If description of mental processes and evaluation of agents and their beliefs are rightly to be considered as complementary concerns on any plausible construal of the epistemological project, then this relationship cries out for explanation. For the complementarity of these concerns is hardly straightforward: One cannot epistemically evaluate a belief without knowing how it was formed, a largely if not wholly a scientific question; on the other hand, epistemic norms are and must be used to evaluate our scientific beliefs and theories, and apparently then require a basis at least partly independent of scientific matters of fact.

This aporia pits the proponent of naturalized epistemology, which asserts a continuity of epistemology and the cognitive or "special sciences," against the non-naturalist as the traditional champion of epistemology’s normative functions. To be sure, there is middle ground in this dispute, the prospect for a pragmatic or "normative naturalism." Consider the naturalistic turn in epistemology from a historical perspective. Various solutions to the Fact/Value problem have been posed under the rubric of "epistemology naturalized." Some are eliminative of epistemology’s normative tasks, and those that are not eliminative hold that the development of normative epistemic standards depends closely on psychology and the cognitive sciences. Eliminative solutions are not new. Before the turn towards naturalized epistemology, the logical empiricists had their own eliminative solutions. They attempted to reduce epistemology not to psychology, but to logic; not to the synthetic, but to the analytic. The rational reconstruction of scientific decision-making was not thought of as a normative undertaking, but was rather glossed as simply part of "the logical analysis of the language of science." The turn away from logical analysis in the 1960s and towards a naturalism that entailed a conception of the "supervenience" of normative criteria on natural fact arose partly in response to explicit forms of non-naturalism. But it responded more directly to the failure of logical empiricism to arrive at criteria of scientific theory-choice by way of the analytic method of conceptual analysis. As has been the case many times already in the twentieth century, the Fact/Value problem was transformed with this newer, naturalized conception of the epistemological project, but reappears in a new form.

Most of us think of ourselves as good naturalists now, and while the naturalistic turn was at first identified with eliminative approaches for some thinkers, most of us may find it hard to think of those ‘closet intuitionists,’ the logical empiricists as naturalistic. The point that they, too, sought to eliminate evaluative language from "the logic of science" should, then, serve as a reminder that the logical empiricist legacy may still exude a residual influence over us. This is an influence by which we tend to perceive the only viable naturalistic models as those that, if not eliminative of normative tasks of epistemology, at least show epistemic evaluation to be sharply contrasted with other evaluative uses of language. At the dawn of a new century, we may still labor under the residue of deep "metaphilosophical" assumptions of a sharp separation between two kinds of judgment, and therefore also between their corresponding forms of philosophical analysis. In the logical empiricist system of thought, the objectivity of epistemological rational reconstructionism, and of "the language of science" in general, was in no small part predicated on its contrast from the presumed subjectivity (or non-cognitive status) of ethical judgment and "the language of morals." David Carr has recently made a closely related point when he wrote, "Thus it is well-known that whilst prescriptivists are willing to concede the possibility of a descriptive sense of goodness or value in the non-moral realm they nevertheless insist that these terms must have a non-descriptive ‘evaluative’ sense in the moral sphere." In still broader terms, this disparity is what Alan Gewirth, in identifying a primary dogma of empiricism, thought of as being behind the unholy combination of a "normative science and a positive ethics"; it is similar to (and also motivates) the combination of "metaphysical realism and moral relativism" that Hilary Putnam has identified and criticized in thinkers such as Bernard Williams.

