 the epiphenomenal argument for symmetry to reality inference. So this is kind of one of the central arguments that I'm making in my in progress book length project on symmetry to reality inference. So I'm excited to share it with you all. Now, I'm gonna start with kind of a rough gesture at the formal properties of the stuff I'm gonna be talking about. As I mean, especially as anyone who has read Bella's paper on symmetry and equivalence will know there are a bunch of ways to fill in the details of the sketch that I'm going to point to here. And I'm not going to kind of like commit myself in this talk to a particular way of filling in those details. But you know, I'm just I'm the assumption I'm going to make is that there is going to be a satisfactory way to fill in those details and that it may be in certain ways kind of dependent on which theoretical framework we're working in so that different frameworks will require different precise definitions of a symmetry. But here's the kind of general features that I'm gonna take such a definition to have. So a symmetry as I'm using the term a dynamical symmetry is going to be a transformation from possible states of a theory to other possible states of the theory, right? It's going to have the further property that it takes dynamically possible histories to other dynamically possible histories which is the same as saying that it commutes with the dynamical laws. That is to say, if I take a state I transform it using a symmetry and then I run the dynamical laws I kind of get the same result that I would if I first ran the dynamical laws for the same period of time and then implemented the symmetry transformation. There's a complication with how do you apply this criterion in the case of time-dependent symmetries like boosts or time reversal. There's a nice discussion of this in David Wallace's paper on symmetry. So I'm gonna kind of again set aside those niceties for our purposes. Another assumption I'm going to make is that symmetries act smoothly continuously and differentially on the theories, what we ordinarily think of as the theories independent and dependent variables and its state space, right? That's going to be an assumption that as we'll see is kind of important to getting any mileage out of the stuff that I'm gonna say. And so at the end of the talk, assuming I have enough time I will refute an objection by Shamik Dasgupta to this smoothness assumption that I'm making. So what's symmetry to reality inference, right? Well, Dasgupta coined the term for a thing that we've been doing for some time. I would trace it at least to like, well, I mean, Leibniz is kind of doing it, right? But I would trace the kind of contemporary version of it to Vile's book on symmetry where he talks pretty explicitly about the special relativistic revolution taking this form, right? So just a sec. Sorry, I think I was doing something that was slowing me down. I think I'm fine now. Okay, so here's the way the inference goes. If I have a quantity Q that's not invariant under some symmetry of my theory, T, right? I start out with the observation that Q is not invariant under the symmetry. That is, it changes when I apply the symmetry. Then the symmetric reality inference pattern goes, well, that gives us a strong reason to prefer an interpretation of that theory according to which that quantity, that non-invariant quantity is not robustly real, right? So, Vile uses the term objectivity is invariance under the group of automorphisms of the theory, the symmetries. I'm taking that to mean something like, well, when I say objectivity, I mean what's really real, right? And nicely, we've already seen some of Jill's thoughts that are kind of related to this, right? I take real real realness, robust realness, as I call it, to be something basically like Jill's notion of structure, right? What is, which parts of the theory are a better match for the way that reality is in itself versus which parts of the theory might be conveying accurate facts in a certain sense, but are a poor match for the way that reality is in itself, right? So, but you know, you can fill it in with whatever your preferred notion is of something's being physically significant or highly fundamental according to a theory. And I don't think that's a distinction that anyone can really do without in our business. You know, I'd be willing to explain why I think that is, but I think it's fairly obvious. A thing to note about symmetry reality inference is that it's a norm for starting with a partial interpretation, right? That is a theory where you know some things about what the theory is saying about reality and coming to a conclusion about what a more complete interpretation of that theory or perhaps a fully complete interpretation of that theory ought to be. That is, you know, a more complete story about what that theory is telling you about reality. You need some ingredients in order to do symmetry reality inference. You need to, well, you need enough ingredients basically to apply a dynamical definition of what, you know, symmetries you're talking about. So you need enough at least definitions of what's going on in the theory to describe the theory's dynamical laws at a certain, you know, in a certain broad way at least. So this isn't going to get you from a completely uninterpreted theory to an interpretation of a theory, but it will allow you to take a partial interpretation of a theory and end up with a better interpretation of it. So here's my goal for this talk. One thing we have already plenty of is kind of parsimony based arguments for symmetry reality inference, right? Arguments that say, well, you know, quantities that aren't invariant under symmetries, they're not detectable. They are surplus structure. Occam's razor in some sense tells us to do without surplus structure. So let's do without those quantities that fail to be invariant. I like those arguments just fine. You know, I totally buy them, but I think there's a further argument that you can make, which is different in a way. So the older parsimony based argument says, here's a theoretical virtue Occam's razor that non-invariant quantities do badly according to that virtue. I'm gonna try to point out, well, here's another theoretical virtue that non-invariant quantities do badly according to, right? That is, I'm gonna argue that quantities that aren't invariant have some pathological properties and hence we have kind of extra