“How to Preserve Your Virtue while Losing Your Perspective,” in Sosa and his Critics, John Greco, ed. (Oxford: Blackwell, 2004). JOHN GRECO HOW TO PRESERVE YOUR VIRTUE WHILE LOSING YOUR PERSPECTIVE Ernest Sosa’s “virtue perspectivism” involves two central ideas. The first is that knowledge is grounded in the intellectual virtues of the knower, where an intellectual virtue is defined as a stable disposition to arrive at truth and avoid falsehood in a relevant field, when in relevant circumstances. The second is that reflective knowledge requires also a perspective on one’s truth-conducive dispositions: “For the exercise of virtue to yield knowledge, one must have some awareness of one’s belief and its source, and of the virtue of that source both in general and in the specific instance.”1 In this regard Sosa maintains a distinction between “animal knowledge” and “reflective knowledge”: One has animal knowledge about one’s environment, one’s past, and one’s own experience if one’s judgments and beliefs about these are direct responses to their impact—e.g., through perception or memory—with little or no benefit of reflection or understanding. One has reflective knowledge if one’s judgment or belief manifests not only such direct response to the fact known but also understanding of its place in a wider whole that includes one’s belief and knowledge of it and how these come about. (KP, 240) There is a tension in Sosa’s thought regarding the above distinction, however. In some places Sosa maintains that animal knowledge is true knowledge, while reflective knowledge is a still higher, more rare achievement. For example, in one place Sosa writes: This is not to deny that there is a kind of ‘knowledge,’ properly so called, that falls short in respect of broad coherence—‘animal knowledge,’ as we might call it. It is rather only to affirm that beyond ‘animal knowledge’ we humans, especially those of us who are philosophical or at least reflective, aspire to a higher knowledge.2
In the same spirit, Sosa compares his distinction between animal and reflective knowledge to Descartes’ distinction between unreflective cognitio and reflective scientia, where the latter results only from Descartes’ hard philosophical work.3 Sosa’s more common position, however, is that animal knowledge falls short of properly human knowledge. Hence he calls it “mere animal” knowledge, and even suggests that the label is only “metaphorical.” (KP, 274-5). Here is a passage illustrating this view of the distinction. How then can one rule out its turning out that just any true belief of one’s own is automatically justified? To my mind the key is the requirement that the field F and the circumstances C must be accessible within one’s epistemic perspective. (Note that this requires considering servomechanic and animal so-called ‘knowledge’ a lesser grade of knowledge, or perhaps viewing the attribution of “knowledge” to such beings as metaphorical, unless we are willing to admit them as beings endowed with their own epistemic perspectives). (KP, 274-5) Thus we can find in Sosa’s writing two positions regarding the animal knowledge/reflective knowledge distinction: a. Most human knowledge is non-reflective or “animal” knowledge. But humans are capable of reflective knowledge, which is a special and relatively rare achievement. b. Human knowledge is reflective knowledge. Non-reflective knowledge falls short of properly human knowledge, although we may speak of animal “knowledge” in an extended sense of the term. It seems clear that Sosa’s considered position is the second. First, the passages that reflect the position in (b) are far more numerous than those that reflect the position in (a). More telling, however, are Sosa’s motivations for adopting the distinction in the first place. His two most important motivations for doing so are to address two problems for generic reliabilism: the generality problem and the meta-incoherence problem. I will say more about these two problems below, but for now the important point is this: both show that something must be added to the conditions that generic reliabilism lays down for knowledge. In response to these objections, the force of which Sosa acknowledges, he adds the requirement of an epistemic perspective. But then having such a perspective must be viewed as a general condition on knowledge. That is, it will not 2
help to solve the generality problem or the meta-incoherence problem to say that sometimes humans have such a perspective when they have knowledge properly so called. On the contrary, Sosa’s strategy for solving the two problems requires him to say that human knowledge is reflective knowledge; i.e. that an epistemic perspective is always required for human knowledge. Once Sosa’s position is interpreted this way, however, an obvious question presents itself. How plausible is it that we humans actually have the required perspective on our beliefs and our faculties, whenever we have human knowledge properly so called? On the face of things, it seems not at all plausible. It seems that, in typical cases of knowledge, people do not have any beliefs about their beliefs, or beliefs about the sources of their beliefs, or beliefs about the reliability of those sources. Let us call this the “Psychological Plausibility Objection” to virtue perspectivism. Two strategies suggest themselves for addressing this objection, and Sosa adopts both of them. First, he stresses that one’s perspective need only be very general: “For ordinary knowledge one needs only a very sketchy and general perspective on one’s own beliefs and their derivation.”4 Second, he stresses that one’s perspective need only be implicit: “We need to distinguish first between fully conscious and subconscious