It is often said that the logical empiricists were foundationalists. But it is worthwhile noting that they were also sceptics, only quite selectively so. They utilized many arguments taken directly from the tropes of the ancient sceptics, and advanced them against the idea of moral knowledge, or of an "ethico-cognitive parallel" (Reichenbach). The rhetorical point of such contrasts was to show that scepticism undermines claims to the objectivity or rationality of ethical judgment, but not to that of scientific judgment. I am concerned with the upshot of these metaphilosophical assumptions, and we need not rehearse in any detail the arguments that contributed to the breakdown of this logical empiricist worldview and the rise of naturalized epistemology. Let it be enough to say first, that these arguments came both from inside and outside the empiricist camp, with figures such as Thomas Kuhn and W.V. Quine sitting ambiguously but intriguingly on its borders; and second, that these arguments involved divulging the implicit intuitionism of the logical empiricists in their treatment of rational reconstructions of scientific decision-making and other aspects of epistemic evaluation. The upshot, as I understand it is closely akin to what Putnam calls the ‘companions in guilt’ argument against the above-mentioned combination of metaphysical realism and moral relativism. More specifically, a sceptic’s slant on that breakdown and the lessons to be taken from are essentially this: Tu Quoque! You yourself do it! You (filth Anglo) logical empiricists are our ‘companions in guilt,’ for your rules or ‘algorithms’ of rational reconstruction are really, as Kuhn argued, ‘values at work’ in science; and if your sceptical arguments succeed, as you claim, in undermining the objectivity of ethical judgment in the choice of normative ethical theory, then they prove far too much for your own good, for they equally well impugn the objectivity of scientific judgment in theory choice!

Quine’s neo-pragmatic approach to language led him to see that Carnap’s distinction between the analytic and the synthetic, another edifice of this logical empiricist program, was contextual, not absolute. Quine’s way of handling the normative dimension of epistemology was generally eliminative, as is famously captured in his description of epistemic normativity as merely ‘a branch of engineering,’ implying its reducibility to a purely instrumental form of reason. But this view, part of Quine’s conception of epistemology reduced to psychology, may simply beg many of the interesting questions raised in Kuhn’s work on "cognitive values." Might it not invite the equally trenchant reply of the Deweyan pragmatist—that the distinction between instrumental and intrinsic value is also (and for many of the same reasons) not absolute, but a matter that can be determined only contextually?

One point more to make before we can relate this discussion back to the Fact/Value problem. Barry Stroud, an epistemologist and noted scholar of the history of scepticism, writes as follows:

If it is correct that scepticism makes its greatest challenges when so addressed to particular spheres, this highlights the poverty of the sharp separations and asymmetries in thinking about evaluative language that characterized the metaphilosophical views of twentieth century logical empiricism. The scientistic strategy of division—of persuasive definition and of contrast between scientific epistemology and ‘mere’ evaluative language, whether in the positivist or even the Quinean naturalistic mode—is a dangerous game. For it plays into the hand of the most serious forms of scepticism. When that strategy can no longer be carried through, when it becomes apparent that an unacknowledged intuitionism has been present from the start, the sceptical upshot remains. Not only does it remain, but it reflects back with still greater force upon the one whose strategy it was to employ the skeptic’s tropes in order to affect the desired contrast. And not the recognition of evaluative components in scientific reasoning, but the perceived philosophical implications of that recognition, is part and parcel of our era’s excessive scepticism about the viability of epistemology as a normative discipline. The question of the ‘place of normativity in naturalized epistemology’ is just the contemporary version of the older unsolved positivist quandary about ‘the place of values in a world of facts,’ and our current inability to come to grips with it, I suggest, reflects equally poorly on our ability to understand knowledge and valuation under a single, unified philosophical perspective.

To summarize my point now in a positive manner, while the science/non-science distinction remains of importance, there is distinct advantage in a metaphilosophical approach that redresses the excessive asymmetries characterizing empiricist thought regarding the relationship between epistemic and other forms of evaluative discourse. This relationship between epistemic and other kinds of evaluation must be of deep concern in any naturalistic philosophy broad and pragmatic enough to address itself to the human agent as both a reasoner and a valuer. If there is to be a less utopian approach to the Fact/Value problem than eliminativism, it can only take place around a new consensus on the need for a more symmetrical treatment of epistemology and ethics as central normative subdisciplines of philosophy.

