From Philosophical
Studies 106, 3 (2001): 227-258.
Reprinted with kind permission of Kluwer Academic
Publishers.
7
Against Explanationist Skepticism Regarding Philosophical Institutions
1 Introduction
Contemporary analytic philosophy is based upon intuitions. Intuitions serve as the primary evidence in the analysis of knowledge, justified belief, right and wrong action, explanation, rational action, intentionality, consciousness and a host of other properties of philosophical interest.[1] Theories or analyses of the properties in question are attacked and defended largely on the basis of their ability to capture intuitive judgements. The force of a putative counterexample, for instance, entirely depends upon an intuition contrary to the claims of a given theory. Consider, for example, the appeals to intuitions about reliable clairvoyants, fake barns, flagpoles and their shadows, monolingual English speakers manipulating Chinese characters, experience machines, and arctic explorers making promises that populate contemporary philosophical discussions.
Explicit concern about such a reliance on intuitions has increased of late and a growing number of philosophers appear doubtful of the epistemic probity of appeals to intuitions.[2] This paper responds to the arguments of some contemporary philosophers who are skeptical about the evidential worth of intuitions; skeptical, that is, that intuitions can properly be treated as evidence for anything other than our contingent psychological constitution. In Section 2, I outline and analyze arguments offered by Gilbert Harman and Alvin Goldman in support of skepticism about the evidential worth of intuitions in moral theory and metaphysics. Harman’s explanationist argument against moral intuitions is well-known. However, many philosophers seem unperturbed by his argument, believing, perhaps, that the argument tells only against moral intuitions. Goldman’s more recent argument is probably less well known, but it demonstrates clearly how Harman’s argument can be generalized to undermine the use of intuitions in other domains of philosophical inquiry. Indeed, I conclude Section 2 by showing how a similar argument could be mounted against the use of intuitions in epistemology. In Section 3, I argue that Nicholas Sturgeon’s influential response to Harman’s argument is unsuccessful and, furthermore, insufficiently general to serve as a reply to other arguments of the same form. However, in Sections 4 and 5, I show that the main premise of the skeptical arguments is epistemically self-defeating and so those arguments cannot succeed in undermining the evidential worth of philosophical intuitions.
2 Explanationist Skepticism
In this section, I present a number of skeptical arguments and show that they share a common structure. They all advance an explanationist principle of justified belief and utilize it to declare intuitions of various kinds evidentially worthless for any sort of inquiry other than empirical psychology. I begin by demonstrating that Gilbert Harman’s skeptical arguments regarding moral intuitions (1977; 1986) and Alvin Goldman’s skeptical arguments regarding metaphysical intuitions (1987; 1989b; 1992b) are informatively viewed as variations on a single argument schema. These philosophers argue that we are not justified in using intuitions as evidence in philosophical inquiry because the best explanation of the occurrence of such intuitions is not one that invokes their truth. As a means of introducing the fundamental problem with such arguments, a problem I develop in detail in later sections of the paper, I conclude this section by presenting a parallel skeptical argument regarding epistemic intuitions.[3]
2.1 Harman on Moral Intuitions
In the first chapter of The Nature of Morality (1977), Gilbert Harman presents a widely discussed argument for the conclusion that moral theory suffers from severe epistemic difficulties. His central claim is that moral theories are epistemically problematic because they cannot be tested and confirmed in the way that scientific theories can. While he admits that we may ‘test’ general moral principles against our intuitive reactions to particular actual and hypothetical situations, he argues that we are not thereby testing our moral theories or principles against the world. Instead, we are merely testing them against our ‘moral sensibility’ or against our tacitly held moral views.
Initially, Harman is concerned with the possibility of moral observation. Taking observation to be our basic mode of epistemic access to the world, Harman questions whether we can test our moral theories against something more than our intuitive reactions to hypothetical cases. Are there, he asks, moral observations against which we might test our moral theories? Claiming that all perception is theory-laden, Harman goes on to present an analysis of observation on which we clearly do make moral observations. Observation, on this extremely liberal view, occurs whenever a belief is formed as a ‘fairly direct result’ of perceptual stimulation. Since such immediate perceptual beliefs may be about moral matters, he maintains that there are moral observations.[4]
In spite of the fact that we make moral observations in his liberal sense of ‘observation,’ Harman claims that there is an important difference between the use of observations in empirical science and the use of observations in moral inquiry which renders moral observations unable to provide evidence for moral theory. The alleged difference is that ‘you need to make assumptions about certain physical facts to explain the occurrence of the observations that support a scientific theory, but you do not seem to need to make assumptions about any moral facts to explain the occurrence’ of the moral observations (1977, p. 6, emphasis added). On Harman’s view, when a person makes an observation, that observation confirms a theory insofar as the best explanation of the fact that the observation occurred must invoke the truth of the theory. The confirming role of an observation, then, is assigned to the occurrence of the observational judgement and not simply to the content of the judgement. So, in Harman’s much-cited example, a physicist’s observation confirms a physical theory to the extent that it seems reasonable to explain the occurrence of the physicist’s observational judgement by referring to something about the world ‘over and above’ the assumptions we might make about his psychology.
Such cases, Harman alleges, stand in stark contrast to cases in which we make moral observations. When, for example, one makes a spontaneous moral judgement that it would be wrong for a doctor to kill an unconsenting healthy patient to save the lives of five other patients, or when one sees a group of children setting fire to a cat and one spontaneously judges that what they are doing is wrong, ‘an assumption about moral facts would seem to be totally irrelevant to the explanation of your making the judgement you make’ (1977, p. 7). All that we must assume in order to arrive at what appears to be an entirely adequate explanation of a person’s making a moral judgement is that the person making the judgement tacitly accepts a set of moral principles, a set of principles which she unconsciously applies to particular cases in order to arrive at an intuitive determination of their moral status.
The key argument seems to be that the
observation of an event can provide observational evidence for or against a scientific theory in the sense that the truth of the observation can be relevant to a reasonable explanation of why that observation was made. A moral observation does not seem, in the same sense, to be observational evidence for or against any moral theory, since the truth or falsity of the moral observation seems to be completely irrelevant to any reasonable explanation of why that observation was made. (1977, 7)
It is important to note (as does Lycan 1988b) the way in which Harman is here using the term ‘observation.’ He is quite clear that his sense of observation is a sense in which an observation that p is the event of your thinking that p. It is not simply p, the content of your observational judgement. Using Sellars’s (1968) helpful terminology, we should distinguish between intuitings (the act or event of intuiting that p) and intuiteds (the proposition one intuits). In a similar manner, we might distinguish between observings (the occurrence of an observational judgement) and observeds (the content of such a judgement) and, even more generally, between judgings and judgeds.
Harman explicitly concedes that moral principles could explain moral observations or intuitions in the latter sense. That is, they could explain moral observeds or intuiteds. If action X is wrong, then a general moral principle might well explain why it is wrong (by entailing the particular judgement when combined with the particular facts of the case). But such principles cannot be tested against the world as they are of no apparent help in explaining the fact that you made the judgement that you made—in explaining your observings or intuitings. So, while observations are evidence for empirical theories because the theories explain the occurrence of your observations, moral observations or intuitions are not evidence for moral theories in the same way since moral theories cannot contribute to the explanation of your judgings.[5]
Harman seems here (as elsewhere (1965; 1973)) to be offering a criterion of justified belief which holds that an observation is evidence for the theory that best explains its occurrence. Though Harman is not as explicit as one might like, the overall argument seems to appeal to something like the following general epistemological principle as its first premise:
[1] Aside from propositions describing the occurrence of her judgements,[6] S is justified in believing only those propositions which are part of the best explanation of S’s making the judgements that she makes.
Notice that this does not require that a moral proposition figure in the best explanation of specifically moral observings or intuitings in order for S to be justified in believing it. Instead, moral propositions (or any other) are observationally confirmed if they are required to explain any of our observings or intuitings.
