like to read on it. Okay, okay. I'm sure the legal practice would have serious research if we want to leave out of discipline.

Yes, there are, but I mean the majority of them that I have noticed are kind of like and method learning principles that are not really about any of these kinds of they don't they don't deal with any of these kinds of philosophical concerns or epistemological concerns uh but i might be wrong i might be wrong anything any question I guess I'm still just kind of grappling with what exactly how exactly the counterfactual works because it seems like in some sense okay okay but very new to this but here I'll just I'll just say

okay go on go on like I'm gonna go outside it's raining I'm going to get wet or right Because it's raining, I'm going to get wet. Right, right. Piece of material influence, right. Right, so the counterfactual is if I stay inside, I don't get wet, right? No. Or... A good diffuser in your counterfactual universe or universes would be that if you go outside, you might not get wet. That's an anomaly upon which the Oedipus of Science feeds. So, I mean, how does that serve the hypothesis, the not counterfactual hypothesis of rain causing

my wetness or something like that? Well, precisely because it's a negative hypothesis. It's a negative hypothesis in the sense that it already postulates the possibility that if you go outside while it's raining you are not going to get wet. In fact it and why is it important? Precisely because it posits a class of possible data, Moodle or counterfactual that up within a context a set of other countries in which the scenario of getting wet will never happen

So, I mean, maybe just to... You can, in fact, get this kind of phenomenon quite regularly in very, very microscopic, microclimate scenarios in which it is raining at this small location, I'm not talking about microscopic location, it is microscopic location and it is like and also all around it but But there is also this kind of small region where you don't get wet.

So this is basically what you might call to be a trigger for science to look into the explanation of why is that this is the case. Hence we have to refine our principles, our invariances, our regularities, our so-called laws. it just seems like I that's I'm guessing I'm trying to determine what's the range of counterfactuals that you might use in it in so we don't know we don't know I don't think that you see this is the whole idea that a great deal as on Korea had said it a great deal of science

is to pose a new hypothesis which can defeat, namely counterfact, what is the established order of experience, right? And in that sense, there is no a priori law for us to be capable of arriving at a set of canonical, counterfactual scenarios or hypotheses. It's just that we have to do it. It's like more a kind of way of a kind of a struggle of an enterprise where you have to constantly, against all of your evidences, posit a counterfactual scenario or hypothesis that might, as a matter of fact, contradict the kind of consequences you got from your premises.

within your factual scenario, your so-called factual scenario. Can I just ask one more thing about it then? Yes, absolutely, absolutely. It just seems like, okay, so... I guess what I'm trying to do is see how the counterfactual serves the building of the sort of structural scientific understanding or whatever. So you have this patch where you won't get wet and going outside. But that counterfactual argument is still kind of framed within the basic hypothesis that's put forward.

Right, right, right. It is essentially, it's absolutely, yes, absolutely. The counterfactual scenario that you are positing is definitely going to be hypothesized, postulated within the established order of your theory within which you are hypothesizing, right? Okay, that's all good and fine. But the thing is that the aim of counterfactuality is to arrive at what you might call to be in both how you regularize or single out invariances from your observed data

or how you structure them. It's essentially what you might call to be a a a methods of yielding anomalies diffusors i guess i'm confused by the use of the word anomaly here because the well i guess i maybe i just need clarification about what that would mean because the patch where it's not raining and everywhere else is raining it seems like that's not so much an anomaly to the hypothesis of getting wet in rain it's no it is not but you see the the point of hypothesis is

always a counterfactual in a counterfactual sense is always the task of it is to in fact posit an anomalous scenario like the match actually being lit in vacuum. Now, of course it might not, but nevertheless, this is a kind of a logical vocabulary or a structuration, structure that you have to take seriously in order for you to actually put brackets brackets refine the bracketing we find the constraints of your regularities your laws within the theory in which you have made such a

hypothesis you And it is of course by any means, it is absolutely a messy problem. That's why science cannot be thought in terms of this kind of empirical versus rationalism rivalry. It is essentially, it's quite from a point of perspective of scientific method, it's very much tangled with logical messiness, logical uncertainty, and also empirical uncertainty, empirical anomalies.