The issues that I have raised in providing this retrospective look at the Fact/Value problem are issues of a metaphilosophical nature. This is the level at which I see contemporary virtue theory beginning its reconstructive account of human agency. Although it was a backbone of his critique of logical empiricism, Kuhn’s term "cognitive values" is rarely heard today, and most epistemologists are concerned with knowledge and justification generally, rather than scientific theory choice. But Kuhn’s essential lesson has not been lost, only transformed during the turn to naturalistic epistemology. Increasingly in recent years, there is growing mutual interest and overlap between those doing work on ethical and on intellectual virtue. Many problems central to ethical agency, problems that have been addressed in virtue ethics such as the relationship between "responsibility" and "luck," are reappearing across subfields of philosophy as problems equally relevant to a conception of epistemic agency. Virtue theorists understand the justification of beliefs to be related to intellectual virtue in a way substantially analogous to the way that the rightness of acts relates to moral virtue. If these analogies between ethical and epistemic evaluation prove strong, it will invite if not demand an account of human agency unified enough to be applicable to fields.

In the language of virtue epistemology the search for an unified account of ethical and epistemic evaluation is simple the search for a unified account of the virtues. Or at least the latter is a substantial step towards the former. Some have suggested these as primary goals of contemporary virtue theory. But the perceived prospects and advantages of a unified theory of the virtues are matters of contention, even among virtue epistemologists. I believe that its prospects are strong, and that there can be grounds for a meeting of minds on the sound form such an account should take. Though I may not get far in developing such an account here, I hope to clear some ground that allows virtue epistemologists of various persuasions to see the advantages of such a "unified account" as part of a response to the perennial Fact/Value problem. To begin this ground-clearing project, the next two sections of this paper will address some of the issues that virtue epistemologists debate among themselves, and some of the ways that dispute becomes exasperated. Examining how oppositions about the definition of intellectual virtue are understood by virtue epistemologists will help to clarify which differences are substantive, and which are likely explainable as the result of divergent interests in the complex topic of epistemic agency. In the final two sections we will then reverse course, exploring the extent to which there is hidden complementarity behind oftentimes sharply expressed differences. We will there examine the shared goal of a "mixed externalist" account of epistemic justification, and how this goal can itself be seen as one step in establishing the form and function of a viable unified account of the virtues.

The best-known version of virtue epistemology, that of Ernest Sosa, is a form of epistemic reliabilism, and that is the identification that virtue epistemology basically has for many philosophers. But reliabilism has been criticized from within the fold of virtue theory for assuming a quantitative, maximizing goal, and for not promoting a "genuinely aretaic" approach. To see how this objections arise, consider Sosa’s proposed definition of intellectual virtue as "a quality bound to help maximize one’s surplus of truth over error." Here intellectual virtues are defined in terms of their truth-conduciveness, so that the aim of truth is taken as conceptually prior to and independent of the concept of a virtue. Jonathan Dancy’s characterization of the reliabilist view as "virtue consequentialism" seems quite fitting. This characterizations, however, and especially Sosa’s one-time description of his account as "epistemic rule-utilitarianism" (which term he borrowed, perhaps unfortunately, from Roderick Firth) have raised questions in the minds of others who take virtue theory as a non-consequentialist approach.

Sosa’s broader virtue theory, which to date remains largely implicit, endorses a strategy for defining both intellectual and ethical virtue. To quote again, "Intellectual virtues are ...ingrained modes of intellectual procedure that tend to give us truth, just as practical virtues are ingrained modes of practical conduct that tend to yield what is good." Sosa has defended this view by arguing both its support in classical Greek virtue theory, as well as its advantageousness and desirability from the perspective of contemporary philosophic naturalism. But as we will see, he is challenged on both these fronts of classical support and contemporary desirability. These challenges are usually framed and developed in the work of a strain of virtue epistemology that I’ll term "virtue responsibilism" in order to clearly mark it off from the Sosa’s branch, which I’ll hereafter refer to as "virtue consequentialism" or (to may make the contrast with responsibilism clearer in some contexts) "virtue reliabilism." Sosa intends his truth-linked definition of the intellectual virtues to respond to a demand for normative criteria of justified belief, as beliefs formed through the instancing of one or more intellectual virtue. This demand for normative criteria is part of the perceived incentive for a consequentialist or goal-linked account of the virtues, since virtue theory is often considered more difficult to maintain if virtues are treated as primary rather than as derived. Indeed, it is often claimed by consequentialists in ethics that if Aristotle did intend to make the virtues of character primary, and definitive of the end (eudaimonia) itself, then his account suffers from circularity.

