For a treat during winter break, I am reading back issues of Philosophy of Science and have been pleasantly surprised that there are articles that I am actually interested in reading! Vol. 74 (2007) has a number of articles that are not so terribly specialized as POS articles have sometimes been in the past. This is not to say that there are no super specialized articles, for those of you who are fans of such, just that there were several articles that were more like the articles that I enjoy reading.
In particular, Margaret Morrison has an article in which she suggests that the recent interests that many philosophers of science have developed in models has wrongly led to ignoring theory (Number 2, April 2007, 195-228). She argues that there is still an important role for theory in philosophy of science and sketches an account of the relationship between theory, model, and the world that shows this. Her account of models differs from the semantic view, which identifies theory as a family of models.
Gabriele Contessa also has an article on models, but his focus is more squarely on the question of how models represent the systems they are models of (Number 1, January 2007. 48-68). Since I am currently working on a paper that deals with representation, I found this article particularly intriguing and so I am raising some questions about it here.
Contessa argues that it is possible to give necessary and sufficient conditions for a model representing a system. He contrasts this view to that of Mauricio Suarez, who denies that this can be done. Contessa’s argument rests on several distinctions and is worth reading the paper, even if only for an overview of several of the key positions that are currently being defended in this discussion. I won’t go into detail about these distinctions but will focus on the one at the core of Contessa’s account of representation.
Contessa claims that much of the confusion over representation comes from a failure to distinguish between representing and representing faithfully. If we (wrongly) think that representation requires faithful or even partially faithful representation we are likely to draw the conclusion that Suarez does. Contessa makes the case that in order for a vehicle to represent a target it need not be a faithful representation. Of course, we are interested in what makes a representation faithful and he acknowledges that a full account of representation would include an account of faithful representation. What he does in this paper is give necessary and sufficient conditions for representation simpliciter (what else is required for faithful representation will have to wait, he notes at the end of the paper). In order to represent, the vehicle must be used by someone as an interpretation of the target system. (I am not doing full justice to the account because I haven’t discussed his definition of interpretation and its connection to surrogative reasoning, a notion that he takes from Chris Swoyer (1991).)
The account that Contessa gives solves several problem and has some intuitively compelling features. However, what may not be quite so intuitive is that at first blush the account seems to commit Contessa to the view that it is possible for anything to represent anything else. There do not seem to be any inherent constraints on what the features objects that become models must have in order to be used as interpretations. On the one hand, this makes sense. Suarez’s point that there isn’t anything that we can point to in the model that would give us the necessary and sufficient conditions for it to function as a representation is vindicated. But Contessa asks us to look more closely at the use of the model to understand what it is to represent. On the other hand, I feel uncomfortable about the idea that a model that is not at all (even partially) faithful could be meaningfully said to represent a system because a user’s interpretation turns it into a model of that system. Why am I uncomfortable with this? Is it because I am confusing faithful representation with representation as Contessa contends?
I am not sure, but here is my first pass at the answer. I am wondering to what extent someone can use an object (a model) as an interpretation of a particular system if there are no features of the model (object) that bear any relation to the system in some way that is independent of the user. If this is right, Contessa’s account begs the question because when a model can be used as an interpretation it already requires that it have some other features in virtue of which it represents. It seems pretty clear from his article that he does not see this as being circular, so either I have not fully understood his account of interpretation, or I am slipping into that confusion of representation and faithful representation. All the force of my worries would then be addressed in the account (yet to be given) of faithful representation.
I am not sure what is going on here but this discussion reminds me of another similar one. Philip Kitcher criticized van Fraassen’s pragmatic account of explanation because the way van Fraassen sets up the relevance relation between the explanandum and the explananda allows anything to be an explanation of anything else under the right circumstances (as long as the right relevance relation holds and what counts as the right relevance relation is dependent on context, so given the right context anything could be an explanation). The similarity between these two cases seems to be connected to the fact that both accounts depend on how something is used by someone. Van Fraassen did not seem to be terribly worried by Kitcher’s criticism, because this was indeed what he intended. Kitcher later acknowledged that the flexibility had been intentional. (Sorry that I don’t have the references here. I will fill them in later.) Another similarity between Contessa’s account of reference and van Fraassen’s account of explanation is that Contessa’s argument precedes by asking us to draw a distinction between representing and representing faithfully. Van Fraassen’s discussion depends on a similar distinction between having an explanation and having a good explanation.
