Monday, June 1, 2009

Draft: Structure and Representation

I finally got around uploading a draft of my my monstruously long two-part paper 'Structure and Representation', which some of you are already familiar with (Part I can be downloaded from here and Part II from here).

In the paper, I defend an (I think rather unorthodox) version of the so-called structuralist account of representation. I'd really appreciate comments (either here or by e-mail) from anyone who finds the time and the courage to go through it, especially since the paper will form the basis for one of the chapters of my book on "scientific" representation, which is now officially under contract with Palgrave Macmillan as part of their new series New Directions in Philosophy of Science edited by Steven French.


  1. I really like the London Underground example, which gives you a rich source of illustrations.

    But I've been away from philosophy of science for awhile and I have to say I'm a puzzled by the opening paragraph's separation of problems related to truth, models, and representation. For one thing, I always thought representations were, at least very often, truth vehicles, i.e., whenever a representation represents a "fact" it become "truth-apt". (I'm guessing that role is now reserved only for propositions?)

    Certainly, models represent and, yes, in ways similar to maps. (Actually, similar is too weak a way of putting it. Maps just are models of territories.) But I don't see anything wrong with the idea of a "false map". It seems to me that some of the examples you construct (the outdated map and the completely random one) are simply "false" and that the other kind are "true" (on the relevant points). Like all represenations, they don't capture the "whole" truth, but surely the map of an rail network can get the location of a station right or wrong.

    So, coming at this somewhat from the outside (but possibly as a reader of your book ... I don't know what the intended audience for the series is), I would like to see those "two basic points" fleshed out a bit. There must be some special constraints on the application of "is true" in the context of this field that I'm not getting.

  2. Hi Thomas,

    One way to characterize the difference between the "syntactic" and the "semantic" views of theories is that according to the former theories are collections of sentences or propositions while according to the latter they are collections of models and one of the only characteristics of models everyone seems to accept is that models, unlike sentences or propositions, are not truth-apt. But if models do not relate to the world by being true or false of it, how do they do so? The most promising answer to this question seems to be that models do so by representing aspects or portions of the world--just like maps or (some) pictures do. I take that the so-called "problem of scientific representation" is exactly the problem of explaining how models represent real-world systems without being truth-apt. If models were truth-apt, the problem would not arise but then the distinction between syntactic and semantic views of theories would also seem to collapse.

    Now, you mention that there is nothing wrong with the idea of a "false map". If this is a point about ordinary language, I don't have much to say except that my intutition is still that talking of the old map of the London Underground as false would constitute a category mistake--maps just aren't the sort of thing that can be true or false. Of course, as I explain in my paper, this does not mean that the conclusions of the inferences users draw from the map can be false and that the map can misrepresent the network. But one of the points of the old map example is that mirepresentation unlike falsity is not an all-or-nothing concept. Mirepresentation comes in respects and degrees and this is particularly important for models, which usually misrepresent to some degree some aspects of the real-world systems they are used to represent but nevertheless they are partially faithful representations of those systems. So, I would just deny that the old map (or, say, the Copernican model of the solar system)is "simply false".

    Btw, one could think that truth also comes in degrees and that the old map is approximately true, but I don't know of any successful account of approximate truth and I hope that the notion of partial faithfulness may be a better candidate for the jobs approximate truth was designed to perform.

  3. Thanks, Gabriele. How old is that distinction between "syntactic" and "semantic" views of theories?

    I'm guessing my pragmatist intuitions are getting in the way here. I would probably argue that sentences (unlike propositions) aren't any more truth-apt than maps. It is only on a interpretation that attributes propositional content that the true or false verdict becomes relevant. So, once you read the map of the underground as saying that a particular station is on a particular line (or further along one line than another station is from the station that you are at). That's where your inferences come in, so I guess we'd agree.

    Also, I very much agree that the Copernican model (and even the Ptolemaic) is not "simply false". Not, however, because I have some notion of "approximate" truth. I think I would defend the (always) "partial" truth of theories. They get some things right and some things wrong.

    I strongly agree that one thing represents another if the former supports valid inferences about the latter.

  4. Thinking about it some more ... I think my inclinations go towards a coherence view that would have truth always cash out in terms of valid inferences on an interpretation of the "representation". That is, I would want to avoid positing an entitity (like a proposition) that is unambiguously "apt" in regard to truth. We don't get anything more apt than representations that must be taken to be true ... one part at a time.

    You got me thinking about the bombing of the Chinese embassy in Belgrade. NATO apologized, saying they had been given an old map. Now, we can say that the map was neither a true nor a false proposition about Belgrade. But we could also say that it was a false (in a strong sense) map of the location of the Chinese embassy. Likewise, the diagram of the underground is not a map of the underground so much as a map of the relative position of stations and the routes bewteen them. It maps each of these (parts) truly or falsely.

    I suppose, however, I am essentially skeptical of a non-ordinary, technical notion of "truth". As you say about approximate truth, I'm not aware of a successful account truth that moves very much beyond "getting the facts right", with all its attendant issues and problems (to be settled, in each case, pragmatically).

  5. Hi Thomas,

    About your comment at 1:21 AM:

    Since, in the case of what I call partially faithful epistemic representations, some inferences will lead to true conclusions and some to false conclusions, I don't feel comfortable talking about the representationa as whole being true or false and, as I say below, I still prefer to say that a certain vehcile represents accurately or faithfully a certain aspect of a certain target rather than saying that it is true about that aspect of the target, but that seems to be largely a
    stylistic/terminological choice.

    About your comment at 3:27AM

    I'm not sure what you mean when you say:

    We don't get anything more apt than representations that must be taken to be true.

    As for the Chinese embassy example, I would tend to say that the map represented the location of the Chinese embassy inaccurately, but again this is probably just a terminological choice.

    I'm not a skeptical as you seem to be about the notion of truth, but I don't think anything I say commits me to any particular account of truth. (But I don't this was your point anyway. Right?)

  6. I'm not skeptical in any philosophical sense, just a bit pessimistic about formalizing the concept of truth within the terminology you propose (i.e., to limits its use to mark the distinction you are interested in). I think you're right that we're haggling about words, not really disagreeing on substance.

    BTW, I think I found out why I didn't get the "epistemic"/"syntactic" distinction right away. Jason Stanley (in Knowledge and Practical Interests) says that there is some variation in the way epistemologists use the word "epistemic" and that some use it identify "truth-conducive" factors. You seem to define "epistemic" in terms of factors that are exactly NOT truth-conducive (but nonetheless representational). I'm not sure how the two projects relate at all, however.

  7. We agree on this:

    "Since ... some inferences will lead to true conclusions and some to false conclusions, I don't feel comfortable talking about the representationa as whole being true or false"

    But I would go further:

    Even when we get down to the individual inference (conclusion drawn from the representation) it will represent only a partial truth (or parts of the true situation), and will, in fact, also license further inferences, some of which will be false.

    I don't mind the infinite regress here because I think coherentism and fallibilism are perfectly okay.

    We represent states of affairs (with models, sentences, what have you) and interpret those representations to imply propositions. These proposition then represent more or less faithfully ... and how faithfully depends on how useful any further inferences are. The regress stops when we have achieved whatever practical end we had set for our self. Like when we get off the train at Paddington.