Naturalism Without Mechanism

As there is occasional interest in the relation between science and metaphysics here, I thought I’d share this article: “Metaphysics of Metamorphosis“, by the philosopher of science John Dupre. Dupre argues that metaphysics that takes science seriously — what he calls “naturalistic metaphysics” — will give us a very different picture of reality than what we get from traditional a priori metaphysics:

This project of science-based metaphysics, sometimes referred to as ‘naturalistic metaphysics’, has been surprisingly controversial. The philosophers James Ladyman at the University of Bristol and Don Ross at the University of Cape Town offered a forceful defence in their book Every Thing Must Go (2007). As that book illustrates, the debate can be technical and vitriolic. Consequently, I won’t defend naturalistic metaphysics from its critics so much as show you how it helps us inch towards an answer to one of the oldest chestnuts in the history of philosophy: is reality made up of things that somehow change over time, or are things just temporary shapes that our perception plucks out from a flux of unruly, unfolding processes?

Dupre argues that adopting scientific metaphysics as our method of doing metaphysics will yield a very different account of what is real: that what is fundamentally real are not things but processes. (In history-of-philosophy terms, one can imagine this as the triumph of Heraclitus over Parmenides and all of the post-Parmenideans: Democritus, Plato, and Aristotle.)  But giving up on “things” and “thingishness” as the ultimate constituents of reality means re-thinking what we are talking about when we talk about mechanisms:

‘Thingness’ has a very real impact on scientific work by motivating the search for mechanisms. A mechanism is a precisely arranged set of stable things whose interactions generate a phenomenon of interest. Scientists often see uncovering mechanisms as the gold standard of scientific insight. This approach certainly has its benefits: not everything can be examined at the same time, and science depends on careful attention to well-defined parts of the whole.

The upshot of process metaphysics isn’t that we should refrain from talking about mechanisms, but that we need to be very careful in specifying the kinds of contexts in which mechanistic explanations make sense. In other words, naturalistic metaphysics turns out to be antithetical to mechanistic metaphysics. This may seem odd, but as Dupre sees it, this is precisely what happens when we make our naturalistic metaphysics responsive to biology and not just to (classical) physics:

There’s room for both epistemologists and metaphysicians under the tree of knowledge. But when it comes to the living world, both kinds of philosopher could do with making room in the shade for the humble scientist. Learning from this third colleague, we see that biology is not fashioned from bits of Lego, carefully sculpted to play a precise part in exactly one bit of machinery. Rather, living things are processes that are capable of assuming many protean forms: dynamic, ever-changing, but balancing, for a time, on just the right side of chaos.

I’ll add that Dupre doesn’t talk about physics, since his specialty is philosophy of biology. But from what little I know of philosophy of physics, it’s processes all the way “down” — and, in cognitive science and philosophy of mind, it’s also processes all the way “up”.

 

 

 

 

 

 

 

 

59 thoughts on “Naturalism Without Mechanism

  1. Mung,

    But Dupre envisions the relation between science and metaphysics as dialectical, with neither foundational to the other. So it’s not an objection. What a foundationalist may see a ‘bug’, the anti-foundationalist sees as a ‘feature’.

    Sellars:

    If I reject the framework of traditional empiricism, it is not because I want to say that empirical knowledge has no foundation. For to put it this way is to suggest that it is really “empirical knowledge so-called,” and to put it in a box with rumors and hoaxes. There is clearly some point to the picture of human knowledge as resting on a level of propositions — observation reports — which do not rest on other propositions in the same way as other propositions rest on them. On the other hand, I do wish to insist that the metaphor of “foundation” is misleading in that it keeps us from seeing that if there is a logical dimension in which other empirical propositions rest on observation reports, there is another logical dimension in which the latter rest on the former.

    Above all, the picture is misleading because of its static character. One seems forced to choose between the picture of an elephant which rests on a tortoise (What supports the tortoise?) and the picture of a great Hegelian serpent of knowledge with its tail in its mouth (Where does it begin?). Neither will do. For empirical knowledge, like its sophisticated extension, science, is rational, not because it has a foundation but because it is a self-correcting enterprise which can put any claim in jeopardy, though not all at once.

    To this, Dupre may be read as saying (correctly, in my view) that metaphysics and epistemology are a dimension of the self-correcting development of human understanding, along with science, neither being “foundational” to the other but strictly complementary.

  2. “Naturalistic metaphysics” just sounds like a way of saying physics that we don’t understand. Or physics without physics.

    Also suspiciously like teleology without teleology.

  3. phoodoo:
    “Naturalistic metaphysics” just sounds like a way of saying physics that we don’t understand.Or physics without physics.

    Also suspiciously like teleology without teleology.

    Is this a response to anything that John Dupre wrote or is free association about what that phrase sounds like to you?

  4. There’s room for both epistemologists and metaphysicians under the tree of knowledge. But when it comes to the living world, both kinds of philosopher could do with making room in the shade for the humble scientist. Learning from this third colleague, we see that….

    I think this claim ought to be extended to the “non-living world” as well. Philosophy always has to yield unto science. I just want people to understand that that doesn’t make philosophy science. In fact, it’s not really right to call it “naturalistic” I don’t think–even though it’s an enterprise that takes place in nature. It might be the case that determinations regarding the “winner” of the process v. thing war show that science can “cut away” at philosophy’s edges, but that kind of question has to be put a certain way to make it philosophical in the first place, I think. Do people have to “cut up” the world into thingishness in order to comprehend it, regardless of how it’s “actually” made up? That seems more like philosophy/psychology than physics to me.

  5. …is reality made up of things that somehow change over time, or are things just temporary shapes that our perception plucks out from a flux of unruly, unfolding processes?

    Surely there’s a good reason why metaphysics and ontology are (somewhat) distinct areas of philosophical inquiry. Metaphysics is more preoccupied with establishing the categories such as existence vs non-existence and reality vs non-reality. Ontology concerns questions about things (objects, processes, etc.) that make up reality.