belief; also, secondly, between what is and what is not verbalizable or symbolizable by the believer… and, finally, also between belief that is manifested through acts of episodic acceptance of a proposition somehow present to one’s mind, even if not symbolically present, from belief that is manifested only in other ways.” (VP, 47) In the remainder of this essay I will argue that these strategies cannot help Sosa to overcome the Psychological Plausibility Objection. First, to address the generality problem and the metaincoherence problem, Sosa must require that one’s epistemic perspective is highly detailed rather than “sketchy and general.” Second, once we are clear about the level of detail required, it becomes highly implausible that humans have even an implicit perspective (in the required detail) on their beliefs and the sources of their beliefs. I end by suggesting that the requirement of an epistemic
3
perspective is not needed anyway. Rather, Sosa’s virtue reliabilism has resources for addressing the generality problem and the meta-incoherence problem without invoking the idea of an epistemic perspective. 1. The Generality Problem and the Meta-incoherence Problem Consider first the generality problem for generic reliabilism. According to generic reliabilism, (GR)
S knows that p only if 1. p is true, and 2. S’s belief that p results from a sufficiently reliable process.
The problem is that any process token is an instance of several process types, and it is not clear which process type is relevant for evaluating reliability in the context of (GR). Consider that for any true belief B(p), we can describe both a) a process type that gave rise to B(p) and that is perfectly reliable, and b) a process type that gave rise to B(p) and that is unreliable. Clearly (GR) needs to say more about which process types are relevant for knowledge. As Sosa recognizes, a version of the generality problem arises for reliabilist theories that invoke the notion of an intellectual virtue or faculty. Sosa writes, One has a faculty only if there is a field F and there is a set of circumstances C such that one would distinguish the true from the false in F in C. But of course whenever one happens to have a true belief B, that belief will manifest many such competences, for many field/circumstance pairs F/C will apply. How then can one rule out its turning out that just any true belief of one’s own is automatically justified? (KP, 274) Sosa’s answer invokes the idea of an epistemic perspective: “To my mind the key is the requirement that the field F and the circumstances C must be accessible within one’s epistemic perspective.”5 But of course specifying F and C too finely is only one half of the generality problem-- one must not specify them too generally either. Such restrictions must heed a two fold objective: (a) that F and C not be made so specific that one is always perfectly reliable and justified whenever one’s belief is true; but also (b) that they not be made so generic that one cannot explain how a 4
subject could have two beliefs both derived from the given faculty (e.g., from his sight, or, more generally yet, from his sensory perception), though one is justified while the other is not. (KP, 284) It is this second consideration that gives rise to the Psychological Plausibility Objection. Given the role that Sosa assigns to an epistemic perspective in solving the generality problem, it cannot be that such a perspective is “sketchy and general.” On the contrary, a perspective must be specific enough to pick out dispositions that are sufficiently reliable to ground knowledge. Next consider the meta-incoherence problem. Here is how Sosa describes it. In the first place, there is the possibility of reliable mechanisms for noninferential acquisition of belief other than introspection, memory, and (ordinary) perception. Clairvoyance, for instance, would be one such mechanism. If magic or surgery suddenly gives one such a gift, however, one finds oneself predicting things ‘out of the blue,’ with too little discernable connection to the rest of one’s beliefs. Such a prediction might turn out to be right, of course: and not just by luck, either, since by hypothesis it derives from the operation of a reliable faculty. But I could not easily think it epistemically justified simply on that account. For anyone similarly minded, this raises a doubt as to the adequacy of mere reliability to induce justification.6 Once again, Sosa’s solution is to invoke the idea of a perspective on one’s beliefs and their sources. The problem of clairvoyance relative to a normal human today is, I think, its failure to meet the challenge of doxastic ascent. If a normal human today were suddenly to receive the gift of clairvoyance he would be helpless to explain the source of beliefs delivered by his new gift and how that source operates. There is much that we know at least implicitly about memory and its mode and conditions of reliable operation. And it is largely this that enables discrimination in favor of memory and against suddenly endowed clairvoyance. (KP, 94-5) The general idea is that an appropriate perspective on one’s belief and its source affords one a kind of internal justification that the clairvoyant lacks: specifically, one is allowed to see one’s belief as reliably produced, and so not just “out of the blue.” It might appear that the Psychological Plausibility Objection is less threatening here than it is regarding the generality problem, since one might easily attribute one’s belief to memory or perception, and this seems enough to distinguish it from that of the suddenly endowed clairvoyant. But this appearance is deceptive, owing to an ambiguity in the notions of memory and perception.
