 So, welcome everybody for this monthly All of Us seminar. Today we have the chance and the privilege to receive Roman Frick from LSE on a representation and after a short break, after a start, there will be a short break and a short comment from Cantère-Hurain from... Comple... Comple... Comple... Comple... Comple Intense University, Madrid. Roman, the floor is yours. Well, thank you very much for inviting me. It's a pleasure and a privilege to be here. The title slide says I'll be talking about scientific representations. This is joint work with my colleague James Nguyen, who wasn't Notre Dame for a while, but is in Stockholm now. And what I'll be saying is mostly based on this book here that came out in the middle of the pandemic, so probably no one noticed. Modeling Nature, in particular chapters 8 and 9. But parts of the talk are also based on my most recent book called Nomads and Theories, just came out a few months ago. And this has at least one thing to recommend it, namely that you get it for free. So it is open access and you can just download it from the publisher's website and it's obviously linked up on my personal website, so you just click the button there. Well, it's probably carrying calls to Newcastle here in this audience but still I'm to make my starting point clear. So I just take this simply scientific fact that models matter as my starting point. So there is hardly a scientific achievement that does not involve a model in one way or another. Just think about planetary motion, the theory of heat, the statistical mechanics of gases, nuclear and atomical structure, DNA structure, the Higgs boson and so on and so on. And this obviously throws you back to the question of what is a model? So now I do what the model scholar does, you hack your question into Google and on the first page you find this thing here, that says a scientific model is a representation of a particular phenomenon in the world. And that's not bad as a definition although we'll qualify that in the talk, but I use this as a motivator for the question I'm really concerned with, namely what is representations? So that's what I want to talk about. And I called my talk representing representation because I'm obviously not the first to ask this question. There's a number of accounts available, so it's conventionalism of various kinds, similarity accounts, isomorphism account, inferentialism, direct fictionalism and the Deci account. And so for all of these here, I would refer you to the book I'm not taking you through all this. I mean we would never get out of this room at the useful time. And what I want to do is focus on the Deci account. I don't think this is a slightly critical name, but you will understand a bit later what the acronym means. When I started working on this suddenly a friend of mine sent me this photo here and said you've already got fans. I said okay, that's interesting. So obviously people get personalized number plates. So here we go, at least someone seems to like the account or can at least be interpreted as doing so. Anyway, it's a very liberal interpretation. So before I delve into the details, I want to give you a preview. So I want to tell you what the Deci account is. And this is sort of encapsulated in this diagram. So you have this big box here called M, that's the model. And the model has an internal structure that consists of an object X, an interpretation, and then I will explain what I mean by that. That gives you a so-called Z representation. The model denotes a target system T. The model exemplifies properties P1, P2 and so on. These properties are connected with a key to another set of properties and these are imputed to the target system. And now sort of what I had grayed out here is basically a linguistic part. So D stands for description. And so what you have on that side here is you have a model description, you have a description of the interpretation, you have a description of the object and then the description of the target. So that is just to recognize that this would play a role in some context that models don't live in sort of a language-free space. They usually introduce and study in a particular language. Now if all of this is sort of a bit abstract to make sense of, here is a concrete example. So take a model of the solar system. So you start with an object here. This is spheres that attract each other gravitationally. You interpret these spheres in terms of the sun and planets. That turns this sort of object into a solar system representation that model exemplifies certain properties, for instance that planets move in elliptic orbits. That properties keyed up. And again I will say more about keys later. This is really just to give you the panorama of what we're going to talk about in this hour. With another property, namely that planets move in an ellipsoid. So this is something that's somewhat like an ellipse but not quite. And that's imputed to actual planets. And obviously you do this using all kinds of descriptions that are formulated in a particular language. So that gives us the plan. I want to explain what motivates this particular way of thinking about models and representations. I want to explain to you how we get there, because there's a lot of boxes and concepts flying around. So why should one think about it in this way and how do we get there and want to flesh out details? Now in doing so you immediately run up against a problem. Namely there seem to be two different kinds of models, at least from an ontological point of view. So one are concrete objects. These are physical things you can put on the table. We'll talk more about this funny object later. Does someone recognize this? I'll tell you more about it. Is it the water model from your paper? Yeah, yeah, yeah. I've never seen a picture of it. I only read you writing about it. So that's the Philips Newlin machine. I'll show you a video of it later so you see it in action. So that's a physical thing. It's about two and a half meter high and a meter and a half wide and it rattles and pumps. It's really quite interesting to see. But then on the other hand you have obviously sort of non-concrete models that Ian Hacking sort of nicely described as something that you hold in your head rather than in your hands. And many models are of that kind when you do the bore model of the atom. You have nothing to put on your table. So you sort of caught in this and there was deliberately vague about this just in the previous slide when I showed you this sort of solar system object here. So we have to pay some attention to that. So the update that's planned then for the talk is that I want to first talk about representing with material models or things like the Philips Newlin machine because a lot of things that I want to say are more naturally said about physical things than about non-physical things. But in the second part, I want to tell you how this carries over so you don't get in a kind of representation that only works for rattling water pumps. This should work for all models. But most work will be in part one so don't get nervous if half an hour in and still there. Because the idea is that most of it carries over to non-material models in a relatively straightforward way and I will tell you how this goes. Okay, part one then. Representing with material models. And I want to start with some background here. The background here is the theory of representation as that's been developed by Nelson Goodman and Caldria Elgin. The Dedecki account is really a sequel. It is a further development of that account. And therefore it's helpful to go over that account first and then I'll tell you where different pieces are changed, inserted and improved. The representation as is something you're familiar with from Caricature's. So here is a classic. Margaret Thatcher represented as a boxer. So that's a warm-up example for word representation as it means. Or if that's a bit too far back for your taste. So here's a more recent one. Although I mean one can't update these things quickly enough. I mean that's already the past too. So here is Liz Tross as a puppet on strings. It's also a case of representation as. But I'm always interested in scientific models at the end of the day. So here is one. Now you may think that this is sort of my feeble attempt of making props for a horror movie, but no, this is the staff of Nobel Prize. So that's myoglobin, represented as a plasticine sausage. And research on that model won Kendra the Nobel Prize in 1963 for myoglobin structure. And this is the one I've already shown you. So this is the Guatemalan economy represented as a system of pipes. And I'll come back to that, as I said. So here's a bit of notation. In what follows I'll use X as the variable denoting the object that does the representing. And that may sound a bit roundabout, but there's various objects that can do representing. There can be machines, plasticine sausages, drawings on paper. So whatever is doing the representing in the case that you're interested in, that's X. Then T is the target system in the world. That's the thing represented, either some unfortunate politician or myoglobin or the Guatemalan economy. So I hope that these two are sort of intuitive enough. The next on the list is one that needs a bit more explaining and to come to that. So the Z is the kind of the representation. In the caricature, the beginning, this would be the box of, for instance, because the caricature attached would be a box of representation. And so that's the third element of the count. Now here is what Catherine Elgin says about representation. She says when X represents T as Z, this is because X is a Z representation that denotes T as it does. X does not merely denote T and happen to be a Z representation, rather in being a Z representation, X exemplifies certain properties and imputes those properties to or related ones to T. Now I don't expect you to digest this immediately. I will spend the next 15 minutes disentangling that. And to help us with this, I made a graphical representation of this. So we have X. That's the object that Darcy went presenting. That is also a Z representation that denotes the target system, that exemplifies certain properties associated with Z and it imputes these on the target system. So with the Tature caricature that works as follows. The caricature is a boxer representation. It denotes Margaret Thatcher. It also exemplifies certain properties that will not associate with boxers like being brutal, being ruthless, being