Remarks for Opening Class on Content and Its Role in Explanation - Hartry Field
I’m not going to say much about the details of Stephen’s discussion, but will focus more on sketching a different way of thinking about the issues.
Stephen’s focus is on skepticism about the whether propositional attitude properties play a causal-explanatory role. Actually he sometimes puts it differently, in terms of skepticism about whether propositional attitude properties play an ESSENTIAL causal-explanatory role. But I’m not sure that there are any philosophers, and certainly there aren’t many, who would flatly deny that propositional attitude properties (or for property-nominalists, propositional attitude predicates) play a causal-explanatory role. Where the philosophical differences are, I think, is over the way in which propositional attitude properties (or predicates) are causal explanatory.
This way of putting the issues isn’t wholly discontinuous with Stephen’s, for cerain views about the way in which the properties (or predicates) explain could easily lead to the slogan that those properties aren’t ESSENTIALLY causal-explanatory, or maybe even to the slogan that in some sense they aren’t genuinely causal-explanatory at all. But the slogans by themselves aren’t very illuminating; they are likely to be optional ways of putting the underlying views about the explanatory role of the properties; and they are likely to disguise differences over what exactly the ("inessential" or "ungenuine") explanatory role of the properties is. Better then to make that the explicit focus.
But someone might say: "Whadya mean WHAT KIND OF causal explanatory role? Propositional attitudes are appealed to in true ‘because’ statements. What else is there to say?"
To see that there is something more to say, consider a different kind of property: physical magnitudes. Why did the quarterback ricochet backwards after colliding head on with the defensive back? Because while the quarterback was going 15 miles an hour and the defensive back only 10, still the quarterback had a mass of only 200 pounds and the defensive back 400, and 15×200 is less than 10×400.
That’s a perfectly good causal explanation by normal standards. But someone might be puzzled by it. For it invokes relations (speed in miles per hour, mass in pounds) between physical objects and numbers. If we had to view this on the model of the way in which relations to physical objects typically enter into explanations, I think the puzzlement would be legitimate. Consider the different question: why does a rock on the moon weigh less than a physically identical rock on the earth? Here the answer turns on the relations between the two rocks and the earth and the moon. And in this case, the relevance of those relations must be explained in terms of causal interactions among the related objects (e.g. the forces between rock A and the moon and between rock B and the earth). But clearly nothing like that obtains between the quarterback and the numbers 15 and 200. We need a different model of how the explanation in terms of numbers works than the model that is appropriate in the simplest cases when only physical objects are involved.
And of course we have such a model (the outlines of which were provided by various mathematicians around 1900). The basic idea is that relations between objects and numbers are derivative from more fundamental relations, either between physical objects themselves or between physical properties. In the physical object version, the fundamental relations might include such things as the part of relation and the sameness-of-mass relation. We then represent the masses of the objects by nonnegative real numbers: this involves assigning numbers to objects in a way that reflects these basic relations, in that certain representational constraints are met; e.g.
Such conventions on representation, together with the intrinsic facts about the quarterback and the defensive back (and their parts), are enough to constrain that the mass-number assigned to the defensive back will be twice that assigned to the quarterback. (A third constraint, e.g. on the number assigned to a certain object, is needed to "set the pound scale" so that the numbers in question will be 400 and 200.)
This demystifies the relation between the quarterback and the number 200. It also suggests a sense in which the relation between quarterbacks and numbers isn’t explanatorily basic. Let’s assume–this goes a fair bit beyond what I’ve said, but I think it’s plausible–that we could write the whole explanation of why the quarterback ricocheted backwards in terms of the fundamental mass-relations and fundamental speed-relations, without invoking numbers. In that case at least, it would be natural to say that the explanation above in terms of numbers, while perfectly good, isn’t basic; rather, it is epiphenomenal on the fundamental explanation, which involves physical objects only. Citing the number 200 as the quarterback’s mass is simply a quick way to allude to the mass-properties that are ultimately of physical relevance. In a sense, numbers don’t enter ESSENTIALLY into the explanation of the quarterback going backwards; but this slogan isn’t very helpful, what’s helpful is the positive account of how they do enter into the explanation.
The above was offered simply as an example of an illuminating account of a kind of causal-explanatory role that differs from the simple straightforward sort of causal explanatory role that relational properties often have. I certainly didn’t mean to suggest that intentional properties have the kind of explanatory role that numerical properties have. That has sometimes been claimed, but it seems to me to be completely wrong. The reason it is wrong is that the objects of intentional relations are representational in a way that numbers aren’t. ‘Has mass 200 pounds’ is, in a sense, representational: it represents the mass of the quarterback. In a similar way, perhaps, ‘believes that Malcolm X was the brother of Pope Pius X’ represents the state of the person who has that belief. But in addition, it represents the world as being in a certain way, and this last crucial aspect has no analog in the case of ‘has mass 200’. The analogy to numbers leaves out the most crucial feature of intentionality.
[Another way to put the matter: If intentional relations were like magnitude relations, then we could use "pure" abstract objects–maybe numbers, maybe more structurally complicated objects, but objects that are "purely" abstract in that they are not in any way constructed out of physical objects and physical relations–as the objects of all intentional attitudes. But (barring irrelevant coding tricks) we obviously can’t.]
Of course, it could well be the case that an account of how intentional relations do explain will license conclusions analogous to those that are licensed by the account of how magnitude relations explain. I’m inclined to think, for instance, that a proper account of explanations involving intentional relations would make such explanations "epiphenomenal" in something like the way that explanations involving magnitude relations are. Indeed, I’m inclined to think that some of the arguments that Schiffer discussed under the guise of arguments for skepticism about the explanatory role of intentional relations can be turned into arguments that their explanatory role is in some sense epiphenomenal. (This is not part of a view that explanations in terms of functional properties are "epiphenomenal"; that seems dubious. But intentional properties have crucial features that functional properties in general do not share, as Schiffer points out.)