Harman’s second premise seems to be the following:
[2] Moral propositions are not part of the best explanation of S’s making the judgements that she makes.
These premises jointly entail the following skeptical conclusion:
[3] S is not justified in believing any moral propositions.
2.2 Goldman on Metaphysical Intuitions
Alvin Goldman’s recent discussions of the use of intuitions in contemporary metaphysics (1987; 1989b; 1992b) seem to reveal a skepticism regarding the evidential worth of metaphysical intuitions for non-psychological purposes. This skepticism is apparently motivated by an explanationist argument quite similar to Harman’s.[7]
Goldman begins by appealing to a distinction he borrows from P. F. Strawson, a distinction between descriptive and revisionary (or prescriptive) metaphysics. Strawson held that ‘descriptive metaphysics is content to describe the actual structure of our thought about the world; revisionary metaphysics is concerned to produce a better structure’ (1959, p. 9). On Goldman’s account, descriptive metaphysics aims to describe our common sense or ‘folk’ ontology.[8] Moving beyond Strawson, Goldman also charges descriptive metaphysics with the task of scientifically explaining why we have the various metaphysical intuitions that we do. Revisionary or prescriptive metaphysics, on Goldman’s account, aims to ascertain the objective truth about the ultimate structure of reality, or, as he puts it, ‘it tries to identify the real constituents of the world, the ones to which we ought to be ontologically committed’ (1987, p. 528). In the service of this aim, prescriptive metaphysics is prepared to depart from our intuitions or our common-sense thinking about metaphysical matters, should such a departure prove warranted.
Goldman’s objection to the work of many contemporary metaphysicians is that they fail to notice the distinction between descriptive and prescriptive metaphysics and, consequently, fail to ask what the evidential role of our intuitions ought to be in each endeavor. Contemporary metaphysics relies quite heavily on appeals to intuition. In fact, it often seems as though a collection of intuitions is the only evidence to which metaphysicians are responsive in their theorizing. In Goldman’s view, this practice is seriously misguided if the aim of such metaphysicians is, as it certainly seems to be, to determine the objective truth about metaphysical matters. Intuitions, Goldman claims, are evidentially relevant if our focus is folk ontology—if we are doing common sense or descriptive metaphysics. If we want to ascertain what people believe regarding some metaphysical matter, or how they conceptualize metaphysical matters, or if we wish to determine what psychological mechanisms people contain, then obviously their judgements and intuitions are excellent evidence.
However, if we are engaged in genuine or prescriptive metaphysics—if we want to know about the structure of reality and not merely what the folk (including ourselves) tacitly believe about that structure—then it is much less clear how intuitions could provide us with good evidence. Goldman raises his central question in the following passage: ‘Why should we regard intuitions or feelings as reliable indicators of genuine metaphysical facts? ... What licenses the inference from these intuitions to the putative metaphysical facts? Why should intuitions be relied upon?’ (1992b, p. 57).[9]
The credibility of intuitions as evidence for metaphysical claims is undermined, Goldman maintains, if we can find good explanations of our various intuitions that do not invoke objective metaphysical facts. Goldman claims, then, that an adequate explanation of our intuitings that does not invoke facts of the relevant sort would be a good reason to think that such intuitings are not good evidence for their contents. Furthermore, he takes cognitive science to be particularly well placed to offer such deflationary explanations, as it is in the general business of offering scientific explanations of mental phenomena. Goldman’s psychologistic philosopher would use cognitive (and perhaps biological and social) science to ‘explain the phenomenological or judgemental events without positing objectively detected properties’ (1989b, p. 46). She would thereby, on Goldman’s view, undermine the use of metaphysical intuitions as evidence for non-psychological or objective metaphysical theories.
Goldman does not merely assert that such merely psychological explanations are possible. He offers some (admittedly quite schematic and speculative) examples of how cognitive science might explain various of our intuitions about the essentiality of origin, haecceitism, Sorites paradoxes, the identity of persons and objects over time, and the apparent ‘flow’ of time. In connection with intuitions appealed to in order to support theories of personal identity over time or of object identity over time, Goldman suggests that our intuitings can be explained in terms of ‘conceptual’ Gestalt rules regarding good continuation. The fact that we are easy victims of Sorites puzzles can, Goldman maintains, be explained by our disposition to create binary, ungraded categories and our disposition to co-classify objects that differ only minutely. Finally, the occurrence of intuitions appealed to by Saul Kripke (1980) to support an essentialist thesis about the origins of individual objects and by Robert Adams (1979) to support haecceitism,[10] might, Goldman avers, be plausibly explained by appeal to certain algorithms governing the mental representation of individuals or ‘our cognitive bookkeeping habits.’
After offering these and other examples of psychological explanations of our metaphysical intuitings, Goldman concludes that we have no good reason to think that intuitions are reliable indicators or ‘detectors’ of metaphysical facts because ‘[t]he availability of a purely psychological explanation of the intuitions ostensibly provides a systematic alternative to such detectivist explanations’ (1992b, p. 61, emphasis added). More precisely, Goldman holds that the explanationist argument succeeds if metaphysical facts are not needed in a complete explanation of our judgings. Nonetheless, this complication does not much affect the skeptical argument since it is less than clear how metaphysical facts of the sorts just mentioned could be any part of a scientific explanation of our psychological dispositions to make certain judgements.
Though Goldman does not as clearly endorse a general explanationist principle such as [1],[11] we may, I think, profitably treat his argument as having the same structure as Harman’s:
[1] Aside from propositions describing the occurrence of her judgements, S is justified in believing only those propositions which are part of the best explanation of S’s making the judgements that she makes.
[2′] Propositions about metaphysical topic X are not part of the best explanation of S’s making the judgements that she makes.
[3′] S is not justified in believing any propositions about metaphysical topic X.
2.3 Goldman* on Epistemic Intuitions
My final example of this kind of skeptical explanationist argument is derived from Alvin Goldman’s recent work on epistemology. This argument is not explicitly formulated by or endorsed by Goldman. Nonetheless, the similarity of Goldman’s epistemological and metaphysical projects, and the role of the scientific explanation of intuitings in each, certainly suggests the skeptical or deflationary argument I shall present. In light of the fact that Goldman does not offer the argument suggested by his work, I shall attribute the argument, not to Goldman, but to Goldman*. Consideration of this argument will allow us to begin to see why all arguments of this sort must fail.
In his recent work, Goldman has invoked a distinction between descriptive and prescriptive epistemology. This distinction is precisely analogous to the one drawn in metaphysics. Where the descriptive metaphysician’s task is to describe the metaphysical views of the folk and to scientifically explain why the folk embrace such an ontology, the task of the descriptive epistemologist is to describe and to scientifically explain the epistemic judgements of the folk. Where the prescriptive metaphysician seeks to improve (when warranted) upon our naive folk ontology, the prescriptive epistemologist seeks to improve (when warranted) upon our epistemic folkways.[12]
As in the case of metaphysics, Goldman attempts to provide a psychological explanation of intuitings rather than a standard philosophical theory accounting for the intuiteds. In his discussion of knowledge in ‘Psychology and Philosophical Analysis’ (1989a), Goldman characterizes his approach as ‘trying to give a plausible, parsimonious explanation of people’s (linguistic) intuitions.’ His goal is to utilize scientific psychology to explain why people have the intuitions about knowledge that they do—to demonstrate how certain psychological mechanisms might generate ‘the observed interpersonal uniformities in intuitive judgements of “knowledge”’ (148).