This is all very incomplete, but I offer a final thought here. Could it be that the ambiguity is in “use”? To use and to use successfully are different are different in the same way as the other two concepts vary. It seems to me that making this distinction can only be done as a matter of degree however. There is some point when it no longer makes sense to describe what is going on as using A in order to do B. So, for instance, I can use a hammer to remove a screw, in the sense that I can pick the hammer up with the intention of taking out the screw with it. I can take the hammer and touch the screw with it and so on, but I cannot remove the screw with it. So I cannot successfully use the hammer to remove the screw. I think it would be reasonable for someone to describe what is happening in the following way: “She thought that she could use the hammer to remove the screw but in fact she cannot use it that way.” It is not just that I was using it and failed, but rather that I wasn’t really using it though I believed that I was. In a similar vein, I might think that I can use Newtonian physics to explain iridescence, but in fact, all explanations that I give in this way will be bad. So am I explaining? I may believe that I am but I will be wrong and since I can only be wrong we might reasonably say, at least at some point during my attempts, that I am not explaining at all, not just that I am explaining badly. Finally, to come back to representing, I might use the salt and pepper shakers on the table to represent the structure of the atom. This would mean, according to Contessa, that I am interpreting the salt and pepper shakers as a representation of the atom. The problem here is that I do not know how I would determine what sorts of things might appropriately represent others, that is, whether it is even plausible to claim that I or anyone else could use these objects in this way. Of course, how I do use any of these things in each of these examples (as a tool, as an explanation, as a representation, via using it as an interpretation) depends on background knowledge and so perhaps my worry about whether it is possible to make a distinction between something being a good tool, good explanation, or faithful representation and being simply a tool, an explanation, or a representation respectively can be resolved by specifying that all judgments about use are relative to background knowledge. But doesn’t this just push the issue of the distinction between a good x and an x into background knowledge and so not eliminate the problem just move it around?
Well, these are some of the puzzles that I have had while thinking about Contessa’s account of representation. I worry that it may be circular and so wonder if it gets us anywhere, but I like that it is so based in use. I hope to have more to say about this as I get clearer on what I want to do with the issue.
Swoyer, Chris (1991), “Structural Representation and Surrogative Reasoning,” Synthese 87: 449-508.
A model that is independent of the user? By that can I assume you mean by this a model that others accept as appropriate and the non-independent model –no one but the user thinks is appropriate. The use issue, seems to me may be subsumed under appropriateness. And the representation issue too. But appropriate by whose criteria? Well, the hammer would be appropriate—if it was ok to gouge or bang the screw out with it–it would really destroy the wood in so doing– is that appropriate or not?—an acceptable result to some and not to others. And take physics— Is light a wave or a particle? If you are a partisan of quantum physics then light is a quanta more than it is a wave or a particle– So what is right? After all the quantum theory–dealing in probabilities rather than strict causality–is the most successfulpredictor ever in physics. But matter as commonly conceivedis ignored by quantum physics–instead you have abstractions. Is that appropriate to the world of matter? Is quantum really telling us about matter then–or just about probabilities and about something statistical? anyway—can of worms.
I was thinking whether models might need to have features that are in some way independent of the user, by which I mean that I acknowledge that one of the things that is required for a model is that it is a model for a user (or group of users) but I was wondering whether to be used it had to have some features that were independent of the goals and interpretations of the users. What I have in mind is that there might be constraints on what could serve as a model, just as there might be constraints on the interpretation of a poem. There may be many ways in which it could be interpreted but there are some interpretations that just won’t work. (Perhaps because they cannot be made consistent, for instance.) So when you ask “Appropriate by whose criteria?” this is what I am suggesting, that there might be some ways in which appropriateness of a model is not just relative the some individual’s criteria. It may well be the case that the only way to sort this is relative to a context, which would include goals. So the context is one in which the goal is the gouge the wood, then, yes, this is an appropriate tool.As for this being a can of worms, well, yes, but I do not despair of sorting it if the key elements of the context can be identified. I think that one of the most interesting things that comes out when we think this through though is that what we value, what our goals our, is surely a vital piece of the context, but is there anything else that constrains models besides the features of the context?