Three other points. First, in several places Schiffer asks why someone might think that there are special problems about intentional explanations that aren’t shared by all special science explanations, e.g. explanations in a computational psychology; why isn’t it only explanations in basic physics that are free of the alleged problems? I think that part of the answer is that typical explanations in computational psychology aren’t epiphenomenal–or at any rate, are epiphenomenal only in the way that explanations that appeal to ordinary pure abstract objects like numbers are epiphenomenal. As I’ve mentioned, propositions or internal sentences represent the world in a way that numbers don’t, and what requires a special account is how this peculiarly representational aspect of them is used in explanation.
Second, Schiffer gives a positive account of the causal relevance of intentional properties that seems to me rather weak on its most obvious interpretation. Basically, the causal relevance of Joan’s desiring to end up in San Francisco to her being on the Frisco-bound plane consists in
But at least on the most ordinary construal of counterfactuals, this does not suffice for causation: we all know that parental phenotypes aren’t causes of the phenotypes of the offspring, but rather, have the parental genotypes as their common cause; still, most of us are happy to say that Joan wouldn’t have had brown eyes if both her parents had had blue eyes. The ordinary use of counterfactuals is a much poorer test of causation than Schiffer’s discussion suggests. (It may be adequate as a test of acceptable ‘because’ statements, but we all know that not all ‘because’ statements purport to state a relation of cause to effect.) Having said this, I reiterate that it does seem perfectly appropriate to say things like ‘A cause of the quarterback’s being knocked backward was that his mass in grams was only 200’ and ‘A cause of her being on the plane was her standing in the desire relation to the proposition that she end up in SF’. But some of the tenor of Schiffer’s final discussion (and his earlier opposition to Jackson and Pettit) seems to suggest that the counterfactual test is supposed to show a kind of causation incompatible with anything epiphenomenal being involved; and I think it is clear that if this is his intention the test fails to satisfy it.
Third, I’m inclined to think that there is more of a connection between causal explanation and laws than Stephen says, though I must admit to a lack of confidence about what it might be. One idea might be that a decent causal explanation must be decomposable into causal links, each of which is believed to be underwritten by some rough law (perhaps statistical, ceteris paribus, etc.). That avoids the implausibility of requiring a law corresponding to the entire causal claim. For a biological explanation, it needn’t be required that the laws underwriting all or even any of the components are biological; laws of physics or chemistry would do. (Of course, this requires a description of the biological entity, e.g. the cell, in physical and chemical terms.) Analogously in psychology. But for intentional psychology there is a prima facie problem: if, as Schiffer assumes, it involves relations to propositions, then since no laws of lower level disciplines like physics or chemistry involve propositions, it seems we need intentional laws to underwrite some of the links in the explanation. So if there are no essentially intentional laws–that is, if any intentional law can be shown to result from a computational law by gratuitously associating intentional components with the computational entities–then that would seem to give grounds for saying that no explanation is ESSENTIALLY intentional. Of course, there are many ways this conclusion could be avoided, but many of them involve pointing to features of intentional explanation not shared by other kinds of explanation; so again, the idea that something special needs to be said about the explanatory role of intentional properties is by no means silly.
There’s much more that could be said about the topic of the explanatory relevance of content. One important issue that Stephen mentions is the need to explain why, if we can explain something in intentional terms and also give a complete explanation of it in non-intentional terms, one explanation doesn’t exclude the other; or, what is almost the same thing, we need to explain how the explanations "mesh". I applaud Stephen’s remark that we can’t just say that the former supervenes on the latter, but must say something to explain the supervenience. (I also applaud the quote from his earlier self on Moore, and wasn’t clear how what he said afterward was supposed to undermine it.) I expect that several different attempts to explain this supervenience will be offered by our contributors to the seminar. (It’s too bad the number analogy isn’t better: I’m not clear whether the story sketched above should be regarded as providing a "reduction" of having mass 200 pounds to the underlying mass-relations among objects, but it certainly explains the supervenience of the former on the latter. We should hope to do as well in the intentionality case.)
Another topic that I hope some contributors will discuss is why in intentional explanations we insist on using sentences that we understand in ‘that’ clauses. Suppose we believe that every (context free etc.) sentence in every language expresses a proposition. If we know that ‘Der Schnee ist weiss’ is such a sentence, it seems that we ought to be able to extend the normal use of ‘that’ clauses to ‘She believes that der Schnee ist weiss’. This ought to be completely intelligible to us even if we don’t understand ‘Der Schnee ist weiss’, for it ought to amount to saying that she stands in the belief relation to the proposition that ‘Der Schnee ist weiss’ expresses. So it would seem that for any explanation of her behavior in terms of a belief that snow is white, an explanation in terms of a belief that der Schnee ist weiss would be equally good–again, even if we didn’t know that ‘Der Schnee ist weiss’ means that snow is white. Of course that’s absurd, but why? (Schiffer’s presentation contains some remarks on the connection of explanation to prediction that may bear on this little puzzle.) I’m inclined to think that reflecting on this puzzle will lead to the realization of a peculiarly first-person aspect of intentional ascriptions that some current theories don’t do justice to. I hope that this will come up in some of the presentations in the seminar.
 There may seem to be an obvious way to avoid this: simply say that what’s wrong with the "explanation" in terms of believing that der Schnee ist weiss is that we don’t know which proposition she believes. But in a sense we do know which proposition she believes: we know that it’s the proposition that der Schnee ist weiss. What we don’t know is which English sentence this corresponds to.