Text: Ernest Sosa, "Knowledge and Intellectual Virtue" (1985)
In the 1980s, a new type of account of knowledge appeared on the scene. The inspiration for it was taken from ethical theories that revolve around virtue. Just as such theories are called "virtue ethics," we now talk of "virtue epistemology." For a good summary of virtue epistemology, see the entry by Joseph Greco in the Stanford Encyclopedia of Philosophy. A 1998 paper entitled "What is Virtue Epistemology?" can be found here.
One of the founders of virtue epistemology, and perhaps its most prominent spokesman, is Ernest Sosa, professor at Brown University. The text for the present lecture was one of the seminal papers in the area. It works out a suggestion made in Sosa's 1980 paper, "The Raft and the Pyramid" (Midwest Studies in Philosophy V, 1980)
Sosa's Motivation for Virtue Epistemology
In the 1980 paper, Sosa was concerned with the dispute between foundationalists and coherentists. In the process of working out the details, he used an anti-foundationalist argument with the following premise: "a(iii) If a mental state provides no guarantee against error, then it cannot serve as a foundation for knowledge" (p. 6). Sosa invoked an analogy from ethics to argue that this premise "seems more likely false" than obvious.
The moral evaluation of an action can be based on one of two factors: "their being optimal (best of all alternatives, all things considered) and their being (subjectively) justified." Take the case of the doctor who delivered the baby Adolf Hitler. It could be argued that "if he had acted less morally, that would have proved better in the fullness of time" (p. 8). But the doctor would not have been morally justified in undertaking an action that would have ended baby Adolf's life.
Similarly, Sosa argues, believing only on the basis of a mental state that provides a guarantee against error would produce optimal results. But in fact, it is reasonable for us to have beliefs that not only are fallible, but are in fact false. "That seems to me not just a conceivable possibility, but indeed a familiar fact of everyday life" (p. 8).
Suppose this is right. Then what is the basis for justified actions and reasonable beliefs that fall short of optimality? The answer is: stable dispositions to act. The doctor has a stable disposition to try to help all of his patients and to harm none of them. This disposition is a better one than a stable disposition which would allow the doctor to bring harm. It is a justified or virtuous disposition. Sosa calls the justification of a disposition "primary justification." It is based on the "greater contribution of value when compared with alternatives" (p. 23).
Although he does not invoke him in the article, Sosa could have noted that Aristotle judged the superiority of dispositions by their moderation (Nichomachean Ethics, Book II, Chapter 6). Thus courage is superior to rashness and cowardice, which are extremes of excess and deficiency, respectively. For Aristotle, a virtue is an excellence in discharging a function. "Every virtue causes its possessors to be in a good state and to perform their functions well" (Chapter 5). So courage, which has what Sosa calls "primary justification," is a virtue, an excellence in action, while rashness and cowardice are vices.
Sosa goes on to say that "secondary justification" pertains to actions. An action is justified when it has as its source a virtuous disposition. So the doctor is secondarily justified in helping the baby Adolf due to the fact that he acted out of a primary disposition which is virtuous.
Now this can be applied to a particular sort of action, the acquisition of belief. A belief is secondarily justified insofar as it has as its source a stable disposition which is primarily justified. This would be an intellectual dispositions, rather than a disposition of character (as Aristotle called courage and the like). What makes an intellectual disposition primarily justified is the degree of its "contribution toward getting us to the truth." Thus he called his position "reliabilism," but we can see that it is more specific than the Goldman's version of reliabilism we discussed in an earlier lecture.
The 1985 article with which we will be concerned here begins with an assumption about what an intellectual virtue is: "a quality bound to help maximize one's surplus of truth over error." The reason this is only an assumption is that there may be other facets of intellectual virtue, such as "generality, coherence, and explanatory power," which might be not be explained in terms of truth-maximization. But this further issue is summarily dropped. Intellectual virtue is considered strictly as a teleological or end-directed concept.
More generally, a virtue (viewed teleologically) must lead to its end in a way that is not accidental. Which intellectual virtues meet this strong requirement? Sosa suggests "rationalist intuition and deduction," as what "would always lead us aright." Let us take Descartes as a paradigm rationalist. If he was right about "rationalist intuition" (what he called in the Meditations "clear and distinct perception"), then belief formed on that basis must be true, and therefore the use of that faculty would maximize one's surplus of truth over error. Logical deduction, if carried out properly, cannot lead a person from truth to falsehood. But Sosa departs from Descartes in asserting that these two faculties these have very narrow scope, and we have to ask whether any other human faculty might be an intellectual virtue.
