Friday, June 11, 2010

Cian Dorr on Every Thing Must Go

Cian Dorr just reviewed Ladyman and Ross (et al), *Every Thing Must Go* at:

Not surprisingly, he spends much time on chapter 1 (in which analytic metaphysics is attacked), especially on the 'Principle of Naturalistic Closure' (PNC). [It echoes one of the criticisms I made on this blog: Dorr then moves to challenging L&R, and offers the following gloss on the methodology of analytic metaphysics:

"To the extent that analytic metaphysicians have been willing to engage in debates about ontological priority, their substantive conclusions have been wildly divergent. If there is any consensus, it is merely that those who want to defend claims about ontological priority should articulate these claims in a certain kind of detail. It is not enough simply to announce that Xs are more fundamental than Ys: if I want to defend this claim, I am supposed, at a minimum, to (i) introduce a language in which I can talk about Xs without even seeming to talk about Ys; and (ii) make some kind of adequacy claim about this language, e.g., that it can express all the genuine facts that we can express using Y-talk, or that all the Y-facts supervene on the facts stateable in the language. For example, if I want to maintain that spacetime is less fundamental than the spatiotemporal relations between bodies, I must describe a language for characterizing these relations, and explain how it can adequately capture, e.g., claims about the global topological structure of spacetime.
The authors do not submit to this discipline in any sustained way."

So, far so good. (It is nice to see the linguistic turn defended so nicely.) Dorr then moves to offering (quite elegantly) four different possible ways of characterizing possible positions that might have been available to L&R. He then points out the moral of his analysis: "But one cannot simply announce that such disputes are to be dissolved: one must earn the right to do so by describing a fundamental language within which no corresponding questions can be formulated"

Dorr has moved from the non-trivial (but eminently defensible) methodological demand that metaphysical claims be made in a clear language to the more strict claim that one must formulate the properties of a fundamental language before one can rule out certain questions (and views). Now, the pluralist in me likes this incredibly high demand because it probably means very few questions can ever be ruled out of philosophy. But against Ladyman & Ross this move is begging the question because why would they have to accept that meta-philosophical [sic] questions need to be settled by appeal to a (hypothetical) fundamental language? It is especially unfair because in analytic metaphysics the deployment of fundamental language is generally used to articulate views that are dogmatically realist (without having to take the philosophy and history of science seriously at all [recall the debates on this blog: ) while L&R are, in part, motivated by taking pessimistic meta-induction argument seriously.

There is more to be said about both Dorr's criticism and his defense of analytic metaphysics. Maybe later.


  1. I found Dorr's review extremely helpful in a number of ways. But even after reading it, I still find L&R's views mystifying. They have a theory that relations are prior to individuals, and that science is about relations not about individuals.

    This theory runs into some rather obvious difficulties. One of these difficulties is that it seems to entail that science cannot distinguish between isomorphic structures -- even to deny that one exists. (For instance, consider a Twin Earth in which water was made out of two twin-hydrogen and on twin-oxygen atoms -- L&R's theory seems to imply that we can't say that this Twin Earth doesn't exist, and can't.)

    L&R respond to this kind of difficulty by saying: "logical variables and constants [are] mere placeholders which are used for the definition and description of the relevant relations even though it is the latter that bear all the ontological weight." I can't figure out what this means. It is a part of what logical variables mean that they can be bound by quantifiers and replaced by constants. How then can you say that the relation "bears all the ontological weight", when binding a variable makes an ontological commitment? L&R are using a language that is superficially familiar to me, but upon examination turns out not to be so.

    I took Dorr to be asking L&R to explain what they meant in, as you put it, "clear language". It's possible that L&R will respond to this reasonable request by accusing Dorr of a kind of methodological fundamentalism. In other words, they might say that Dorr's request is, at bottom, a demand for explanation in e.g., set theory. But, they protest, they have rejected set theory.

    Do you sympathize with this response? Are you, as a pluralist, inclined to allow L&R their claim to a set of concepts that cannot be explained in terms that you and I understand?

  2. This comment has been removed by the author.

  3. I also found Dorr's review useful, but primarily because he offered a novel and revealing (I think) defense of the utility and value of analytic metaphysics, but about that some other time.