aggressive, something along those lines. And it imputes these two Margaret Thatcher. So that's how representation as on the Goodman and Elgin account works. And here is again the general scheme. And now we have to do some work. So we have to explain what a Z representation is. I hope you have some intuition from the examples I showed you, but we have to say a bit more. We should say something about denotation and we have to say something about exemplification. That sort of sets the agenda. Okay, let's get started. Let's start with denotation. Denotation is the two-place relation between a symbol and the object to which the symbol applies. I mean, you're familiar with that paradigmatic example here. There's some proper names. So for instance, the name Louval and Earth refers to this town here. And the crucial point here is that denotation is the core of representation. So the representation is the representation being about the target system and the aboutness of the representation is given to us by the representation denoting the target system. And Goodman and Elgin say that X is a representation of T if and only if X denotes T. And the crucial preposition here is the off because I start by saying I give you an account of representation as. And so these little prepositions matter. So this is representation of which will be a part of representation as but obviously representation as will be more. And this is also an important caveat here namely that denotation presupposes existence. So only something that exists can be denoted. Now this has an immediate consequence namely that things like pictures of unicorns do not denote anything at all because there are no unicorns. And so the consequence is that pictures of unicorns are not representations of anything at all. Now that is a somewhat counterintuitive consequence. So recently that scene in the London Underground there's a lovely flock of unicorns running through a shallow lake and that's an ad for an energy company trying to sell you an energy account. Surely that company didn't achieve the flock of unicorn because the picture represents nothing at all. So what are we to make of this? How do I square the view that this is not a representation of anything with the fact that this obviously has some representation of content? And this is where an important move happens and that goes back to Goodman's languages of art. And so there Goodman says we misled into believing that something is a representation only if there is something in the world that it represents. So we have to distinguish between a picture of a unicorn the off here again and a unicorn picture which is an unbreakable predicate indicated by the hyphen here or more generally a representation of a Z or a Z representation. And what we have in the unicorn case is of course that this is a unicorn picture but it's not a picture of a unicorn because these are two modes of representation that you have to keep separate. One does not imply the other. Some Z representations denote the Z and others don't and some representations of a Z are Z representation and others aren't. At this point you sort of may secretly think this is just the other trickery and this philosopher is having a bit of fun here but I want to try to convince you that this is actually a helpful distinction in trying to make sense of many images that we are very familiar with. You know that. That is a map of Europe and we can say this is a territory representation that portrays a territory and it is also a representation of a territory namely of Europe. So this is a simple case in which they coincide. This is both a Z representation and the representation of a Z. But now look at this here. Someone recognized that. Any fans of TV shows in the house? Yeah, so that's the world according to Game of Thrones. So I hope you would agree with me that this is a territory representation just as the map of Europe before was but it's not a representation of a territory which there is no Z thing as the world of Game of Thrones it's of course an imaginary construct so you see that you can have a territory representation that's not a representation of a territory. Well look at this here so that's a slightly personalized version of a map sort of the British Isle sorry the language farting down ships towards France. I would say that this is a territory representation but not a representation of a territory it's a representation of national attitudes and military ambitions probably. Sadly this goes back to James Gilray in 1793 to say something's never changed but there we go. So here's another case of a representation that's a Z representation but not a representation of a Z. And now look at this it's the word Europe that's clearly a representation of Europe because the expression denotes Europe but it's not a territory representation. So here we can now sum up what we have in the vertical you have whether something is a Z representation or not and in the horizontal you have whether something is a representation of a Z or not which yes yes is the map of Europe no and yes it's Game of Thrones and the caricature something that's a representation of but not a Z representation it's the word Europe and if something is not a one or the other then it's just a mere object like this glass here that has no semantic function at all. So I hope I have convinced you at least that this is not just sort of philosophical trickery to make this distinction and that something useful could come out of it but now you will ask immediately okay what makes something a Z representation? I've just put up these images and said that's a boxer representation or that's a territory representation and that's a very good question and I will come back to it later now in the context of visual representations that has sparked a huge literature the entire PhDs are written on it but boiling it down to essential I think one can say there are two accounts one is perceptual accounts let's say a picture X is a Z representation if on the normal conditions an observer would see a Z in X for instance Lopes is a proponent of this account or Goodman and Elgin have their own accounts the genre accounts of pictures belong to certain genres and are recognizable as such we just do recognize territory representations because we're familiar with the genre now I don't want to get into this because neither of these accounts whatever their intrinsic worth in their domain they don't work for scientific ones they're just not going to help us so I would ask you to keep operating with an intuitive notion of what a Z representation is for 150 minutes or so then I give you an account of what this would mean in scientific case okay let's sum up that's our graph we had initially so we've shed some light on the notation of a Z representation now we have to say something about exemplification now why do we want exemplification before getting into it I want to tell you what motivates getting into it so what we really want to understand scientific models is that models are objects with an internal structure so they're not like words so you can twist and turn the word you well enough as long as you want you won't learn anything about the city but models are not like this models must be objects that have properties these properties are important in their representational function and you want to understand how this works and that's essentially the problem to which exemplification answers or Reguse has this nice phrase where he says models are objects that have an internal structure they have a life, an internal life of their own but that's what we have to understand so exemplification you're all familiar with that you want to paint your living room you go to a paint shop and show paint samples and you choose the color you want to use to paint your room now this a paint sample sort of paint swatches you have here they represent by exemplifying their own color so that's an example of representation by exemplification now what does that mean? intuitively an item exemplifies a property if it represents by instantiating the property and one can coin a formula here and say exemplification is possession plus reference or instantiation plus reference and if you want examples beyond that I mean samples of all kinds are examples for this sort of this kind of representation if you go to the market and you try the cheese before you buy it, the little bit of cheese you get this cheese sample and that would be sent by exemplification because the sample doesn't just instantiate the flavor of the cheese and it represents the flavor of the cheese I've noticed that lexicographical science don't do this so words don't have that property the word London doesn't instantiate any properties that you would associate with the city of London excuse me I think that you should say by instantiating the same property because drawings for example they instantiate properties and properties of Margaret Thatcher and exemplified by the drawing but some properties of the drawing do not correspond to properties of Margaret Thatcher some of them do but you've just anticipated the slide so it's sameness but the truth that's exactly the point you're making thank you is that exemplification is selective in the sense that exemplification implies instantiation but not vice versa so not as you said not every property that is instantiated is also exemplified so exemplification is highly selective now take again our color swatches they exemplify their colors but they don't exemplify the geometrical shape the thing could be round and it would still do the same thing so rectangularity is not represented here so exemplification is selective and the selection a lot can be said about how this happens but basically it's a function of the context so if you've shown this in a paint shop it's obvious that the swatches exemplify color if I somehow negate and bring it to geometry class and hold it up and say rectangle then it may come to exemplify rectangularity but it doesn't do that in the context of the page distorted sorry? distorted they can distort as well I will come to that later so far it's really the same same property a certain shade of red is exemplified and it presents exactly that shade of red but this is exactly a point of want to relax later on so again you see how this answers to what I said before we want an internal structure so you can study the paint sample the paint swatch and you also have exemplification that requires epistemic access you can look at the color swatch if the color