reason to reject them even over and above the parsimony or Occamist based reasons that we have to reject them already. So here's the kind of starting point for why I think the non-invariant quantities have a pathological feature. They are, as I'm gonna say, epiphenomenal, right? That is to say, what do I mean by epiphenomenal? I mean they kind of, they live in their own separate shadow world is kind of like an evocative way to put it, right? They notice the non-invariant quantities kind of notice themselves and they also notice the invariant quantities, but the invariant quantities kind of don't notice what's going on with the non-invariant quantities, right? So the non-invariant quantities are kind of like this extra shadow that's cast by the universe of the invariant quantities and doesn't really do anything back to the invariant quantities. And I'm gonna say that's a pathological feature that they have. So here's the kind of example that I think is relevant to motivate this. The non-invariant quantities can't be reliably correlated with invariant quantities. And John Roberts in a really cool paper about symmetry from BJPS shows how this can be used to kind of establish the common proverb that quantities that aren't invariant under symmetries are not measurable, right? So the kind of example he's thinking of is the following. Well, suppose that I take my prized possession and it has some absolute velocity and I think I've got some kind of absolute velocity detector, right? So I hook my prized possession up to the absolute velocity detector and the little like digital readout on the absolute velocity detector reads out some number. I think it's saying that it's going like 60 miles an hour at absolute velocity. Well, that readout, of course, is a readout in terms of invariant quantities, right? The display shows you the kind of relative positions of the little LCD things on the readout screen. And now those don't change when I apply a Galilean boost to the entire system of my prized possession and the detector. So as a result, right? The detector's output number and invariant fact, right? In terms of the kind of relative positions of the LCD blips, that's going to be left unchanged by that Galilean boost but the absolute velocity of my prized possession is not going to be left unchanged, okay? So of course this example shows that quantities that aren't invariant are not measurable. What it also shows is that you can't set up your non-invariant quantities in a way so that the invariant quantities kind of care what their values are, right? There's no way to set up a kind of counterfactual dependence or a kind of like dynamical dependence of the invariant quantities on the non-invariant quantities because you're just going to be able to present another argument like Roberts's argument for whatever symmetry it is that you are talking about that these quantities fail to be invariant under. And so they can't be reliably correlated by the dynamical laws with the invariant quantities. There's also a nice discussion of this in, you know, again Wallace's very rich recent paper where he goes into a nice formal account of this. Okay, so now I want to draw a parallel between this epiphenomenal trait of the non-invariant quantities in a theory with symmetries and qualia in an epiphenomenal dualist picture of the mind, right? Cause there's something that these two things have in common. Think of the dualist qualia. So think of some view on which, well, when you see a red object there's all the physical stuff that happens and then additionally via some psychophysical law that connects your brain with the immaterial or dualist properties that represent consciousness. You also end up experiencing red qualia, right? But those red qualia don't have any physical effect on the stuff that's going on in your brain kind of by assumption, right? The view is they're epiphenomenal, okay? So again, you have this kind of shadow world of the qualia, right? They are acted upon by the physical stuff. You know, you point yourself at the apple and red qualia will show up as a result of those physical things happening but nothing physical is going to happen as a result of the red qualia showing up, okay? So the physical properties all go about their business as if the dualist epiphenomenal qualia did not exist but the physical properties affect the dualist qualia, right? Because what you're looking at is going to be dynamically related to which qualia you are experiencing. In the same way, right? The quantities that fail to be invariant under a symmetry are dynamically isolated from the invariant quantities, right? The invariant quantities kind of go about their business. So the display number on the quote unquote absolute velocity meter, it goes about its business without actually noticing what absolute velocity my prized possession has. And there's no way to kind of create a nomologically reliable correlation between that number on the display and the absolute velocity of the object. So there's a kind of dynamical isolation, one-way dynamical isolation between the invariant quantities and the non-invariant quantities. Now the view, right? So here's my new-ish at least argument for symmetry-to-reality inference. Again, one thing you're gonna think is, well, Occam's razor, you can get rid of these extra epiphenomenal properties and that's true, right? You can get rid of them without empirical cost of the theory. So that's a good thing, getting rid of kind of excess stuff. The other further thing I want to say is that it's actively a vice to posit epiphenomenal properties that are dynamically isolated from the rest of what's going on in the universe in this way. And I think that the example of dualist quality kind of illustrates why this is, right? Now epiphenomenals about the mind, I'm not hardcore enough to say it's obviously false, right? But the reason why I'm not that hardcore and I'm not willing to go so far out on a limb as to say that there obviously aren't epiphenomenal qualities just because the problem of consciousness is so hard to solve that you can kind of get yourself inside the frame of mind as saying, well, there's no way to solve it except by saying that there are these extra quality of properties that live in their own shadow world, right? But clearly if you could do without those shadow properties that are acted upon by the physical but don't act upon the physical themselves, you would have a better