    Dupre thinks he is saying something revolutionary about metaphysics, but actually he is only doodling with ontology.

    The question about mechanism (how mechanical is reality? and, given that, how naturalistic?) is somewhat interesting, but it’s a well-known theme on the same old question: What is ontologically fundamental and why?

  6. Erik,

    As far as I understand it, ontology is part of metaphysics.

    I think that Dupre is needlessly verbose and his message, is there’s any, is completely lost in the convoluted diatribe he managed to write.

  7. From the article KN links to.

    A good place to begin is with the question of essentialism.

    Dawkins has railed against essentialism, here, for example. (Sorry to bring him into the discussion, KN!)

    In the living world, at least, a metaphysics of ‘things’ is hard to sustain. Where once we had discrete and distinct ‘proteins’ and ‘organisms’, all we are left with are highly dynamic processes.

    This is a good point.

    ETA that I forgot to add that not just living organisms are better regarded as processes rather than things. Stars, for instance.

  8. Alan Fox: I forgot to add that not just living organisms are better regarded as processes rather than things. Stars, for instance.

    Oh Alan. So Stars are not things but a process? What process is a star? Does that process ever come to an end, and at the end of that process does the star cease to be a star? What does it then become?

  9. This project of science-based metaphysics, sometimes referred to as ‘naturalistic metaphysics’, has been surprisingly controversial.

    I wonder why. Maybe they should call it ‘stultified metaphysics’ or ‘mechanistic metaphysics’.

    Have we forgotten that ‘science’ is a human endeavor?

    …science aims to discover truths about the world…

    🙂

    As long as those truths are limited to mechanics.

    An ancient philosophical tradition dating back to Plato and Aristotle sought to discover essences, the defining properties of things…

    I think he does a disservice by starting with essences. An even more fundamental question is how things can both change and not change, or even how to understand change in a way that makes sense.

    Instead of searching for things with fixed essences based on form and function, naturalistic metaphysics suggests that we need to move to a picture that’s much more dynamic – in which any ‘thingness’ is strictly temporary.

    I thought that modern science had rejected essences long ago. Is he saying that they were rejected in name only but not in actual scientific practice? Who knew.

  10. Mung: Oh Alan.

    What?

    So Stars are not things but a process?

    Not quite what I said. Stars are evolving. What we see in the sky at any moment are just snapshots.

    What process is a star?

    A giant nuclear fusion process

    Does that process ever come to an end, and at the end of that process does the star cease to be a star?

    If the star is big enough, a supernova

    What does it then become?

    Depends on its size.

  11. walto: It might be the case that determinations regarding the “winner” of the process v. thing war show that science can “cut away” at philosophy’s edges, but that kind of question has to be put a certain way to make it philosophical in the first place, I think.

    That seems quite clearly right to me.

    <blockquote?Do people have to “cut up” the world into thingishness in order to comprehend it, regardless of how it’s “actually” made up? That seems more like philosophy/psychology than physics to me

    That’s a question in the metaphysics of perception, perhaps?

    Then again, I tend to think that Gibson was basically right: what an animal directly perceives are affordances. Paul Coates has some good objections to this view, but I think I have an answer to them.

    In any event, there are lots of process ontologies around the world that are just as ‘grounded’ in the manifest image as thing-ontology is in Aristotle’s manifest image. (Consider Aztec metaphysics and some schools of Buddhist metaphysics.) So a thing-ontology is optional and contingent even with regard to the manifest image, let alone whatever the scientific image ends up telling us.

  12. Kantian Naturalist: That seems quite clearly right to me.

    <blockquote?Do people have to “cut up” the world into thingishness in order to comprehend it, regardless of how it’s “actually” made up? That seems more like philosophy/psychology than physics to me

    That’s a question in the metaphysics of perception, perhaps?

    Then again, I tend to think that Gibson was basically right: what an animal directly perceives are affordances. Paul Coates has some good objections to this view, but I think I have an answer to them.

    In any event, there are lots of process ontologies around the world that are just as ‘grounded’ in the manifest image as thing-ontology is in Aristotle’s manifest image. (Consider Aztec metaphysics and some schools of Buddhist metaphysics.) So a thing-ontology is optional and contingent even with regard to the manifest image, let alone whatever the scientific image ends up telling us.

    Right. And I’m with you about Coates. Wouldn’t be too worried about anything he says.

  13. Dawkins has railed against essentialism,

    Alan Fox: Dawkins has railed against essentialism,

    Ernst Mayr wrote books railing against essentialism. Before Dawkins, I believe.

    Two that I know are What is Evolution and What is Biology.

  14. petrushka:
    Dawkins has railed against essentialism,

    Ernst Mayr wrote books railing against essentialism. Before Dawkins, I believe.

    Two that I know are What is Evolution and What is Biology.

    This is Biology in 1997 and What Evolution is in 2001. I did find a video of Mayr blaming essentialist thinking for retarding progress in science.

  15. Curious about “affordances” and how useful they’ve proved as a concept in, say, biology. What is essentially (heh!) different between an affordance and an aspect of an adaptive landscape or niche environment?

    PS and why should we discount anything Paul Coates says?

  16. Alan Fox:
    Curious about “affordances” and how useful they’ve proved as a concept in, say, biology. What is essentially (heh!) different between an affordance and an aspect of an adaptive landscape or niche environment?

    There is no ‘essential’ difference between an affordance and an aspect of the niche. Gibson pretty much says that an organism’s niche or environment (he borrows the term Umwelt from Jakob von Uexküll) is the set of all affordances. The conceptual difference is that affordances are what an organism directly perceives. Affordances are a bridge-concept between ecology and the psychology of perception — hence the coinage ‘ecological psychology’.

    PS and why should we discount anything Paul Coates says?