swatch is too small to see then it's simply not a color swatch and that sort of gives you the epistemic access to properties that you want in scientific models again coming back to the entire diagram here so we now also have said what the exemplification is and so we have the whole thing in place once again you had this already but just to see how it works the caricature that's a boxer representation denotes stature it exemplifies the properties and it imputes these to her ok so far for Goodman and Elgin that's the background now we want to apply this account to scientific representations ok the first move you make is just you take the caricature out of it and you stick the scientific model in so that's what you want to do so you would get something like the philips newlin machine representation that denotes the economy of Guatemala it exemplifies properties like for instance there being an anti correlation of interest rates and inflation and it imputes these to the Guatemala economy just quickly why do I pick Guatemala it's not just because I always wanted to go on a holiday there but so when we researched this paper we really followed up on minutes and reports from central banks of course central banks on the toes are secretive about what they do so it was a time when basically every central bank had one of these machines the Bank of England had one the National Bank of New Zealand had one and so on but the Malans are the only one who have gone on record as actually using it for policy making for their economy so they were quite open about this I don't know maybe the Bank of England but they just don't tell us but that's why we picked the case of Guatemala because they only know that it's actually used in this way but now we have a bit of explaining to do what does it mean to say that this thing is an economy representation what exactly are we asserting here how does that thing exemplify property like there being an anti correlation between interest rates and inflation to impute these properties to the Guatemala economy and so that's what we have to explain now now let's turn to Z representation in science and so the question here really is what turns this thing here into an economy representation and the answer James and I are engaged is given an interpretation of corporate water pipe properties in terms of economy properties and that gives us the Z representation and rather than arguing for this in the abstract I want to show you how the machine actually works and some commenting on it yeah I'm going to try to do some calculations some actual calculations to make some predictions and see if we can get the right answer I'll just close some things down taxes I lost the budgets some very simple calculations today and what I'm essentially going to do is shut down the foreign sector completely I'm just going to close all of ours down here and we'll just concentrate for a moment on the banking sector and the government sector for a really banking sector is what we're really going to look at today and see if we can do some predictions do some calculations this is one of the big advances that Felix made he told that there were quite a bit of disagreement between Keynes and Robertson about exactly what it was that said interest rates and apparently if a pair of them had lived to see this they would have looked at Philip's solution to this and they would never have argued essentially it's treated like a a stock but one of the clear things about this machine is the way it makes a clear distinction in what's a stock and what's a flow at the bottom there's a stock of money here in the bank there's a stock of money held abroad in the foreign exchange and everything else is a flow and I'm going to try and do some calculation ok so I stop it here so you get the gist of this here's the guy standing in front of this water tank and say here's the foreign sector there's money in the bank there's the treasury and so on and I think that's exactly what happened with x that has certain x properties in this case this is a water pipe system that has hydraulic properties then you have a domain that we're interested in here and that's an economy and what happens is in an interpretation you just pair up x properties with z properties and one can refine this indefinitely I guess the basic level of what happens is you pair up with other sort of properties like having a central bank is paired up with having a big water tank in the middle or having a foreign sector is paired up with having a certain part of the on the right hand side of the machine and then mass terms are also equipped with a mass correlation function there was a lot of talk of money and whenever there was money at stake you point it to water and so the amount of water is correlated with the amount of money you can just say one liter of money corresponds to one million of the model currency whatever you want to say here so that's effectively what you do that then allows you to say given time interval for instance 5 million of the model currency flew through a certain flown through a certain bank so now we are equipped to give a definition of a z representation the z representation is a pair x i where x is an object like that water pipe system you've seen and i is an interpretation and now I'm going really out on a limb giving you a definition of a model I would say a model is simply a z representation where x has been chosen by a scientist or a scientific community to be a model so that is what a model is, I mean in the scientific context I mean I know for logicians that something different the kinds of models I'm talking about that's why I started saying I mean just in scientific models so that is the definition now some of you will immediately be tempted to cry foul and say well that doesn't require a target system and yes that's a conclusion I want to embrace so being a model does not require a target system and I will say more about this later so this is not just a remark I make in passing but that is exactly what we should say and so we see that the google definition is most correct so a model is not just a representation full stop it is a z representation so we can qualify it in this way so this is the example we have seen that already in the video this philips newlin machine becomes an economy representation if it's described as a water pipe object and water pipe properties are correlated with economy properties notice that nothing forces this on you you could be interested in the education system and be interested in how students flow through an education system you can take the same machine and interpret the amount of water as the number of students and then turn this into an education system interpretation by sorry representation simply by changing the interpretation here so none of these choices are in any way intrinsic to the objects themselves so now we can start building up the general account and we start building up that diagram that I had at the beginning in the preview so we see the model is the object x with an interpretation sometimes we call it an oz interpretation because it's the object of the z and that makes the model now that model can denote the target system in some cases sometimes the fillet newlin machine was used to denote say they got them out of an economy sometimes it doesn't and that can also involve part part denotations so it's not the case that only the whole model denotes the whole target you can obviously have part part denotation that's perfectly fine and that sort of subsumed and I take it that we set what denotation is and we leave it at that for the moment so now what about exemplification you recall exemplification requires instantiation but now we sort of in a little pickle here because we want to say something like that the machine instantiates economy properties but water pipes don't instantiate economy properties so that seems to be a category mistake but that seems to have an easy solution because we can introduce the notion of an instantiation under an interpretation so if we interpret water as money we can say the economy that's represented in the model has a million euros there if it has a little water there and that is perfectly good to ground exemplification claims so we can then say I exemplification is I instantiation plus reference and I mean there's absolutely nothing deep here this is just philosophical housekeeping so we're not saying inconsistent six so we understand what that arrow here means so the model can I exemplify particular Z properties but now we call that image again so we then just said these properties are just imputed to the target of the representation now in scientific context it's too simple so scientific models don't usually portray their targets as having exactly the same features as the model itself and hence the properties exemplified by the model are not the same as the one that imputed to the target so we need to transform these properties somehow and the job is done by what James and I call a translation key and to introduce that idea I want to take you to a familiar example namely maps so this is a map of Switzerland and so just imagine for the sake of the example this is an old-fashioned paper map that you have in front of your desk you can measure distances on it so I measure that the distance between the top and the bottom is 22 centimeters that's a property that is exemplified by the maps because on distances are important for maps but obviously you would have misunderstand the representation if you saw a map that says that Switzerland is 22 centimeters from north to south I mean the place is small but it's not that small so something else has to happen and likewise with the dot that has written in a yellow area you shouldn't entrap from that that the city of Chur is yellow so the properties imputed to the target are that the north-south extension is 220 kilometers at the city of 400-600 meters above sea levels and so there is a conversion taking place and that's given to you by the legend or sometimes the key of the map because you just look at what scale the map is that you look up either at the bottom or on the back or at the color cards and so that motivates the introduction of the notion of a key so when you need other for a more scientific example as I started something simple you use litmus paper to see how acidic the solution is so at some point the paper exemplifies property P is red the litmus paper doesn't tell you the solution is red the litmus paper tells you that the solution is