Another point to actually, you know, this is like a kind of a classical Brandomian critique sense that how can you in fact posit an empirical regularity a law a natural law within a theoretical system either you say that the red dots follow these kinds patterns and hence by virtue of observation there is this regularity which according to Brandon and Sellers is just basically naive empiricism, the myth of the given

or you try already implicitly, implicitly, whereas models do it explicitly, you implicitly analyze these patterns of red dots by virtue of deploying some moda vocabularies. The whole task of explanation, I just don't think that you can get rid of the idea of modal vocabularies. There is no such thing as explanation without using modal vocabularies.

Of course, we might quibble that, well, these are not modal vocabularies all the time or logical vocabularies in a counterfactual sense. But nevertheless, you should have some sort of counterfactual scenarios, hypotheses, what Adam said, threshold of inaccuracy, meta-learning, meta-heuristics, so on and so forth. But the counterfactual is not, it's not counterfactual scenarios that would somehow challenge your theory or… No, they don't challenge your theory. They absolutely don't challenge your theory. It's just that postulating such counterfactual scenarios is essentially what you might call

to be a logical means of determining such and such regularity, okay? Refining its parameters, refining its threshold. But however, if you turn this counterfactual scenario into a real model in the way that we have been talking about, you might actually arrive, as a matter of fact, across some anomalous set of data which can no longer simply be plugged into our notion of law. law cannot simply be applied to these sets of data and there are quite numerous examples

of this in the history of science when the hypothesis is posited as a model dealing with a real-world phenomena as a counterfactual in a counterfactual vein and then we see that Oh well, you know, the movement of celestial bodies was not on the circle path. It actually had a very irregular oblique path. And that fundamentally changes the movements and the kind of equations of motions that we had. Artemis, I don't think that all hypotheses, surely hypotheses can be also understood as

thought experiments, but I don't think that all thought experiments can be understood as these kind of counterfactual scenarios. I think the concept of thought experiment in philosophy yes, it has this idea that it kind of follows the logic of the modal vocabularies if this would have been, then what would have happened? I was actually wrote through experimentation with actual experimentation, like thinking, putting the constraint in the hypothesis and try, try, try,

because we don't know exactly, as you said it. Yeah, okay, if that's the case, I mean, those kinds of thought experiments, yes, well, of course, You see, I wouldn't simply equate thought experiments with the notion of a counterfactual hypothesis. You know, the notion of the concept of thought experiment in philosophies and science is quite broad, whereas counterfactual hypotheses, even though they are broad, but nevertheless are narrower, precisely because they have some sort of constraints, either in terms of application, scope, assignment, and so on and so forth. You know, for example, okay, let me give you an example. Like, you know, it is reported that Michael Faraday experimented a thought experiment,

basically, possibly a thought experiment in which he put himself in the position of an electrical current within a magnetic field. Okay? So this is also a thought experiment where basically this little guy, what is usually called the Faraday-Banom, little man, sees that the electrical field, Faraday law of the three fingers, that how he rotates in the electrical field always actually

corresponds to the vector space of this kind of configuration, x, y, z, you know, rotational figures. So this is actually a thought experiment. This thought experiment, however, does not yield new kind of data. It actually describes a kind of phenomenon that we couldn't even arrive at. It's just like a descriptory. So not all, I would say that not all thought experiments are like counterfactual hypotheses in which they

are going to yield anomalies. They can be actually purely experimental. They can be purely descriptive. for example, Einstein putting himself in the place of a photon, you know, the train ride, where he talks about, you know, the kind of a photon at a greatest speed bends, you know, by virtue of the gravitational field. So, you know, these are not all, I would say that not all thought experiments have a counterfactual scenario in the way that we have been talking about. Surely they are in fact counterfactual in a kind of a rudimentary sense, but their

aim is not to in fact arrive at a set of anomalous data, change the principles of the current theory, so on and so forth. It might be just purely sets of heuristic devices, experimental heuristic devices, through which a scientist, a mathematician, or an engineer can arrive at a specific solution, whatever that solution might be with regard

to whatever solution the problem is. I mean, whatever the problem is. So yeah, I would say that thought experiment, not all thought experiments are counterfactual scenarios in the way that we've thought about. But yes, it is yet to be determined how much thought experiments are in fact can be explained and described in terms of counterfactual and modal scenarios and to what ends. Unfortunately, I don't know it. But yes, history of science is full of these thought experiments, but not all of these thought experiments are actually models

or change a theory or lead to an anomalous set of data, so on and so forth. Can I ask something? Absolutely. Yeah, I was thinking if there were like some similarities to the paradigm of manipulative abduction, because as I'm referring especially to the example offered by Adam, because, yeah, there is an exploitation of an environment, in this case, in the First World War, the field itself.