    I can’t speak to what walto had in mind. I myself respect Coates’ work immensely. I found his book on the metaphysics of perception to be very helpful in making clear the arguments in support of critical realism (i.e. an organism is immediately aware of its own sensations and conceptualizes those sensations in terms of the objects that typically cause sensations of that kind).

    I don’t entirely agree with critical realism, because I think it rests on a conflation between what’s happening at the level of the whole organism and what’s happening at the level of one of its parts (e.g. the retina-thalamus-cortex circuitry). But my own understanding of direct realism is far improved by having considered Coates’ criticisms of it.

    In particular, Coates argues that ecological psychologists like Gibson, Alva Noë, and Anthony Chemero have been too eager to talk about direct perception in terms of affordances, which are ways of possible movement, and haven’t had enough to say about the role of sensations or what is actually perceptually present to the organism. I think that’s a fair criticism. I’ve been trying to think about how to incorporate sensations and affordances into a more comprehensive model, but without much success (so far).

  17. Kantian Naturalist: In particular, Coates argues that ecological psychologists like Gibson, Alva Noë, and Anthony Chemero have been too eager to talk about direct perception in terms of affordances, which are ways of possible movement, and haven’t had enough to say about the role of sensations or what is actually perceptually present to the organism. I think that’s a fair criticism.

    I’m pretty sure that Gibson’s view was that pereception is prior to sensation. Gibson saw perception as providing (or picking up) information.

    My take on that, is that sensation doesn’t actually present anything. Rather, sensation is the experience of perceiving (of getting the information).

    I don’t entirely agree with Gibson’s view of information. He thought that information was out there to be picked up. I think that’s nonsense. As I see it, what is out there is evidence, not information. And the perceptual system is constructing information, and using that evidence in the construction of information. But there is far more to it than picking up what is already out there.

  18. Neil Rickert: I’m pretty sure that Gibson’s view was that perception is prior to sensation.Gibson saw perception as providing (or picking up) information.

    Yes, that’s my reading of Gibson as well.

    My take on that, is that sensation doesn’t actually present anything. Rather, sensation is the experience of perceiving (of getting the information).

    I didn’t mean “perceptually present” as “something else that the sensations present to the system” but “the sensations themselves as what the system is actually aware of”.

    In philosophy and psychology of perception, ‘perceptual presence’ contrasts with ‘perceptual absence’. One can be perceptually aware of something without directly seeing it — for example, the unseen hidden surfaces of a three-dimensional object. You see (are perceptually aware of) the whole object, even though what you see of it is the surface that faces you.

    It seems plausible to me that perceptual awareness of absence involves affordances (what one would see if one were to rotate the object or move one’s body) and perceptual awareness of presence involves sensations + affordances.

    I don’t entirely agree with Gibson’s view of information. He thought that information was out there to be picked up. I think that’s nonsense.

    I’m inclined to agree. It does make the metaphysics rather cumbersome, to put it mildly. But to my knowledge Gibson never really worked out what he thought ‘information’ was.

    As I see it, what is out there is evidence, not information. And the perceptual system is constructing information, and using that evidence in the construction of information.But there is far more to it than picking up what is already out there.

    I’m with you on the constructivism. One of the main reasons I’m interested in the predictive processing theory of cognition is that it offers a nice resolution of the debate between constructivists like Kant, Helmholtz, and Marr and direct realists like McDowell and Gibson. The basic idea is that brains are generating endogeous models of the environment that function as map-like representations of the affordances, and those models then convey to lower-level neuronal assemblies what sensations to expect. If the activation of sensory receptors (i.e. sensations) generates information that is too much at odds with the predictions, then the resulting prediction errors are used to revise the models that generate the predictions. In this way the endogenous models (themselves constructed by interacting neuronal assemblies) are corrected when and as necessary for the organism to maintain a grip on the affordances that comprise its niche.

    I find this to be a highly compelling way of thinking about what brains do and the role of brains in maintaining mostly successful behavior.

    The really hard question is how to understand objectivity, truth, and related epistemic concepts in terms of predictive processing as a theory of animal cognition. That’s my current research project.

    The key idea that I’m pursuing, based on Nietzsche, Sellars, and Davidson, is that objectivity requires coordination between differently embodied-and-embedded cognitive systems. If two or more systems are to successfully coordinate their actions, then each needs to minimize the discrepancies between its representations of the shared situation and those of the animal with which it is coordinating. There’s a lot of detail to work out, but that’s basically the beginning of the appearance/reality distinction, awareness of the possibility of error and hence of truth, etc.

  19. Kantian Naturalist: There is no ‘essential’ difference between an affordance and an aspect of the niche. Gibson pretty much says that an organism’s niche or environment (he borrows the term Umwelt from Jakob von Uexküll) is the set of all affordances. The conceptual difference is that affordances are what an organism directly perceives. Affordances are a bridge-concept between ecology and the psychology of perception — hence the coinage ‘ecological psychology’.

    That makes sense! 🙂

  20. Kantian Naturalist: I myself respect Coates’ work immensely. I found his book on the metaphysics of perception to be very helpful in making clear the arguments in support of critical realism (i.e. an organism is immediately aware of its own sensations and conceptualizes those sensations in terms of the objects that typically cause sensations of that kind).

    I am not at all familiar with Paul Coates. I was just intrigued by walto’s dismissal, especially when a quick Google indicated Coates is a fan of Wilfred Sellars

  21. Alan Fox: I am not at all familiar with Paul Coates. I was just intrigued by walto’s dismissal, especially when a quick Google indicated Coates is a fan of Wilfred Sellars

    Yes, Paul Coates is one of those philosophers who has spent a lot of time trying to show that Sellars’s views are much more compelling than would seem on first pass. I suppose I’m one of them, too, though I don’t want my career to be established on the basis of an ordinary potato.*

    * that is, a common tater.

  22. Kantian Naturalist: The basic idea is that brains are generating endogeous models of the environment that function as map-like representations of the affordances, …

    This is where we look at things very differently.