theory of the mind, right? Other things being equal if you could do without those epiphenomenal properties. I think clearly a theory is better that is to say more plausible if it doesn't posit any epiphenomenal stuff. And this goes above and beyond the betterness of a theory being parsimonious, right? It's not just that a parsimonious theory is better. A theory that doesn't have these ineffectual epiphenomenal quantities is also better, right? An analogy that I kind of make here is think about other traits that we think of as like physically pathological, right? Like if you have a theory where there's kind of like, you know, totally unmediated action at a distance instead in the sense of like transmitting energy without a field as a mediator. If you think of a theory where it fails to have a well-posed initial value problem so that you can't predict either probabilistically or deterministically the future from the past, reality could be like that, right? Reality could have laws that meet those two criteria but yet we tend to say it's a pathology of a theory if it's laws have those features, right? What I take that to be an indication of is we're kind of as theorists glomming onto some reasons to think that it's not plausible to suppose that reality has those features. And I also think it's implausible to suppose that reality has epiphenomenal features. Maybe it does, but we should assume that it doesn't unless we need to suppose that it has epiphenomenal features. So it's kind of implausible to start with that reality includes a parallel shadow world of epiphenomenal. A thing to notice is that actually in the dialogue about the interpretation of quantum mechanics this issue has come up before actually. So here's an old objection to Bohmian mechanics by Adnor Shimoni that has been also opposed by David Albert a few times, right? Think of the particles or the core puzzles in Bohmian mechanics. Those dynamically depend on the wave function via the guidance equation, but the Bohmian particles don't do anything back to the wave function, right? You know, as Shimoni put it, there's action of the wave function on the particles without reaction of the particles on the wave function. I mean, I think it's telling about this example actually that the Bohmians, you know, Durer Goldstein, Zange and their co-authors have responded to this objection not by saying, look, we don't care about this. Instead, this is one of the reasons they cite as a motivation to interpret the wave function as a physical law rather than a concrete physical thing that that gets rid of this problem about the particles being in a sense epiphenomenal, right? So even the proponents of the Bohmian mechanics view realized that this is kind of a theoretical disadvantage or a prima facie pathology of their theory. Now, the reason why the Bohmians can pull this move, they can say, well, actually only the stuff that you think of as quote unquote epiphenomenal are real concrete physical quantities and the wave function is itself a law. The reason that they can do that is that in Bohmian mechanics, the thought goes the experimentally measurable data supervenes on the positions of the particles and in particular, like the particles are measurable, right? So that allows them to get rid of the, what we might say the non epiphenomenal quantities instead of the epiphenomenal ones. But in the present case with symmetries we have kind of the opposite situation only the invariant quantities, the ones that don't live in the shadow world, right? Only those ones are measurable and the ones that are epiphenomenal the quantities that fail to be invariant, the variant quantities you might say, those ones fail to be measurable. So I think we should kind of pull the opposite move that the Bohmians do in the case of the theory with symmetries and get rid of the epiphenomenal quantities to kind of sum up the argument like meme form, right? You can kind of break down the properties that a theory with symmetries is talking about into two classes. So there are the variant ones, right? And the invariant ones. So the invariant ones, some of them are measurable, right? Some of them are essential to the recording of past measurement results. And so it looks like you would be making a big mistake if you developed an interpretation of the theory that did without those quantities. On the other hand, there are the variant properties on the left side of this diagram. And well, none of those are measurable. That follows from the argument that Robert's made. They dynamically depend on the invariant properties but they don't dynamically influence the invariant properties themselves. So they are both surplus in the in the Achimist sense, but they also, I would say, have a pathology in the sense that they don't have any dynamical influence on the whole rest of the world, right? So those properties we ought to do without in our interpretation of the theory. Now, I said before that I was going to defend this smoothness assumption that we make when we do symmetry to reality inference. Without this assumption, you kind of can't get anything because the definition of a symmetry becomes so potentially anything goes, right? Cause you could have a symmetry that just interchanges two states, like let's say it just interchanges the vacuum state and the three particle state or something and nothing else if you didn't require the symmetry to be continuous in terms of the basic variables. And then you'd have way too many symmetries and symmetry reality inference would be obviously wrong. So Dasgupta though is very down on the idea that the smoothest requirement can be an assumption that we make at the start of the process of symmetry to reality inference. I'm gonna need to show that that's not true, right? Because I think we do need to make that assumption from the start of the process of symmetry to reality inference. Well, here is Dasgupta's objection to doing that. He says, look, suppose for example, that you have some classical mechanical theory and you require your symmetries to be continuous as transformations on state space or continuous in terms of the position variable, for example, right? He says, well, a definition like that is objectionably arbitrary cause you're picking some physical features, right? You're picking the geometry of the state space or you are picking the position variable and you're saying, well, it's gonna follow by the definition of symmetry that you