Can Rationality Be Naturalistically Explained? Jeffrey Dunn Abstract: Dan Chiappe and John Vervaeke (1997) conclude their article, Fodor, Cherniak and the Naturalization of Rationality, with an argument for the claim that we could never naturalistically explain rationality. I first clarify this argument and the conclusion that it is meant to establish. I then argue that once clarified, the argument fails. Introduction Dan Chiappe and John Vervaeke conclude their article, Fodor, Cherniak and the Naturalization of Rationality, with an argument for the claim that we are epistemically bounded from naturalistically explaining rationality. There are different things that one might be attempting to do when explaining human rationality. One might be trying to explain what a rational set of beliefs or a rational action is like. Such explanations will take the form of substantive theories of rational belief and action, and are usually formulated by decision theorists and logicians. Alternatively, one might be trying to give an explanatory story about how the human brain evolved to its present state. Such explanations are scientific and will be provided (if they are provided) by evolutionary biology. Either kind of explanation seems amenable to the naturalist: a naturalist could, perhaps, offer an instrumental account of what it is that makes these beliefs or actions rational,(1) and could simply adopt the scientific explanation of how the brain evolved. When trying to explain human rationality, however, one might be after something slightly different than either of these: one might want to explain how it is that humans
manage to be rational. Thus, one is after not just a scientific theory of how the brain evolved---an explanation of how the brain got to be like it is, is distinct from an explanation of how humans manage to be rational, even if we humans are rational in virtue of our brains being how they are.(2) Nor would one be after a substantive theory of rational beliefs and actions---such a theory purports to tell us what a rational belief set is, but now how we manage to realize this. Instead, one might be after an explanation of how it is that we manage to be rational, when we are. It seems to be this kind of explanation that Chiappe & Vervaeke think the naturalist could not possibly give. They write: we will give reasons for believing that we cannot provide an explanation of how rationality can come to exist from purely natural processes (p. 811). In this paper I will evaluate their argument for this claim. Before doing so, however, let me briefly note why such an argument is important. Above I said that we may get a substantive account of rationality from logic and decision theory. But one may be unhappy with this. One may think that a large part of rationality consists not only in the resulting beliefs and actions, but in the processes and procedures that bring about rational behavior and beliefs. If so, then explaining this is important if we are to understand human rationality. Thus, if it is impossible to give a certain kind of naturalistic explanation of rationality, then a naturalistic theory of this aspect of rationality is impossible. The Argument Chiappe & Vervaeke tell us that they are specifically concerned with methodological naturalism. They state the view as follows: Methodological naturalism claims that we 2
should explain the mind using processes that do not presuppose rationality (p. 812). As stated, this only commits the methodological naturalist to a claim about our obligations. But the methodological naturalist is plausibly committed to the claim that we should and can explain the mind using processes that do not presuppose rationality. Since Chiappe & Vervaeke are only concerned with the possibility of so-doing, we can initially state methodological naturalism as the following: (MN) It is possible that we scientifically explain the mind using processes that do not presuppose rationality.(3) Since we are specifically concerned with giving a naturalistic explanation of rationality, it will be helpful to have a more specific instance of MN: (MN-Rat) It is possible that we scientifically explain rationality using processes that do not presuppose rationality. The conclusion of Chiappe & Vervaeke s argument is that there can be no naturalistic explanation of rationality. This corresponds to a denial of MN-Rat. This denial is to follow from considerations about methodological naturalism, together with claims about scientific explanation made by both Carl Hempel (1988) and Hilary Putnam (1985). Hempel and Putnam both argue that scientific theories, when they explain, presuppose rationality. Given this, the basic argument given by Chiappe & Vervaeke is: 1. If MN-Rat is true, then it is possible to scientifically explain rationality using processes that do not presuppose rationality. 2. When scientific theories explain, they presuppose rationality. 3. Thus, MN-Rat is not true. 3