In treating the other major epistemological analysandum, Goldman’s new two-tiered reliabilism (1992c) continues in this psychologistic vein as it aims to explain why people judge that certain beliefs are justified or unjustified. His proposed explanation is that ‘the epistemic evaluator has a mentally stored set, or list, of cognitive virtues and vices.’ When presented with an actual case of belief acquisition or a description of a possible one, the evaluator (unconsciously) compares the process which produced the belief to the internally represented list. If the producing process matches (or is sufficiently similar to) a vice, then the evaluatee’s belief is classified as unjustified. If the process matches (or is sufficiently similar to) a virtue, then the evaluatee’s belief is classified as justified. Goldman claims that this new reliabilism accounts for the main alleged counterexamples to reliabilism. Again, however, it ‘accounts for’ them by offering a psychological explanation of why evaluators might produce the problematic (from the perspective of previous reliabilisms) judgements.
In both of these papers, Goldman’s focus is on offering psychological explanations of the occurrence of people’s intuitions.[13] Goldman does not himself offer a skeptical lesson regarding epistemic intuitions. Instead, he explicitly sets aside the questions of whether a given person in a particular case really has knowledge (1989a, p. 148) or is really justified in believing p (1994, p. 315). Still, it is clear that all the ingredients are present for a skeptical argument precisely parallel to the others we have already examined and consistency would seem to require its consideration. The use of intuitions in epistemology seems, prima facie, no different from the use of intuitions in moral theory or metaphysics. Furthermore, Goldman’s psychological discussions support an analogous second premise according to which our judgements (judgings) about justification and knowledge are best explained by something other than the truth of epistemic propositions. Thus, absent reason to suppose that there is a relevant difference between the domains, the following argument seems as well motivated as the arguments I’ve already outlined in this section:
[1] Aside from propositions describing the occurrence of her judgements, S is justified in believing only those propositions which are part of the best explanation of S’s making the judgements that she makes.
[2²] Epistemic propositions are not part of the best explanation of S’s making the judgements that she makes.
[3²] S is not justified in accepting any epistemic propositions.
The peculiarity of this conclusion is, I shall later show, indicative of the reason for the failure of all of the skeptical arguments so far considered.
I have, thus far, presented a series of contemporary arguments for skepticism about the evidential use of intuitions in various areas of philosophy. They all have the same structure and rely on the same epistemological main premise. The argument form is quite important to address in light of the widespread use of intuitions throughout philosophy noted at the beginning of this paper. Should the general argument succeed, the practice of using intuitions as evidence would be shown unjustified in any domain wherein the truth of the propositions intuited is unnecessary for the best explanation of our intuitings and this would, I believe, reduce much philosophical investigation to nothing more than armchair psychology.
Given the validity of this argument form, there are two
main routes available to those who would deny its soundness. The first is to
concede the explanatory requirement of the first premise and to argue that particular instances of the second
premise are false by demonstrating that propositions of the type in question are part of the best explanation of our
making various judgements. An example of this strategy will be considered and
criticized in the next section. The other option, and the one I shall pursue in
the remainder of the paper, is to attack premise [1], the premise common to all
the arguments. I shall do so by arguing that (i) we are not justified in
believing [1], (ii) [1] cannot be argued for in a manner consistent with its
content, and (iii) the explanationist cannot, given her other views, be
justified in maintaining belief in [1]. If I am right, then all of the
explanationist arguments fail.
1 Sturgeon’s Argument Against Premise [2]
The best example of the first strategy—that of accepting the explanatory requirement and defending the explanatory relevance of a set of propositions or properties—is found in Sturgeon’s (1984) rejoinder to Harman. According to Sturgeon, ‘Harman is quite wrong, not in thinking that the explanatory role of our beliefs is important to their justification, but in thinking that moral beliefs play no such role’ (1984, p. 53). In this section, I’ll argue that Sturgeon’s attempt to undermine the argument in this way fails.
The quote from Sturgeon just offered clearly suggests that Sturgeon seeks to undermine the case for premise [2] while accepting premise [1]. That this is Sturgeon’s purpose is further evident in his comment on Thomas Nagel’s response to the explanationist argument. Nagel had claimed that ‘it begs the question to assume that explanatory necessity is the test of reality in this area. The claim that certain reasons exist is a normative claim, not a claim about the best explanation of anything’ (1980, p. 114). Sturgeon’s assessment of Nagel’s rejoinder to the explanationist argument is that ‘this [i.e. Nagel’s] retreat will certainly make it more difficult to fit moral knowledge into anything like a causal theory of knowledge, which seems plausible for many other cases’ (1984, p. 55, emphasis added). Sturgeon’s apparent goal, then, is to show that moral propositions can meet the explanatory requirement and be fit into ‘something like’ a causal theory of knowledge.
Recall that [2], in Harman’s version of the general argument, claimed that moral propositions were explanatorily irrelevant to the explanation of the occurrence of our observations and intuitions. Sturgeon attacks this claim by proposing a sufficient condition of explanatory relevance and attempting to show that moral propositions or moral facts satisfy this criterion of explanatory relevance. Sturgeon’s proposed criterion of irrelevance is the following: ‘if a particular assumption is completely irrelevant to the explanation of a certain fact, then the fact would have obtained, and we could have explained it just as well, even if the assumption had been false’ (1984, p. 65). Since Sturgeon appears to be offering a necessary condition of explanatory irrelevance, and since a principle is relevant if and only if it is not irrelevant, we may take Sturgeon’s statement to advance (at least) the following sufficient condition of explanatory relevance:
[CT] A particular proposition P is relevant to the explanation of a fact Q IF éif P had been false, then it would not be the case that fact Q obtainedù.[14]
With this sufficient condition of explanatory relevance in hand, Sturgeon argues that many moral propositions are relevant to the explanation of our judgements. Consider Harman’s example in which, seeing a group of children pour gasoline on a cat and set fire to it, you spontaneously judge (or ‘observe’) that they have done something wrong. Sturgeon claims that the following counterfactual might be true: éif the children had not been doing something wrong, then you would not have thought that they were doing something wrongù. If so, then assuming [CT] is correct, the wrongness of their act is explanatorily relevant to your judging that their act was wrong.
Sturgeon arrives at this result by claiming that since what makes an act wrong (and he here assumes the existence of moral facts in order to determine what sort of explanatory role they would play if they existed) is its possession of various natural properties, the counterfactual situation relevant for the evaluation of the antecedent of the conditional éif the children were not doing something wrongù, is one in which the children’s action is sufficiently different in purely natural or non-moral terms as to make the action morally permissible. More precisely, if moral properties supervene on natural properties, if the possession of certain natural properties is metaphysically sufficient for the possession of certain moral properties, then in order for the children’s act to be morally permissible it must differ in some purely natural respect from the actual act. On this view, as Sturgeon claims, the question becomes: ‘Even if the children had been doing something else, something just [naturalistically] different enough not to be wrong, would you have taken them even so to be doing something wrong?’ (1984, 66). The answer to this question will obviously depend on the specific situation and the particular moral appraiser in question, but it seems fairly clear that the answer will often be ‘No.’ Thus, the moral fact of the children doing something wrong could be explanatorily relevant to the occurrence of a person’s judgement that they were doing something wrong.
Though I am inclined to think that [CT] has a number of unacceptable implications,[15] I shall here focus upon Sturgeon’s application of it to particular moral judgements. As it stands, [CT] is unacceptable on account of its implications regarding necessary truths. On one standard semantics for counterfactuals, counterfactuals with necessarily false antecedents (so-called ‘counterpossibles’) are uniformly evaluated as trivially true (Lewis 1973; Stalnaker 1968).[16] In conjunction with [CT] as a sufficient condition of relevance, this implies that any necessary truth is explanatorily relevant to any fact whatsoever. This is surely false and in the absence of a new semantics for counterpossibles it seems we must restrict the scope of [CT] to contingent truths. Let us call such an amended account [CT*].