Generation and Transmission Faculties
Logically, there are two kinds of faculties whose output is belief. A "generation" faculty has non-beliefs among its inputs, while a "transmission" faculty has nothing but beliefs as inputs. Thus rationalist intuition would be a generation faculty and rationalist deduction a transmission faculty. We can think of reason itself as a single faculty with both generation and transmission sub-faculties. There are other things that are traditionally recognized as faculties of the human mind, including non-deductive inference as well as "perception, introspection, and memory." If these can be counted as intellectual virtues as well, then the scope of belief from virtuous faculties can be increased tremendously.
Memory can be regarded as a transmission faculty, and a good one at that. Beliefs one actually remembers (as opposed to apparently remembers) are exactly the same as the original beliefs. In this way, memory is even more virtuous than deduction, since in deduction a new belief can be formed, so the process is a little more complicated and perhaps not as reliable. Sosa concludes that a rationalist should count memory as a virtue.
Descartes himself thought that memory could be defective, and that we need to know that it is not defective before we can trust it entirely. But if there is a defect, it is not with memory as Sosa defines it. Rather, the problem is with apparent memory, which may come "out of the blue" and is not a reproduction of an earlier belief. So Descartes's question would be how we can reliably distinguish merely apparent memories from real memories.
The same problem affects what Sosa here calls intuition, which
poses the same kind of problem as memory, as Descartes was aware.
In many cases there is the appearance of clear and distinct
perception when in fact the perception is unclear or indistinct.
Sosa describes intuitive reason as the ability to grasp simple
necessary truths and accept them as true upon considering them. A
person with a highly virtuous intuitive reason might still believe
what he does out of the blue, due to some cause other than
intuition. The lesson here is that if even rational intuition
allows room for error, a rationalist must allow memory room for
error as well.
Perception seems to be like intuitive reason in the sense that it functions as a generation faculty. It is broken down into two types: "external perception of the senses" and "internal perception of introspection." Both are capable of generating beliefs out of non-belief states (or belief states at a lower level). So, sight generates beliefs "about the colors and shapes of surfaces seen fully, within a certain range, and in adequate light." The problem raised earlier recurs with the possibility of a person with good vision but subject to hallucination. The virtue of the faculty, internal as well as external, seems not to be removed by the possibility of error, as long as the virtue of memory and intuition are not threatened by error.
It is possible to think of perceptual belief as based on an
experience as if p. If
perception is then tied essentially to experience, we would not be
able to say that it could remain virtuous if it frequently leads
one astray. That is, it may be that things are not the way that "as
if" experience says they are. But this possibility does not touch
memory or reason, because we do not always have an "as if"
experience when using those faculties.
Virtuous Faculties and
What makes a faculty intellectually virtuous is the powers it has. The use of that faculty must lead to a favorable ratio of truth over error as compared to "feasible competitors." It is not a lack of virtue when one has false beliefs about what the faculty delivers, say believing of what he remembers that he did not remember it.
We evaluate generative faculties by the truth-to-error ratio of their total output of beliefs. But with transmissive faculties, virtue should be measured only the truth-to-error output relative to true inputs. So propositional memory and reasoning (inductive and deductive) "deliver a truth/error differential apparently undiminished by falsehood. But the generative faculties of perception do seem to be diminished by falsehood.
In the case of introspection, it seems that we can err with respect to the identification of what mental state we are actually in, as with beliefs about the number of sides in an image we are imagining. And external perception flirts more dangerously with error, as it depends on factors external to us. So a perfectly virtuous faculty of external perception could still be subject to error.
Even though introspection is not subject to this kind of error, it is still subject to error. We can, through practice, reduce the chances for error, but the risk remains due to the possibility of abnormality. "Such abnormality could derive from a variety of causes including hypnosis, brainwashing , and neural engineering.
Some people might object that the faculty of pure introspection could not introduce error. If we get the number of sides in the figure wrong, it is due to the process of counting. If we are hypnotized into having a belief about what is going on in our minds, it is the hypnosis that causes the problem. In this way, introspection is seen as just as immune from error as rational intuition. But then the same could be said for external perception.