The conclusion establishes what Chiappe & Vervaeke want: that there can be no naturalistically acceptable explanation of rationality. Premise 1 is true because of their definition of methodological naturalism. Premise 2 requires a bit more defense. The idea is that scientific theories do not explain on their own, only theories which presuppose relevance in some way. They write: theories by themselves do not explain anything. It is only theories plus their implicit assumptions of relevance that explain events (p. 815).(4) They then note that one aspect of human rationality is the ability to judge relevance, from which we get premise 2.(5) Premise 1 Before evaluating this argument we must clarify a few things. Consider, again, the statement of methodological naturalism: (MN-Rat) It is possible that we scientifically explain rationality using processes that do not presuppose rationality. Although this statement of the view allows us to easily formulate the argument, it is doubly ambiguous. First: does it claim that when giving explanations we must not use processes that presuppose rationality, or that explanations must not appeal to processes that presuppose rationality? Second: what does it mean to say that a process presupposes rationality? The answer to these two questions are critical to understanding the argument. In giving their account of methodological naturalism, Chiappe & Vervaeke appeal to Gary Hatfield s (1990) account. Hatfield tells us: naturalistic explanation requires more than simply speaking of laws of thought as any theorist of the soul might; the naturalist must reduce the activities 4
of the mind to processes that do not include irreducibly rational or judgmental operations. Reason and judgment must themselves be explained by appeal to laws that do not include bare cognitive powers (p. 25)(6) It appears, then, that we have an answer to our second question. A process presupposes rationality when no matter how we reduce it to other processes, at least one unreduced rational or judgmental process remains. This is further supported by looking at one of Chiappe & Vervaeke s own examples. Christopher Cherniak, according to Chiappe & Vervaeke, attempts to (partially) naturalistically explain how humans are able to realize rational thought. Part of his explanation appeals to the fact that humans store different information in different mental compartments, which are sorted by relevance. Chiappe & Vervaeke object: Cherniak has not explained how the compartments are created the process of organizing information into compartments itself requires the skill [rationality] to be explained (p. 810). The problem with Cherniak s explanation is that it appeals to a process that itself includes an irreducibly judgmental operation: the ability sort information by relevance. In Chiappe & Vervaeke s terminology, his explanation appeals to a process that presupposes rationality. With our new understanding of the phrase presupposes rationality, our first question is: does methodological naturalism claim that when we give explanations we must not use processes that include irreducibly rational or judgmental operations, or that explanations must not appeal to such processes? Methodological naturalism, of course, is a methodological thesis. It tells us what explanations should look like. But it hardly matters to such a thesis what procedures we use when we give such explanations. The 5
real claim of the methodological naturalist is that the explanans shouldn t appeal to processes that include irreducibly rational or judgmental operations in explaining the explanandum. It would be no objection to MN-Rat to show that we must use some process which is irreducibly rational when giving explanations, so long as we never need to appeal to this process to explain rationality. Given this, we are in a position to state a more precise version of MN-Rat: (MN-Rat) It is possible, when giving a scientific explanation of rationality, that the explanans appeal only to processes that do not include irreducibly rational or judgmental operations. The original argument, then, will need to be reworked. Premise 2 says that scientific theories presuppose rationality, while premise 1 says that we can explain rationality only by appealing to certain kinds of processes. Thus, we must either add further premises or rework premise 2. On the face of it, there isn t much that can be done to shore up the argument. To see why, first consider premise 2. For this premise to be true, the phrase presuppose rationality must have something to do with the way in which theories get applied during the act of explanation. This is made clear, when Chiappe & Vervaeke argue for the truth of premise 2. They tell us: theoretical statements are expected to hold only under relevant conditions, which conditions are not determined by the explicit content of theories Theories are not algorithms that determine there own conditions of application. Instead, they presuppose a rational homunculus whose judgements of relevance can 6
complete theories by determining the relevant conditions for their application. (p. 814) Premise 2 is concerned with the act of explanation. However, we ve just seen that premise 1 is concerned with what is appealed to in the content of the explanation. From the fact that it is possible to explain something without appealing to irreducibly rational processes, it doesn t follow that we can explain that thing without using rational processes, and applying theories in a generally intelligent way. Thus, as it stands, their argument does not support the desired conclusion. Rethinking the Argument Chiappe & Vervaeke are committed to the claim that there is something especially pernicious about purportedly explanatory naturalistic theories of rationality in contrast to other theories in the natural sciences (see, for instance, p. 817). They make a point of emphasizing that any explanation of rationality will be objectionably circular, in a way that other explanations will not be. They write: explanations of how a phenomenon comes to be must not postulate an ontology that is itself constituted by the phenomenon to be explained (p. 812). For example, you can t explain solidity by postulating solidity. So, the worry is that there is no explanation that we could give of our own rationality which would be non-circular. If we are careful in spelling out how scientific explanations are to presuppose rationality (that is, premise 2), I think we will be able to get a better picture of the objectionable circularity that Chiappe & Vervaeke are worried about and thus a better representation of their argument. 7