[CT*] still allows us to say that Sturgeon’s original counterfactual involving the children is true because it is surely a contingent fact that the children are setting fire to a cat. Sturgeon is quite aware, however, that others will think the relevant counterfactual a different one. On this alternative, the children’s action is kept fixed and it is the moral status of the action which is varied in violation of the supervenience of the moral. The counterfactual would then be éif the children were not doing something wrong (not because they were doing something different, but because setting a cat on fire for fun was not wrong), then S would not believe that they were doing something wrongù. This counterfactual has (as Sturgeon admits) a necessarily false antecedent and so [CT*] will not apply to it. While Sturgeon claims this reading of the counterfactual is not ‘the natural one,’ it seems to be the relevant one if the supervenient moral properties of that particular action are to be shown explanatorily relevant to the observer’s judgement. This is clear when one focuses on the fact that the most basic moral judgement which one might make when confronted with the children is not that they are doing something wrong, but that what they are doing is wrong. (Indeed, this is how Sturgeon (following Harman) occasionally formulates the judgement.) One’s reaction is to the particular action in which the children are engaged. The natural counterfactual with respect to this judgment is éif what the children are doing were not wrong, then S would not believe that what they are doing is wrongù. This judgement, however, clearly holds the action constant and supposes a variation in its actual moral status and so is just the sort of counterpossible which cannot be treated by [CT*].
The failure of [CT*] to demonstrate explanatory relevance is even more clear in connection with the contents of intuitive judgements about hypothetical cases which serve as the primary evidence for moral philosophers because those judgements are more clearly judgements regarding necessary propositions. Consider, for example, the hospital case discussed by Harman in which we are asked for our intuition regarding the morality of a doctor’s sacrificing the life of an unconsenting patient with a minor illness in order to save the lives of five patients with life-threatening maladies. When considering this case, most of us have the intuition that such an action would be morally wrong. The relevant counterfactual seems to be éif the doctor’s action were not wrong, then S would not believe that the action was wrongù. However, in making our judgement about this hypothetical case, we are clearly forming a judgement regarding a particular case the nature of which is fixed by description. Given that we fully understand the case in question, little sense can be made of the antecedent of the conditional requiring that the doctor’s action in this (merely hypothetical) case have a different moral status because that would require that we hold the description of the case constant and vary its actual moral properties which would be a plain violation of supervenience.
Perhaps it will be claimed that the relevant counterfactual is one in which the nature of the hypothetical case is varied enough so that the action of the doctor is not wrong. In other words, it might be claimed that, on analogy with Sturgeon’s preferred treatment of the cat case, the relevant counterfactual is éif the case presented were different enough so that the action of the doctor was not wrong, then S would not believe that the (different) action was wrongù. This counterfactual seems true and has a contingent antecedent so it may be properly treated by [CT*]. However, this proposal would have us determine whether the fact that an action type is wrong is explanatorily relevant to our judgement about the action by considering whether we would judge an entirely different action to be wrong. If we treat the content of our judgement as a claim about a certain type of action, then it seems clear that the explanatory test should be applied to it by holding the nonmoral description of the action constant and varying the moral status and this procedure generates a counterfactual with an impossible antecedent and so [CT*] cannot be used to demonstrate that the fact that the doctor’s action is wrong is explanatorily relevant to our judgement about that case.
Even if the relevance of Sturgeon’s preferred counterfactuals could be plausibly defended, a second difficulty with Sturgeon’s argument concerns his appeal to the strong supervenience of the moral upon the nonmoral in his case for the explanatory relevance of the moral. Sturgeon’s case for the explanatory relevance of moral properties or facts depends essentially on his assuming the strong supervenience of the moral on the nonmoral. However, there seems to me a serious difficulty with Sturgeon’s justification for assuming this doctrine in the context of his argument with the proponent of an explanationist epistemology.
A typical formulation of a supervenience claim runs as follows:
[S] Let A and B be families of properties closed under Boolean operations: A strongly supervenes on B just in case, necessarily, for each x and each property F in A, if x has F, then there is a property G in B such that x has G, and necessarily if any y has G, it has F.[17]
In the case of moral properties, the A family would be the set of moral properties and the B family would be the set of natural properties.
Even if Sturgeon is within his rights in assuming (for the purposes of argument) the existence of moral properties possessed by particular actions, what justifies his strong supervenience claim? Recall that, on the criterion of justified belief with which Harman is working, belief that p is justified only if p plays a role in the best explanation of our judgings. Since [S] is a doctrine of modal metaphysics, it isn’t at all clear how [S] could be a part of the best causal explanation of some judging of ours. Indeed, [S] is, if true for moral properties and nonmoral ones, a necessary truth and I have already argued that such propositions are not properly treated by Sturgeon’s [CT*]. Sturgeon justifies his assuming [S] by claiming that it is ‘widely held’ that if there are moral properties, then they satisfy [S]. This, of course, is no justification at all by the lights of the explanatory criterion and Sturgeon had claimed that he aimed to respect that criterion.
This shows that his claim, which I quoted at the outset of this section, about the importance of fitting our moral knowledge into a causal theory of knowledge is misleading since he does not actually seem to hold any such theory for philosophical knowledge generally. Sturgeon’s attempt to show the explanatory relevance of moral properties, succeeds, if it does, only by appeal to premises that fail the very epistemological principle which Sturgeon seeks to show that moral propositions satisfy. As will become clearer below, this was to be expected since most propositions of philosophical interest will fall afoul of [1].
Let me conclude my discussion of Sturgeon’s attack on [2] by noting that even if Sturgeon’s arguments were, contrary to what I have argued, successful, they would serve only to rebut Harman’s claims about moral facts. It is far from obvious that Sturgeon’s arguments for the explanatory relevance of moral facts can be straightforwardly adapted to other sorts of philosophical facts. However, Sturgeon’s failure to rebut the general skeptical argument form need not worry those who treat intuitions as evidence, as I will show in the next two sections that the explanationist argument, and hence all of its particular instances, can be shown to fail because of difficulties with premise [1], the premise common to all of the skeptical arguments outlined in Section 2.
2 Premise [1] Is Unsupported and Methodologically Self-Defeating
In light of Sturgeon’s failure to deliver a compelling general argument against the form of argument outlined in Section 2, and given the independent plausibility of the appropriate second premise in the case of many philosophical propositions, an evaluation of premise [1] is quite urgently required. We are, it would certainly seem, owed a reason for thinking that each of the premises of the explanationist argument is true, especially now that their sharply skeptical implications are clear. However, in the writings of the explanationists, little argument is given for [1], taken, as it must be if the argument is to succeed, as a fully general requirement applying to all our beliefs. No doubt the requirement has some plausibility as a requirement on the formation of empirical belief, provided that ‘explanation’ is construed broadly enough. But why should we suppose that it applies universally? In particular, why should we assume it holds for the domains with which we have been concerned? If the explanationist is to provide us with good reasons for her skeptical conclusions, she must justify acceptance of [1].
However, there seem to me to be only two ways that the explanationist might go about justifying belief in [1]. She could argue either that it is simply intuitive, that [1] is simply an intuitively obvious general truth about when a belief is justified, or she could argue that it is inductively supported by our intuitions regarding particular cases of justified belief. In the remainder of this section, I’ll argue that not only are both of these claims substantively mistaken, but these two ways of arguing for [1] are methodologically inconsistent with the content of [1].
Consider an explanationist who takes the first course and argues that [1] is simply intuitive and, on that ground, that we ought to accept it as true. Such a philosopher claims that her justification (and, presumably, ours) for accepting [1] is the fact that it seems intuitively true. Against this, I can report that it isn’t at all intuitive to many of us that [1] is the correct criterion of justified belief. This is especially plausible once one sees what seem to be its skeptical implications. Furthermore, there is good evidence that the explanationists with which this paper is concerned don’t themselves think it is intuitively clear in itself. When these explanationists have tried to argue for it, they have often done so using an inductive argument based on intuitions about particular cases of justified and unjustified belief.[18] They have, in other words, taken the second course.