acidic so effectively that representation comes with a key that translates colors into acidity levels and usually you get a nice cardboard box where you get all the shadings and you just hold it to it or you can have a tolerance threshold so the model can be plus minus 5% or there can be various limit relations James and I discussed this in the paper or simply idealizations in general can be seen as processes that inform what key you use for a model so that gives you that axis here so when you have the piece yes, one question is that do you mean litmus paper to be a model for acidity? it's a representation I mean litmus paper is a representation it represents properties of the solution that you stick it in would it also be a model? no, I wouldn't want to call it a model is there something that excludes it from the model? given definitions yeah, there is no interpretation there so when the litmus paper has a property that you key up but that thing isn't there so there is no interpretation like in the philips unit machine that's what I think but I think you pointed to an important area because the boundary between the Watson model and the Sonovon model is quite fluid but I think it is there is no conventionality in exemplification because the user you know, Bas van Vrasen insists in his theory of representation that representation is always used by someone, a subject a human subject so when there is an exemplification you like the color of the litmus paper the user has to see the app the text in which the user finds himself or herself has to make clear that this is an exemplification I completely agree with that and van Vrasen's hopes of representation is something like there is no representation without the user and yeah I'm in complete agreement with that I think van Vrasen and I have a slightly different idea of representation in detail means I completely agree with that and then we have implementation here and so we have the whole diagram and I've just made it a bit smaller to put my descriptions here and that also relates to the point you just made I want to be explicit that this is done in language we talk about this, this is done by actors so this is not a theory of representation representations somehow live in a physical space by themselves representations is done by users and now you see why it's a decking account that's just acronym for the key ingredients namely denotation, exemplification keying up and imputation that's where this comes from now I want to give you a few corollaries so we want to say that representation is faithful if T indeed has the properties that the representation attributes to it or that imputes to it now that this is the case is not built into the account the representation can be completely wrong someone amazing that Margaret Thatcher wasn't brutal and ruthless at all and the caricature completely misrepresents her still the caricature does represent her as such that's really all that representation tells you whether a representation is veridical or not it's not part of the semantics of a representation that's sort of a point that somehow often leads to confusion when people want to build more into an account of representation than should be in it so that representation is veridical it's not part of the representation itself just like sentences don't come sort of with the label of this leaves it when I'm true so that the objects have the properties that the representation describes to them is not part of the representation corollary 2 and I think that speaks to your question so the scientific model is not a synonym for scientific representation so not all models are representation of some things or some models can be just representation so Michael Weisberg discusses this nice example of four sex populations I think the three sex don't work anymore they found some crazy creatures somewhere that needs three sexes but for all I know at least please correct me if biology have moved on I'm not aware of any species in the world that have four sexes but still there are four sex models so these are models clearly but they're not models of something and that's why I didn't want to redefine models such that they require a target system and in the decade count you have a nice explanation of that you can say these models they are that representation so Weisberg's four sex models they are four sex model four sex species representations but they're not representations of such but vice-versa not all representations are models of graphs, diagrams, litmus, paper and so on so I should say clearly that this project has no imperialist tendencies I don't want to say it absolutely everything's a model there is a conventionality as you said about more counts as a model or not we can have a discussion about what convention we should adopt what James and I want to make is just a conditional claim that if we have something we want to call a model and that something would present also in the sense of decade that's the claim we would want to make corollary three is the decade account explains how learning from models takes place which is a good making feature thing so you study the model you see what's exemplified in a given context then you ask what is the key that the model is based on and then you know how you learn from a model and finally I want to emphasize that the decade account operates as a certain level of generality so we talk about objects keys exemplification obviously that needs to be concretized in every particular instance so this is not an account that straightforwardly applies to every object that's a model but you have to concretize it so you have to say what's the key what's the exemplification relation and so on extremely hurt but do you take your decade account of representation if characterization of representations something is a representation if it's a decade representation yeah we would want to say that so I know that is obviously a bold claim but yes we would want to say that at least once we agreed what the model is and we say that the model is representation then the model represents if and only if it satisfies the decade conditions okay but then I suppose in the properties when you talk about properties you take properties in a very general sense you take into account also relational properties oh yeah at this point we have made no assumption about what there are many people who talk about representation they talk about structure and the structure is probably it's of course captured by the notion of relation of properties and then okay whatever in a long version of the talk but we have so much time we would specify this further but no at this point no assumption about the nature of properties there can be monadic properties or polyadic properties it's not even assumed that they're all independent so there can be all kind of dependency relations between properties so I take this, this comes back to the previous caveat exactly one of the issues that has to be sorted out in particular cases I don't believe that there is a one size fits all accounts so that's part of my belief with structuralism that fits some cases nicely but typically my colleagues in philosophy stare at me in this belief and I tell them about structural representation because I say well none of our ones work in this way and I take it as an advantage of this account that it can accommodate that properties can be structural properties nothing wrong with that but it's not committed to them but I come to that also right now so as I mentioned at the beginning some models of material objects like the philips newly machined but obviously not all the Newtonian model of the solar system you hold it in your hand or in your hands the ball model of the atom the shilling model of social segregation and so on and so on so what are we doing about these I've just talked about material models so far and I think the good news is that this entire scheme can remain in place that's why I invested so much time into it we just have to say something about what this object here is and there are these schools of thought but before we come to these let us reflect on what we actually want let's work it backwards so we want that the objects and I put them in scare codes now because they are not objects of a physical kind they must have properties that we can study so for instance model planets must have certain features they must be right and wrong in the models the fact that something is not physical doesn't mean that it's arbitrary there must be claims that are correct in the model there must be claims that are incorrect in the model and we must have an epistemology for this we must be able to find out because if models say are sort of abstract objects we never have any access to that's not helpful so we need to have an epistemology for that but these are the three conditions of adequacy for any ontology two candidates here so one school of thought thinks that models are fictional entities something along the lines of Sherlock Holmes or Middle-Earth as Peter Gottfried Smith's favourite example is the other school of thought says that models are mathematical objects set theoretical structures graph theoretical structures or something of that sort there's a lot to be said about this I don't want to discuss this here although I have a favourite here what matters for the current talk is that Deckey is indifferent towards this choice as far as the Deckey account of representation is concerned you can basically take your pick whatever choice you make depends on further philosophical commitments you have but it doesn't depend on that account of representation and you can then see these descriptions that I have put there just specifying these objects either the mathematical objects or the fictional objects that's also one of the motivations to have the language always there now with an eye on time I skip some of the details here you can ask me later about these if you want and I just want to wrap up by saying I think this account for representation both for material and non material model you just have to make a choice for what you think non-material models are mathematical structure fictional objects or something else this account doesn't care and it can deal with all of them for instance if you're a structuralist and you go back to this you can say the object X is a mathematical structure the mathematical structure becomes a model you can write it in one way or another to take a mathematical equation you interpret one of the terms as a population