The problem with "purism" of this type is that we would then have to credit someone with virtuous perception despite a great prevalence of error. The sense of "perception" that is relevant to epistemology is not pure, but includes factors that could be called "external" to the pure perceptual faculty. It will not do, from the standpoint of knowledge, to say that a person with excellent eyesight has a virtuous faculty of perception when he is subject to frequent visual hallucination.
This holds against another defense of the virtue of faculties whose apparent output is false beliefs. You could say that erroneous memories, for example, are not real memories, but only "ostensible" memories. One never goes wrong when one has real memories, so memory is always virtuous. Sosa counters that "it still seems absurd" to credit someone with a virtuous faculty, good memory in this case," when a person is frequently "duped" by bad ostensible memory.
We get a clue as to the absurdity of such a claim by noting that we are talking about virtue "in an epistemically most relevant sense." So then we could think of being mostly right in normal circumstances as a necessary, but not sufficient, condition for intellectual virtue. What would be sufficient is the same kind of thing the Stoics required, that when conditions are normal, only the property that the faculty is supposed to detect will yield the sensory images of the faculty.
This requires that we attribute to the faculty "traces" (sensory images) of the "impact" perceptual objects have on it. The presence or absence of the traces would then be the guides for perceptual belief. The extent to which they indicate the presence of their object would be the measure of the virtue of the faculty.
But there is nothing like traces in the other faculties. Memory is the most promising case, in that we might view "ostensible memories" as traces. However, Sosa has already claimed that these are not generally memories but beliefs about what is retained in memory. He says that this kind of consideration also holds for introspection, intuition, and deduction.
This raises the question of whether error is possible in the faculties other than external perception. It seems that for this to be the case, there has to be some "sort of belief-guiding pre-belief appearance" to be the source of error. Sosa uses the example of memory to argue that there is no such thing. Anything we can call such an appearance can be due to some other cause than an element of memory itself.
For example, mis-remembering a phone number might be due not to an operation of memory but instead to guessing. Even if it feels like there is a remembering there, that feeling could have been induced by hypnosis, for example. The feeling might just be a feeling of confidence. Sosa concludes that the best response to this situation is to ask why not instead consider memory and the other faculties as infallible.
One sub-faculty that might seem prone to error is inductive or explanatory (abductive) reason. Sosa calls these both functions of "ampliative" reason. Through the use of this sub-faculty, we arrive at a coherent and comprehensive view of the world. And because this view can be a false one, the faculty is fallible. But since neither intuition, deduction, introspection, nor memory builds a world-view, they are not fallible in this way.
Virtue and Justification
A final reason we might take these faculties to be fallible is
that their use seems to issue in false beliefs that are nonetheless
justified. "This fits with our notion that a belief's justification
derives from the endowments and conduct that lie behind it. (This
is what was discussed at the beginning of this lecture.) We call a
false external perceptual belief justified when there is some
external factor which prevents it from being true. This does not
detract from its virtue: there is "no defect or misconduct in the
It may seem that there is also justified false belief in the
cases of memory, deduction, and intuition. Sosa gives examples in
each case. A person with excellent memory believes that David Hume
wrote An Essay Concerning Human
Understanding, which in fact was written by John Locke.
"Given his excellent memory proven right on countless occasions, is
he not justified in accepting what he ostensibly remembers?" If so,
this seems to allow us to say that memory is fallible.
Sosa counters that in each of these cases, there is something other than the faculty in question that gives rise to the false belief. In the memory case, it might be some sort of association that triggers the belief. (Sosa does not make this specific suggestion; in the case of an error in deduction, he says that it is just as plausible to suggest "blundering through inattention" as the source of error as to blame the faculty of deduction.)
This gives rise to the question of why we might think beliefs formed in this way are justified. The answer is that ampliative reason is the source of justification. The way to the belief is "roundabout." The subject believes that the faculty is reliable and justifiably attributes the belief to that faculty. This is what sustains the new belief, say that Hume wrote the Essay.
So the source of the belief is not to be found in the faculty which is its apparent source. Instead, it is "ampliative, coherence-seeking reason." An objection to this claim is that people do not go through this sophisticated kind of reasoning. Sosa responds first that the premises of the reasoning are only "implicit." He then appeals to a distinction he has not yet made, between animal knowledge and reflective knowledge.