can't run a symmetry reality inference to get rid of that quantity that you're treating as privileged, right? Because you're requiring the symmetries to be continuous in that quantity. So you kind of have to assume that it's real in your process of symmetry reality inference. I mean, what you're doing effectively is taking the differential structure of the theory and saying that is objective structure before I allow myself to engage in symmetry reality inference at all. And Dasgupta says, well, look, what's so special about the differential structure of a theory or anything else you could come up with? Why are they by definition immune to being rejected as unreal on the basis of the symmetry reality inference, right? And Dasgupta's idea goes, well, there's no principled way to pick out those quantities as the special ones that you shouldn't be able to get rid of when you apply symmetry reality inference. And indeed, right, he goes on to say in another part of this paper, well, we would have said prior to Einstein that things like absolute simultaneity were part of the privileged structure that symmetries need to respect, but of course that turned out to be false, right? And it turned out that we got rid of some of those features with symmetry reality inference. Well, I think that Dasgupta is being a little bit idealistic here, a little bit too idealistic. So by way of defending these smoothness requirements I'd first like to point out, right? Symmetry reality inference, as I said at the start of this talk has to begin with a partial interpretation of the theory that we are working on, right? I can't give you a theory, for example, and tell you, oh, I'm not telling you which variables in this theory represent time, right? And then expect you to be able to look at the theory and figure out what are the dynamical symmetries of the theory, right? That's not possible because like dynamical symmetry involves time as a kind of input to it, okay? So there's some amount of interpretation that needs to be done on the theory to begin with, right? And so one of the assumptions like that that we need to make in order to engage in symmetry reality inference is to assume that metrical properties of our basic variables have some physical significance. If we had dynamical symmetries that were discontinuous in those basic variables in the way that Dasgupta is entertaining, then the partial interpretation, the starting partial interpretation of the theory would treat states as arbitrarily, that are arbitrarily similar in these metrics as being qualitatively different, like totally different, okay? And what that would mean is our starting partial interpretation is all wrong, right? The theory is not in good shape to apply symmetry reality inference. We need to go back to the drawing board and develop a new starting partial interpretation before we can even begin symmetry to reality inference, right? So it's kind of like, if you're going to do symmetry to reality inference at all, you need to begin with the assumption that you have at least a minimally satisfactory partial interpretation, right? Among other things, it might need to identify the time variables that you can talk about dynamics. Another thing that we've seen that it needs to do is that it needs to have some definition of continuity for the basic variables so that you can apply a smoothness requirement. And it's too bad that we can't start with nothing and engage in symmetry to reality inference, but that's just not possible. And Dasgupta is trying to suppose that we can start with nothing, well, we can't. So the nature of this man's mistake, I would say, is that he's lost track of the fact that symmetry real inference only makes sense at all once we already have an empirically adequate partial interpretation of the theory, right? If we don't have that to begin with, then the process can't get started in the first place. And so, yeah, we need to make some assumptions in order to suppose that there's an empirically adequate partial interpretation of the theory. And there's no other way to go about this. So the complaint against a view like mine that I need to make some substantive assumptions about a starting partial interpretation, that's really what everybody would need to do if they want to engage in symmetry inference at all. All right, so that is what I have for you folks today. Thank you very much for listening. Thank you, and good to that. I have enough time to discuss your talk. Okay, so the two of us will be the respondent. Hey, can everyone hear me? Yes, so thank you for being here on time and for going to make a response besides the talk. So you have up to 10 minutes and then David Baker will try to reply. Hey, my pleasure. So thanks to David for a characteristically interesting thought-provoking set of ideas and also for the transform of Shavdell, for those of us who are geeks on the subject as those children. So I've got, I guess, four quick things broadly flow together. So the starting thought is coming off that does smoothness concern, which I take it links back to the initial shout out to Bellard. And my concern here is that absolutely, as you obviously recognize, we have to have some kind of constraint on what kind of transformations count as dynamical symmetries to which these arguments can be applied because otherwise they trivialize. But the worry is that, as you know, merely requiring kind of topological or differential topological notions like smoothness isn't anything like strong enough to deliver that to us. I mean, it's easy enough to construct a smooth but otherwise arbitrary bijections of the solution space and extend them smoothly but otherwise arbitrarily to everything else. It's easy enough to be in Hamiltonian dynamics and construct an arbitrary smooth canonical transformation of the data and then propagate it forward as a time-dependent symmetry. And even if we don't wanna play those kind of formal games, we have, you know, be a physically relevant examples like the lens range of symmetries of the two-body problem where manifestly on our normal interpretation is there is the symmetries are transforming a leaving variant quantities that we wanna take as physically significant. So in the lens range of your case, for instance, the shape of the orbit is not invariant under the symmetry transformation, but the shape of orbits are not physical than what it is. And I sort of, I mean, in some ways, this is prefacing things