    To me, a map is neither a model nor a representation. Compare a road map of your city (or any city) with an aerial photograph of that city. They are very different. The aerial photograph is a representation. The map isn’t.

    The aerial photograph is presumably true, or was true at the time it was taken. The map is not true. The map is iconic, exaggerated and oversimplied. For example, the map does not show trees, lamp posts, etc.

    If you were to try using the aerial photograph while driving, you would find that hard to do. The map is far better for that purpose.

    We might say that the aerial photograph is veridical while the map is pragmatic. And I see that distinction as very important. Yes, I expect that we do have something like a map in our heads, but we do not have anything like an aerial photograph. That is to say, knowledge is pragmatic, not veridical. Facts are veridical, but we usually don’t fill our heads with facts. (And I guess that’s my disagreement with “justified true belief” as a characterization of knowledge).

    We use maps as a guide. In some sense, the map is part of the solution of the intentionality problem. The map allows us to connect linguistic exression (such as addresses) with the real world. We might even say that the map is something of a standard for solving that kind of intentionality problem.

    If the activation of sensory receptors (i.e. sensations) generates information …

    This is where I disagree on information. The activation of sensory receptors does not generate information and, by itself, could not be the basis for sensation. At best, it could provide what William James described as a blooming buzzing confusion.

    Compare to what science provides. We use a thermometer to measure temperature. The thermometer depends on being calibrated with a temperature scale (such as the Celsius scale or the Fahrenheit scale). The
    temperature scale is something like a map. The scale is our guide for connecting linguistic expression with reality, with the instrument (the thermometer) mediating that connecting.

    We cannot have information about the world until we have developed some kind of pragmatic internal map or standard or scale. Learning, by a child, must involve the formation of internal maps or standards. Developing those maps fits with what Eleanor Gibson described as “perceptual learning”. It isn’t acquiring of facts. It is acquiring the ability to appropriately discriminate so that we can perceive detail. And for a child, listening to the story of “Little Red Riding Hood” can be helpful in developing those internal standards. This is why knowledge can come from fiction, even though that fiction has no justified true beliefs.

    Activation of sensory cells could not be information. But activation in coordination with the application of internal standards could generate information.

    The really hard question is how to understand objectivity, truth, and related epistemic concepts in terms of predictive processing as a theory of animal cognition.

    My current take is that objectivity requires some kind of agreement within the community. I’ve suggested that we develop internal pragmatic standards. But we do that autonomously, so we should expect everyone to start off with different standards. Because we are an eusocial species, it is to our benefit to coordinate with others. So we tend to recalibrate our internal standards, so that they are reasonably consistent with what others seem to be using. And it is that recalibration toward a common standard (perhaps in the form of social norms), that makes objectivity possible.

    I see truth as conformance with public standards. If there are no standards, there is no basis for truth. The reason that I see scientific theories as neither true nor false, is that there are no standards for scientific theories, though an accepted theory then becomes a standard for data collected under that theory. Likewise, there are no standards for standards. In both cases, there are pragmatic requirements, but those fall far short of being a standard.

  23. Kantian Naturalist: I can’t speak to what walto had in mind. I myself respect Coates’ work immensely. I found his book on the metaphysics of perception to be very helpful in making clear the arguments in support of critical realism (i.e. an organism is immediately aware of its own sensations and conceptualizes those sensations in terms of the objects that typically cause sensations of that kind).

    I thought his book on perception was pretty bad, myself. He actually thinks he can “prove” that indirect realism is true.

    Poor bloke.

  24. Neil Rickert: To me, a map is neither a model nor a representation. Compare a road map of your city (or any city) with an aerial photograph of that city. They are very different. The aerial photograph is a representation. The map isn’t.

    What is the map?

    Neil Rickert: The aerial photograph is presumably true, or was true at the time it was taken. The map is not true. The map is iconic, exaggerated and oversimplied. For example, the map does not show trees, lamp posts, etc.

    There are plenty of things that the photograph does not show either. Both the photograph and the map have this in common: They show some elements/aspects and omit other elements/aspects about something while they (photograph and map) are not that something. This is the definition of representation.

    The map and the photograph are both representations. You will have to try draw some solid definitive distinctions to show you have managed to bring yourself to think that the map is anything else. What else is it?

  25. Neil Rickert,

    I was shocked to hear you say that maps aren’t representations, though perhaps now I better understand the deeper points of our disagreement over the years.

    As i see it, maps are paradigmatic representations because of the structural resemblance between the relations between objects in one set (the territory) and the relations between objects in another set (the map). It’s a second-order relation or relation between relations. Maps can convey spatial relations, political relations, more abstract relations (e.g. export-import ratios), correlations, etc. — and even a sound recording might be thought of as a “temporal map”.

    I do share your conviction that the function of maps is pragmatic, but to my mind that does not undermine their representational character — on the contrary, because on my view, to be a representation is (among other things) to have the pragmatic function of transforming relations between objects in the mapped set (territory) by virtue of manipulating the relations between objects in the mapping set (map). And a map has to be just-good-enough in its accuracy and precision for it to be used as a map. Adequacy comes in degrees, and what makes a map adequate (and to what degree of adequacy) depends on what the map needs to be used for.

    I take these points about how maps represent to apply to how cognitive states represent — they represent by virtue of being integrated into a functional map of environmental affordances and organismal states. (We can think of the representations as instantiated in looping networks of top-down predictions and bottom-up prediction errors across a multi-level hierarchy of neuronal assemblies.)

    In short, there’s no reason at all (except bad philosophy) for thinking that Gibsonian ecological psychology can’t make common cause with representationalistic cognitive science.

  26. That was one of the best ever posts by Neil. Can’t we just let it alone for a while and reflect on that? It could be months before he can once again rise to the level of such lucidity.