What if ostensible intuitions, for example, could always not be real while being at the same time indistinguishable from true intuitions? In that case, what would be the value of rational intuition for theory of knowledge? This is precisely the key question for Descartes. He would respond that we can tell the difference. Sosa describes that process as one that that combines intuition and introspection ("intuition-com-introspection"). This process itself can be fallible.
But suppose there sometimes cases in which the process works very badly, or even that in most cases it malfunctions. The defender of intuition might respond that these cases are not relevant. One reason would be that this is fallibility concerning the sources of the belief and its justification. It does not impugn the faculty from which the belief actually arises or the justification. If it is required that one have justified belief about what makes his belief justified, an infinite regress could be generated.
Sosa is not compelled by these arguments. He questions whether it is not absurd "to allow that someone can know that p because his belief that p does derive from intuition even though his ostensible intuitions almost never turn out to be real. But if we base knowledge and justification solely on a virtuous source, this is going to have to be allowed.
There may be a way to understand why virtuous faculties are not sufficient for knowledge and justification, and that is by denying their infallibility. Sosa presents an alternative to the account of memory he had given at the beginning of the paper, where he had defined memory in terms of the preservation of belief.
We can instead first define what the faculty of memory is, and "only then define a retentive memory belief as a belief deriving causally from the exercise of that faculty. In this way, we might be able to say that the cause of a memory of a belief that p is something other than an actual earlier belief that p. This would leave open the door for fallibility of memory.
A faculty is a kind of disposition, which is an ability or a power. Sosa lists several examples of dispositional properties: being an "incline-roller," being a "level-roller," and being a "round-snug-fitter." A basketball and a bicycle tire would have all of these properties. What grounds its having these dispositions is "some intrinsic property or some set of intrinsic properties." It could be the roundness of the basketball and the cylindrical shape of the tire.
We can think of faculties either as dispositions or as intrinsic properties which ground dispositions. Sosa chooses the former, since if it were intrinsic properties, there would be two different faculties, one in the basketball and one in the tire.
Given that faculties are thought of as dispositions or powers, we can point to the way the powers are exercised. A faculty is defined as the ability to achieve accomplishments of a certain sort. Whether they can in fact achieve them is relative to circumstances. Our ability to detect the color of an object depends on its size, its distance from it, the absence of intervening obstacles, our subjective state, etc.
All that this requires, in the case of "species-wide accomplishments of a certain sort," is a cognitive faculty. And this description does not bring in any considerations of infallibility. The only thing that is needed is "a good success ratio." Common sense is not in a position to formulate "underlying abilities that are in fact infallible."
Sosa next makes a comparison between someone exercising an ability and the dance of a puppet. What really makes things work is not the powers of the puppet, but the abilities of the puppeteer. This is the agent or the responsible agent in the process. If we are hypnotized, we are like a puppet, not an agent. Our beliefs display some of our own powers, but the active agent is elsewhere. And this is not enough for knowledge. (Sosa does acknowledge that in some unreal circumstances, we could gain knowledge through hypnosis.)
The powers of human beings that give them knowledge and justification are those that allow us "to distinguish the true from the false in a certain subject field, to attain truth and avoid error in that field." Generally, in our ordinary habitat (or that which is ordinary when we form beliefs), "one would believe what is true and not believe what is false, concerning matters in that field."
How the fields are to be defined depends on our empirical investigation of human cognition. They must not be narrowed to achieve "artificially a sort of pseudoreliability." The fields should be "natural," having a "place in useful, illuminating generalizations about human cognition."
Sosa goes on to compare the reliability of a faculty with its degree of generality. We can make more reliable judgments when the faculty is more broad. So we have a faculty of discerning a surface to be a polygon: a shape with three or more sides. A more refined faculty is that of discerning the specific shape of a polygon, say that of having exactly seven sides (being a heptagon). One is better justified in making the judgment that the shape is a polygon because of the greater reliability of the more general faculty.
Pure rationalism, with perfectly reliable certainty, "is a failed epistemology." Reliabilism is a "more realistic successor." Sosa defines it as the claim that a belief is epistemically justified just in case "it is produced or sustained by a cognitive process that reliably yields truth and avoids error." He notes that there are two main objections to reliabilism. One questions whether reliability is necessary for justification, and one questions whether it is sufficient for justification.
It is dubious whether reliability is sufficient for justification because someone may in fact be reliable but believe that one is not. So, a person may actually be clairvoyant, but "considers the deliverances of his gift to be inexplicable superstition, and has excellent reasons for this dim view." It may not be a necessary condition because an unreliable demon-victim may at the same time be allowed "in fairness" the same justification we allow for ourselves.