I'll say myself, but my diagnosis, what is worth of what's going on there is that there needs to be a more intimate connection here between which symmetries you wanna take seriously and the actual idea you're using to tell us what we should read from symmetries when we do take them seriously, which is we really wanna connect to the idea that symmetry transformations leave somehow have a dynamical separation, somehow there's a salient way which is unobservable. And that ought to be something we're getting at whether we ought to be able to get up formally. But whether or not that's the right diagnosis, I think there's a need in this kind of program to say something to trim down the dynamical symmetries that's these be much stronger than, I think I totally agree with your position on description in terms that seem totally reasonable, but as I say, and as you know, mere smoothness is just only the beginning of constraining that group. Okay, so that flows into the second issue I wanted to raise. I mean, the way you were describing the symmetry transformations is very much the universe or of closed systems, isolation, others. And I wanna hear the word subsystem here. I mean, partly I wanna hear it because it's a shout out to the various conference theme, but even independent of that, I wanna hear it on kind of methodological grounds. So if I take the Galilean symmetries we were talking earlier, I mean, we don't want to ask or I claim we shouldn't want to ask what would happen if we performed a Galilean symmetry transformation of the universe. Why shouldn't we ask them that? Well, A, because we can't do it as a practical matter. It's relatively straightforward with Optimus Prime in uniform motion relative to wealth as prime was previously. It's significantly more difficult to have the universe. But I mean, it's not just a practical matter if I can say just the universe probably doesn't have the Galilean symmetry group as part of its symmetries. It probably don't have any translation symmetry group as part of our symmetries. It probably has no global symmetry at all. So it sort of feels to me that if the game is we wanna understand, for instance, whether boost properties are a variant, we ought to be asking about boost properties or subsystems, boost Optimus Prime. And that's the framework I wanna be thinking in terms of. I mean, there's an alternative route that says if I'm gonna play with toy universes where these really are the exact symmetries that I just get nervous about what we could infer there. I sort of would like to keep our symmetry inferences concerned to the actual symmetries we see around us and the things we can get epistemic handles on. And then that sort of links to the third thought, which is that is it so clear that we correctly want to say the variant properties are not real? I mean, maybe this turns on this issue of quite what real in the most fundamental sense or something which I'm kind of a little nervous about. But in some more ordinary sense, it seems that the velocity of a moving object is measurable. It's something where if it was moving in different speed, the speedometer was showing something different as there's counterfactual dependence, for instance, on the properties of the absolute between the invariant, the velocity property of my car and the invariant properties of the speedometer. It's true that to calibrate that, I need to just pick an arbitrary choice of velocity, but it's still true that if my car had been moving 20 meters per second less quickly in as an absolute factor and all other facts were held fixed, then the speedometer would have shown a different kind of number. It's also true that a lot of our descriptive machinery in lots of ordinary places seems to be used varying quantities. So we seem to, I mean, Enrique Gomez brought up the example of the metric and the chat a minute together. I think that's a nice example. Feels as if we want to talk about metric features of certain subsystems. And again, I can see how if one wants to move to the level of the universe, one starts thinking, well, I would use all these things to relational properties, but then I worry a little bit again about how cosmological we're pushed to being. And I think part of this is perhaps that's a legit methodology, but maybe it shows that there's a substantial set of methodological questions as to what we're doing in your framework that it might be interesting to make explicit. It might be interesting to think whether we're thinking primarily of synergies of the universe. And then if so, what's the methodology there and what's the connection to sort of more down to work, bits of sort of physics data. And the last observation I want to make is in a slightly different direction, which is the epiphenomenal observable point. And I think it's something really important here. And I think a shout out to Roberts is right. I think Roberts points something important out. It's something I've thought about. I want to wonder though, whether it's really the case that we should think about the dialectic here as the epiphenomenal properties are disconnected from the observable properties. Sorry, let me rephrase that. The epiphenomenal argument is disconnected from the observable argument. Because I kind of want to suggest that what makes us want to say that epiphenomenal properties shouldn't be in our theory is precisely that they are in principle on observable properties. And we could have no reason to have them. And conversely, the only grounds on which we'd ever have reason to say something is in principle on observable is if it's provably epiphenomenal. I mean, what's the form of the argument that we actually can't observe symmetry, varying properties or we can't record symmetry, varying properties in symmetry and varying properties? And ultimately it's about the fact that I have a, as you say yourself, a closed system dynamics for the invariant properties. And precisely because I have a closed system dynamics for the invariant properties, it's never going to be the case that differences in the varying properties can sharpen the difference of the invariant properties. So the argument that some variables are epiphenomenal with respect to the rest of the variables just seems to be the same argument as the argument that those variables are observable. With respect to those other variables. And so then if I ask like, why don't we want epiphenomenal variables in theories? The kind of arguments