On Sosa’s account, intellectual virtues are any truth conducive cognitive faculty or stable cognitive disposition, including genetically-granted ones such as the various perceptual faculties, and the transmission faculties of deduction and memory. Virtue responsibilists dispute such definitions. The responsibilists’ departure from Sosa comes in thinking of virtues as accruing to their possessor rather than to the faculties themselves. If justified belief is in part a matter of epistemic responsibility, as right action is a matter of ethical responsibility, then attributions of virtue are more appropriately assigned to persons than to faculties. Linda Zagzebski like several other critics of Sosa including Jonathan Dancy, points out that the account of intellectual virtues as faculties like memory, sight, etc. runs counter to Aristotle’s account of virtue. Virtue for Aristotle is a hexis, a state or disposition acquired either by habit (ethical virtues) or teaching (intellectual virtues); it is one that is "in accordance" with but not "by" nature. This should be enough to dispel such an identification of genetic faculties or natural capacities with intellectual virtues, Dancy and Zagzebski have both argued.

Sosa’s way of responding to these objections is to argue that such a definition, while perhaps not Aristotelian, has a sense that is recognizably "still Greek." Plato is an obvious source for a broader conception of virtue. Since this is one recognizes non-moral goods, even prescriptivists (as Sosa once described himself as being) can see in it a possible descriptive sense of goodness or value. But Sosa can point out that intellectual virtue is presented quite differently than ethical virtue even in the context of Aristotle’s thought. The latter are seen as dispositions to choose the mean, or to deliberately desire the mean. By contrast, intellectual virtue for Aristotle, is a dispositions to hit the truth, and Aristotle therefore has a straightforwardly consequential definition for intellectual virtue, whatever be his views about moral virtue. So to quote Sosa’s direct response, "as Dancy himself acknowledges, if one tries to oppose the consequentialist answer and to more closely assimilate the relation of the good and the virtues of character to the true and the intellectual virtues, we commit ourselves to some form of the doctrine of the unity of the virtues....But it does not seem as if Aristotle would have wished to assert the unity of the intellectual virtues. They are not defined in terms of hitting a mean, and the account he gives of them (some concerned with things necessary, others with things that may be or not be) makes it difficult to suppose that in the absence of one, one lacks all the others" (chapter 8).

Dancy (chapter 7) concedes that when one turns to Aristotle's account of the intellectual virtues, "he does seem to link them to truth in the way that Sosa wants. Aristotle does say that an intellectual virtue is a state by which one will most hit the truth" (NE 1139b12-13). But Dancy then makes the same move as Zagzebski and the virtue responsibilists: he opts to pass over Aristotle’s apparent asymmetries between the two kinds of virtue, and to treat the intellectual virtues in the way that he sees Aristotle treating the ethical ones. For Dancy a genuinely aretaic epistemology requires a pluralistic and holistic conception of the goals of our intellectual life. The consequentialist retorts, ‘What makes all these things virtues, if not the goal which they promote?’. There must be some characteristic unitariness to the intellectual virtues, or we could not know that in identifying virtues as such, we are not merely ‘picking out a motley crew,’ as Sosa puts it. Dancy’s answer: "There is an apparent unity in Aristotle's account of the moral virtues, but hidden diversity...The source of unity is the way in which a good character is a well-knit one, not some end which good character would promote." Somewhat like John McDowell, Dancy is saying that praise and blame, for a true virtue theorist, will be mediated by consideration of the entire sort of life that surrounds this failure or that success: "there will be a holistic aspect to our moral and to our epistemic assessment."