First, according to Chiappe & Vervaeke, a circular explanation of x is one that postulates an ontology of x-ness. To see how a naturalistic explanation of rationality might postulate an ontology of rationality, we need to say something more specific about how scientific explanations are to presuppose rationality. Chiappe & Vervaeke focus not on rationality, in general, but rather one aspect of it: the ability to judge relevance. Since the ability to judge relevance is part of rationality, it is acceptable to focus on this. Their specific claim is that scientific explanations presuppose relevance. But how is it that scientific explanations do this? In a typical deductive-nomological (D-N) explanation,(7) we cite the relevant law, then cite the fact that the case under consideration falls under the law. If this entails the event in question, then we have explained the event.(8) Consider, then, what a simple D-N explanation for some event, E, might look like: 1. It is a law that if A, then E. 2. It was the case that A. 3. Thus, it was the case that E. The claim made by both Hempel and Putnam (and endorsed by Chiappe & Vervaeke), that scientific explanations presuppose relevance, can be seen as modifying the D-N explanation. They propose that explanations in science really look like this: 1. (In relevant circumstances) it is a law that if A, then E. 2. It was the case that A. 3. The circumstances were relevant. 4. Thus, it was the case that E. 8
Given this, any scientific explanation will somewhere appeal to a law that has the form of the law in 1 and a fact like the one mentioned in 3.(9) We are now in a position to see how an explanation of our ability to judge relevance might be unacceptably circular. In virtue of the law in 1 and the fact in line 3, scientific explanations postulate an ontology of relevance. More perspicuously: the law holds only in certain circumstances, circumstances that have the property of relevancy. But this means that any explanation appealing to laws like this postulates that there are these things: circumstances that have the property of relevancy. If, then, we re scientifically explaining our ability to judge relevance, the explanation itself appeals to the property of relevancy. This, one could claim, makes the explanation circular. However persuasive this line of thought is, it can t be right. Remember that according to Chiappe & Vervaeke, a circular explanation of x is one that postulates an ontology of x-ness. Thus, a circular explanation of our ability to judge relevance would be one that appeals to an ontology which includes the property ability-to-judge-relevance. But this is not what happens in scientific explanations of our ability to judge relevance. Instead, the explanation of our ability to judge relevance appeals to an ontology of relevancy. In general, however, this does not make for circular explanations. For example, an explanation of our ability to judge spiciness (of food) is not circular because it appeals to an ontology that includes the property of spiciness. So, is there anything objectionably circular about explaining our ability to judge relevance? An explanation of our ability to judge xness obviously postulates the existence of xness (otherwise, there is no such ability to explain). But, as we ve seen, this is not yet circular. One thought is that such an explanation is objectionable because the property of 9
relevancy is understood solely in terms of our ability to judge relevance. Thus, an explanation that appeals to relevancy tacitly appeals to our ability to judge relevance. But this is implausible: we don t make things relevant to each other by judging them to be so. An alternative way of pushing the idea that this kind of circularity is objectionable is by noting that, according to Hempel and Putnam, whether or not some law, L, is relevant to some situation, S, is not a feature of L and S themselves. Rather, as Chiappe & Vervaeke tell us, it requires a rational homunculus whose judgements of relevance can complete theories by determining the relevant conditions for their application (p. 814). On this view, relevance is a relation that holds between rationally interpreted laws and situations. Thus, what explains our ability to judge relevance is a rationallyinterpreted law, L, and the appropriate conditions holding. This, however, is not a problem. If we grant the Hempel/Putnam point, then for L to explain it must be rationally-interpreted. But from this it does not follow that such an explanation is objectionably circular. The rationally interpreted laws explain our ability to judge relevance. Again, we must be mindful of the distinction between using our ability to judge relevance and appealing to our ability to judge relevance. We use our rational ability to interpret laws in the kind of way that allows us to explain rationality. There is nothing circular here. An alternative thought about what is objectionable is that an explanation of our ability to judge xness requires us to explain xness.(10) That is, if you re going to explain our ability to judge spiciness, then you must be able to explain what spiciness is. And, the thought goes, you can t do this non-circularly when it comes to our ability to judge relevance. For example, perhaps we explain spiciness as: the property that things have, 10