one tends to come up with tend to be either, if you like, they're directly epistemically unsatisfactory because we know we can observe the relevant variable so it can't be epiphenomenal. Or we don't feel quite so confident there but we feel that the unobservability of something is a problem in its own terms. So if I think about the full philosophy of mine case, what's the argument to gain some mental properties being epiphenomenal? I mean, the standard kind of arguments from a philocyte functionalist like me is, well, don't mental properties seem to have causal properties? Doesn't your belief that I have mental states sometimes connect to the fact I'm telling you about it, isn't there a link between my believing something and my searching that I believe something? Isn't there some sense in which my being in pain is connected to the fact that the renewal signals passing through me? And if you can solve those problems, if you've got a, which I think is utterly hopeless, but if you can solve those problems and somehow make sense of the fact that I can, we can somehow have knowledge of mental states and have semantic content about mental states and so on, despite the fact that my nose is connected to mental states and physical, it's not obvious to me that there's a residual argument against their phenomenality. And then similarly with the symmetry case, if the epiphenomenality seems to establish unobservability, but it seems to be so misanonymous with unobservability, why ever would we want unobservable things in our theories? If one bites the bullet and totally unobservable things in theories, it's just not obvious to me there's a separate argument or root for epiphenomenal things to be doing what they want to do. And I think if you look even at the bone case, then the sort of physicist objections here tend to be very much along the lines of, well, the particles have nothing to do. They can't matter to anything observable. And then as you say, don't even say no, you're completely misunderstanding the theory, but certainly that seems to be the form of how you talk to physicists. These are just useless. They're not doing anything. You can't get at them, you can't measure them, you can't observe them. So I'd want to ask whether we should think about the epiphenomenality case as a new argument rather than a way of really sharpening and getting tight on what the old argument was. Okay, I think that's all I got. Thank you. Thank you, David. That's plenty. Yeah, lots of great stuff. So let me, I'm not gonna be able to get to all of it. I'll probably have to not talk about the first thing. So like, let me talk about the last thing first because that way I'll remember better. So like, yeah, I think it, I think they are separate concepts. I think they come apart in a number of examples, right? So like, I mean, I would say the Bohmian example is one, but I don't think that's the only one, right? So like, for example, you can have quantities that are like unobservable in a certain sense because they're like behind a horizon or something like that. And that's not a case of epiphenomenality. You know, you can have quantities that are like unobservable because it's impossible in principle for measuring devices to kind of like be fine grained enough to notice them or something. And I wouldn't, I think you could have examples like that. Like, you know, for example, like, you know, these kind of arguments that one sometimes sees made about like string theory, you know, speculatively obviously about like, oh, we can't probe distances smaller than like the string characteristic length or something like that, right? It's not obvious to me that that will turn out to be properties that are epiphenomenal in my sense, even though they would turn out in that theory to be in principle unmeasurable. So I think the concepts do come apart. But like, ultimately I suppose the question of their extension is actually a little bit, I mean, I don't think it's completely separate, but there's the further question of whether there's kind of like, even if the two properties are extensionally the same, is there a further reason from epiphenomenality over and above unobservability? In a sense, it's quite difficult to kind of like, argue about that question because it's a kind of bedrock question about norms of theory choice, I think in a way, right? It's like, so what I was trying to do with the Bohmian example and the Qualia example was like pump people's intuitions and say, hey, like maybe you do kind of start out endorsing, you the listener, start out endorsing a norm that says epiphenomenal properties are not good properties to posit other things being equal. If that intuition pumping is failing in your case, then like, it's very difficult to know what else to say. I haven't really trusted my intuitions pumping. Yeah, well, yeah, right. So like, I mean, I mean intuition pumping in quite a broad sense in the sense that I'm trying to, so like a thing that we sometimes do with examples in science is say, look, people seem to have made an inference for the following reasons. We think it was a good inference. Therefore you think inferences of this sort are justified, right? That's the sense in which I'm intuition pumping here. So, okay. So I think that's the kind of dialectic of that case. Okay, so subsystems, yes. I think this is the crux of a lot of where you and I see things differently where otherwise I think on this kind of issue, you and I are pretty close to each other in a lot of ways in our views. So like, so one of the parts of the book that I haven't fully written yet is a long discussion of your views about subsystems versus total systems. I think that is kind of central to a lot of the stuff that's going on here. I'll just say a little bit about my approach. So the way that I see what we're doing when we, and this connects with the stuff that you were talking about as well about the real universe probably in the precise sense lacking rigid symmetries, for example, right? So I think that what we're doing a lot of the time when, so as you point out yourself many times, what we're doing whenever we interpret physics these days as opposed to like in 100 or 500 years or something is we're interpreting a theory that's not perfectly fundamental and isn't yet finished in a certain sense, right? So the way that I tend to see that process is as kind of a process and steps where you say, well, what would things be like if