    I think we want to distinguish between three issues here. First, must L&R explain their positions and concepts in pre-existing (clear) language? I think there the answer is a qualified, "no." They could demand that we enter their system and learn it, as it were, from inside out. (Think Kant, Hegel, Heidegger, etc. [They do toy with the idea of adopting the Hegelian Stance, 62].) But of course we would have to be convinced of the advantages of their new language (e.g., Dennettian real patterns, locators, and their own peculiar way of distinguishing between formal/material mode). I have to admit that I am not persuaded that they have given us a road map to how such a language (or family of languages) is supposed to work. This difficulty is partly due to their having a kind of dialectical way of arguing in which the authors of journal articles stand for known positions. This allows them to be very inclusive toward others, but (as in Continental philosophy) it demands a great deal of familiarity on the part of their readers. But it also means it is very unfair to quote any single, technical sounding sentence out of context. The sentence you cite coheres with their general suspicion of the Quine-ean approach to reading ontology of our regimented languages. (Interestingly, Lewisian metaphysicians reject Quine's ontology, but the "fundamental language" is supposed to offer us some economic way to make ontology and ideology transparent to us. So, there is a kind of mutual begging the question going on.)
    Incidentally, I know I am being unfair to L&R by comparing them to continental philosophers, but Dorr's response to L&R is our standard analytic response to Continental philosophy (when we're polite): there might be something there, but it's unclear to us.
    Second, I don't think L&R reject set theory tout court, but it has something of an instrumental role for them (e.g., 119, but I may be getting the direction between material and formal mode wrong here).
    Third, as far as your Twin-Earth example, I think the PNC rules it out of consideration. (This has problems, but they do offer an argument.)
    But maybe I have misunderstood L&R.

  4. Eric, as you are fully aware, making the L&R thesis radically non-decomposable makes it quite unappealing to most. Perhaps, there is a mutual begging the question at some level, but that is so whenever two people disagree. I mean there is some level at which, for instance, you can't refute Berkeley, but at most beg the question against him. The question is: does this give me a reason to say that there are only relations, no relata (or only derivatively relata)? Surely not.

    Not sure why PNC rules my T-E example out. My example relates physics to chemistry.

  5. Only having read the review I have to say that Dorr has a point because this sort of structuralism is hard, if not impossible, to make coherent. I will admit it has a certain attraction to me. There are an infinite number of different axioms you could use to set-up Euclidean geometry (most commonly few use Euclid's original parallel axiom and use instead Playfair's version of it), but I think few would demand on this basis that we believe that each axiomatization is a distinct geometry (all the same theorems being provable or being axioms in all these geometries). So the temptation to see the structure as all that matters is clear to me. However, the problem remains you still actually have to state axioms to be able to do stuff in Euclidean geometry. You just won't get a proof without them.

    For a physical example, special relativity requires there be no preferred reference frame, but you still need to use a perspective from some reference frame to describe anything. That all references must in some sense be on a par does not mean that there must be some reference-frame-less description that is the true or best one. Nor does it mean that there is a real state without a reference frame.

    So I think the point (or suppressed premise) is there is some substantive argument (although inductive I'm not sure I can deductively show it) against the position for a purely relational ontological language on the grounds of incoherence that requires counter-evidence of coherence.

    I would say that its not clear that we have a fully coherent robust rigorous etc. language for talking about things in general. Natural language probably lacks all that we want in terms of rigour and my sense is no formal language yet produced is capable of replacing natural language. However, in so far as we have working languages that refer to individuals we have an inductive case that such languages can be extended or modified to work, but it seems there is no such a case for a pure structuralist language shorn of the taint of individuals and in my opinion perhaps a case for the opposite claim that attempts to do it suggest it is impossible.

    Allan Olley

  6. I think L&R wish to dispense with the requirement that metaphysics must be first and foremost about a particular (and definite) language and its structure. They want it to be about the structures that sciences say there are. They take physics to be the fundamental science, and , so they may expect that the mathematical language that does justice to it is in some sense privileged. But I don't think they need to have a unified and final language for this. (They certainly think the world is unified, that is to say, there are "universal real patterns" about which measurements carry information.) This is why Dorr's review, which raises important questions, is incapable of doing justice to L&R's project of "rainforest realism," which is about patterns (articulated in terms of projectability and information encoding. see p. 233, especially for their definition.)

    Mohan, your example relates physics to chemistry. But it is not a counter-example to existing science and the questions it raises. L&R have what they call a "Piercean Verificationaism" which allows them to rule out objections that are generated from logically possible and consistent with contemporary sciences.

    [Again, I am not sure I get L&R right and I am the best spokesperson for their view.]

  7. Eric,

    You say L&R deny that metaphysics is about a particular language. I agree: for me, it's about individuals, properties, and relations. Allan says that as far as physics and geometry are concerned, structure is "all that matters". I don't agree, but I understand the thought he is expressing.