This claim clearly draws upon Aristotle’s ethics as its model, extending this model to include intellectual virtue; it draws more specifically on Aristotle’s indication (again explicitly only within his ethics) of an internal rather than an external relationship of means to ends, which serves to distinguish ethical virtue from skill, and moral and evaluative from productive reasoning. A skill is causally separable from the outcome it is used to produce, but an ethical virtue cannot likewise be conceived as separate from the ends of truth and justice; these latter are part of what we mean by that human flourishing which honesty or fairness serve to promote; they are constitutive of those ends rather than just causally productive of them. Responsibilists like Dancy and Zagzebski have also noted that the reliabilists’ epistemic goal of maximizing true beliefs and minimizing false ones is not as easily discernible as it may appear: it may well be two goals, at times in tension with each other. We need to find a proper balance of strength and security in our practices of belief acquisition, at times perhaps even committing ourselves to brave error in order to enhance understanding. Maximizing true beliefs (‘Jamesian reliability’) will depend upon the virtue of intellectual courage, while minimizing false ones (‘Cartesian reliability’) leans on the virtue of intellectual caution. If so, this would certainly support Dancy’s claim that there is hidden diversity in the epistemic as well as in the ethical goal.

Both reliabilist and responsibilist virtue epistemology is sometimes presented to us as being "strongly Aristotelian." Unfortunately, Aristotle’s thought may be misused as a means of arguing and mediating this ‘in-house’ dispute among virtue epistemologists, just as it often is in disputes among contemporary virtue ethicists. Yet what we have found is that for reasons of contemporary concern, virtue reliabilists and virtue responsibilists both depart from Aristotle in more ways than they admit. Consider that neither group splits the field of virtues in the Aristotelian manner, offering a consequentialist account in one area and a non-consequentialist account in the other. Both, in other words, want more symmetry in their own accounts than Aristotle’s appears to provide. They both take leave from Aristotle’s account of intellectual virtue, yet in revealingly different ways. The one attempts to understand all virtues as he sees Aristotle understand the intellectual virtues, and the other prefers drawing the model from Aristotle’s ethics or virtues of character.

To be more specific, the virtue reliabilists utilize the account they find in Aristotle’s treatment of intellectual virtues, yet they broaden the notion of virtue beyond states acquired either by habit or by teaching. They do not distinguish teleology form consequentialism, and while they draw analogies between ethical and epistemic evaluation, these analogies are typically quite thin because unlike responsibilists, they rarely express their own views on ethical evaluation directly. The responsibilists, on the other hand, express a robust sense of the primacy of the virtues, and this makes virtue theory in their eyes a more unique philosophical alternative. Their account of intellectual virtue exhibits a strong affinity with Aristotle’s account of the relationship of virtues and eudaimonia in his ethics. In fact they sometimes argue for subsuming the intellectual virtues under the moral virtues (Zagzebski), or at least for a far more holistic conception of our epistemic goal than Aristotle’s own account supports. Sometimes for instance it has been suggested that eudaimonia should be thought of as common goal of both the ethical and intellectual life. As a mark of their similarity with ethical virtues, several responsibilists including Zagzebski and Montmarquet have tried to provide a conception of the unity of the intellectual virtues analogous to Aristotle’s conception of the doctrine of the mean. Along with these differences from Aristotle, they also sometimes build into their conceptions of intellectual virtue the same kind of strong motivational requirement that Aristotle makes for the ethical virtues. Here again the responsibilist’s treatment of the relationship between the virtues appears to evoke a stronger symmetry between the ethical and intellectual than Aristotle intended. If consistently developed, the model here is not Aristotelian but neo-Aristotelian, in that it creates a subclass of "ethical virtues pertaining to intellect." The Greek thinker would not have understood the idea of a kind of ethical virtue that pertains to the intellectual sphere (intellectual courage, conscientiousness, fairness, trustworthiness, etc, all intellectual analogous of acknowledged ethical virtues).