the theory were exactly true? Now, what sort of approximation relation can we see as existing between the theory and reality? And on the basis of that, what conclusions do we draw about the reality we inhabit on the basis of the theory that we're interpreting? So the way that I would tend to see things as well, like I start out with a theory maybe that has a global symmetry. I conclude that were that theory exactly true, then reality would be completely invariant under that global symmetry. And I proceed then to draw some conclusions about what reality is approximately like in the sense that like, of course all of this is gonna be necessarily fuzzy because we don't have the deeper, fully deep underlying theory, but we come up with some justified conjectures about the sense in which it might be a good approximation to the truth to say that the non-invariant traits of the theory are not real. And I guess the form that that would take to me is so like at different scales or different degrees of approximation, we can talk about which quantities are approximately fundamental at that scale, right? So like in the domain of large N, right? There's a sense in which temperature is kind of close to fundamental or is as robustly real as any of the quantities that we're talking about in thermodynamics. So in that sense, an interpretation of thermodynamics has to be based on temperature as one of the basic quantities. Similarly, I would say an interpretation of Newtonian physics should be based on relative position as one of the basic quantities and not on absolute position as one of the basic quantities. And then we've got all kinds of questions to ask about like what's the connection between that and our world? I think there are a lot of parallels there, but that's kind of the way that I would tend to proceed by kind of like holding on to some of these idealizations while we are interpreting the approximately true theory and then at the stage of saying what's the approximation relation between the approximately true theory and our reality? That's when I bring in, yeah, that's when I tend to wanna say now you kind of put forward your caveats about the approximately true theory being wrong basically speaking. And I think that kind of fits pretty well with what the practice of science kind of looks like when it comes to these things. So about the fact that we use a lot of variant properties and relational properties versus invariant ones and the kind of thing that Gomez brings up in the comments. I mean, I think this is one of the tough things. And so keep in mind, similar to reality inference here, Dasgupta and I share this view. It's not supposed to be a 100% knockdown reason to interpret a theory as being invariant. It's supposed to be a strong protanto reason to interpret the invariant quantities as real. And that might be counteracted in some situations by the conclusion that you can't come up with a good interpretation of the theory on which the only real stuff is the invariant stuff. And then you might retreat to a view on which, oh, like I allow some covariant quantities in as well or something which are kind of as invariant as we can get. In the case of GR, I mean like, there's of course a long running debate about to what extent does the sophistication approach, which has kind of come back into the dialogue with Neil Dewar's work, to what extent does that approach allow us to view something like the metric as in a sense totally invariant under the diffeomorphism symmetry of the theory because if you say that the points don't have hexaities, there's a sense in which you can talk about the metric as being something that's left unchanged by the symmetries if you apply to it that metaphysical interpretation. So that would be the kind of direction that I would want to attempt to take things. And then if that failed, then I would retreat to the idea that, well, you know, symmetry reality inference just provides protanto reasons to interpret a theory this way. You might need an interpretation that involves variant quantities, but if so, you should do so only grudgingly, right? And to the minimum degree necessary in order to make the theory work. So I guess that's kind of the useful stuff that perhaps I have to say right now in response. Okay, if you're satisfied enough, then we will pass the questions by other persons, if there are any. Yeah, anyone else? Yeah, I mean, I would at this time that others should have a chance first. He looks like Jacob has a question. Yeah, but he is not, he choose that was a confidence. John, do you have some? Yeah, thanks. This was really, really cool. I have a question about one of the things that you said in the beginning about the really real thing. And so I was wondering, I mean, you know, some empiricists seem also interested in something like the symmetry reality inference, where I'm thinking of doing a process, I'm thinking of some of the things, I mean, the paper with this mail and some of the papers this mail has on our own also sort of suggests these things. And so obviously, or maybe obviously, it seems to me anyway that they can't be interested in the symmetry to reality inference because they're not realized about it anyway. And so do you think that it's just like misguided? Is this one of the things you think we need this notion of really real reality for? Or do you think there's just sort of two separate things going on that should be treated differently? Yeah, I mean, that paper is difficult to interpret in a number of ways. And I can send you my discussion of it that I've written up for more details on it. But I mean, one of the things that's kind of going on there, of course, is that Van Frossen has his own definition of what it is to interpret a theory and what makes an interpretation interesting, which is quite different, I would say, from the definition that David Wallace and I share in common. So what David and I are more concerned with is how do we take a theory and draw conclusions from it about this world that we inhabit? Van Frossen, when he's interpreting a theory, is asking how could that theory be possibly true? What would, so like, I, Van Frossen, Boss would say, I, Boss, do not believe that this theory is describing reality accurately. But if I did believe that, what would I be believing? That's the kind of thing that he's asking about. So in that sense, when he's interpreting a theory, Van Frossen, the empiricist, takes on a sort of