We can fix the first problem as Goldman did, by requiring that for justification, there must be no undermining belief that is caused by a reliable process. Sosa interprets this as meaning that there is no reliable process that would interfere with the generation of the belief. This could include information about the reliability of the faculty in question.
But this fix does not deal with the demon-victim case. And this case carries with it an additional problem, that the high truth-ratio could be caused by external forces. One's memory, for example, would take a correct input and yield an incorrect output, except for the interference of "extremely rare and random inversion-preventing radiation." It would be reliable but would not seem to amount to knowledge.
What is missing here? Sosa makes a comparison with a Gettier case and Goldman's barn case. In both kinds of cases, one has a true belief only by virtue of luck. The problem for reliabilism, on Sosa's assessment, is not that it is a bad theory of justification, but rather that the kind of justification it yields is not sufficient for knowledge, when combined with true belief.
Sosa gives an example of a case where one believes for a bad reason (reading a horoscope) but is right because it is no accident that the horoscope says what it does (the reading being planted by someone hoping to shape the person's belief). The person's belief is reliable in a straightforward way, but he lacks knowledge because he lacks justification. "his reason for trusting the horoscope is not adequate--to put it kindly."
The Value of Virtuous Intellectual Faculties
Being produced by a virtuous intellectual faculty provides prima facie justification, which can be over-ridden. But it is practically important for a person to know "how reliable and trustworthy his own judgments are in various categories." This also practically important for other members of the community, because of our reliance on testimony.
This explains the importance of justification. "What interests us in justification is essentially the trustworthiness and reliability of the subject with regard to the field of his judgment, in situations normal for judgments in that field." This is also supposed to explain why external factors do not matter for justification. "What we care about in justification are the epistemic endowments and conduct of the subject, his intellectual virtues."
Two Kinds of Knowledge
At this point, Sosa draws a distinction between two kinds of knowledge:
Reflective knowledge is better than animal knowledge because the addition of the understanding makes the belief more likely to be right.
Humans have knowledge that is more sophisticated than that had by non-human animals. The reason is that we have automatic functions for monitoring our background information and sensory input for contrary evidence. And when we are responding to sensory stimuli, we go for what is the most coherent hypothesis. "Reason is always at least a silent partner on the watch for other relevant data, a silent partner whose very silence is a contributing cause of the belief outcome."
Both kinds of knowledge "require a true belief whose justification by its source in intellectual virtue is prima facie but not overridden." To override such a belief requires appeal to a wider intrinsic state of the subject. The conditional probability of the truth of the belief given the wider state is less than the probability given just the virtuous state.
Defeaters can be either opposing or disabling. If the belief in question is that one sees a pink surface before one, an opposing overrider is that there are no pink surfaces in the room. A disabling overrider is that there is a red light shining on the surface. What makes acceptance of the overriding testimony something that defeats knowledge?
The reason is that it destroys "rational coherence," which we regard as "the best overall guide." It is an irony, Sosa claims, that justification is lost even when someone has a true belief based on the operation of a virtuous faculty and the contrary testimony is false. The person is no better off than someone without that contrary testomony. If mature enough to appreciate the testimony, "one errs not to give it due weight, which may betray a flawed epistemic character." This affects not only others, but one's "own higher self."
Some incoherence does not reveal flaws in one's epistemic character. We do recognize generally that we are capable of error, as in the case of someone who notes in the preface to a book that it undoubtedly contains errors. Nor is the skeptic guilty when he believes things in the course of a normal day. Neither the careful author nor the careless skeptic is guilty because the consequences of withdrawing belief in the face of skeptical arguments or the possibility of error are directly opposed to "one's prospects towards getting in the best relation to the truth."
Sosa concludes by summarizing his position. Knowledge is the outcome of the exercise of a virtuous intellectual faculty. And there are two sorts of knowledge. But reliability is not sufficient for justification in either animal or reflective knowledge. A reliable "pure" process of true vision can be undercut by external interference in the production of belief, producing spurious perceptions. So he does not have the ability to discern truth from error over a whole field of beliefs. His faculties are not reliable in the broader sense required for the kind of justification relevant to knowledge.
The last paragraph refers to other material in a book of essays in which this paper originally appeared. We cannot pursue these matters here.
[ Course Home Page | Menu of Lectures ]