It is fine that virtue responsibilist wants to throw emphasis upon this kind of disposition, a kind that Zagzebski rightly points out seems to have "fallen into the gap between ethics and epistemology" and been neglected to some extent in both fields. But the concern comes when responsibilists insist that this newly recognized class of dispositions is definitive of intellectual virtue. There are good reasons for thinking this approach to defining the virtues is overly restrictive. Lets call habits and dispositions of this kind (ethical virtues pertaining to intellect or better, to intellectual activity) CT virtues, for critical thinking. Thinking is, after all, an activity. It seems misguided to define intellectual virtue in terms of this class of CT virtues, as responsibilists do when they insist that only acquired traits can count as virtues, or when they place strong motivational constraints on all intellectual virtues. Why? Theoretically, such a definition displaces the broader, reliabilist class of faculties and dispositions to hit the truth (we could call these "cognitive virtues," retaining "intellectual" as the cover term for both kinds); at the same time, Zagzebski’s model places all CT and the specifically ethical virtues that they are analogues of under a general conception of virtue as a disposition to act morally. This is the position that subsumes or assimilates the virtues, the model of "a virtue epistemology based on virtue in the ethical sense." In essence, such a model is not neo-Aristotelian, but rather non-Aristotelian.

By contrast, the non-assimilationist model of virtue that my suggested terms would indicate—with "cognitive" virtues (‘passive’ faculties such as the perceptual) on the one side and "specifically moral" and "CT" virtues on the other—could likely be seen as neo-Aristotelian. But as I said above, the is nothing (except hopefully clarity) to be gained by such labels. The more important point is the simplest and most practical: that in order for virtue epistemologists to share a broad range of interests, they need to begin with a broad enough definition of intellectual virtue. I think this is also the case with respect to definitions of virtue theory itself, and it is to debate over that issue we can turn next.

In Virtues of the Mind (1996) Zagzebski’s main objections to Sosa are that "his model of moral theory is act-based, and his definition of virtue is consequentialist." By these charges she has alleged that Sosa’s approach does not really constitute a virtue theory. I will briefly rebut those arguments here. The latter of these charges breaks down into two separable issues, the relationship of virtues to rules, and the relationship of virtues to goals, both of which are discussed by Zagzebski. I will briefly take the three desiderata of a virtue theory in what seems the most straightforward manner.

The most sound criteria for distinguishing a virtue theory that Zagzebski proposes is in terms of relationships of priority between virtues to rules. This is a potentially strong desiderata, in that it gives us a typology based upon deep-level formal relationships. On this construal, all and only virtue theories in ethics define rules (rightness) in terms of the source of an action in an ethical virtue; and by analogy, all and only virtue theories in epistemology define rules (justification) in terms of the source of a belief in an intellectual virtue. This desiderata is very closely related to the change in the direction of analysis, that Greco uses to define virtue theory. He says virtue theories in epistemology define justification by the source of belief in intellectual virtue, rather than defining intellectual virtue as a disposition to believe rightly. By this criterion, however, Sosa’s approach surely constitutes a virtue theory, since his definitions of knowledge and justification in terms of their ground in intellectual virtues demonstrates the conceptual priority he gives to the virtues in relationship to rules.