5
One thing we might mean by memory and perception is reliable memory and reliable perception. But this meaning is not relevant in the present context, since it would be empty and unhelpful to have, as part of one’s perspective, that one’s reliable memory is reliable. For the notion of memory to do any work in the present context, it must be defined independently of reliability, so that one might helpfully think that one’s belief B has its source in memory, and that one’s memory is reliable in present circumstances. But now we are back to the second half of the generality problem. That is, we are back to the problem of characterizing a relevant disposition, with relevant specificity, so as to pick out a disposition that is sufficiently reliable to generate knowledge. One needs exactly the level of specificity to solve the meta-incoherence problem as is needed to solve the generality problem. 2. The Psychological Plausibility Objection Renewed How plausible is it that human knowledge always involves a detailed perspective; i.e., a perspective detailed enough to pick out dispositions sufficiently reliable to generate knowledge? Remember, this amounts to the claim that every time S knows p, S also believes truly 1) that S believes p, 2) that p is in some field F, 3) that S is in some set of circumstances C, and 4) that S’s belief B(p) results from a highly reliable disposition D(F/C). Again, it would seem that this is not plausible at all. On the face of things, in the typical case, people seem not to have beliefs about their beliefs, nor beliefs about the sources of their beliefs. The claim that they do is more plausible, of course, if we allow the second-order beliefs to be very general. It is (perhaps) not completely implausible, for example, that whenever one has perceptual knowledge, one has an implicit belief that one’s knowledge derives from perception. But as we have seen above, this kind of very general belief will not do the work that Sosa needs it to do regarding the generality problem or the meta-incoherence problem. For that we need something much more specific, and now it becomes highly implausible that one has beliefs with the right level of specificity, even implicitly.
6
Or at least this is what I will argue below. Before getting to that, however, it is necessary to consider a suggestion that would help Sosa to answer the present concern. In reply to a related objection, Sosa has suggested that we might treat our cognitive dispositions as if they were highly implicit beliefs. For example, given S’s disposition to form a belief B(p) in response to experience E, we might attribute to S the belief that E is a reliable indication that p is true. [T]he habit of moving from ‘looks round’ to ‘is round’ is strictly correct if, in the relevant circumstances, anything that looked round would in fact be round . . . . [T]here is some motivation to view such habits as implicit beliefs (that can be correct and justified) in the corresponding conditionals, as suggested above. (VP, 42-3) The present suggestion would allow that we have the required beliefs about our reliable dispositions, since it simply identifies such dispositions with beliefs affirming their reliability. However, there are good reasons not to collapse the distinction between a) implicit beliefs and b) habits or dispositions for forming beliefs. One reason is that often there are such dispositions where there are no such beliefs. For example, simple pattern recognition in perception involves highly complex dispositions to go from perceptual cues to beliefs about external stimuli.7 But it is implausible to attribute beliefs about such perceptual cues, and about their connections to external stimuli, to perceivers.
That is, it is implausible to attribute such beliefs even to the most
sophisticated adult perceivers, not to mention small children and animals. But all perceivers, small children and animals included, have the relevant dispositions to form perceptual beliefs. Moreover, there is a second reason for Sosa to insist on the distinction between implicit beliefs and dispositions for forming beliefs. For without it, his distinction between animal knowledge and reflective knowledge collapses. Recall that even animal knowledge requires a source in reliable cognitive dispositions. If we say that simply having such dispositions is sufficient for having an epistemic perspective, then there will be no difference between animal and reflective knowledge.