Consider, then, the second option for the explanationist, an inductive argument for [1] based on our intuitions about particular cases of justified and unjustified belief. To follow the route here envisaged, the explanationist would have to argue that [1] is supported by our intuitions regarding when particular actual and hypothetical beliefs are justified or unjustified. It is clear, I think, that the method of beginning with our intuitions about particular cases of justified belief will not support [1] since many of what seem, intuitively, to be our most justified beliefs run afoul of [1]. After all, many (though, of course, not all) of our particular epistemic beliefs, moral beliefs, and modal beliefs seem, intuitively, no less justified than our empirical beliefs. Indeed, some of them seem more justified. Since it seems implausible that all of these propositions are required in the best explanation of the occurrence of our judgements, it seems that [1], which requires such a role, derives no evidential support from such cases of intuitively justified belief.
The proponent of [1] is, then, unlikely to find support for [1] in a truly representative set of our particular intuitive judgements about justified belief. Indeed, any argument that is even prima facie successful in inductively supporting [1] is likely to be guilty of unwarranted generalization from an unrepresentative set of our intuitions about justified belief, our intuitions about particular cases of empirical belief. So, there is no intuitive justification (and certainly no all-things-considered intuitive justification) for [1]. If this is right, and only intuitions can be appealed to as the evidential basis for [1], then [1] is simply inadequately supported by the evidence. This fact alone suffices to undermine the explanationist argument for those of us who begin by relying on intuitions.
Even if I were wrong in maintaining that there simply is not sufficient prima facie intuitive support for [1], there would remain a deeper problem with both of these attempts to argue for and clarify the content of [1] on the basis of intuitions about justified belief and explanation.[19] The problem is that they involve treating intuitions (intuiteds) as evidence in a manner inconsistent with the content of [1].[20] Such attempts to substantiate [1] are, in this regard, methodologically self-defeating.
The explanationist, it seems to me, can only justify acceptance of his main premise by appeal to evidence which that very premise takes to be without value. According to [1], S’s belief in p (where p does not report the occurrence of one of S’s judgements) is justified only if the proposition believed plays a role in the best explanation of S’s judgings. No such showing of explanatory relevance is involved in the two methods of arguing for the truth of [1] just discussed. Rather, each approach would take the mere fact that the principle is the content of an intuition or best explains the content of a set of particular intuitions (i.e. intuiteds) as sufficient for justified belief in that principle. This is to treat intuiteds as supporting evidence for a principle allowing only intuitings to count as evidence. No such argument could possibly succeed in providing us with reason to think belief in premise [1] was justified. Since the only possible evidence for [1] is evidence that the principle itself deems inappropriate, it seems to me that any attempt to argue for [1] must treat intuitions as justifying evidence in order to argue that they are not properly treated as evidence. I conclude that the explanationist’s practice must be inconsistent with her professed methodology.
In response to these problems, the proponent of the explanationist argument might stubbornly continue to maintain that [1] captures a truth about when a belief is justified. I’ve shown that we who rely upon intuitions have been given no good reason to accept [1] and that the explanationist could have, by his own standards, no good reason in the (putative) intuitions for accepting [1]. Perhaps the explanationists will not be much impressed by these considerations. Setting aside their past practice, they might reply that intuitions are not a terribly worthwhile guide to the truth. Even if our particular intuitions about modality, epistemology and morality are no worse off, intuitively, than our empirical beliefs, they may claim that it begs the question against them to claim that that fact means that they are justified. The explanationists may maintain (without any good reason, so far as I can see, once they eschew the appeal to epistemic intuitions) that [1] captures a truth about justified belief.
However, even if this is a consistent stance for the explanationist to adopt, my arguments thus far suffice to justify those of us who require positive argument for our epistemological principles in rejecting their premise [1]. The explanationists could endorse [1] and repudiate intuitions from the outset, but as long as we begin by relying upon intuitions to provide epistemological guidance, the explanationist argument cannot rationally motivate any departure from our practice since the explanationist cannot justify her premises as only intuitions would seem to provide epistemological guidance and they do not support [1]. While I think that the explanationists should be much more concerned about what sort of coherent and dialectically acceptable positive justification could be provided for [1], it may seem as though a stalemate looms between the explanationist skeptics and their opponents. While this is not necessarily an epistemological problem for those of us who require some positive justification for our epistemic principles, it is a somewhat (but only somewhat) unsatisfactory dialectical situation.
We can avoid such a situation, however, because we can show that not only is [1] impossible to justifiably argue for, it also cannot be coherently maintained by the explanationist skeptics. In the next section, I’ll argue that [1] cannot be justifiably believed by reflective proponents of the skeptical arguments that I have considered because [1] itself runs afoul of the necessary condition for justified belief which it states, and this fact decisively undermines their attack on intuitions as evidence from within and thereby breaks any potential stalemate.
Before continuing, however, I ought to discuss a possible objection to my argument that utilizing intuitions to argue for [1] is self-defeating.[21] It might be replied, on behalf of the explanationist skeptic about intuitions, that appealing to intuitions in order to support [1] is simply part of a reductio of the evidential value of intuitions. What is suggested here is a purported demonstration that treating intuitions as evidence is self-defeating since intuition undermines its own value as evidence.
[a] Suppose (for reductio) that one is justified in believing p when p is intuitive.
[b] [1] is intuitive.
[c] So, one is justified in believing [1].
[d] If [1] is true, then one is not justified in believing p when p is intuitive.
[e] So, one is justified in believing that one is not justified in believing p when p is intuitive.
[f] So, if one is justified in believing that p when p is intuitive, then one is justified in believing that one is not justified in believing p when p is intuitive.
[g] If one is justified in believing that one is not justified in believing p on the basis of evidence e, then one is not justified in believing p on the basis of evidence e.
[h] So, if one is justified in believing that p when p is intuitive, then one is not justified in believing p when p is intuitive.
The conclusion in [h] is that the view that intuitions are justifying evidence is self-defeating. While one might dispute [g], even the provisional conclusion reached in [f] is an unacceptable result. This argument would release the skeptic about intuitions from any need to actually accept [1], much less justify her acceptance by appeal to intuition, since she would merely advance it in the context of a reductio of the view that intuitions are justifying evidence. As a result, my complaints about the skeptic about intuitions appealing to intuitions to justify her own views could be turned aside.
Whatever the merits of this approach, I see no reason whatsoever to think that this argument lies behind Harman’s or Goldman’s skepticisms about moral or metaphysical intuitions. Both of these philosophers appear to genuinely endorse [1] (or some near relative) and, as I noted above, occasionally provide intuitive considerations in their favor. In addition, even if this argument were successful in supporting a skeptical conclusion about intuitions, it would leave the explanationist position captured in [1] entirely without support. The explanationist, after all, is the proponent of a particular epistemological position, one according to which the necessary (and perhaps sufficient) conditions for justified belief in certain propositions depend upon explanatory considerations. If this reductio argument worked, then one could conclude, on its basis alone, that intuitions are not good evidence. If that is true, then an explanationist plainly cannot justify his belief in [1] on the grounds that it is intuitive (directly or inductively) and I cannot see what other grounds he could have. While this doesn’t matter to the success of this purported reductio argument, eschewing intuitions seems to deprive the explanationist of any grounds for thinking that various of our empirical beliefs are justified and the explanationists with which we have been concerned, in spite of their skepticism regarding various philosophical domains, are not at all skeptical about the vast majority of empirical beliefs.