density for instance it becomes a population model and all the rest just runs as it did before and I hope this is not just sort of consolation for the specialist as Fyroff and Grunz called it so I hope that this has somewhat short for scientific practice checklist here that practitioners could use so be clear on what your model entity X is and what properties it has make sure your interpretation is unambiguous and explicit so you have very clearly laid down what your interpretation is make sure it's clear what the target system is and what the notation of the target system means if there is no target system make that clear too so not having a target doesn't disqualify what you do as non-modeling or the single study is not deprived of the status of a model on this account it's just a model that doesn't represent a particular target system say if the Guatemalans had never bought that machine and this had all just been sitting around in the basement of the bank of England they were a bit but never used it for actual policy setting in the United Kingdom that still would be an economy representation and an economic model it just would not be a representation of the UK economy and it's important to be clear on these things then regarding 2 and 3 never confuse a Z representation with a representation of Z that sounds trivial I put in the abstract but the mistake is made and some of you may have read papers that claim that the world literally is a Z-automaton just because it's represented by a Z-automaton that's exactly the sort of confusion they have in mind a model can be based on the mathematical structure of a Z-automaton that's great but that doesn't mean the world is a Z-automaton not even if you want to say that this model is an accurate representation which it may well be but that doesn't mean that the world is just the same as the model object be explicit about the properties you take to be exemplified that's the point you highlighted so not every property instantiated is also exemplified it's highly contextual so be careful what you take at least in physics that's often left largely implicit so the use of physical model is often not very well explained in that part the world is just somehow like the model and you live it like that so ideally we really would want to know what the key is because that is what generates your epistemic claim do you have some notion of idealization at work a certain approximation of what's going on here and then say which properties are imputed to the target there can be properties that are exemplified but you don't want to impute them you have this option so say which ones you actually impute and then check the accuracy of the representation and that's the Sony question of how truth or trust is established as I said before the fact that a property pops out of the model doesn't mean that the target has the property so that is just a mistake so we need to have one way or another to establish that the target really has that the traditional method is of course just go to the laboratory and make a test but in some cases you may not be able to do that if a climate model gives you a result for the levels of global warming in 2050 you can't just stick the global climate into the laboratory and see what happens in 2050 you need sort of other measures to come to a decision of what's trustworthy and what not but a theory of representation is nothing to say about this and I don't mean that this question is not important it's extremely important it is just a separate question and sort of loading an account of representation with the demand of solving that question too which is modest of orders in a way that we should ok and on this note I say thank you so much take five and return with the comment of Count Andrew I pointed at the other screen so you don't normally have hybrid commentaries but this is cool this is with a new a new test I like this I'm happy to have a discussion with the comments of Cantin please you have the floor if there's a problem with the sound I will tell you ok can you hear me yes so thank you for all this very interesting questions so you presented this caricature of my attention this is a representation of as a boxer and then you talked about I don't remember the name of the machine the water pipe machines and I was expecting you to say this is the representation of the economy as a water pipe and I was surprised when you say no it's an economy representation so it's a representation of the economy as an economy and so I was surprised and I wonder why would you need because you would have why would you need this internal thing between Z Z and X and then the keys and it seems to me that there is some kind of redundancy and that you would get all the job that the interpretation does just by having the keys to the relationship between the object and the document so you would have a representation of the economy as a water pipe and it has it exemplifies the probability of having one liter of water and this and you have a key that translates one liter of water into the water so I was just a bit puzzled why do you need the interpretation inside the language and there's the keys and why not just one kind of relation that is given by the keys the other first question I don't know if I would you prefer to answer one after the other or both at the same time I can't hear you please please continue the microphone is muted but the other one is not connected to you too I can't hear anything I have to find where is my moment where is it wait a minute so he knows one second I think please please ask all your questions first please and the second question is so what is the count is that it's very liberal so it can come from maybe Canada's representation so it's maybe you think about I guess you know this theory which is very influenced by Canada who say representation is just stipulation so it's a bit more sophisticated so it doesn't have the same problems of Canada it's just stipulation but I thought maybe it has one it could have one problem maybe you have a solution but there is a paper by a post who criticized it for not taking into account the kind of comfortable aspects so basically or to say it maybe differently something can be negatively representation because of it seems it could be representation of something because of not of use because so particular user takes the model to be representation so for example I can have a map of a woman and a man in my pocket and I'm not using it I'm not taking anything of the map to be to represent but then you could say well this is a map of a woman and a man another thing because it's the normal use it's kind of a community but I don't know if you have something to say about community not of representation if you want to reduce everything to use particular users particular users or if you think that there is another story to be told about not so that's my three questions okay do you want to project again? not necessarily I don't know I may need to get at some point but thank you very much for that these are really good questions and thank you for drawing attention to these let me start with your first question so you ask why do I not say that this is representing the economy as a system of water pipes so in effect you're asking what is the Z here so why do I not say this is just a water pipe representation the point is it comes out of an analysis of how these models work in practice if you talk to economists what they really do is they say well this is a Keynesian economy because what is enshrined in the way the flows work the system is set up is effectively the principles of Keynesian economics well in the Hicks version but so you're short changing what the model does if you say well that this is just pipes with water flows and so it's really important to have that mechanism in place and if you spell it out in detail what should really say is that this is a water pipe system that is a Keynesian economy representation all hyphenated so that's the Z and it represents say Guatemala economy as a Keynesian economy and well one this is important for the practice of how these things are used but also for our understanding of targetless models I think it's important to me that I think I've emphasized that numerous times during the talk that we're not making the representation of an actual target in any way a condition for being either a representation or a model and say I mentioned that in the talk even if the Guatemalans had never used the model in this way it would still be an economy representation but it wouldn't be an economy it wouldn't be a representation of any particular economy and that point becomes impossible to make if you lump the key and the interpretation into one because if you don't have these as separate steps and you don't have a target system there's nothing to key properties of it and then suddenly you're there with that water thing but no actual target to which it relates then you don't really know how to do the semantics anymore so keeping the interpretation and the keying up separate helps you solve this problem but I do accept that this introduces a little bit of attention where the heuristic examples of the caricature at the beginning don't really have that separation very clearly but that is also because you don't really have an account for the Z representation is explicitly in there you just look at the caricature and say ah it looks like a box now if you if you insert in a proper account of Z representation there for instance take the seeing in account that's associated with Gombre or Lopez or Wolheim you would say well this is a is an array of lines on paper that is such that informed spectator would see a boxer in it and the boxer is then keyed up with a particular subject in one way or another so that would make this two step process more visible but I definitely think in the scientific case it's worse having both to make sense of targetless models and to do justice to scientific practice and if I maybe can ask another example here the one that I had on the slides that I skipped over at the end because I wanted to leave time for discussions so to take the Newtonian model of the solar system the model itself is is an imaginary model it's a big sphere it's a small sphere that are placed in otherwise empty space and the interaction is just gravity between them that becomes a solar system or rather a sun-earth system model only once you say the big sphere is interpreted as the sun the small sphere is interpreted as yes there's