And they're also trying to, as the manipulative abduction does, trying to select some hypotheses and deliberately negating some of them. And so there is this access to new ones and there is access to unavailable information. I mean, through this negating of hypothesis, there is this access to some unavailable information. They were unavailable before doing this specific case, the shoot. So what I'm asking is, can be the counterfactual compared to manipulative abduction?

So may I repeat, please again, do correct me if I'm right. If you are saying something like this, so basically, for example, in the political terms, the capitalist real subsumption allows us to think in one way rather than others and hence this is just like what you might call an extreme case and in that for example certain kind of social counterfactual experimentation that we do are already abiding by the logic of capitalism and hence would it be a way to in fact postulate counterfactuals

or alternatives to the very platform system which constrained the way how we conduct such experimentations yeah yeah probably yes I believe that probably Jovan can help me because we I know the other class with the French on Francois Schmidt elaborated this method that the of the without and it's I believe it's sort of similar to what I'm trying to say. Sorry.

I spaced out a little. Yeah, sure. I mean, yeah, I mean, they're all, they're all this counterfactual and also this is a paradigm of manipulative abduction. I believe they have like some, and also the method of without elaborated in the last class. I believe there are some features, there's similar features and yeah I also saw the connection you were referring because as Tony mentioned in the text in that class action has the the experimental characteristic in itself and it promises something

new let's say so i also saw the connection of the counterfactual method with abduction but yeah the a huge part of the counterfactual method is when we talked about like the porn star that does not fuck or like how do you if you if you subtract an invariant from a discipline like if you subtract the invariant from let's say philosophy and you have all these anomalies but you don't recognize these anomalies as you don't give them a new paradigm but you you you work with them you work with those anomalies and and figure out stuff um and so then the question is how do you figure out these stuff like what kind of what kind of uh i think we should go back to meredith's uh comment about how there is a

subsuming a kind of non-existence into epistemology. So the question is, how do you make a kind of like non-existent space within epistemological grounds that you put all these anomalies together that... Yeah, I think it's... Right, okay. Okay, this is getting way too difficult, but yeah, superbly exciting. Maybe Theo can chime in a little bit or link any person would be appreciated. Well, can I just ask Joven, I think this is kind of the question that I was trying to

the old time i i really wish i would have known about counterfactuals before this but um but that's counterfactual scenario too um i guess the thing that i'm trying to articulate is just like how would it be that you would come up with a counterfactual scenario it seems like all that the counterfactual scenario is is kind of um it's just like the inverse of your positive scenario that you're hypothesizing or something but i'm not sure if i don't know the language that well so i'm not sure if i'm even understanding but you see this is where that's i would say that within the realm

of logical counterfactuals, modal vocabularies. Isn't it the case that Artemis said, it's just let's play with the toys. The toys are the counterfactuals. Essentially, as many constructs, counterfactual constructs that we can make, the better we have a chance to revise our notion of object, our notion of a certain phenomenon, regularities, patterns, so on and so forth. Coming back to the very early Kantian question, logic as a canon, namely the entire task of logic

should be about application to real world phenomenon or logic as an organon. Namely, the very constitution of all sciences, in which, in that realm, we can think about models, studying other models, basically toy models, all sorts of models, which are basically giving us different range of counterfactual scenarios, which we wouldn't have been even to think about, to postulate, to explain, if we were not capable of thinking about other kinds of scenarios, no matter how bizarre they may be.

like the three sex biology versus two sex psychology biology or like Eddington that the whole idea of modeling just doesn't bottom up in the actual. Sometimes we need to think about non actual scenarios in order to determine our definition of actuality. That's the toy model, the toy approach to logical construction and hypothesization, or what you might call to be, you know, counterfactuality. It's the idea of taking counterfactuals as a class of logical vocabularies very seriously

and play with them as much as you can. and then crystallize them into applicable models. See what will happen if you actually deploy them. What I think would be maybe... What I'm thinking is, I think I understand well, I think I understand in a certain way the question Theo is asking, is how can you orient yourself in thought in order to come up with the counterfactual you want?