Regardless of whether this argument is true to the writings of Harman and Goldman or consistent with their non-skeptical aims, it fails to support skepticism about intuitions. The first reason is that [1] is not, as was noted earlier, intuitive. While the proponent of this reductio argument need not treat intuition as evidence for [1] since she need not actually believe [1], the argument still requires that [1] be intuitive. As it is not, the argument cannot even get started. The importance of this fact ought not be underestimated because the force of the argument as a reductio of the view that intuitions are evidence depends on the extent to which the proponents of that view accept the other premises required to arrive at the absurdity.
Even if [1] were
to some degree intuitive, however, this argument fails to justify any sort of
skepticism about intuitions because intuition provides (in light of the
paradoxes) merely a defeasible
justification for belief and once the supposition in made (in [a]) that
intuitions justify belief, there are other
intuitions which must be taken into account before arriving at such a
skeptical conclusion. As noted above, the fact is that most of us have a host
of strong intuitions that various particular intuitive beliefs are justified or
evident. Furthermore, most of us also have the general intuition that
intuitions are evidence. Even if [1] were intuitive, these contrary intuitions
surely provide (in conjunction with plausible principles) defeaters for [1].
Finally, the argument of the next section shows that [1] can be used to show
that [1] itself cannot be justifiably believed. This result seems to further
undermine any prima facie support had
by [1]. In light of all these defeaters, it is clear that the reductio argument does not ultimately
succeed. 5. Premise [1] is Reflectively
Self-Defeating
The last section demonstrated that the explanationist can give no good reason for accepting [1] since the only apparent means of supporting [1] goes, according to the content of [1], no distance toward showing belief in [1] justified. In this section, I will argue that [1] cannot be justifiably believed by the explanationists even if they eschew the need for any positive support of it. I will show that [1], when conjoined with propositions the explanationist skeptic already accepts, generates its own defeater and is therefore epistemically self-defeating. So, even if one could rationally come to believe it, one could not reflectively and rationally maintain such belief.[22]
Since [1] contains a necessary condition for the justified acceptance of any proposition not about the occurrence of an observation or intuition, acceptance of [1] itself is justified only if it satisfies the very requirement it articulates. Since [1] is obviously not a doctrine reporting the occurrence of an intuition or observation, if [1] is true, then in order for us to be justified in believing it, its truth must be required in the best explanation of our observings or intuitings.
Unfortunately, there is no more reason to suppose that [1] has any such explanatory role than there is to suppose that the moral, modal, and epistemic propositions attacked by Harman, Goldman and Goldman* have such a role. Indeed, because [1] is a normative proposition about when a belief is justified, rather than a proposition of psychological or natural science, it is difficult to see how its truth could play any role in the explanation of the occurrence of any of our experiences or intuitions. However, if [1] does not satisfy the explanatory requirement, then, if [1] is true, it would seem to be something we cannot be justified in believing.
This shows that any argument relying on [1] is self-undermining and so all of the arguments surveyed in Section 2 must fail. There are two, roughly equivalent, ways of seeing why this is the case. First, one might begin by noting that [1] is required by all of the explanationist arguments I’ve outlined. Since, as I’ve just argued above, [1] (with the aid of other plausible principles) implies that we cannot be justified in accepting [1], and since that premise is the cornerstone of all the particular explanationist arguments, all of those arguments fail.
Alternatively, one might take the approach I took in Section 2.3 when I outlined the skeptical argument of Goldman*. Here, one may simply note that propositions about epistemic justification seem, as a class, as likely to fail the requirement of premise [1] as the other classes of propositions attacked by the explanationists. This means that the general argument schema outlined in Section 2 can be instantiated using epistemic propositions. The conclusion of such an argument would be that S is not justified in believing any epistemic propositions. However, if so, then because premise [1] is an epistemic proposition, S would not be justified in believing it. The conclusion, if true, undermines justified belief in the premises. Without such a premise, the argument would fail to justify belief in the conclusion.
Let me try to put my point a little more precisely. Consider the following propositions:
[4] [1] does not assert the occurrence of a judgement.
[5] [1] is not part of the best explanation of the occurrence of some judgement.
[6] No one is justified in believing [1].
Since the explanationist asserts [1] in her argument, we may assume that she believes [1]. I take it that all parties will admit that [4] is obviously true. In addition, [5] seems true as well. At the very least, [5] seems as justified as the second premises ([2], [2′] and [2′]) in the various skeptical arguments I’ve presented. If the explanationist balks at [5] and attempts to find an explanatory role for [1] in order to render it self-supporting, then she is likely to allow just as plausible an explanatory role for other epistemic propositions, and probably for moral and metaphysical ones as well.
Since [1], [4], and [5] entail [6], the explanationist is in the peculiar position of believing a set of propositions which entails that no one, including herself, is justified in believing one of those propositions. Since the explanationist is thereby committed to the claim that her own belief in [1] is not justified, it cannot be used in an argument meant to justify belief in a conclusion. This fact, by itself, doesn’t show that [1] is false, nor that the conclusions of the explanationist arguments are not true. It does, however, show that the explanationists cannot consistently and reflectively take themselves to be justified in believing [1], [23] and that should be sufficient to undermine entirely the skeptical arguments of Section 2.[24]
3 Conclusion
I’ve presented some contemporary arguments for skepticism about the evidential value of intuitions in particular philosophical domains and shown that they are versions of a single argument schema with potentially great skeptical import. I have, however, argued that the main premise of that argument schema is not and cannot be positively justified for those of us who already rely upon intuitions. More importantly, I’ve shown that the explanationist doctrine is self-defeating in that it undermines its own justification. Therefore, the explanationist argument cannot justify the skeptical conclusions which it was invoked to support. If there is something epistemically illegitimate about relying upon intuitions as evidence, we have yet to see what it is.
No doubt it will be claimed that the failure of this particular skeptical argument does not amount to a positive defense of the evidential worth of intuitions. However, I think some elements of such a defense are in place. Intuitions seem necessary for the justification of epistemological claims and so seem necessary for the reflective construction of any epistemological doctrine. This means that there is an incoherence in attempts to argue that intuitions never provide us with good reasons for accepting certain propositions while wishing to justifiably maintain any epistemological doctrine. This, of course, leaves open the possibility of a more selective skepticism about some specific subclass of intuitions. An argument capable of justifying such a selective skepticism, however, would have to be based, unlike the explanationist arguments of Harman and Goldman, on claims that do not apply equally well to epistemic intuitions.
Acknowledgements
I am grateful to Keith Lehrer, Robert Cummins, William
Lycan, André Ariew, Michael Rea, Michael Bergmann, Rahul Kumar, David Silver,
Thomas Bontly and audiences at The University of Arizona,
References
Adams, R. (1979). ‘Primitive Thisness and Primitive Identity.’ Journal of Philosophy 76: 5-26.
Bealer, G. (1993). ‘The Incoherence of Empiricism.’ In
Steven Wagner and Richard Warner (eds.), Naturalism:
A Critical Appraisal. Notre Dame:
DePaul, M. and Ramsey, W. (Eds.). (1998). Rethinking
Intuition: The Psychology of Intuition and Its Role in Philosophical Inquiry.
Gibbard, A. (1990). Wise Choices, Apt Feelings.
Goldman, A. (1976). ‘Discrimination and Perceptual Knowledge.’ The Journal of Philosophy 73: 771-791. Reprinted in Goldman (1992a).
Goldman, A. (1979). ‘What is Justified Belief?’ In George Pappas (ed.), Justification and Knowledge. Reprinted in Goldman (1992a).
Goldman, A. (1986). Epistemology and Cognition.
Goldman, A. (1987). ‘Cognitive Science and Metaphysics.’ Journal of Philosophy 84: 537-544.
Goldman, A. (1989a). ‘Psychology and Philosophical Analysis.’ Proceedings of the Aristotelian Society 89: 195-209. Reprinted in Goldman (1992a).
Goldman, A. (1989b). ‘Metaphysics, Mind, and Mental Science.’ Philosophical Topics 17: 131-145. Reprinted in Goldman (1992a).