  27. Kantian Naturalist: As i see it, maps are paradigmatic representations because of the structural resemblance between the relations between objects in one set (the territory) and the relations between objects in another set (the map).

    Okay. But here are two things to keep in mind:

    1: I understand how human cognition works (or at least I claim to), and you don’t…

    2: I am not bound by the traditions of philosophy, but you are.

    You talk of a “structural resemblance”. But you probably cannot give a clear definition of what that means.

    You are saying that the map is a representation because of its structural resemblance. I am say that the only reason there is a structural resemblance, is that human social convention makes it so.

    When teaching, I would make chalk marks on the blackboard. The chalk marks might be words or diagrams. The chalk marks are representations. The blackboard is the representation system (or framework) on which we make those representations. I am saying that the map is more like the blackboard than like the chalk marks. If we are planning a trip, we might make marks on the map, to detail our plans. The marks are representations, while the map is the framework on which we make those representations.

    Think back to some war movies. You see people planning military actions. They have a wall map. They write on that map, or they attach notes with thumbtacks. The map is the representation system, while the notes that they attach are the current information about their plans.

    Here’s what I see as the important distinction:

    There is a truth requirement for representations. The usefulness of a representation depends on it being approximately true, for otherwise it does not represent.

    There is no truth requirement for representation systems. There is only a pragmatic requirement. Roughly speaking, a representation system has to be a good enough fit so as to allow it to accomodate the kind of representations that we want to make.

    Now look at maps. Is there a truth requirement? I don’t see that there is. For world maps, we can have a Mercator projection. Or we can have a conic projection. Or a Mollwiede projection, or a sinusoidal projection, or a stereographic projection. There are many different forms for maps. They cannot all be true. But they can all be useful. That’s why it seems to me that there is a pragmatic requirement for maps, but no truth requirement. Or, said differently, maps are defined primarily by conventions, rather than by empirical data. So maps are more like representation systems than they are like representations.

    To me, the distinction between representation systems and representations is very important. Let me give an example that I find useful.

    Think of quantum mechanics. Roughly speaking, it describes everything about reality in terms of discrete entities. So we should be using discrete mathematics to describe it. We should not be using differential equations.

    Yet physicists do use differential equations. So how could that possibly work?

    The answer is fairly simple. They are using the mathematical structure of the representation system. And the representation system that they use is mostly that of Einstein and Newton. It’s a continuous representation system, so we can use continuous mathematics (including differential equations), even though we are representing discrete events.

    Put that in terms of human cognition. In our thinking and planning, we might be said to use representions of our hypotheticals. It is the structure of the representation system that constrains those representations. And without those constraints, thought processes would be far less useful.

    For me: information is representations, while knowledge is the representation system. There’s a kind of duality (in the mathematical sense of duality) between them.

  28. Neil Rickert: Okay. But here are two things to keep in mind:

    1:I understand how human cognition works (or at least I claim to), and you don’t…

    Well, it’s true that I make no such claim on my own behalf. But I do try to keep up with cognitive science and neuroscience. I don’t know of any cognitive scientists who claim that they truly understand how human cognition works. There are some very promising models, and there’s plenty of room for improvement.

    2:I am not bound by the traditions of philosophy, but you are.

    Oh, I’m not bound by those traditions at all. I just find a bunch of texts useful for my own thinking and I love teaching all sorts of philosophies. There are very few constraints on philosophical practice.

    You talk of a “structural resemblance”. But you probably cannot give a clear definition of what that means.

    Here’s a definition I found recently that seem promising to me:

    Suppose SV = (V, RV) is a system comprising a set V of objects, and a set RV of relations defined on the members of V. [… ]We will say that there is a second order resemblance between two systems SV = (V, RV) and SO = (O, RO) if, for at least some objects in V and some relations in RV, there is a one-to-one mapping from V to O and a one-to-one mapping from RV to RO such that when a relation in RV holds of objects in V, the corresponding relation in RO holds of the corresponding objects in O.

    Notice that maps can satisfy this requirement regardless of whatever distortions, idealizations, and abstractions they introduce: it requires that (for example) the difference between the distance between Beijing and Paris and between Paris and Berlin in a Mercator projection of the Earth correspond to the difference between those distances on the actual planet. (Likewise for all other projection systems.)

    In any event, the real question isn’t “are cartographic maps themselves human artifacts?” (since that is obviously true, and not worth debating) but rather “is animal cognition importantly map-like?”. The fact that human beings construct maps doesn’t entail hat nothing can be map-like without being a social convention. There could be ‘natural maps’ as well as artificial maps.

    I find the distinction between “representation” and “representational system” to be somewhat forced, as well as the distinction between veridicality and pragmatic usefulness. I like what you say about “representational systems”, and I find very little to disagree with there — though I would also say that animal cognition generally involves representational systems.

    As I see it, maps, diagrams, and theories are specifically human ways of augmenting and shaping our representational systems. I think of them as basically “cognitive prostheses” because they are artificial maps that augment and extend and amplify our own natural maps. On your view, I take it, maps and theories are tools; I think of them as prostheses. There’s a much more intimate connection with our specifically human type of animal cognition.

    I would drop the distinction between representations and representational systems and just say that representations are components of representational systems. And while representational systems have pragmatic functions, I don’t see any clear-cut divide between usefulness and veridicality. Rather, I think that generally speaking, a representational system will be more useful to the extent that it is more adequate — but that there are always trade-offs against adequacy.

    One representational system could be more adequate than another, yet the less adequate system could be just good enough most of the time, be easier to use, yield faster results (e.g. less complicated neurocomputational processes for updating actions based on sensory stimuli), require less energy, or be developmentally canalized as a result of past natural selection. I suspect that, generally speaking, natural selection will favor more adequate representational systems if these constraints can be overcome.