I think it will be something like this. Correct me if I'm wrong. Yes, okay. control. So in that way, the counterfactual will be working as a way to enhance explication.

It seems you have to choose in the phrasing, within the phrasing of the counterfactual, within your hypothesis, some way it's kind of embedded in it. So the way in which the answer you can give to it will decide upon the matters at hand. So it's kind of difficult to come up with a recipe or something like this. You see, here I think the question becomes a little bit topsy-turvy. and I genuinely don't have a great answer to this, but the way that I see it is that counterfactuals, in fact,

are not about the matters at hand. They are not about the study of a particular phenomenon in the world. They literally, absolutely are disconnected from a facial with any real phenomenon. essentially at the level of counterfactual and this is Brandon's thesis yes you have to take the idea of language itself seriously without positing the metaphysical correlation or relation between language and the world model and the target system coming back to Hegel again so at the level of language and again Karna, how many new structural resources can you generate

in order to possibly enrich the application of language to the world, the model to the real world phenomenon. And I think the counterfactuals in that sense should be thought more with regard to the diversification of linguistic structures and by that I do not mean just natural language, any kind of formal language, language in general, which might possibly once apply to real world phenomenon and which are you know the kind of insights we had garnered from the application

of one theory to a real aspect of the world. However here a question arises and for which I do not yet have an answer and that is the question, the preneal question, is that then how can you really avoid arbitrariness of counterfactual scenarios? In the sense that, well, I can say that, well, God also created the entire universe and also all possible worlds. Also, I can say that, you know, within the Darwinian theory there might be such and such counterfactual scenarios. What would

then be the criteria or the standard of distinction right from wrong that can be applied to these two possible counterfactual scenarios such that I actually can productively bracket my range of counterfactual scenarios without succumbing into all sorts of theological mischiefs. That is absolutely, I don't know. I don't think that… So like the counterfactual scenarios need to be factual in some way but deliberately not arbitrary?

There needs to be factual prima facie because that is that's the whole point. There are essentially logical entities. Logical entities couched in terms of modal vocabularies. Now the thing is that it's exactly like the model so you see for example the three sex biology where we basically positive without having any actual basis for it there is no such thing as the three sex biology some sort of non actual entity nevertheless we we are hypothesizing about it precisely because it gives us some insights but then at the level of what you might call to be epistemological deduction

namely uh what determines the objectivity of our claims then if that is really the case then how can we say that god as the ultimate creator or any sort of uh intelligent design scenario is inferior to, for example, a scientific counterfactual scenario? Well, that is actually a question. Maybe perhaps an answer to that would be precisely the idea of the kind of lines of inferences that we draw in each scenario, in the theistic intelligent design scenario

and in the counterfactual scenario. I don't know. I need to think about it. Fantastic, polemic, superb, really excellent. I would say, going back to Brandon, yes, absolutely the range of the model vocabulary is not talking about the world itself, itself, it is actually in the Brandomian construal, it is talking about the expressions you use in the descriptive vocabulary, right? So it's fixing the rules for using the terms that

are present in the descriptive sentences. But nevertheless, it seems there is a somewhat may be indirect responsiveness to the way things are in the world. Once you fix a vocabulary, through its ramified model claims, some vocabulary can be participating in. So you have like a set of all the, if it's possible, idealized set of all the possible model claims that are taken to be true about a vocabulary. Taken to be true meaning it is permissive to use the vocabulary that way.