By can of worms, I meant only that there are potentially lots of tangled complexities involved insorting this issue–I didn’t mean it shouldn’t be attempted. Safe to say generally that we are constrained as human beings–to act as we do.All models, for instance, as with all descriptions, can be said to have the basic form —-an x is a y. Scientific modeling is just an extension of description and inquiry common to us all–asking, “is what is described actually the case?” and saying “If x does Y under such and such conditions it seems it should do P under this and that circumstance.” And then determining if it is the case. One can say that there are cultural constraints: rules or limitations of a societal sort—yes? Scientific conventions constrain a scientific model– a model is valued generally in science if it shows elegance, simplicity, consonance with previous thought on the matter (models that are too different will be rejected generally—-science polices its theoretical borders),ability to explain as many phenomena as possible, has predictive power, has testability. In science though, the models–the descriptions– aren’t as important as the quantitative measurements–and the descriptive mathematical equations. The models are just enough to form a basis for the calculations. The models themselves, you may have noticed, can be a little vague and unworkable—e.g. Einstein’s idea of space warping under influence of mass–supposedly accounting for gravity– doesn’t really tell you what it is for space to warp. Scientists use the analogy of a trough made by the sun’s gravity that the earth has fallen into—but it remains vague and doesn’t work to describe the generally uniform gravity around a sphere as with a planet— so what is warping–is it like a congealing of space that grabs the planet in its grip–like a kind of cosmic flypaper? Why would warping of space result in this? Another example: expansion of the universe—indicated supposedly by the red-shift observed in galaxies–but this model becomes very unclear once you try to get specific about where all the galaxies are going in relation to each other and just exactly why the space is also not increasing between planets and stars within galaxies (supposedly it is the mass/gravity that prevents this–but what prevents space from separating stars at the periphery of a galaxy–where gravity is much diminished?)There seems to be lots of room to ask central questions that go unanswered. In molecular biology the amount of data is immense—and proceeds generally by discovering one tiny step at a time in cellular processes—–each discovery is like a little dot and they are often widely scattered—with lots and lots of dots to fill in between. In this case–the models are constrained by the sheer lack of knowledge. But despite incompleteness— models or mechanisms may still be very useful. Mendel’s rules of heredity mention nothing about genes or DNA —but despite the lack of these fundamental conceptions—he could predict generally the colors his peas would take. And California indians—despite having no inkling of the Generally the progress of science has been a progressively more detailed description –on finer and finer levels– of common phenomena. with resulting greater control over processes. People have been tanning animal skin to make leather for thousands of years using certain chemicals. All the tanners knew is that if you add such and such to the skin–you get leather— but detailed chemical knowledge of the process has increased dramaticallyin the last two hundred years andimprovements have followed.
Sorry, forgot to complete the comment that the California indians used to chew willow bark for headaches— thefact that they did not know there wasaspirin in it–didn’t prevent the cure. If their model was thatcertain “spirits” in the bark created the benefit. Presumably the indians came to associate willow bark with no headache and over centuries perhaps, confirmed the association. And Jonas Salk, the creator of the polio vaccine–did so at a time when knowledge of viruses was much less detailed. SOme detail is just not needed to achieve good results— which is not to say thatknowledge in greater detail couldn’t lead to improvements. To say Salk had a model but the indians didn’t–seems an uncharitable thing. After all, our current models may appear so comparatively primitive to scientists 300 years from now that they wouldn’t label them scientific models.
I am not sure that I would say that the correct characterization of models was that they have the basic for “an x is a y”. I think that part of the point of Contessa’s article, and I agree with this point, is that a model is a three part relation and so has the form “x is a y for A” where A is some person or group. In other words, there is an inherently pragmatic element in the notion of a model, since it has to be a model for someone (for some purpose).I also think that what you say about the constraints on a model are more accurately thought of as constraints on theories, or perhaps, better than constraints, the sort of virtues that we hope or strive for our scientific theories to have. Models can sometimes be heuristics and as such may not display these virtues at all. I think that I will be doing another post on models and representation before long, so maybe that will make how I am thinking about this clearer. For now though, by models I mean everything from physical models like the Watson-Crick DNA model to mathematical models and so some of the things that you say about mathematics I would interpret as being about mathematical models. So when you say that the models are not so important as the quantitative measurements, well, the mathematical models would have the particular quantitative measurements embedded in them as empirical substructures.Finally, on chewing willow bark and polio vaccines, I agree and would say that both are operating with models. The point that you make about not needing detail to be useful is right and I think particularly important and interesting. To what extent do the successes of our models count for their truth? And to what extent does it matter?