    Now, it is true that in a strict sense, if we are to accept the principle of bivalence for natural languages, then we are forced to say that an assertion is either true or false — nothing in between. That’s why I’m talking about adequacy rather than veridicality. But adequacy is not a matter of sheer social convention; what makes one representational system more adequate than another is the fineness of grain of the mapping relations, and that’s not something that is arbitrarily stipulated.

  29. Kantian Naturalist: Oh, I’m not bound by those traditions at all.

    A quick response to that. Those traditions are deeply engrained, because of your training in philosophy. So it isn’t easy for you to break with tradition, or even to recognize the extent to which you are bound by tradition.

    I have not had such training. And when I started looking into human cognition, I deliberately avoided the literature from philosophy of mind, because I didn’t want those traditions to be able to mislead. Even then, I could see that I was making a break with tradition.

    I have, of course, been reading some of the philosophy literature since then, mainly to try to work out how to bridge the communication barrier.

    Here’s a definition [of “structural resemblance”] I found recently that seem promising to me:

    Okay. But I don’t think it can work. We are really talking about human cognition. And human cognition has to get going well before you are able to frame or pose that definition.

    In any case, let me try to make my point a different way.

    I take a thermometer. Let’s say it is a Celsius thermometer. And I measure the outside temperature. As a result, I can say “The outside temperature is 5 degrees C.” That’s a representation.

    If, however, I say “The melting point of ice is 0 degrees C,” that isn’t a representation. I don’t even have to read the thermometer to come up with that. It is an analytic truth, because that is part of the definition of the temperature scale. Similarly, “The boiling point of water is 100 degrees C” is an analytic truth.

    Those two analytic truths are important for a different reason. They anchor the thermometer scale to reality. If I did not have those anchors, then the melting point of ice could be 5 degrees today, 23 degrees tomorrow. It would all be meaningless. We need anchors. For otherwise we could not make reference to reality at all.

    Now get back to maps. I purchase a map at a local store.

    If the map is a representation, then the representation system is a blank sheet of paper. Looking at the map, I see New York near the right side of the page. But it should be at the left side, so the representation is false. Or, more generally, New York could be anywhere. The whole map is meaningless.

    So I’m saying that what we see on the map are anchors that show us how to attach the map to reality. And if they are anchors, they are analytic truths so aren’t really representations.

    Now, for sure, you could say that part of what is on the map consists of anchors. And part consists of representations. That’s probably about right. But what’s an anchor for you might not be an anchor for me. This is the same problem as “What’s an analytic statement for you might be synthetic for me.” We have to live with that problem.

    We were really concerned with cognitive maps, rather than maps that we buy at the store. And it seems to me that a cognitive map consists almost entirely of anchors. The parts that we don’t really know, so have to consult as representation — those are unlikely to get into a cognitive map.

    Finally, a note on analytic statements. The discussion in the philosophical literature is very frustrating. There seems to be little or no recognition, that analytic statements are important for anchoring our representations to reality. Without them, we would lack intentionality (aboutness).

  30. Neil Rickert,

    I very much like your emphasis on cognitive “anchors”, and I agree that what works as an anchor for you will probably not be an anchor for me — not if anchors are produced through bottom-up individual learning experiences, anyway. (That’s not to say that there cannot be “shared anchors”.) I’m inclined to think that one of the main functions of public language to keep our anchors from diverging so much that collaboration becomes impossible.

    A somewhat deeper problem I’m wrestling with right now could be put as follows: are there non-linguistic concepts? Notice that this is not the same as “are there non-linguistic mental representations?” Arguably the problem with Fodor is that he rightly sees that there have got to be mental representations, but then conflates mental representations with concepts, and then — noticing the intimate link between concepts and language — concludes that there must be private language, or a language of thought.

    Whereas if one is persuaded by Wittgenstein’s argument against the very idea of a private language — as I am and Fodor is quite obviously not — then the conclusion should be either that there are non-linguistic concepts or that there are non-conceptual representations. For a long time I took the former view, but lately I’ve been wondering if that’s the right move. The question is whether it’s possible for an animal’s cognitive representations to satisfy Evans’ Generality Constraint for concepts in the absence of intersubjective discourse to ensure that it’s representations allow for re-identification and singular thought.

    I’m not too sure what to say about “analytic truths”. If we adopt a strictly Kantian view of conceptual content, then an analytic truth is just one that somehow exhibits that content. But do concepts have the kind of structure that Kant assumes they have? I think there’s pretty compelling reason to think that they don’t. Sure, we can always stipulate definitions, but what’s gained by saying that definitions are “analytic”? I’m not saying that Quine was right to reject analyticity — I think his argument is question-begging — but I’d still like to have a better sense of what is actually important about the analytic/synthetic distinction.

    And I do appreciate Morton White’s unfairly neglected “The Analytic and the Synthetic: An Untenable Dualism”, the upshot of which there’s no way to clean demarcate between what words mean and how the world is. Quine makes that point as well, but in Quine the deeper point is muddled because his extensionalism makes me allergic to the very idea of linguistic meaning. I’m fine with intensional semantics; I just don’t see any way of cleaving apart meanings and facts.

    I do think there’s a need for a distinction between statements that make explicit the constitutive rules of a language-game and statements about the world that are governed by those rules. That’s the best interpretation I can put on the distinction between “synthetic a priori” and “synthetic a posteriori” statements. But even this distinction is quite rough-and-ready and isn’t carved in stone. The axioms of Euclidean geometry were a priori until they weren’t; likewise with Newtonian stipulations about movement.

  31. Kantian Naturalist,

    I pretty much agree with everything in your first paragraph.

    are there non-linguistic concepts?

    That’s a difficult question, because it depends on what we mean by “concept”.

    My view is that yes, there are non-linguistic concepts. But the concepts that we use in language are greatly enriched by that way of using. If that richness is part of what you mean by “concept”, then it could be reasonable to deny that there a non-linguistic concepts.

    For me, I see concepts arising from how we carve up the world. But we carve up the world far more finely than we can express in language. And how finely we carve varies from person to person.