But it seems that the fixing of these rules is somewhat responsive to the way things are behaving in the world as well it seems like you can um i think there is in reason in philosophy the first chapter he it's he says uh to understand the phrase the cat is on the mat you have to know how to use the terms but the usage of these terms is in answers in some way to how cats behave and how Matt behaves. Right, right, yes. It seems there is, nevertheless, even if this is a degree further from the direct realism, there's degrees further from that,

but it is nevertheless responsive to... Bottoms out at that level, yes. And that's why I said, I think that, again, I'm most probably wrong but I actually take side with Carnap again on this point. Precisely, I would say that linguistic resources, absolutely, in no way whatsoever, have no primary correlation with the real, with the world. That is not job of the language to be correlated with the world. that would in fact be a metaphysical bloatware

to posit the idea language has something to do about the world. No, in fact, we only talk about the world to the extent that we can talk about language in itself and being capable of triangulating at a very rudimentary level the protocols of observation with the syntax of our linguistic calculus. And in that sense, I would say then, the idea of counterfactual scenarios are purely linguistic.

It is just that they become hypothetical in the scientific sense when we are capable of deploying them at that rudimentary bridging level between protocol observational statements and the syntax of a linguistic calculus. calculus. But of course, then here another question arises again as if we weren't in the swamp already. So what would be this bridge between the protocol observational sentences like the red dots in your visual fields and the real number predicates which geometrically

map them, you know, the registers of these red dots at the level of a linguistic calculus. Well again coming back to the very idea of we have been talking about, it seems to me that philosophy literally at this point is struggling at basically that fine line between naive empiricism and naive rationalism. And the only way to get out of this swamp is for us to be capable of explaining, describing coherently at the most basic level how pure observations

can be mapped into pure minimal syntax. The place or location of symbols within a language or a calculus. us. Sorry that this class is becoming a course in skepticism instead of modeling. It's just that for many, many philosophers and scientists, even in Weisberg, you know, Supes and so on

and so forth, it's just that, you know, they have already taken for granted some of the more inflated claims or inflated presuppositions, which actually enable their systems of thoughts. Of course, we are not going to get into these kinds of stuff precisely because these are not really part of our class, essentially these are, I would say that in order for us to talk about

these problems coherently, we cannot just talk about modeling, we should actually talk about as I mentioned, the ultimate question, rationalism and empiricism. by that I don't mean complex rationalism versus complex empiricism, I mean naive rationalism, pure arbitrary concepts versus pure sensory data. if we cannot find a bridge, a tissue of how observations, sensory observations, and symbolic

design, the calculus of symbols, the syntax, are related to one another, then we are in in fact, in the business of naive empiricism and naive rationalism. And to that extent, I would say that Carnap was absolutely right. Carnap's logical structure of the world, in fact, the entire edifice of logical positivism was constructed in order to tackle this problem. It failed them. I don't think that coin actually give a good solution to this either.

And you see this problem surfacing again and again throughout the centuries. One of the latest one is speculative realism. Okay, let me just quickly get back to similarities stuff. So very, very quickly, you see, coming back to Schelling's idea, Schelling's segregation

model. You know, Philadelphia, as you know it, is a racially diverse city. For example, according to the 2010 census, the population was about 44% African-American, 39% Caucasian, 12.5 Latino and 5.4 Asian. However, when one looks more closely at census tracts, one sees a very typical urban demographic pattern, racial clustering by neighborhood, essentially Schelling's model. There are probably many factors driving this demographic pattern,

But an interesting model of one possible factor was developed by Thomas Schelling in 1970s. He wanted to know if it was possible for a city to segregate even if its members did not have as strong preferences, like racial preferences, to live in segregated neighborhoods. So to investigate this question, Schelling then constructed an agent-based model consisting of two racial groups and a grid representing a city. So in the original version, the model was physically instantiated on a chessboard with dimes and nickels representing two types of individuals, A and B, and the squares on the chessboard representing a spatial location.

Apart from the individuals and their initially random spatial layout, the model also contained a utility function and a movement rule. The utility function said that each individual prefers that at least 30% of its neighbors be of the same time. So the A's want at least 30% of their neighbors to be A's and likewise the B's. Schelling's neighborhood were defined as a standard Moore neighborhood. A set of nine adjacent grid elements, an agent standing on some grid element E, can have anywhere from zero to eight neighbors in the adjoining elements.