7
Let us return, then, to what seems to be Sosa’s considered position: that an epistemic perspective is constituted by one’s second-order beliefs regarding one’s first-order beliefs and their sources. And let us remember that the second-order beliefs in question must not specify one’s sources in a sketchy or general way, but rather in a way that picks out sources that are sufficiently reliable to generate knowledge. The Psychological Plausibility Objection now amounts to this: It seems implausible that such second-order beliefs exist in typical cases of human knowledge, even subconsciously, and even in ways that are not verbalizable by the knower. It is hard to see how one could have good evidence in favor of such beliefs if, as Sosa suggests, they are not “manifested through acts of episodic acceptance of a proposition somehow present to one’s mind,” but are rather “manifested only in other ways,” such as in one’s patterns of inference. For in that case, it will always be simpler to attribute only a disposition to make the relevant inferences, rather than a perspective on one’s inferential dispositions. More importantly, there is good evidence against our having the required epistemic perspective in the typical case. Let us say that one has an epistemic perspective on cognitive disposition D(F/C) only if either a) one has an occurrent representation of D(F/C) (either consciously or subconsciously), or b) one has a disposition to have an occurrent representation of D(F/C) (either consciously or subconsciously), where D(F/C) picks out a disposition that is sufficiently reliable to generate knowledge. There is good evidence that the reliability of our cognitive dispositions is a function of factors that are not represented at all in the typical case, either occurrently or dispositionally.8 And if that is so, then one will not have the required perspective on such faculties in the typical case. First, consider our perceptual faculties, which take us from various kinds of inputs to beliefs about external stimuli. The problem here is that not all of the inputs relevant to reliability are represented by the perceiver, either occurrently or dispositionally. Occurrently, we do not typically
8
have any beliefs at all about inputs—one forms only beliefs about the objects perceived. But what is important for present purposes is that we may lack even a disposition to form representations of relevant inputs as well. For many of those inputs go unnoticed by the perceiver, and therefore are not available for occurrent representation at a different time. A particularly drastic illustration of this occurs in the phenomenon of blindsight. There are now several documented cases of people who are blind in some part of their visual field, but who can nevertheless discriminate size, shape, location and/or orientation of objects in the blind part of the field. Since damage in blindsighted subjects is to the visual cortex rather than to the eyes themselves, it is hypothesized that information from the eye still reaches the brain, although bypassing the mechanisms normally responsible for conscious visual experience.9 Of course blindsight is relatively rare, and not very reliable in known cases. Nevertheless, it is a dramatic illustration of how factors that are not represented or even available for representation by the subject can affect the reliability of our cognitive dispositions. What makes the example of blindsight dramatic is that, at least in some cases, it seems to operate entirely on inputs that are not conscious at all. But it is now thought that less dramatic forms of nonconscious acquisition of information are common in normal human cognition. In what sense is such information acquisition “nonconscious”? Sometimes what is meant is that important initial inputs go unnoticed, perhaps because exposure time is below the threshold needed for conscious detection. Other times new information is acquired via nonconscious cognitive processing, as when a perceptual system uses algorithms and heuristics that are too complicated to be used in conscious inferences. In either case, the cognitive dispositions involved are not available for representation by the subject, either occurrently or dispositionally, because essential elements of them are not so available.10 Again, such inaccessibility is thought to be a pervasive aspect of human cognition, involving both our non-inferential and inferential faculties. Lewicki, Hill and Czyzewska write,
9
this lack of access to the nature of these processes (which are essentially responsible for most of what we see, experience, and feel) is not limited to the so-called low-level processes that support only the consciously controlled cognition (e.g., pattern recognition). People have no access to processes as high-level as those involved in playing chess, feeling love, forming impression of people, or problem solving and creative thinking…11 I have been arguing that the reliability of our cognitive faculties is partly a function of factors that are not available for representation by the subject, either occurrently or dispositionally. That characterization of human cognition requires only the following assumption: that our cognitive functioning is importantly sensitive to information about the world without needing to represent that information, and without needing to have it available for representation.