I could say that a model is a kind of description—or is every description a kind of model? A description is a kind of metaphor–in fact language is a kind of metaphor if we take metaphor to mean one thing standing for another. Certainly a model is something that stands for something else–assuming you think it is possible to clearly separate thing from description (as the statement “separate thing from description” is itself ambiguous as to whether a thing, state of affairs or a description)Scientists will often hedge on their verbal model by saying that the mathematical description is more important and consistent and so the incompleteness or contradictions of the verbal model can be ignored for the time being. Quantum Physics is a very accurate predictor mathematically—but, according to EInstein, is a profoundly incomplete theory verbally–conceptually. “Has to be a model for some purpose. ” So what is a purpose? On a lark–does that qualify? Is anything then a purpose?Is there then no such thing as doing without a purpose? I might or might not make up a model with a purpose in mind—why would a purpose be a necessary thing? If I come up with a theory for my own pleasure—and I do do this–is this a purpose? Can one infer motivation from behavior? Lots of problems with that. It seems to me you can say that x is a y–that is, x phenomena conforms to y category or model or some such, sans any reference to a user of the model. What, it isn’t a model unless some group or individual claims it? Is this what you are saying? By x is a y for A I assume you mean that might not be a y for B. That is, another group might have other ideas. But what is pragmatic–what works? Depends on your definition–and definitions vary among folk.If pragmatic may vary among folk–then what model is appropriate may vary among folk.Are there some elements in a model which are independent of the people who use the model? What, in the sense that their ability to question the model in whole or in part, or imagine another possibility will be somehow removed from their repertoire of behaviors?– Are you saying that there is some kind of indubitable definition of “pragmatic” or of “model” or some such—seems to me a longshot. Group consensus may place limits on the form of a model—but eliminate imagination in some way?–again a longshot. What is pragmatic surely will vary among folk. If your goal is to create a vaccine you surely are constrained to some extent in your procedures–and what is pragmatic –what works–in relation to your goal. If you just want to know something more generally about viruses then you are perhaps less constrained– and what is pragmatic –what works– is perhaps different for you. Or is one equally constrained in each situation–but in different ways? If what is pragmatic can vary and pragmatism is at the core ofmodeling–modeling for a purpose–(as a group of users has a purposes–yes?) and purposes change–then how will you say that any one group must be constrained by certain rules which they are given and may not alter? What would such rules consist of? OR are you saying that models–a sub category of description I can say–are constrained to communicate –must follow certain grammatical and syntactic or visual presentation rules. Could say most people will have such expectations of conformity towhat they are used to. Is this what you seek to explicate—the general rules most folk expect? What they all seem to have in common? Could be, then, anthropology or sociology or experimental philosophy or the like. Is that your aim? Or are you trying to create a definition of model and argue thatyour definition must be the one and only because of a, b and c? Again, if what model works may vary among folk, it seems just as interesting question to ask–how domodels vary as to ask how they do not.
Simplicity is what is needed here. The accuracy of models is determined by their consistancy being congruant with the acutalities they represent regardless of interpretation. Just as the appropriateness of function is determined by the intrinsic nature of identity regardless of the user’s intentions. “Paradox’s of duality require such considerations and distinctions between the relative aspects of characteristics and the various conditions of circumstance with respect to cause and effect. For example, light is neither a wave or a particle in and of itself, but rather a form of energy released from nuclear collisions. It is the nature of the medium through which it travels that determine whether it behaves more as a wave or a particle. Mathematical predictions work well in quantum physics becuase of the sensitivity of subatomic particles make it impossible to observe them directly without affecting their behavior as it is different when in their natural state unobserved; Einstein’s black box. So how can one be certain that a model and representation is accurate? Well, if it works, then go with it.