Goldman, A. (1992a). Liaisons.
Goldman, A. (1992b). ‘Cognition and Modal Metaphysics.’ In Goldman (1992a).
Goldman, A. (1992c). ‘Epistemic Folkways and Scientific Epistemology.’ In Goldman (1992a).
Goldman, A. (1994). ‘Naturalistic Epistemology and Reliabilism.’
In P. French et. al. (eds.),
Goldman, A. and Pust, J. (1998). ‘Philosophical Theory and Intuitional Evidence.’ In DePaul and Ramsey (1998).
Harman, G. (1965). ‘Inference to the Best Explanation.’ Philosophical Review 74: 88-95.
Harman, G. (1973). Thought. Princeton:
Harman, G. (1977). The Nature of Morality.
Harman, G. (1984). ‘Is There a Single True Morality?’ In David Copp and David Zimmerman (eds.), Morality, Reason and Truth. NJ: Rowman & Allenheld.
Harman, G. (1986). ‘Moral Explanations of Natural Facts—Can Moral Claims be Tested Against Moral Reality?’ Southern Journal of Philosophy Supplement 24: 57-68.
Kim, J. (1984). ‘Concepts of Supervenience.’ Philosophy and Phenomenological Research 45: 153-176.
Kripke, S. (1980). Naming and Necessity. Princeton:
Lewis, D. (1973). Counterfactuals.
Lycan, W. (1988a). Judgement and Justification.
Lycan, W. (1988b). ‘Moral Facts and Moral Knowledge’ In Lycan (1988a).
Markie, P. (1996). ‘Goldman’s New Reliabilism.’ Philosophy and Phenomenological Research 56: 799-817.
Moser, P. (1985). Empirical Justification.
Nagel, T. (1980). ‘The Limits of Objectivity.’ In The Tanner
Lectures on Human Values.
Pressler, J. (1988). ‘Is There any Basis for Moral Scepticism?’ Australasian Journal of Philosophy 66: 354-367.
Quinn, W. (1986). ‘Truth and Explanation in Ethics.’ Ethics 96: 524-544.
Sayre-McCord, G. (1988). ‘Moral Theory and Explanatory
Impotence.’
Sellars, W. (1968). Science and Metaphysics.
Sorenson, R. (1988). Blindspots.
Sosa, E. (1993). ‘Proper Functionalism and Virtue Epistemology.’ Nous 27: 51-65.
Stalnaker, R. (1968). ‘A Theory of Counterfactuals.’ In
Nicholas Rescher (ed.), Studies in Logical Theory. American Philosophical
Quarterly Monograph Series.
Sturgeon, N. (1984). ‘Moral Explanations.’ In David Copp and David Zimmerman (eds.), Morality, Reason and Truth. NJ: Rowman and Allenheld.
Strawson, P. (1959). Individuals.
Wierenga, E. (1998). ‘Theism and Counterpossibles.’ Philosophical Studies 89: 87-103.
[1]I shall assume that the reader has some intuitive grasp of what I mean by ‘intuition.’ See Bealer (1993) for an excellent analysis of the nature of intuitions. Though I do not believe that intuitions are beliefs, rigorous observation of this distinction creates needless complication and so I shall not distinguish between intuitions and intuitive beliefs in what follows.
[2]In addition to the authors discussed below, see a number of the articles in DePaul and Ramsey (1998).
[3]Neither Harman nor Goldman is entirely skeptical about the use of intuitions as evidence. Their skepticism is limited to treating philosophical intuitions as evidence for a non-psychological theory. Harman explicitly concedes that moral intuitions provide evidence for a psychology of moral judgement. Goldman repeatedly claims that intuitive judgements provide good evidence for psychological hypotheses. In his most recent work (Goldman and Pust 1998), he attempts to make a virtue of necessity and defend (at least in part) the practice of appealing to intuitions by arguing that intuitions provide excellent evidence for claims about our concepts or tacit theories, where these are taken to be mental representations or psychological mechanisms of some sort. Though I was co-author with Goldman on ‘Philosophical Theory and Intuitional Evidence’ (written in 1996) the present paper should indicate why I no longer find that earlier defense of the evidential use of intuitions entirely satisfactory. My change of view was adumbrated in footnote 14 of Goldman and Pust (1998).
[4]There are some difficulties with this account of observation. The intuitive judgement one arrives at when confronted with a linguistic description of a merely possible case does not seem to clearly count as an observation on this criterion. While this is acceptable if Harman means to stipulate that only immediate judgements about actual, concrete, visually presented cases are to count as observations, this account does not encompass all of what Harman himself admits we normally use as evidence for moral theories. Indeed, one of his own examples involves an intuitive judgement produced in response to a description of a merely hypothetical case—one in which a doctor takes the life of an unconsenting healthy person to save the lives of five other patients. This case is not distinguished from other, more clearly ‘perceptual,’ cases in Harman’s discussion. I would maintain, then, that the issue of whether we make moral ‘observations’ in any standard sense of that term is irrelevant to Harman’s argument. We free ourselves from unnecessary complication and do not much affect the argument if we allow any non-inferential perceptual or intuitive belief to count as an observation in Harman’s sense of that term.
[5]As Harman elsewhere summarizes the argument:
‘Naturalists see an important difference between our factual beliefs and our moral beliefs. Our ordinary factual beliefs provide us with evidence that there is an independent world of objects because our having those beliefs cannot be plausibly explained without assuming we interact with an independent world of objects external to ourselves, objects we perceive and manipulate. But our having the moral beliefs we have can be explained entirely in terms of our upbringing and values; without any appeal to an independent realm of values and obligations. So our moral beliefs do not provide us with evidence ... ‘ (1984, pp. 32-33, emphasis added)
See Gibbard (1990, 121) for a very similar argument.
[6]Though not discussed by Harman, this first clause is necessary to provide for the explainees that serve as the foundations of explanationist epistemology. If a proposition’s justification depends on what it explains, then there must be something for it to explain in the first place. I here set aside the question of what, if anything, justifies our believing propositions regarding the occurrence of our judgements. For more on these topics see Lycan (1988a).
[7]This similarity is noted by Goldman (1992b, p. 66).
[8]Though Goldman invokes Strawson’s distinction, he seems to have a different conception of the nature and relative importance of the two disciplines. Strawson held that descriptive metaphysics was the more fundamental discipline and that whatever ‘philosophical utility’ could be ascribed to prescriptive metaphysical systems could only be ascribed to them ‘because there is another kind of metaphysics [that revealed by descriptive metaphysics] which needs no justification at all beyond that of inquiry in general’ (1959, p. 9, emphasis added). Revisionary metaphysics was ‘at the service of descriptive metaphysics,’ and the core aspect of our conceptual scheme revealed by descriptive metaphysics was, in Strawson’s estimation, ‘indispensable’ (1959, 9).
[9]It is interesting that Goldman, the most prominent proponent of reliabilism, worries about our lack of independent evidence for the reliability of intuitions. A straightforward reliabilism would claim simply that metaphysical beliefs are justified if and only if the intuitions upon which they are based are reliable, but that we need not have any independent evidence that they are reliable. Perhaps Goldman believes that concerns about the lack of a causal-explanatory connection trip the ‘no-defeater’ condition in his reliabilist analysis of justified belief.
[10]Both
Kripke and Adams clearly appeal to intuiteds as evidence. Kripke
famously claims that he doesn’t know ‘what more conclusive evidence one can
have about anything’ (1980, p. 42).