Given the available
empirical evidence, it is not implausible that this assumption is true. Notice, however, that Sosa’s requirement of an epistemic perspective is undermined so long as the assumption is possibly true. This is because Sosa’s requirement of an epistemic perspective is a philosophical claim about the conditions for human knowledge, and so carries necessity. The above argument can therefore proceed as follows, where VP stands for virtue perspectivism, K for the proposition that some person S has human knowledge, and EP for the proposition that S has a relevant epistemic perspective. 1. VP => Nec (K -> EP) 2. Poss (K & not-EP) Therefore, 3. Not-VP Sosa might be inclined to deny premise (2), and to hold that if cognitive psychology is right about the nature of our cognition, then there is little or no human knowledge properly so called. But such wholesale downgrading of our cognitive achievements would be neither necessary nor warranted. On the contrary, it would be more reasonable to conclude that human knowledge does not require an epistemic perspective.
10
3. Preserving Virtue while Losing Perspective At the beginning of this essay I suggested that virtue perspectivism involves two central ideas: that knowledge is grounded in the reliable dispositions of the knower, and that reflective knowledge requires a perspective on those reliable dispositions. I then argued that the requirement of an epistemic perspective is psychologically implausible, given the way that such a perspective must be characterized to effectively address the generality problem and the meta-incoherence problem. In this final section I will argue that the requirement is not needed anyway, since Sosa’s virtue reliabilism has resources for solving both problems without invoking the idea of an epistemic perspective. As Sosa notes, the meta-incoherence problem is a problem about internal justification. But there are ways of defining internal justification within a virtue theory, and without invoking the idea of a perspective. In fact, Sosa gives us an important one of these. Consider the victim of Descartes’s evil demon, whose beliefs seem to be justified in all respects, internally speaking. Sosa argues that we may define a relevant sense of internal justification by noting that there are two ways that a belief can fail by way of reliability. One way is that something goes wrong “from the skin inward.” For example, the subject might fail to respond appropriately to her sensory experience, or might fail to reason appropriately from her beliefs. Another way to go wrong, however, is “from the skin outward.” Perhaps there is no flaw to be found downstream from experience and belief, but one’s cognitive faculties are simply not fitted for one’s environment. It is only in this second way that the demon victim fails: internally speaking, she is in as good working order as we are. But then there is a straightforward sense in which even the victim’s beliefs are internally justified, Sosa argues. Namely, they are beliefs that result from intellectual virtues. We need only add that whether a cognitive faculty counts as a virtue is relative to an environment. The victim’s perception and reasoning faculties are not reliable in her demon environment, and hence are not virtues relative 11
to her world. But those same faculties are reliable, and therefore do count as virtues, relative to the actual world. Accordingly, we have a sense in which the demon victim’s beliefs are internally justified although not reliably formed. It is plausible that the meta-incoherence problem raises the issue of subjective justification as well as that of internal justification, where subjective justification pertains to how things are from the believer’s point of view. Even if there is typically no perspective on one’s reliable dispositions, the fact that a person interprets experience one way rather than another, or draws one inference rather than another, suggests an awareness of sorts of the reliability of one’s evidential grounds. Or at least this is so if the person is trying to form her beliefs accurately in the first place—if the person is in the normal mode of trying to believe what is true, as opposed to what is convenient, or comforting, or politically correct. We may use these considerations to define a relevant sense of subjective justification: A belief is subjectively justified just in case it is produced by cognitive dispositions that the believer manifests when motivated to believe what is true.