    As for Jerry Fodor — my impression is that he sees thought as requiring logic, and logic as requiring language. I get that impression from his “Methodological Solipsism” paper. Incidentally, many AI people have a similar view.

    For myself, I see thought as simulated behavior. Language use is a prominent part of human behavior, so it is prominent in thought. But we also think about actions other than speech, so I disagree with what I take to be Fodor’s view. And I certainly disagree with him on whether there is a “Language of Thought”. By the way, Fodor died recently.

    I’m not too sure what to say about “analytic truths”.

    Maybe I should say a bit more.

    I’m not too troubled about Quine’s argument. I take his objection to be about the use of the distinction in the philosophy of language. And if what’s analytic for me can be different from what’s analytic for you, then he has a point with that objection.

    I’m more troubled with philosophers who give the impression that they saw the existence of analytic statements as an embarassment (because they have no informational content about the world), and are relieved by Quine’s argument which they take as saying that we can dismiss that issue. They seem to not understand that analytic statements will crop up as we find ways to anchor our linguistic expression to reality.

  32. Neil Rickert,

    I don’t understand how analytic truths can function as anchors. How would you respond to Searle’s Chinese Room argument, according to which, one could memorize an entire dictionary–or according to your picture–an entire atlas, and still not understand a single thing?

    In other words, what good is a weightless anchor in outer space?

  33. walto: I don’t understand how analytic truths can function as anchors.

    They can’t. You must have misread something. Or perhaps I didn’t explain well.

    It is our behavioral practices that function as anchors. And, as part of a community, it is the conventions that establish those behavioral practices. I am particularly thinking of behavioral practices that we follow in order to get useful information about reality.

    If, however, we attempt to describe what we are doing in the form of a proposition that purports to be a description of the world, then that proposition will typically be analytic.

    You see this in other threads at TSZ. Creationists are insisting that “survival of the fittest” is a tautology. And that’s because they are viewing it as a proposition about the world. But it isn’t. From the perspective of population genetics, it describes a behavioral practice of how we anchor “fitness” to reality.

    How would you respond to Searle’s Chinese Room argument, according to which, one could memorize an entire dictionary–or according to your picture–an entire atlas, and still not understand a single thing?

    I’ve responded often. The argument is terrible, and doesn’t actually prove anything. It is an attempt (a failed attempt) to prove what is already obvious to mathematicians and computer scientists — namely that computation is based on syntax, rather than semantics.

    I agree with Searle’s conclusion — that you can’t get semantics from syntax — but I deny that he has actually proved that.

    In other words, what good is a weightless anchor in outer space?

    If it is useless, then it isn’t an anchor. I’m pretty sure that when astronauts were on a trip to the moon, they would have seen their radio communications with earth as provide a useful anchor.

  34. Neil Rickert: walto: I don’t understand how analytic truths can function as anchors.

    Neil: They can’t. You must have misread something. Or perhaps I didn’t explain well.

    Maybe I misread this?

    Neil Rickert: Those two analytic truths are important for a different reason. They anchor the thermometer scale to reality. If I did not have those anchors, then the melting point of ice could be 5 degrees today, 23 degrees tomorrow. It would all be meaningless. We need anchors. For otherwise we could not make reference to reality at all.

    So I ask again, how can analytic truths be anchors? Or if they can’t be after all, what were you trying to say in your earlier post?

  35. walto: Maybe I misread this?

    Okay, I didn’t explain well.

    What actually anchors the temperature scale, is the way that we defined 0 and 100 degrees. And that leads to a behavioral practice in how we calibrate our thermometers. But if we treat those as simple propositions, then “the melting point of ice is 0 degrees” becomes an analytic truth. It is our definitions and calibration practices that make it an analytic truth. So to say (as I did), that the analytic truths are anchors was a gross over-simplification.

  36. Neil Rickert: I agree with Searle’s conclusion — that you can’t get semantics from syntax — but I deny that he has actually proved that.

    To be clear, Searle himself doesn’t think that he’s proven that, either. It’s an assumption of his that then motivates the rest of the argument. The Chinese room thought-experiment is designed to illustrate that assumption, not to prove it.

    More specifically, his argument is as follows:

    1. Syntax is insufficient for semantics.
    2. Minds are semantic engines: they can understand meanings.
    3. A computer program is a syntactical engine: it just processes meaningless symbols according to rules.
    4. Therefore, no program could be a mind.

    The Chinese room thought-experiment is an illustration of this argument, and it depends on assuming (1). There’s no argument for (1). It’s just an article of faith.

  37. walto:

    Thanks, Neil. Very interesting post.

    Maybe interesting, but it’s also a disastrous mess.

    Neil Rickert: To me, the distinction between representation systems and representations is very important. Let me give an example that I find useful.

    Think of quantum mechanics. Roughly speaking, it describes everything about reality in terms of discrete entities. So we should be using discrete mathematics to describe it. We should not be using differential equations.

    Sorry, but quantum mechanics is the exact opposite to discrete entities. Ever heard of “quantum field” with its fluctuations?

    A few questions arise from your statement: What the heck do you think quantum mechanics is? Where did you read about it? Or, alternatively, what do you think “discrete entities” are?

    Neil Rickert: Put that in terms of human cognition. In our thinking and planning, we might be said to use representions of our hypotheticals. It is the structure of the representation system that constrains those representations. And without those constraints, thought processes would be far less useful.

    Now a more basic question arises: Have you ever entertained a coherent thought? Tried to at least?

    Here are some of the distinctions outlined in a workable way for you: Representations may form a system. Let’s term such a system “the big picture”. In the big picture, each representation has its place, like a symbol on a map. So, thus far “representation” equates with “a symbol on the map”. However, as we know from everyday language, a picture (as in “the big picture”) is also a representation. A symbol on the map is a representation of something, but so is the entire map. And so is a photograph along with whatever is pictured in it.