Another important consideration, an extension of the same formal approach, comes from looking at the relationship between virtues and goals or ends. In Virtues of the Mind, Zagzebski says that this relationship distinguishes different forms of virtue theory: "…The question of which comes first, end or motive, is the point at issue between a virtue theory that is happiness-based [goal-based] and an agent-based theory of the form I have called motivation-based." Those that give priority to the goal, and define virtues in terms of conduciveness to that goal, are what she calls "Good-based" virtue theories. The priority relationship here would be of the form 1. End (truth) 2. Virtue 3. Rules (justification). Those that reverse the place of the end or goal, seeing virtue as derivative from or definitive of the goal, are what she calls "motivation-based" virtue ethics. Their order of organization would be 1. Virtue 2. Rules 3. End (or possibly 1. Virtue 2.End 3. Rules). For Sosa, the intellectual virtues are defined in terms of their truth-conduciveness, an objective matter viewed as independently attributable to cognitive processes. For Zagzebski on the other hand, truth is treated as a derivative concept in much the same way that the good is a derivative concept in her ethics.

Zagzebski certainly wants to argue that the form of virtue theory she calls motivation-based is best able (or perhaps even uniquely able) to account for the value of the motive to seek knowledge, and the greater value of knowledge when compared with the value of true belief. On Zagzebski’s preferred motivation-based account, the goodness of virtues is based on the goodness of the agent’s motives, and this form of goodness is conceived as intrinsic, not derived. These are interesting questions, but separate, it seems to me, from the question of whether "Good based" theories can constitute a form of virtue theory. Since in her book she allows that there is room within virtue theory to accommodate different views of the relationship between the virtues and the Good, to say as she sometimes appears to that Sosa’s account is not a form of virtue theory because its definition of virtue is consequentialist, would clearly be inconsistent. By her own criteria, "virtue consequentialism" is consequentialism and virtue theory both.

Given what we’ve said about the distinction between Good and Motivation-based virtue theory, Zagzebski’s argument that Sosa’s approach is not a form of virtue theory because "his model of moral theory is act-based" is equally curious. The differences between act and agent-based accounts are interesting their own right, and she is correct that (Chisholmian) deontological and consequentialist theories is epistemology have structural similarities with act-based ethics" (8). These are reasons to think that virtue consequentialism is a belief-based and not strictly-speaking an agent-based theory. But using this as a desiderata to show that Sosa’s virtue consequentialism is not a virtue theory again illustrates how Zagzebski’s categories are inconsistently applied in Virtues of the Mind.

Arguably this way of dividing the field is not exactly correct even in its characterization of normative ethical theories. That question aside, Zagzebski appears to agree strongly with Christopher Hookway in desiring to de-emphasize the centrality of the issue of justification in epistemology in order to make issues of active agency the primary focus. As Hookway puts this responsibilist claim, "The primary focus is on how we order activities directed at answering questions and assessing methods of answering questions; it is not upon the epistemic status of beliefs." I would suggest that it is really matters of interest and emphasis that is at the root of this matter, not the desiderata for virtue theory itself. Once again the conclusion is that the ways of dividing the field under criteria (a) and (c) give us different results. Having begun with the desiderata of section (a), Zagzebski can’t consistently use the distinction between belief-based and agent-based theories to argue that virtue consequentialism is not a form of virtue theory.

In the previous two sections we discussed some of the dividing points between virtue reliabilism and virtue responsibilism, focussing around their respective understandings both of intellectual virtue and of virtue theory. We have found, I believe, that the cause of many of these disputes is primarily the alternative interest and focus of the two strains. This raises the expectation that these ‘in house’ disputes can be mediated, and are due as much to confusion as to substantive philosophical differences. In this section, then, I would like to focus and build upon an important shared emphasis in these two strains of virtue epistemology.

Virtue epistemologists share a broad range of assumptions that bring them together with interest in an intellectual virtue-centered account of epistemic agency. An account of epistemic agency will necessarily involve both descriptive and evaluative concerns; it will also involve consideration of both "passive" and "active" agency. At the most passive end of the spectrum are the functions of processes like those of the human perceptual faculties, and the beliefs they may causally entail; at the most active end are dispositions that are acquired either by habituation or teaching, and which are reflected in "chosen acts" that influence the rational acceptance of propositions. Justification is a particularly difficult issue in epistemology, in part because in discussing it we often cut across divisions that should be kept clear. Some pertinent examples are the division between evaluation of beliefs and evaluation of agents, between the criteria of justification and the conduct of inquiry, between belief and the act of propositional acceptance, between objective evidence supporting a belief and the agent’s own evidence, and between passive and active agency. The problems engendered by cutting across these distinctions haphazardly have been abundantly evident over the past three decades in the epistemological debates over "internalism" and "externalism" in epistemology.