In cases of knowledge such
dispositions will also be virtues, since they will be objectively reliable in addition to being well motivated. But even in cases where the believer is not reliable, she may nevertheless have justified beliefs in this sense, since her believing may nevertheless manifest well-motivated dispositions. Is the belief of the newly endowed clairvoyant internally and subjectively justified, in the abovedefined senses? Only if it manifests reliable and properly motivated dispositions, including dispositions to recognize and respond to relevant counter-evidence. But what if it does? Then such a clairvoyant is a far cry from cognitive agents like us, and I find it less plausible that she lacks justification or knowledge. It would be odd, in fact, to think that only cognitive faculties like our own can generate justification and knowledge. And, of course, Sosa does not think this. The generality problem is more difficult, and I can only point to a strategy that seems promising for solving it. Once again, my claim is that Sosa has resources for addressing the problem within his
12
own virtue theory, and without invoking the requirement of an epistemic perspective. When discussing the generality problem, Sosa often suggests that our cognitive faculties are natural kinds: “Just how fields are to be defined is determined by the lay of interesting, illuminating generalizations about human cognition, which psychology and cognitive science are supposed in time to uncover.” (KP, 236) Sosa also stresses the social importance of our epistemic concepts and practices. For example, he writes: We care about justification because it tends to indicate a state of the subject that is important and of interest to his community, a state of great interest and importance to an information-sharing social species. What sort of state? Presumably, the state of being a dependable source of information over a certain field in certain circumstances. In order for this information to be obtainable and to be of later use, however, the sort of field F and the sort of circumstances C must be projectible, and must have some minimal objective likelihood of being repeated in the careers of normal members of the epistemic community. (KP, 281-2) These passages contain two central ideas: a) that our faculties are and/or involve natural, projectible kinds, and b) that because of this they serve certain of our purposes as informationsharing beings. He then goes on to invoke a third idea: that of an epistemic perspective. For it is through our cognizance of such relevant F and C that we grasp the relevant faculties whose possession by us and others makes us dependable informants and cognizers. What is more, it is precisely by grasping how one does oneself have such animal aptitude over a certain field F in certain circumstances C that one bootstraps up to a higher-level of reflective justification. (KP, 282) Why not keep the first two ideas while dropping the third? In other words, we can say that a disposition D(F/C) is relevant for knowledge just in case it involves an F and a C that are both a) natural and projectible, and b) have some minimal objective likelihood of being repeated in the careers of normal members of the epistemic community. Of course such F and C must also be defined narrowly enough so as to make D(F/C) highly reliable. Which Fs and Cs these turn out to be will depend on the psychological and sociological facts about knowledge production and knowledge sharing. But it will not depend on the relevant Fs and Cs being represented by anyone.12
13
1 “Intellectual Virtue in Perspective,” in Knowledge in Pespective: Selected Essays in Epistemology (Cambridge: Cambridge University Press, 1991) (KP), p. 292.
“Perspectives in Virtue Epistemology: A Response to Dancy and BonJour,” Philosophical Studies 78 (1995), p. 233. Reprinted in Axtell (2000).
2
For example, see “How to Resolve the Pyrrhonian Problematic: A Lesson form Descartes,” Philosophical Studies 85 (1997) and “Post-script to ‘Proper Functionalism and Virtue Epistemology’,” in Jonathan Kvanvig , ed., Warrant in Contemporary Epistemology (Lanham, Md: Rowman&Littlefield, 1996). 3
4 “Virtue Perspectivism: A Response to Foley and Fumerton,” Philosphical Issues, 5, Truth and Rationality (1994) (VP), p. 41. See also KP, 278.
KP, 274. Later in the same paper Sosa suggests that the relevant F and C are those appropriately usable by the knowledge attributer rather than the knower herself. Of course, the two are the same in cases where one attributes knowledge to oneself. See pp. 282 and 290. The objections I raise below are not affected by the choice between these two options. 5
"Nature Unmirrored, Epistemology Naturalized," Synthese 55: 49-72. Reprinted in KP, p. 94. See also “Reliabilism and Intellectual Virtue,” KP, 132.
6
7
See for example J. Hochberg, Perception (Englewood Cliffs, N.J.: Prentice Hall, 1978).
That is, such factors are not represented by the subject. According to some current theories in cognitive psychology, many states not represented by the subject are represented in a sub-personal way, for example in sub-personal mechanisms operative in perception. Of course, it is representation at the personal level that is at issue regarding Sosa’s claims about an epistemic perspective. This qualification is intended in what follows below.
8
9
See L Weiskrants, Blindsight: A Case Study and Implications (Oxford: Clarendon Press, 1986.
For an informative review of relevant literature, see Paul Lewicki, Thomas Hill and Maria Czyzewska, “Nonconscious Acquisition of Information,” American Psychologist 47, 6 (1992): 796-801. 10
11
Ibid
12
I would like to thank Hilary Kornblith for his helpful comments on an earlier draft.
14