nothing intrinsic in two spheres that tells them to be sun and earth and you could make another choice here you could be Niels Bohr and say ah the big sphere is the proton and the small sphere is the electron and then the exactly same model object becomes a different model because it has a different interpretation and again it helps to have interpretations here because it tells you what the representational content of the model is as understood as a set representation now I hope that answers the first question now the second questions was about community norms of what represents what or is it just an individual volunteeristic act I think this is a very good question and this is something I see very much as space for further work in the taking account we have been talking very much in sort of a user centered way someone takes something to represent something else and that's often how it happens at some point Newton said let's take these two spheres and let them represent the sun and the earth that does not imply that all representation all this works in this way that doesn't mean that there are no community norms there could of course be community norms that regiment things in one way or another and sometimes one is even caught in the middle between these things there is a norm that would suggest that you should use a model in a particular way but then you want to use it in a different way an example here is is epidemiological models that tell you how diseases spread they obviously come out of medicine but they have recently been used by the police in crime fighting they sort of they've torn these models out of the epidemiological context and say we're now seeing not how a virus spreads but how crime spreads and so that's something where you had a community norm but then the community norm was changed by force by tearing the model out of the context and as if there's a lot of interesting things to be said about how this works and what happens in such cases I don't think the DECI account has formulated it gives you any of this so it is just silent about it but I don't think these considerations in any way stand in tension or contradiction with the account I mean this is just something that could be inserted into it and could be used to complement the account in this way and I think this would be an interesting thing to do Thank you, Kantein, do you have some further questions? Just a bit I will ask some more precision on my first question so what difference do you make between the water pumps to say it's exemplify having a 1 million dollars in the ILO in that place and then what difference is there between the color because the back exemplify the color but in this game the color is not imputed so I'm a bit cautious about the difference between the two cases why is the color exemplify but not having more later liter why isn't it exemplify the color is a property that you put into the key and that is keyed up I mean as you said at the beginning of your comment it's relatively liberal and that allows you to do a number of things maybe you could give an alternative interpretation of the map you could say the map so I exemplify being 600 meter above sea level by being yellow and then that is imputed to the target with a certain precision key probably just minus 50 meter because the city is sort of hilly I mean I'd be happy with that I mean I don't think there's a problem it just depends how you want to apply the account to a particular case I think you're right in pointing out that there is leeway here but I think the account can accommodate both uses applying my back to you but it's sort of a strange setup to actually see you I have to turn my back to the camera I see myself okay thank you let's now return to the general discussion people online can write their question in the comment there will be a little bit of delay but we will come to your question just first maybe you want to return to your and maybe okay questions first Michel okay thank you very much this is a very stimulating talk I'm very interested in representation so I have several questions but I think I tend to make in my view of representation rather sharp distinction between on the one hand the representing artifact the X the X in your terminology on the one hand and on the other hand propositions propositions I think that representing artifacts are not propositions they are not propositions of course some representing artifacts can be called faithful correct or even true but only in a derivative way it's because the imputation I think that's a nice word imputation by the model of some properties to the target is realized on propositions true propositions for right true true propositions because if you want to representation that is a mean representation representing artifact to be successful in the first place that is to identify the target it is about and second that's the first thing and the second thing is well is the representation of the mean representing artifact correct or accurate is it faithful because in order to to make the representing artifact to be a representation of a specific target it must be true for example taking the example of the economy of Guatemala that there is a banking system in Guatemala or there is a woman Margaret Thatcher who has some face which is on the representing artifact so that's the first thing the second thing is that you need to rely on true propositions to be able to say that for example Margaret Thatcher is represented as a boxer or rather that someone who is a hard puncher she's not a boxer in the context I think that's the point of interpretation in this example so there cannot be a successful representation and further a correct representation or true representation I don't have to be thinking about words here unless you suppose that some propositions are true and I think that's an important distinction is something that seems to me that you don't make clearly enough but at least to me in your account Okay, thank you I think this is at least two aspects to what you're saying this account presupposes that you have means of target identification that are not reliant on the model itself I have said nothing about this so I think this is also something one has to say more so we have to be able to identify the target prior to having the model in place and this is because I don't believe in this sort of magnet series of representation where the model represents whatever is somehow structurally isomorphic to it I think this is sort of getting things things upside down so I agree as you say we need propositions that's why I have a description of the target where we can identify the target there's a good question about how this works and I haven't said anything about it I accept that so that stands outside I think the key issue here about identifying a target which is susceptible to be represented you have to do some kind of abstraction you didn't mention abstraction but you abstract some properties of the target and then you make those properties by which you call the key and correspond to some properties of the representing artefact it seems to me yeah that's a separate step I think we can identify a target through a pointing for instance I could just point to this guy the sun so I don't have to abstract anything but I have to have the means of identifying the sun somehow and these means of identification get more tricky when you get to things you can't observe directly like micro or macro entities so there's a lot to be said there but I don't think that step involves any abstraction or anything like that this is just identifying the object that comes in when you choose a what you call the representing artefact there's nothing in the sun and the earth that forces you to represent it as perfect shears that's a great model but you can have a different model I mean Kepler had the model famously so this is sort of the creativity of the scientists who comes up with a particular model object to take the shelling model of social segregation it's a genius idea to take a checkerboard a dynamic on a checkerboard but there's nothing in social segregation that forces this on you and obviously you can then use these as models only if you can tell a plausible story about what properties they exemplify and how they are imputed but what account you give depends on what you choose and so on and comes into play now for the second kind of proposition I see models as generating proposition because what sort of in a sense you can see this whole diagram as encoding property attribution namely target T has property Q so that's the claim that the model generated I don't think the model itself is but the model generates a proposition T has property Q but I will not say this proposition must be true it's a proposition can be true of course that's an empirical question which is the case if you want to make a model of a specific object like the proton or the atom of course you cannot point to T but what you do then well I suppose that the proton has a mass as a charge rather than this case as a charge they have properties of motion and this is a way to identify the target now whether the atom is truly a planetary system in some sense that you can put into brackets but there is a way to identify the target and then what you do is you attribute properties to the target and then you make statements and then which corresponds to propositions and I think that's the basis the start of any representation yeah I completely agree with that the point I'm making is that the propositions that you make that are true and that you start with as you say they stand outside the model they are pre-model as it were they are in the background series I also saw something I didn't talk about but I don't think that models live in the vacuum so sometimes this model literature has gone overboard a little bit by thinking that well everything is a model it's all this work against a theoretical background and we need that theoretical background to make the model work for instance by saying we identify the proton through a certain mass through a certain charge maybe through a certain spin the proton is just the thing that has these properties and then we can model it and say it consists of walks and whatever else you want to say about it oh yes I think we're in agreement here this is just a part of a theory representation that I haven't talked about today Kevin thank you very much for this talk I agree with you that there was a time in which of models there is not enough