[11]Goldman has claimed, in correspondence, that he does not endorse [1]. Still, his papers which I have focused on here certainly suggest that some sort of explanationist principle is correct. One of the papers explicitly (though tentatively) advances a general principle according to which, if an intuition [intuiting] can be fully explained without appeal to the putative fact intuited, then there is a presumption against the veridicality of the intuition (1989b, p. 53). This is, admittedly, weaker than [1]. However, Goldman clearly thinks the presumption is not overturned in the case of the modal intuitions of the sorts he discusses and so I believe any alternative construal of his view is open to the self-defeat objections developed in Sections 4 and 5 of this paper. Regardless of whether Goldman ultimately accepts [1], the papers on which I have focused show how the proponent of [1] might extend his skepticism to other areas of philosophical inquiry.
[12]In hindsight, the distinction between descriptive and prescriptive epistemology is clearly implicit in Goldman’s response to an objection to the process reliabilism presented in ‘What is Justified Belief?’ In an interesting footnote, Goldman writes:
In reply to this objection, we might simply indicate that the theory is intended to capture our ordinary notion of justifiedness, and this ordinary notion has been formed without recognition of this kind of problem. The theory is not wrong as a theory of the ordinary (naive) conception of justifiedness. (1979, fn 8, 125)
In the same footnote, he allows for the possibility that we may want a theory to ‘do more than capture the ordinary conception of justifiedness,’ and if we are so inclined, we may try to ‘fix’ or ‘revise’ the theory to avoid the problem.
[13]This feature has been noted by critics such as Ernest Sosa (1993) and Peter Markie (1996). In Markie’s view, Goldman’s new theory ‘does not offer any explanation of what makes a virtuous epistemic process virtuous or a vicious epistemic process vicious. It is a psychological hypothesis about how people make judgements of epistemic justification’ (1996, p. 804). Sosa also worries that Goldman’s proposal, for all of its possible psychological virtues, ‘does not tell us what it is for a virtue or faculty to be virtuous, and what is involved in a belief’s being epistemically justified’ (1993, p. 61). Goldman replies in Goldman and Pust (1998).
Paul Moser (1985) anticipated Markie’s and Sosa’s objections by a decade. Addressing Goldman’s statement that ‘What we really want is an explanation of why we count, or would count, certain beliefs as justified and others as unjustified’ and that ‘the reason we count beliefs as justified is that they are formed by what we believe to be reliable belief-forming processes,’ (1979, p. 121) Moser writes:
The requirement suggested here applies not to beliefs which simply are justified, but rather to beliefs which we take to be justified .... It should be obvious that this alternative to the above-discussed reliability requirement would severely change the face of epistemic reliabilism. Given this alternative requirement, epistemic reliabilism would be a view simply about beliefs which we take to be justified; it would have no bearing on beliefs which merely are justified. (1985, 241)
A focus on believed rather than actual reliability is more explicit in the new two-tiered theory (Goldman 1992c) criticized by Sosa and Markie, but something similar motivated the appeal to ‘normal worlds’ in Epistemology and Cognition (Goldman 1986) since normal worlds were those worlds consistent with various of our common sense beliefs about the world. Even as far back as Goldman (1976), we find Goldman trying to explain ‘why people are credited with knowing’ rather than why it is that they know or do not know (1976, 99). So, a concern with the psychology of the epistemic evaluator, and hence a focus on believed rather than actual reliability, has been, in one form or another, a consistent feature of Goldman’s reliabilism since its inception.
[14]Sturgeon notes some exceptions to this test involving cases of overdetermination and ‘backtracking’ counterfactuals (1984, pp. 75-76). He claims that none of his cases are cases of this sort. In what follows, I shall use ‘é ù’ to bracket propositions.
[15]The first problem is that the claim that [CT] is a sufficient condition for explanatory relevance seems to create the possibility of puzzling overdetermination. If moral properties strongly supervene on natural properties, then [CT] yields the result that a natural event’s natural and moral properties are both explanatorily relevant to the occurrence of any event counterfactually dependent upon the event. A second problem with taking [CT] as a sufficient condition of causal-explanatory adequacy is that it rules out the metaphysical possibility of genuinely epiphenomenal properties that strongly supervene on causally efficacious physico-chemical properties.
[16]I realize that this feature of the Lewis-Stalnaker semantics is thought to be a defect by those who believe that some counterpossibles are non-trivially false and others non-trivially true. Here, I can only state my view that the case for this has yet to be made out. For a defense of the Lewis-Stalnaker view, see Wierenga (1998).
[17]This formulation is from Kim (1984).
[18]See Harman’s procedure in Thought (1973). Of that procedure he claims: ‘I will apply the strategy of using intuitive judgements about when people know things in order to discover principles of reasoning that justify belief’ (116, emphasis added), and ‘I deny that there is any way to discover ... irrationality except by way of intuitive judgements about when people come to know things’ (117, emphasis added). See also Goldman’s arguments in his (1976), (1979), and (1986) which involve extensive appeal to intuitions about cases.
[19]Specification of the content of [1] sufficient to determine the real extent of its application requires an understanding of the nature of explanation. That understanding, I would maintain, can only be reached by appeal to intuitions. This is evident in the writings of the explanationists discussed above (see especially Harman 1973). Indeed, every attempt to argue for or against theories of the nature of explanation with which I am familiar proceeds, in large measure, by appeal to intuitions about cases. To cite just one example, consider the flagpole counterexample to the D-N model of explanation. This dependence creates, I think, another reason to think that explanationism is methodologically self-undermining.
[20]This point is also made by George Bealer in his (1993).
[21]This sort of reply was suggested to me by Robert Cummins and William Lycan. They may not endorse my development of the reply.
[22]Other philosophers, including Sayre-McCord (1988), Quinn (1986) and Pressler (1988), have offered somewhat similar arguments. All of them are responding to Harman (1977). Sayre-McCord (1988) argues against a close relative of [1], claiming that, if it is true (as it must be to undermine moral facts), then it presupposes the existence of evaluative facts. In particular, it presupposes the existence of irreducible evaluative facts regarding what makes an explanation ‘good’ or ‘better,’ and so it cannot be used to show that there are no evaluative facts at all (450). He suggests that whatever ontological niche is home to evaluative facts about good explanation will suffice as well for evaluative moral facts. Quinn makes a point more like mine with his remark that ‘philosophy itself seems to be another area on which the requirement [something like my [1]] casts grave doubts, in this case doubts that must apply to the requirement itself’ (1986, 539). However, Quinn does not clearly argue that the requirement is self-defeating in the ways I discuss here. (See his somewhat opaque discussion immediately following the remark just quoted.) Finally, Pressler (1988), though not explicitly attacking the principle as self-undermining, suggests that it implies that we are unjustified in believing propositions about justification. Since he takes the implication to be obviously false, he takes the principle to be false.
[23]We are in the vicinity here of a ‘blindspot’ in the wonderfully apt terminology of Sorenson (1988). A proposition p is in a justification blindspot for an agent s at time t IFF p is consistent and the following condition is met: ~à(p & Jstp & Test) where ‘Jstp’ means that s is justified at t in believing p and ‘Test’ means that s is completely epistemically thorough at t and so is justified in believing all the logical consequences of what s justifiably believes and this justification distributes over material conditionals (1988, 52). If p = [[1] & [4] & [5]], then since p É [6], [6] É ~Jst[1], and Jstp É Jst[1] (by the definition of Test and the logical truth (([1] & [4] & [5]) É [1])), we can derive the impossible (Jst[1] & ~Jst[1]).
[24][1] might also preclude the possibility of justified belief in each of the second premises in the three particular arguments discussed above ([2], [2′] and [2²]). Each of those premises asserts that a certain class of propositions is not part of the best explanation of our making various judgements. While I am happy to accept this, it isn’t clear that the explanationist can reflectively do so. There is little reason to believe that the claim that some explanation E is the best explanation of our judgements is itself part of the best explanation of our judgements. If this is correct, then the argument of the last section might be altered to claim that either the truth of [1] undermines justified belief in all of the second premises required for the skeptical arguments, or it undermines justified belief in itself.