hypothetical realism, which I think would be kind of compatible with the type of assumption that I'm making there. Van Frossen is a quirky figure in a lot of ways. And I think this is one of them and that makes it kind of difficult to keep an eye on where all the pieces of his system fit together. But when he's talking about interpreting a theory, which he doesn't always treat as a useful process, but when he does treat as a useful process, he's thinking of it as a hypothetical activity in which he, in a sense, takes on the guise of the realist for present purposes, right? Okay, yeah, thanks. And yeah, I would like to, I'll email you about it, but I'd like to- Great, great. Awesome job. Joan, honestly. Hi, yeah, thanks, Dave. I was just wondering, so you say that a theory is better to the extent that it doesn't posit epiphenomenal stuff. That's a kind of theoretical virtue. And I was just wondering why that isn't a version of the parsimony argument or parsimony considerations. You say that epiphenomena are surplus, plus they have this pathology, they don't have a dynamical influence on the rest of the world. But that just feels like a way of saying the way in which it would be less parsimonious to posit them. They're not needed to account for the rest of the world. And it felt to me different and kind from the other sorts of pathologies that you mentioned, which I didn't write down, I forget what they are, apologies. So I just wanted to hear more about why this is sort of a new kind of argument for the symmetry to reality and for instance, not sort of a version of parsimony considerations. Yeah, I mean, so one of the kind of other pathologies that I mentioned was kind of like discontinuous action at a distance, right? So I think that one is kind of instructively similar to the epiphenomena one actually, because they're both assumptions that we think are plausible about what kind of like, what ways it is reasonable to suppose that some parts of the world influence other parts of the world, right? And they're defeasible, right? Like they're not, you know, we don't, our brains don't break down, let me imagine a universe with action at a distance is not mediated by fields or only imagine a universe with epiphenomenal stuff. But like, you know, we just, it seems implausible, right? It seems like kind of a goofy conspiracy theory way for the universe to be set up. Is that just the same as parsimony? Yeah, I mean, this kind of gets to the, this kind of gets to the exchange at the outset of my response to David's comments, right? It's really a question about, yeah, if you imagine a, you know, if you imagine a quantity that you can do without, but that isn't epiphenomenal, does that seem, you know, like less of a good quantity to get rid of than one that is epiphenomenal, right? And I think it does to me, and I think that like, I mean, I think that like the, again, I think the Bohmian mechanics example is a nice illustration of this, right? Because there we have a theory that exhibits epiphenomenalism in my sense. And it's quite difficult to get rid of the epiphenomenalism because you can't just get rid of the quote unquote epiphenomenal quantities because they're the particle positions. And yet the Bohmians were like, okay, we got to do what we can to make it so that our universe doesn't have these two kind of quasi-isolated parallel worlds to it, right? Yeah, it's just kind of like, it's just like, yeah, I think it's easiest to gesture at it with metaphors. You know, it's like the idea that the universe contains parallel universes that are disconnected in some ways and connected in other ways. It's implausible, right? You know, ha, ha, ha, good one, David. Yeah, you know, so like, so that's a view and if the examples don't move you in that direction, then they don't, right? You know, it's kind of like, you know, I should say as a general principle, I think that like theoretical as opposed to empirical reasoning in physics, is something where I just feel like it's kind of, it's a bit like normative ethics. We're following norms. We have a really hard time writing down specifically, what are the norms that we're following in detail? And if we do try to write that down, we're gonna miss out on a bunch of truths about what those norms are. But, you know, like we can get a good rough sense of the norms that we're following. And in particular, our judgments about cases can lead us in the direction of understanding when those norms are being violated, right? So that's why I'm trying to motivate the principle with analogies to other cases. That's what I would do if I was writing a paper and applied ethics, right? Okay, so I would like to finish, I can wait a minute. So, Kristian, if you could be very brief and also take a very brief and then we finish. Okay, perfect. Many thanks for the talk, David. It was really very interesting. I mean, I was wondering if you think of the symmetry to reality inference as a sufficient condition for really real properties or only as a necessary condition? Because let me give you an example that you already mentioned, volume mechanics, again. More metaphysical orientations could say that, of course, there are many variant properties of the theory that seem to refer to really real properties, but they actually do not. Really real properties are only position. The rest is just part of the representation, right? There are ways in which we can express, for example, how relative distances vary from one time to time and so on and so forth. But they are not real properties in the same sense. So it seems that this is a case in which the inference is only at best necessary, but it's not sufficient. You need to say something else about why we are allowed to promote in high properties as really real properties. I mean, some women could run this story. So I'd just like to hear your thoughts about it. Yeah, I mean, my thoughts are that you nailed it. You're exactly right. It's a necessary condition, but not a sufficient condition. I mean, indeed, it's not a strictly necessary condition because as I was saying, it's a defeasible norm. There are gonna be exceptions. We can at least come up with in principle, it's a protanto reason to interpret theories this way. Okay, perfect. Thanks, yeah. Yeah, we have a very interesting discussion about we should move on with the program. So thank you a lot for the talk and thanks to the respondent and the other persons who are asking questions.