such that in relevant situations, when things have that property they cause a certain response. If the property of relevancy is explained in an analogous way, then we explain relevancy as: the property that things have, such that in relevant circumstances, things with that property meet some condition. Of course, this kind of explanation of the property of relevancy is circular. Nevertheless, there is nothing to worry about here for the naturalist. First of all, we have shifted senses of explanation. No longer are we giving causal explanations, but rather trying to explain properties, in some different sense of explanation. Second, and more importantly, we should reject the claim that to scientifically explain our ability to judge xness, we must be able to explain in this new sense of explanation the property of xness. For example, it seems possible that we explain our ability to judge distance, without being able to explain what it is for two things to be separated by some distance. So long as we generally understand the notion of distance, this will suffice. We can simply take distance as a primitive, unexplained notion. Perhaps it would be disappointing that we couldn t wholly elucidate the notion of distance, but it would not impugn our explanation of our ability to judge distance. After all, explanations must come to an end somewhere. To make these points clearer, let us consider an example. We can construct cases where an explanation appeals to a law with certain assumptions about relevance that yield bad explanations of our ability to judge relevance. For example, perhaps we want to explain how humans judge whether or not a bit of information is relevant to restaurant situations. Imagine that the proffered explanation appeals to different mental compartments, one of which is the restaurant compartment, where information relevant 11
to restaurant situations is stored. The explanation for how information gets sorted between these different compartments, appeals to the following law: L: In relevant conditions, the agent stores received information in the restaurant compartment. Let us suppose that relevant conditions here just means: situations where the agent receives restaurant-relevant information. If this is the case, then of course, any explanation appealing to L of how the agent manages to judge which information is relevant to restaurant situations is unsatisfying. By understanding relevant conditions in this way we do not seem to gain much insight into how this sorting is to proceed. Nevertheless, it is important to note that such an explanation is not circular. We use our rational ability to judge situations where the agent receives restaurant-relevant information to fill in the content of the law. This is then put to use in explaining how the agent manages to make a certain judgment about relevance. Of course, there is something unsatisfying about this explanation. We d prefer to know something more about how this sorting proceeds. That is, since we don t think that there is a little rational homunculus inside each person s head, we d like to know what features of these relevant situations are being tracked so that they are sorted into the proper mental compartment. But there seems to be no reason that a (slightly) more realistic kind of law couldn t be both put to use in a more satisfying explanation and be naturalistically unobjectionable. Consider a law of the following form: L : [In relevant situations], if conditions C are met, the agent s sorting mechanism stores received information in the agent s restaurant compartment. 12
In L the prefix in relevant situations is merely implicit in the statement of the law, as Hempel and Putnam urge. Specification of conditions C is what will really give such a law explanatory power, and there is no reason that these conditions must be expressed by appealing to restaurant-relevant situations. For instance, conditions C could be completely qualitatively expressed. The relevancy prefix may only serve to rule out really off-the-wall situations, for example, situations when the agent suddenly undergoes massive brain trauma, or is under alien mind-control, or where the agent has a strong desire to act erratically, etc. But such an appeal to relevancy does not seem to render unsatisfying the explanation of our ability to judge whether or not a bit of information is relevant to restaurant situations. Thus, although we certainly can construct circularsounding explanations by appealing to dubious laws, it isn t required by naturalism that we do so. Conclusion We began by looking at Chiappe & Vervaeke s interesting argument to the claim that we can never naturalistically explain rationality. The first gloss of this argument was seen to be flawed for a very fundamental reason: part of the argument concerned itself with the act of explanation, while the other part concerned itself with the content of an explanation. The second gloss of the argument remedied this problem, focusing entirely on the content of purported naturalistic explanations of rationality. The alleged problem with such explanations was that they are circular: such explanations try to explain our ability to judge relevancy while appealing to our ability to judge relevancy. However, we saw that this was mistaken, too. Explanations of our ability to judge relevancy may 13