information I was wondering how are you going to be second am I there to find the presentations and how they work what is the relation between your model and your question if models are not always presentation of something there is no target what do you learn about what do you learn about what is the target of the presentation of your model ok let me do the last question first because that's probably quicker so what do I learn in a representation that's not a representation of well I learn about how the model objects behave so if you study the 4x population and you stipulate certain properties from the 4x population I just learn how that 4x population behaves how fast it can grow how fast it shrinks and you basically learn how the objects behave how useful that is depends on what you want to do if you are not interested in 4x populations it tells you very little so you will just learn how the objects in the model behave and that is often extremely useful that's probably more your specialty than mine Alexander but if you look at early quantum fields here for instance the so called 5 or 5 model it was known very very quickly that this model doesn't describe real particles there are no young particles but the model was studied extensively because the techniques you can use in this model were very very useful so physicists learned how the thing behaves on the renormalization they learned about symmetry breaking they learned about all kind of things like the 5-4 model and if you are interested in these things that's a real treasure trove if not so in brief you learn how the objects behave and in doing so you learn about the concepts because you are learning how your conceptual scheme plays out in certain scenarios and that is extremely important now where do series fit in here again a lot can be said about this but often this goes into the specification of the model object I'm sort of cutting a long story short if I say the Newtonian model is two perfect spheres in empty space that by itself doesn't give you anything nothing follows from that so you have to have two perfect spheres that attract each other gravitationally or at least with a one over all square law and they follow Newton's equation so often to specify what the X is in the theory in particular with non-material models you need series because the series tell you how the objects behave what is the time evolution and that is one of the key that is where series enter that is also where series and models work in tandem I mean Newtonian theory doesn't tell you much about planetary motion by itself you have to sort of model into it but then the model by itself doesn't tell you anything either so you need both of them and so if you specify what the model object is almost invariable there may be some theory that will end there there may be further theory entries when you want to spell out what the key is and things like that but the most important point is the theory entry here is the specification of the model object so if there is an under relation of denudation but under relation of interpretation or zero presentations I would say series first and foremost live inside this orchestra here so Newtonian theory doesn't tell you you have to interpret the big thing as the sum it's just a big ball that moves in a particular way so interpretation here is not determined by Newtonian theory nor what is exemplified but without Newtonian theory in here I mean there's nothing here for us so I wouldn't say that and different models work in different ways there's often these questions that I'm asked so having models and series relate and my answer to that question is that it's the wrong questions in the sense that there's no one relation that models and series have to each other there are many relations that models and series can have and you have very series driven models like Newton's model on one side and you have practically independent models like the Vodka that's a Vodka-Volterra model on the other which doesn't have any biological series in it it has some general mathematics in it but not anything that biologists would recognize as a biological series there's basically everything in between so there is a sliding scale of degrees of series dependence in models and that has a lot to do with having a specific idea yeah this is great stuff I really like this approach I wanted to push on a great area that you mentioned near the end of the talk which is between representations that are serving as mere representations and representations that are serving as models because on the one hand I can actually ask this pretty rapidly I share your intuition because I'm actually playing within another context for a project right now an object that seems to sort of inhabit a gray zone where it's sort of unclear how it's being used in practice whether it's being used in such a way as to be a mere representation or a proper model but then on the other hand when I look at the account full account this account is so heavyweight that it makes it seem as though it would be difficult for there to be a gray area meaning things that are used as mere representations and things that are used as models because something's being used as a model implies a lot of moving parts that we ought to be able to detect in the scientific practice and so I wonder what you think about these cases how do we make sense for you mentioned a bit of conventionality whether we decide to sort of take a representation as a model or take it as merely a sort of process how is this going to work and how do you maybe a related question how do you when you go to apply this account in a new domain of scientific practice how do you go about finding parts in the practice of what the scientists are up to to see how the account maps onto what they're doing first it's a weird way to draw the line I mean as I mentioned I think there's a degree of conventionality in this but I would say if you look at the scientific practice you would call something a model if it is an object in a certain sense that has sort of an internal life it has a dynamics you see how it involves it does something either physically on the table or in the fictional world of the model that you can actually study something on forms in it I think that is a defining feature of a model and if something doesn't have that it's hard to call it a model and then it's the aspect of there being an interpretation so you have model objects or sort of the artifacts that are used representationally and you interpret them in a particular way so you interpret spheres as planets with you more on the structuralist side you interpret certain terms of an equation as particles so that's interesting because the equation has certain properties and if you think about the imaginary scenario that you can create through these equations something happening in them and I think that is really something that is a defining feature for a model now let me be clear this is a pragmatic feature when I extract this from how scientists speak how we want to carve this up I mean there's nothing semantic in this so there's nothing that has any philosophical necessity to tell us about how it behaves and you can find out about how it behaves I mean take a Poincaré 3-body problem-data there's a model with 3 bodies and Poincaré spends many times studying what it does you find out something about the model so you have to be able to look at the object in this way and it must do something and you can watch what it does and I think that is a crucial aspect us calling something a model now how would I go about it and find out in scientific practice I think there's no silver bullet it's just case-by-case and if I have to do this in a new field I just go and read a textbook or two or three and talk to practitioners read research papers check it out figure out what it does and sometimes you find out that well maybe they're actually not really modelling at all maybe they just give descriptions of what's going on and that's fine that's probably the last question how do you deal with other things so here we have a model that does the something controversial claim that they will make and that will be refuted if that's wrong is that this account applies to the old models so you could put an image so you can put the microscope image so take the model around forget about all this just replace this box same with the microscope image that microscope image has all kind of properties but it doesn't accept the effect of being selective you key that up I think at least some visual representations of the science work in exactly the same way it's an interesting case study to be made still by the way we have a lot of stories about black holes that have been pictured from the Harvard group to the pitigalism so you get these these wonderful images but I mean you're not just seeing a black hole the way you're seeing a football so there's a lot going on there so you put that image and you have to spell out very carefully what properties that image actually exemplifies there's a story about a new colour to be told as well which is that they use actually a community conventional colour colour counting so what is exemplified is imputed and thinking about these images in this way I would submit actually helps understand how these images work that's cool Peter? ask a question strong enough because sometimes people on the internet they don't hear well questions that are thank you for coming so thanks a lot for I think it goes to you so this is very appealing but that was a bit well maybe my question has some mightnum undertones and I hope it's interesting beyond so I was a bit surprised with your sort of quite almost actualistic or realist the idea of well the unicorn is another representation of the unicorn and the four sex population is not what is represented but I mean this seems to be very interesting representations of completely fictional or possible objects or even impossible objects and there are still representations of that object even though it is not physical existence and think of mathematics like of course mathematics itself could maybe be a model of course but a square can be a representation a square is not a very interesting model but there is complex mathematics that is represented by very simple stuff as well as mathematical objects that are represented and are modeled I guess I could imagine a model of what's the name of the Harry Potter sports which some complex fictional stuff that has been described but interesting words and