    I don't, however, understand the thought contained in a sentence from L&R that Dorr quotes: "logical variables and constants [are] mere placeholders which are used for the definition and description of the relevant relations even though it is the latter that bear all the ontological weight."

    What exactly does it mean to say that "logical variables and constants" carry none of the "ontological weight"? This seems to deny that there are (i.e. that there exist) individuals that stand in the structural relations described by physics. And that, I think, is incoherent. Unless it's meant to express some kind of idealism, in which case it's coherent -- but metaphysical, and (in my opinion) false.

    Here's the problem. The only way to understand what relation 'is more massive than' expresses is to understand what has to be true when a pair of individuals satisfies this relation. It's true, as Allan says, that physics may not tell us WHICH individuals satisfy the relation. But logic does tell us that the idea of the relation presupposes that there be individuals that could satisfy it.

    I am not making a terribly original point, I take it -- just belabouring what lots of others have said.

    One last point. Allan, special relativity may require that there not be a privileged spatial frame of reference. But it does not (so far as I am aware) require that there be no privileged domain of individuals.

  8. Mohan, the sentence you quote is from a section in which L&R confront the objection to their view that relations are impossible without relata. In that context they concede that without a fully developed language ("framework") for their own favored metaphysics (one that does not privilege individuals over relations) they will treat other notations heuristically (in context they appeal to work by French on quantum mechanics and a historical case study by the terrific Locke and Newton scholar) Domski involving Poincare). So, all they are saying is that they refuse to take their metaphysical marching orders from first order logic.

  9. To be clear I was using special relativity (and geometry) analogically and heuristically, as examples where certain individual things (but not all individual things) are shown to be arbitrary and so in some sense not real (not real as absolutes anyway). So a real preferred rest frame seems to me to be a logical individual that is more real than the relative rest frames of special relativity, but even so rest frames in special relativity are not really so unreal that they are dispensable. I suspect the sort of argument L&R are motivated by has the same sort of catch.

    As to their broader claim it seems like they may just be making a stronger claim than they need to. Imagine physics comes up with two contenders for a final theory, one about particles interacting and one about fields interacting. Imagine further theorems are showing that the particles and all their properties show up as secondary properties in the field theory and likewise the field and its properties are derived as secondary properties in the particle approach. There is clearly a motivation to deny that one is more correct or true than the other. L&R seem to want to say both are true and indeed one and the same. I would like to say they are two perspectives on the same theory, which seems equivalent in some sense, I'm not sure that completely dissolves the difference between them or requires I deny individuals/relata as fundamental.

  10. I think folks are getting misleading impression from Dorr's review and this discussion. L&R are not anti-realists.

  11. I actually was not intending to suggest they were (if I had I would have said they would just take two isomorphic theories as both false). Having skimmed a bit of Ch.3 and 4 on Google books there are some realist-antirealist tensions in their arguments (since they criticize other realists in Ch. 3 for example).

    Realists can still think that some things don't exist. They can believe that some things that seem to exist don't exist, even if people act like they do. In particular L&R seem to deny the existence of relata (or at least relata as fundamentals) even though they admit it can seem as though relata exists and even though some (all?) of us act as if they exist. As Mohan has been suggesting denying relata existence does seem necessarily to lead you to having to conclude that isomorphic theories are identical because how would two views/theories with different relata but the same relations (an isomorphic structure) differ under those rules? I think the case I tried to sketch would be a case of such a failure to differ in relations. Some rejection of a singular description of reality seems to be necessitated by their positing of rainforests over Quinean deserts, there is clearly some overlap between their real patterns (ie the same pattern is being described as being part of two or more different patterns), since one real pattern is made up of other real patterns and so on.

    Allan Olley

  12. Sorry, Allan and Mohan, for being a bit slow on the uptake. I had assumed that folk were familiar with structural realism (which is very popular in parts of the UK -- Leeds, Bristol, LSE). Recall that L&R are responding to the pessimistic meta-induction argument. In response to it, L&R want to claim that science discerns the structures of the world. (Moreover, they do so by building on and then contrasting their own view to Van Fraassen, who on their account accepts under-determination.) Within that context they privilege relations over individuals.
    Now, this means unless science has epistemic access to differences between otherwise isomorphic structures, L&R are happy to be agnostic (I think). You are right that for them it is relations all the way down (although they treat this is a tentative empirical hypothesis.)

    I think one can (must) resist the identification of realism with a singular (final) description of reality. (A scientific realist can accept the deliverance of science without expecting that science will converge on a single description.)