appeal to relevancy, but need not appeal to our ability to judge relevancy. Thus, there is no circularity in such explanations. Accordingly, Chiappe & Vervaeke s argument does not establish the desired conclusion. One might think, however, that though the methodological naturalist need not be upset about this situation, this argument gives the metaphysical naturalist a reason for concern. On one way of understanding the view, metaphysical naturalism does not want to admit an ontology that includes the (irreducible) property of relevancy. But if that is the problem, it is hard to see how the problem really has anything to do with explanations of rationality. Indeed, if we grant the Hempel/Putnam point that all scientific explanations have assumptions of relevance, any scientific explanation is enough to show that naturalism is a hopeless project. Whatever the merits of such an argument, it is not the argument that Chiappe & Vervaeke give. They are clear that there is a special problem for the naturalist when it comes to scientific explanations of rationality. We ve seen that such an argument fails. Notes 1 That is, an account of the following form: A belief/action is rational if and only if it is instrumentally useful for attaining the agents desires. One could then spell out instrumentally useful in any number of ways. Whether or not such an account is plausible is another story. But the project does not seem to be self-defeating in any kind of obvious way. 14
2 Compare: an explanation of how my leg moved as it did is different than an explanation of how I managed to kick the ball, even if my leg moving as it did was my kicking of the ball. Explanations are intensional, not extensional. 3 Since the methodological naturalist takes scientific explanation as his guide, the kinds of explanations that the methodological naturalist is concerned with are scientific explanations. 4 In addition: theoretical statements are expected to hold only under relevant conditions, which conditions are not determined by the explicit content of theories they presuppose a rational homunculus whose judgements of relevance can complete theories by determining the relevant conditions for their application (p. 814). 5 It is important to note that the aspect of rationality that will concern us here is the human ability to judge relevance (p. 812). 6 This account of methodological naturalism is Hatfield s account of what methodological naturalism meant to early modern theorists such as Descartes and Hume. It is an open question whether or not this is the best way to formulate methodological naturalism today. Nevertheless, I will continue to presuppose this account of methodological naturalism since it is the one appealed to by Chiappe & Vervaeke. 7 See Hempel & Oppenheim (1948). 15
8 There are well-known problems with D-N explanations, which shouldn t matter for this illustrative purpose. 9 Note that the same will be true if we adopt a causal account of explanation. The relevant causal statements will, as Chiappe & Vervaeke note, appeal to relevant circumstances in a similar kind of way. Thus, moving to, say, a causal account of explanation merely makes the problem harder to state, it doesn t make it go away. One could, at this point, claim that causal explanations don t admit this kind of assumption of relevance. But I will continue to assume that they do, since I believe that even granting this assumption, there is a flaw in the argument. 10 If one takes this suggestion, the argument might look like the following: 1. If MN-Rat is true, then it is possible to scientifically explain our ability to judge relevance. 2. If it is possible to scientifically explain our ability to judge relevance, then you can scientifically explain what it is to be relevant independent of our ability to judge relevance. 3. It is not the case that you can explain what it is to be relevant, independent of our ability to judge relevance (because all explanations presuppose our ability to judge relevance). 4. Thus, MN-Rat is false. As I explain shortly, premise 2 should be rejected. 16
References Chiappe, Dan & Vervaeke, John. (1997). Fodor, Cherniak and the Naturalization of Rationality. Theory & Psychology, 7, 799-821. Hatfield, Gary. (1990). The Natural and the Normative: Theories of spatial perception from Kant to Helmholtz. Cambridge, MA: MIT Press. Hempel, C. G. & Oppenheim, Paul. (1948). Studies in the logic of explanation. Philosophy of Science, 15, 135-175. Hempel, C. G. (1988). Provisos: A problem concerning the inferential function of scientific theories. In A. Grunbaum & W. C. Salmon (Eds.), The Limitations of Deductivism (pp. 19-36). Berkeley: University of California Press. Putnam, Hilary. (1985). Why reason can t be naturalized. in Hilary Putnam, Realism and Reason: Philosophical papers volume 3 (pp. 229-247). New York: Cambridge University Press. Putnam, Hilary. (1993). Renewing Philosophy. Cambridge, MA: Harvard University Press. 17