you want to get a better model of it you want to like describe how it can have behaved supposing like Harry Potter would be a coherent but the main question is why you won't want to stress that it's actually physical existence I don't know what about the physicality but actually existing stuff that can be modeled why not abstract object or even impossible fictional okay thank you I think you are right to say that I have some actualist bias but I think for the representation account actually nothing much handled so I just took it to be a datum of scientific practice that signals that we take not to exist are somehow the subject matter of models like forcex populations or young mills particles and things like that now if you want to turn around and say no they actually exist they're just sort of mononium objects they don't really exist they are according they might come into existence and that's why it might be scientifically interesting I mean in the Dekki scheme you just reclassified forcex population representation would then also be a representation of a forcex population if you think the forcex population exists so that depends on your metaphysics you have a sort of ritual ontology denying presupposing I don't think anything in the Dekki account stops you from using it in this way if you think the golden mountain exists and you have a model of the golden mountain then it's a model of the golden mountain that's I think by the likes of Dekki that's fine so to make the physics of what exists and what doesn't stand outside it so if I just wanted to make room for non-existence in the sense that at least one might want to have the option of talking about representations in the case of objects that don't exist so you don't want to commit yourself to objects existing just because they appear in a model so it's more sort of creating aspects and obviously things can be reclassified that happens in the history of science take the Dirac sea of electrons for instance Dirac had this negative energy solutions to his equation he first just threw them away in sort of mathematical artefacts and then suddenly they got reclassified above positrons so suddenly what looked like was just in the ball this was real and so the account is flexible to make sense of that you can have your own your own ontology combined question on mine maybe it's my turn yes it is when you described a good modeler I thought you were excessively you asked too much because there's good scientific representation where some parts of this diagrams is opaque let's see the example but there's the philosophy of biology and medicine will correct me an animal model to study some kind of physiological process you don't master well the key necessarily the exemplification is partial however you have good reason to use this model some evolutionary history the mouse is not that far from us so it's not that bad so is it bad modeling or is it maybe it's not a model at all in the accounting to the key I don't know look I would have to look at the cases but at least in the cases I'm familiar with there is a degree of handwriting going on because the key is often not spelled out but it's assumed to be in the background this is a plea to being explicit of course we acknowledge that not the whole model may be keyed up maybe there is an uncertainty to it maybe there's bandwidth as it were that's fine nothing in the key says it has to be precise it is just a plea for saying something about how you think that the properties in the model relate to the properties in the world beyond the way the world is in one way was all about like the model and I think that is what we should demand models think back to the beginning of COVID when we had all these models that said all kinds of things about multiplication numbers how the infection grows orders of magnitude and each group said well the pandemic just behaves like a model and then you're staring at each other so I think this is very unsatisfactory if they had said the key is just one of tendancy we take this model to say that the pandemic will increase or infection numbers will increase but we don't know by how much that would have been extremely helpful and that's a key which is saying we take the sign of the change as the properties we impute but not the actual magnitude and I think it's not enough to demand that and I think that's an integral part of the epistemology of models and that scheme is liberal in the sense that basically any relation can specify somehow qualifies as a key there's no sort of threshold that you have to pass but if you really can't say how do you think the model relates to the world then you have an incomplete representation you still have a model in this view how do you complete representation does that still sound to you in complete in what sense okay I understand your argument about the key even for the mouse representing some model of human process the key, you have to say something about the key of course but the exemplify in the end you're not exactly sure exemplify properties it's not that easy you have good reason to use this model because you would say they are close in the erosionary tree so maybe it's good it's a good test model good button it will always be incomplete except if you have a complete understanding of the process you want to understand okay look it's a loud asking completeness no, I'm not asking completeness I ask for specificity so this can change all the time you can change your model object, you can change what you take to be exemplified you can evolve in time there's nothing wrong with changing a model and changing its understanding what you take to be exemplified all this is totally fine I just think at any given instant of time a model must be able to say now given what we know we take this and this and this to be exemplified and this and this and we don't take it to be exemplified and that's how we key it up and that's currently where we stand there's nothing wrong with changing your mind in the light of better evidence or further insights so what I think we cannot do is remain voodoo so that's really what I think matters it's not that you name it downwards but at any time you can be specific Peter? yeah this might be a kind of far-fetched question I'm already apologizing I don't know whether I would make the idea very clear but I was wondering whether you always have this sort of direct denotation like nondescriptivist I guess that's like the object is there and it's just like the model is then independently of that constructed right but it seems in science often we know the existence of the model or we know we can refer to it via the models we've developed so there maybe if we talk about some black hole or something I don't know enough about cosmology but some black hole very far away I mean we can maybe make a model of that of that black hole but it's always through the bigger model that is our view of cosmology at some point maybe it's not the best example but the feeling that we can sometimes identify something by its place in the way the model describes the world or something like that yeah I wouldn't call this the model I think as I said before I think models were in the context of background series and background series often help you identify identify objects and series are complex constructs that have various layers and so there's nothing simple but a model of this view is a relatively local relatively narrow it's a particular object that's typically a particular way now that cannot identify its own time I think so otherwise you get in this vicious circle that well this model refers to whatever is similar enough to what the model is and I don't think that helps much it becomes almost so you at least to make a miracle okay so I mean obviously mathematical sciences this may be different but if you do physics or biology you have to be able to point out what your target system is and that if identification has to be independent of the particular model you're going to use tomorrow its properties maybe you could associate it with some observations that make it operational even though you cannot really under logically point at something but then maybe you're modeling the observations rather than some object I'm not sure so I might debate that I mean radical empiricists would think all models are just about phenomena not about objects but I mean cosmology is a good example in that sense because just the fact that there is a model that has a black hole in it doesn't lead cosmologists to believe in the reality of their commas there is an awful lot around it that has to do with empirical observations with measurements which what your telescopes tell you and so on all that is part of identifying the target I think that is crucial I mean if someone just comes up with a model that wouldn't be seen as establishing of an astronomical project when how this is done is highly no trivial and I mean as I said I haven't said anything about it so I think this gets us into a whole different way of debating identifying entities and all that and things like that very complicated and obviously this account doesn't solve these problems with a magic land but I think to meaningfully model you must be able to identify the target independently of a model and otherwise you can't regenerate and just be afraid of the collapse or is that that's a claim I am committed to I mean if you're from Rome but that's interesting because you're asking to the background theory to do a lot of stuff that clearly your approach is not able to at least this approach but on the other hand now I realize that you cannot defend the semantic view so you cannot define that the background theory is just a collection of models or something else okay but that's another interesting discussion it's an ultimately now I want to know what is the background theory for you that's why I have all these linguistic things the semantic view we will get nervous so once again this is not an imperious project I don't think this is all the philosophy of science this account doesn't do everything you want to do it's just an account of how models represent there's many other important problems and I'm generally skeptical the accounts I presented cut the woody and knots and certainly all the problems are solved and I don't want to promise anything about this it's a localized account I hope it does what it is designed to do but it doesn't work in isolation it doesn't solve every problem in philosophy humble note thank you Roman thank you for the questions