  13. This comment has been removed by the author.

  14. Eric, you should always take it as given that I am unfamiliar with everything!

    But seriously, all that I said that it was incoherent "to deny that there are (i.e. that there exist) individuals that stand in the structural relations described by physics," UNLESS that is meant to express an idealist position. Yes, since L&R are realists, I can abbreviate that to: it is incoherent.

    On an earlier reply, it seems like a prevarication to say that one is not going to take one's marching orders from first-order logic. (No need for "first-order", by the way: the understanding of "relations" I proposed is not restricted to first-order logic.)

    A lot of the arguments you report seem to me to be refusals to engage with the opponent.

    1. The opponent says "Why can't I say that this Earth is not Twin Earth?" -- which would be metaphysics in L&R's book. L&R respond with "Peircean verificationism". But verification is surely one of the things that Twin Earth examples refute -- if Twin Earth is coherent, verificationism is false.

    2. The opponent asks "What do you mean by a relation if not something satisfied by n-tuples of individuals?" L&R reply: we'll tell you some other time.

    I do appreciate that knowledge-of-structures is a good response to the pessimistic induction. And a clever one, since structuralism does identify something on which say Newton and Einstein agree. But the position that I have (deliberately) called structuralism doesn't demand that you hold that it's "relations all the way down", does it?

  15. I don't profess to be any kind of expert on Ladyman & Ross's views, but I think the relationship between "ontic structural realism" and realism as we know it is somewhat strained. Indeed, in their book L&R dedicate a whole section to distinguishing their views from those of Kant and the logical positivists!

    It might be more accurate to see them as starting from something like constructive empiricism and moving slightly in the direction of realism by positing a "structure" to reality that transcends experience.

    The empiricist says: "There are phenomena. Science finds regularities in the phenomena. But what else is there? We don't know."

    The ontic structural realist says: "There are phenomena. But what else is there? There is a grid of modal facts that in some sense 'govern' the phenomena. This 'modal structure' is what is revealed by fundamental physics."

    I'm not saying this amounts to a coherent position, but it's the only sense I can make of OSR.

  16. It occurs to me that in my description I may have conflated existence and knowledge claims. Realism and anti-realism are usually more debates about what we have reason to say/believe exists than about what exists. Like van Fraasen's claim about unobservables, its not that they don't exist just that science does not warrant a belief in them.

    So, one reading of L&R's position is that they deny we have knowledge of the relata and only of relations and then they conclude from this that relata become less _real_ . Depending on what is at stake in _real_, it seems like the criticism of verificationism come in. Just because something is unknowable (even in principle) is not the same as it not existing or being a mere secondary property or the like. Then again perhaps they do not mean to take such a position and the denial of relata rhetorical excess meant to be a simple denial of specific knowledge.

    Still just because we know nothing about an object does not mean we can not talk about it (as this conversation shows?) and indeed maybe we must talk about it. This seems to me an implication of Mohan's position and Dorr's position in defense of traditional metaphysics.

    Mohan is it fair to respond to the Twin Earth question by saying, well we can't tell whether we are on Twin Earth or not but we either are or not? Likewise with Earth, and not both Earth and Twin Earth, etc... Might L&J just take that root?

  17. I so much want to jump in here but I'm really up against a couple of deadlines. Let me just say two things:
    first, I just don't understand why it is incoherent "to deny that there are (i.e. that there exist) individuals that stand in the structural relations described by physics,". I could, for example, argue that quantum objects are non-individuals in the sense that Decio Krause and I have tried to explicate and insist that non-individual objects can stand in relations. Or I can argue, as Ladyman and Saunders have, that such objects are 'thin' in the sense of having a kind of contextual individuality, which in the quantum context is provided by a form of relational 'weak discernibility' (for fermions at least). Or I could argue, as that madman Steven French does, that there are no objects, thin or otherwise, and that we need to understand physical structures in terms of relations without objects as relata (either by taking the relata to be themselves relations, which seems both formally and metaphysically admissable; or by following Mertz and introducing a different understanding of relation that doesn't beg the question at issue, or ... you get the picture!).
    Secondly, I know its the book thats being reviewed but if you want to understand ontic structural realism and James'form in particular, you should read some of the more recent work. (And just to give a plug there's a collection on structuralism edited by the Bokulich's coming out soon with Springer and another edited by Landry and Rickles coming out in the Western Ontario series - order now while stocks last!).
    And just to go all Spanish Inquisition and add a third thing: not all structural realists subscribe to everything Ladyman and Ross say in that first chapter. Nevertheless, the defensiveness of the likes of Dorr in the face of such polemic does give one pause for thought!