The movement rule can be described as follows. Agents sequentially choose to either remain in place or move to new location. When it is an agent's turn to make a decision, it determines whether its utility function is satisfied or not. If it is satisfied, the agent remains where it is. If it is not satisfied, then the agent moves to the nearest empty location. This sequence of decisions continues until all the agents satisfy their utility function. Essentially, the system reaches a kind of an equilibrium state.

So the dynamics of a modern computer-based implementation of Schelling's model are, you know, basically are these grids where you can see the evolution of, you know, some prefixed position of agents. agents and then throughout time you see that it starts according to this preference and utility functions completely being randomized and then it reaches an equilibrium clustering state simply you know every agent finally moves to some sort of satisfying

neighborhood position now this clearly shows that starting from a random distribution of agents the equilibrium state of the model is segregation Schelling's major result is that the small preferences for similarity can lead to massive segregation. If you remember Adam's response in the second session, if I remember correctly, that it is not based whether these agents are racist or not. You might not, in fact, be a racist. You might just have a minor preference choice with regard to the clustering of the neighborhood.

As long as this preference choice is there and you act on this preference choice, you know, rational choice theory, game theoretic version, the result would be segregation. and of course the result is quite robust across many updating different neighborhood sizes different spatial configurations so on and so forth in fact it is extremely hard to avoid segregation when agents have some preference for like neighbors now of course Schelling King offered his model as a how possibly explanation.

We will talk about what a how possibly explanation is when next session we will talk about toy model and what toy models are. As a how possibly explain it, how possibly it can explain it, rather than as a matter of fact explaining it. One potential mechanism by which neighborhoods could segregate Okay? But what if the mechanism described by this model is actually part of the explanation of how Philadelphia came to have the neighborhood structure that it has now, or in 2010? If this were the case, in what relation to Philadelphia would this highly idealized model of the segregation dynamic stand?

The answer is that the model would be similar to Philadelphia. Highly idealized models are not truthful representations of their targets, nor are they isomorphic to their targets. They are similar to their targets. So what I'm going to talk about is that what is exactly this concept of similarity-based idea of model world relationship? How can it be developed in non-arbitrary, less vague terms?

So, with regard to the concept of similarity, we can say that even though a number of philosophers have advocated similarity as the ideal candidate for the model world relation, for example cataracte similarity has a patchy history in philosophy of science I mentioned Quine criticism of the idea of similarity in the previous sessions While Quine's argument primarily rested on simple chemical examples, he also invoked more foundational arguments from Nelson Goodman, another philosopher I have mentioned a few times already.

One of Goodman's arguments against similarity is that appeals to it merely label something unknown rather than giving a characterization of the relationship in question. A proper analysis ought to give a reductive definition of similarity. However, Goodman argues that no such definition can be found. Another of Goodman's arguments is that similarity is vague concept precisely because it is way too promiscuous of a relationship.

for any three objects he argued there will always be some respect in which two of the objects resemble each other more than the third if we have a green square, a red square and a red circle there is no obvious pair whose members are more similar to each other than the third Goodman uses this second problem to show that there can be no context-free similarity metric. Essentially, similarity is always context sensitive in a Goodmanian sense. Such criticisms launched by the likes

of Quine and Goodman led, you know, Geir and Catroy to restrict their accounts of model-world similarity, such that they told us a model must resemble its target in certain respects and degrees, presumably given to us by background theory and theoretical interests. For example, cataract tells us that the relevant similarity between models and their targets is behavioral similarity, which can also be understood as similarity between causal structures.

similarity for this you know for this reason seem to be the relation that holds between models and the world because it comes in degrees can be used to compare idealized models to target systems or alternatively can relate qualitative features of models to target systems, so on and so forth. However, they give us few details about what similarity supervenes on, how it depends on context, how similarity judgments are to be evaluated,

and et cetera. So I think we are running short on time. We have only 15 more minutes. So I will stop here. And then next session, hopefully for the next 40 minutes, next first 40 minutes, sorry, first 40 minutes of the next session, I will go over a list of, you know, without getting too much details, go over certain kinds of constraints, some of which are theoretical in a general sense, and some of which are within the intention of a modeler or a theorist.