    Can you argue for a categorical distinction between a storm and a wind? (Knowing you by now, you just might.) No, a storm is just a strong wind. A map is a representation of a territory produced by surveying techniques rather than by means of photography, but they both represent the territory and are thus representations.

    It’s not that photographs are representations and maps are something like “representation systems”. If we use map symbols in a systematic way, we can say we are using a “representation system”, but maps are still representations. Photography is another technique with a whole science behind it, so this can be called another “representation system”.

    In yet another sense, representation is an abstract phenomenon or general activity. The structure and nature of this phenomenon or activity is a legitimate object of analysis, like the structure and nature of anything is. Maps and photographs have different structures, natures, and purposes, which makes us call the one “map” and the other “photograph”, but this does not change the fact that maps and photographs are both representations in the sense of being individual instances of representing something.

  38. Erik,

    Also a good post, Erik.

    I have the sense from Neil’s writings on this thread that he’s considered objections of the kind you’ve brought up here, but it’s true that he has the burden of spelling out the distinction he wants to make with more precision than he has to this point.

    In any case, as a fan of the Tractatus, I enjoy these discussions of structural realism, whatever their “result.”

  39. I think I’m not really happy with the distinction — as I thought it being made above? — between anchors and representations.

    If we’re using “anchor” in a slightly metaphorical sense, to mean, “that which relates our models of the world to the world”, then I’m fine with it. But I’d want to stress that what does the anchoring is the organism’s representations, and specifically, it’s lower-level (and possibly non-conceptual) representations, at the parts of the cognitive system that co-vary with activation or inhibition of the sensory receptors.

    Here’s how I think of it: any cognitive system, for it to function as a cognitive system, must guide the organism through the world well enough to achieve the Four Fs (feeding, fighting, fleeing, and reproducing). The cognitive system does by virtue of being a map of the affordances that comprise the organism’s niche (and, for non-sessile organisms, also a map of the spatio-temporal location of the organism in that niche).

    The map is anchored to the world (to the extent that it is) by virtue of the map includes representations that co-vary with the activation of sensory receptors. So the navigation of the environment depends on successful coordination of how the map is guiding action and how the map is updated in light of stimuli.

    With this lightly sketched account in view, I want to say that’s a false dilemma to ask “are the representations veridical or pragmatic?” Representations that aren’t somehow tracking the causal structure of the world aren’t going to be useful for very long.

    Once we accept some kind of realism — that the world has a causal and modal structure which is epistemically accessible to beings like us — that’s all we need for breaking the dead-lock between cognitivism and pragmatism, because then we can show that any cognitive system that has needs must be able to track the causal and modal structure of the world well enough to satisfy those needs.

    (Of course, “unguided” evolution will not equip the organism with cognitive abilities that allow it track causal and modal structures more adequately than is necessary to satisfy its needs. That is the kernel of truth in what Plantinga calls “Darwin’s Doubt.”)

  40. walto: Hahaha.You’re so demure!

    I try to keep my comments at TSZ family-friendly. Besides, that’s Pat Churchland’s version of the joke.

  41. walto: In any case, as a fan of the Tractatus, I enjoy these discussions of structural realism, whatever their “result.”

    I have read the Tractatus. Already 1.11 (“The world is determined by the facts…”) smelled fishy. If the world is the totality of facts (1.1) and in between you have not introduced anything that would explicate what it means to “determine”, then you cannot say whether the world is determined by facts or vice versa. The point 2 with its elevation of “atomic facts” was next annoyance and 2.02 (“The object is simple.”) was a dealbreaker. We just had to read (in 2.01231) that in order to know an object we have to know not its external but its internal qualities, so objects are evidently composed of both internal and external qualities, which means objects are composite, not simple.

    Kantian Naturalist: I try to keep my comments at TSZ family-friendly.

    Why? When your family comes here to read your comments, they will inevitably see the others’ comments too and the family-friendliness is out the window.

  42. Kantian Naturalist: The map is anchored to the world (to the extent that it is) by virtue of the map includes representations that co-vary with the activation of sensory receptors.

    No, that cannot work.

    You fail to understand the problems that a cognitive system must solve. That’s not surprising. The problems are well hidden. But they are hidden in plain view. It’s just that almost nobody sees them, because our cognitive systems do such a fine job of solving them.

    Think of a tree. I see you have brought up the idea of plants having minds.

    If a tree had a cognitive system, it could work about as you describe. Light falls on the tree, and is detected (by photo-synthesis, at least). So the state of a light detector is a representation of something about the world — perhaps not a very precise representation, but still a representation.

    By comparison, consider a human. Light falls on a retinal receptor. That’s a representation of light falling on the retinal receptor. But you cannot say much more than that. You cannot infer much of anything about the world, other than that there is light.

    The difference is that tree is fixed. Its orientation to the world does not change, apart from a little shaking due to the wind. But our orientation to the world changes greatly. We move, our head moves relative to our bodies, and our eyes move relative to our heads.

    The tree is anchored to the world. The tree is its own anchor. But we are mobile. We are not anchored to the world. So we need to create anchors with our behaviors. We cannot do it with representations from the passive receiving of stimulation.

    You want to see what’s out there. But you don’t know where your head is pointing, so you can’t make out what is there. You can try to fix that by aligning with tree or a lamp post. But, unfortunately, you can’t see the tree or the lamp post until you have found a way to see what’s out there.

    That’s the problem a cognitive system must solve. A tree doesn’t need a cognitive system, because it does not need to solve that kind of problem.

    It is easy to see how we could have purely subjective representations (or person-relative representations), though they would be like a blooming buzzing confusion (as suggested by William James). The problem is how to have representations of the external world.

  43. Kantian Naturalist: The map is anchored to the world…

    Neil Rickert: No, that cannot work. … The tree is anchored to the world. The tree is its own anchor. But we are mobile. We are not anchored to the world.

    Aren’t we talking some completely different anchors here?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.