  18. Hi Allan,

    Verificationists don't generally like the Law of the Excluded Middle. They think that P is meaningful if and only if either it or its negation is verifiable (in the long run). Since they think that e.g. The Absolute is weightless is meaningless, they would deny that either it is or it is not. Similarly, I suppose, for Twin Earth.

    My thought was just that Physics seems to tell me that I am not living on Twin Earth.

    Jonathan, I know this isn't a position you would defend, but I am not sure what you mean by "a grid of modal facts that govern the phenomena". What form would the modal facts take? -- Just general facts? What sort of phenomena would they govern? Any singular phenomena?

  19. Ladyman & Ross provide their own `theory of ontology' in Section 4.4. Rainforest Realism. L&R list four concepts needed in their theory: information (Oh Lord!), compressibility, projectibility, perspective. Much more turns out to be needed: locator, computer, physical possibility, pattern, structure, model, classes, memberhip, observation-point, dimension, computation, existence, the gerry-meandered variable `S', and arguably more.
    And whatever results will clarify Ontic Structural Realism? Perhaps. The temperature and humidity of a rainforest makes it difficult to breath, apart from swarms of anoying stick insects. If the theory of ontology of L&R were
    to be wholy or partly formalised, the language needed to do so would be a preferred ontological language. This is why I don't see, contra Eric Schliesser, how L&R are NOT committed to such an `unique and ideal' language, as Cian Dorr at some point suggests in his Notre Dame Review.

    F.A. Muller

  20. Ha, Fred,
    We can just have our reading group on L&R on-line!

    1. I agree that L&R introduce a whole network of concepts that really amount to their own 'language'. [Recall that in response to Mohan I wrote above: "must L&R explain their positions and concepts in pre-existing (clear) language? I think there the answer is a qualified, "no." They could demand that we enter their system and learn it, as it were, from inside out... But of course we would have to be convinced of the advantages of their new language... I have to admit that I am not persuaded that they have given us a road map to how such a language (or family of languages) is supposed to work."

    2. But I see them as resisting the demand of formalization into what you call "some unique and ideal" *preferred* "ontological language." Why would they have to accept this? (Do they accept this demand and have I missed it?) WHy can't there be ontological languages as befits the needs of different sciences?

    3. In particular, even if there is a preferred ontological language, why should metaphysics be about it rather than about the underlying sciences? This was my main motive in blogging about Dorr's review. (I am really agnostic about their relationism.) There is a lot wrong with L&R's book and their arguments against contemporary metaphysics. But much of Lewisian contemporary metaphysics is committed to analyzing some future (and in my position illusory) ideal language; L&R quite rightly insists that it is better to have metaphysics be guided by the sciences. This doesn't rule out formalization and regimented languages, but it should not be the main point of the project.

  21. Re Relations without relata - this criticism has been doing the rounds for a long time now and I survey responses to it in the literature in the Stanford Encyclopedia of Philosophy article on structural realism. I confess to being rather mystified myself as to why people for whom it is so obvious that OSR is incoherent waste their time going on about it. Life is short after all. While I don't want to pretend that everything we said about individuals was as perspicuous is would be desirable, I must point out that our structural realism is motivated by consideration of the putative individuals in Quantum Mechanics and General Relativity. In the context of both theories there are good reasons to deny intrinsic individuation. Whether one denies there are individuals I now see as a partly terminological dispute since one can adopt a logically very thin notion of individual that is compatible with structuralism (see the references in the Stanford article to my paper with Hannes Leitgeb and to my article in the Proceedings of the Aristotelian Society from 2007). I now put things differently and I am still undecided about some matters but I stand by one of the basic claims of ETMG which is that contra Lowe and others the world is not made of 'self-individuating' elements. There is no reason in philosophical logic or metaphysics why the relata of relations cannot themselves be relational structures indeed all the relata we have identified so far in physics have turned out to be so. As Poincare says 'the atom is a world'.

    Re languages - Eric is quite right that we do not take our marching orders from first order logic nor do we recognise the requirement that metaphysical theorising turns on establishing a privileged vocabulary. We engage in speculative metaphysics in the cause of articulating the unity of science. We are crazy enough to hope that our theory might help science progress. If it doesn't then by our lights it is not worth very much. To those who complain about the technical vocabulary we introduced I can only say that we did our best. If you think the basic framework we advocate might be right then why not have a go at articulating a better version of it.

    I have lots more to say about Cian Dorr's review but I confine myself here to commenting on the themes discussed in this thread.