I will elaborate these constraints and, you know, come up with some sort of answer and a conclusion with regard to how the notion of similarity is usually being understood among today's modelers across the board, engineers, theorists, so on and so forth. So with that said, before I move to questions, did we actually ask to people present responses today, Theo?

No, I think you said you were going to post some reading on the classroom. Okay. So basically, we have come, you know, a good deal with regard to some philosophical questions, of broader nature, modeling questions, stages of modeling, so on and so forth. Now, I think that without you, in fact, having your reading material at this point, you should be able to come up with some questions, some responses. It doesn't need to be confirming what I have said. It can be fundamentally skeptical. It can be fundamentally challenging what we have been looking at, so on and so forth.

Whatever. Nevertheless, we need two respondents for the next session. It would be also really nice if we could have some sort of follow-up text from which to think true what we discussed in today's sessions, because it would seem, well, it's a perfect place for everyone. Sure. One second. Let me just get my book here. Yes. Let me get my book. One more time. The concept that we looked at today. Right, right. The actual stuff that we're going to support. even if it's bad examples of your argument just for me to no no i think uh we can still read um

sorry one second So, one thing that I can come up with is chapter 5 of the Wiseman book Target Directed Modeling, but also I have a PDF, I have forgotten the name of the author, I will post it on the

Google Classroom, I just send it to Theo so he can post it, which is basically again coming back to the idea of similarity. But also maybe it's good to have another text, the very essay of Goodman I just mentioned with regard to his criticism of similarity, that similarity is just like, you know, not context-free. It doesn't come for cheap. So these, yeah, well, one second. I'm going to give you the title. One sec. So it is called Seven Strictures on Similarity.

in a book called in uh problems and projects indiana police uh bobs and meryl uh written in 1972 seven strictures on similarity that's that's good man i say ways of world making yes, it has actually some tidbits on this question of similarity in a chapter called, I think,

Puzzles of Perception, if I remember correctly. However, I wouldn't suggest it precisely because reading that chapter requires reading earlier chapters. I don't want any of you to read a whole book for the next session. So I would say a kind of a crystallized criticism of Goodman can be found in that essay that I just mentioned. Yeah, Justin, yeah, well, no, I'm not going to self-reference myself. No, no, no, we will talk about these it's just wrong to put my own text as a reference I am here to

be interrogated not to be the you know the legalist judge well we will get to those points however don't worry we will get to those stuff with regard to the question of world making and logical world construction. Yes, we will get, but it's still, I think, a bit too early. Yes, also, yes, Language of Art. Yeah, that's a superb... I mean, majority of Goodman's books are utterly fantastic.

And he's quite very lucid, very playful, in fact. Two volunteers. Sean. Hey, so I have never read Goodman, but I was talking to JP about his work recently, and I think I would really enjoy it. So I'd be happy to present on that essay. My only concern is I'm going to be out of town on Saturday, and my train is going to be getting back at like half an hour before the session starts. So I'll be able to make the session. But would it be OK if I maybe present like deeper into it?

At the end, yeah, sure. Yeah, absolutely. At the end, yes. Okay, well then I'd be happy to present on the Goodman. I've been looking forward to reading his stuff anyway. Superb, superb. Excellent, excellent. Ah, okay, Mari. Fantastic, excellent. Thank you so much. And yeah, Reza, do you have a copy of that Goodman somewhere? No, unfortunately, I don't have it in PDF. But let me see. Let's check LibGen to see if we can find it there. I didn't see the problems in projects.

there but well just let me let me just see I mean these texts should be online I mean they're quite old I don't know okay one second one second let me let me go to sci-hub Does anyone think where site hub is? DTW?

Yes. Okay, one second. No. I will find it. I will definitely find it for you. um thank you yeah unfortunately it is it appears to be not online i think jp found a copy of this uh strictures and similarities though oh it's unscripted so i think in order to download it you have to have

um to be signed signed up but i doesn't have i have a script account so i can i can download it and send it to everybody okay that's superb that would be fantastic i can do it yeah yeah i can do it thanks excellent superb thank you so much okay i'll do it right now because afterwards i will i will definitely forget about that okay excellent all right i'm gonna go ahead and end the broadcast thus there's any stragglers also still let's think