Virtue epistemologists often stumble in dealing with these distinctions as well. Nevertheless, virtue epistemology may present unique potential for developing an account of epistemic agency that pays proper respect to these divisions. Consider justification as one important aspect of such an account of agency. One of the assumptions virtue epistemologists share which I regard as definitive of their unique philosophical approach, is the need for a "dual component" conception of justification, sometimes also called a "mixed externalist" account of justification. This is one that integrates constraints on an agent’s faculty reliability with constraints on the agent’s responsibility in gathering and processing evidence. So to cite but three instances, Ernest Sosa writes that in his virtue perspectivism, a proposition is evident or reflectively known (from the K point of view) to a subject "only if both he is rationally justified in believing it and is in a position to know (from the K point of view) whether it is true." Linda Zagzebski defines knowledge and justification through a "dual-component" account of intellectual virtue, which builds into the conception of the virtues themselves both "a characteristic motivation to produce a certain desired end and reliable success in bringing about that end." Greco also describes his as a "mixed theory": "The main idea is that an adequate account of knowledge ought to contain both a responsibility condition and a reliability condition. Moreover, a virtue account can explain how the two are tied together. In cases of knowledge, objective reliability is grounded in epistemically responsible action."

Despite their differences, these three thinkers have been most explicit proponents of a "mixed" accounts of justification. All three see this as a characteristic by which virtue theory can help epistemologists overcome the internalist/externalist opposition. Zagzebski’s "dual-component" account of the virtues, for example has twin aspects of success and motivation. Being virtuous involves success –‘getting it right’-, but it also involves the ability to see what that right thing is, which in turn involves the appropriate desire or motivation. One way of understanding the basis for such mixed accounts is to follow Zagzebski’s suggestion that Hilary Kornblith’s way of organizing questions presents a most promising framework for reconciling internalism and externalism. In his "Ever Since Descartes" (1985), Kornblith presents internalism and externalism as views that can be made complementary by paying attention to the different levels of descriptive and evaluative analysis that go into epistemic evaluation.

Kornblith begins by distinguishing two objects of epistemic evaluation, "appropriate processes for arriving at beliefs" and "appropriate actions for agents to take." The former object is of primarily non-voluntary processes, and the latter of voluntary actions that influence beliefs. Each of these objects of evaluation can themselves be divided, depending upon whether we adopt an objective (external) or subjective (agent’s own) perspective on them. So long as we are asking only the question of how we ought, objectively speaking, arrive at our beliefs, a reliabilist answer in terms of reliable processes suffices. But we then neglect whether the belief is properly integrated into the agent’s perspective. If we ask the same question from the position of the agent’s own lights (background beliefs, available evidence, etc.), an internalist answer is called for; but then we neglect the issue of whether the belief is formed in a way that is actually (objectively) truth-conducive. If both questions have a claim as pertinent to a satisfactory account of justified belief, we cannot afford to neglect either one.

With respect to the second object of epistemic evaluation, Kornblith says that "An epistemically responsible agent desires to have true beliefs, and thus desires to have his beliefs produced by processes which lead to true beliefs; his actions are guided by these desires." According to Kornblith, "…epistemic evaluation finds its natural ground in our desires in a way which makes truth something we should care about whatever else we may value." Since he deems the two questions asked from the objective perspective to be closely dependent, we are left with three basic forms of epistemic evaluation.

Internal constraints may be understood in various ways. Reliabilists who have brought aspects of internalism about justification into their epistemologies have often done so by focussing on consideration of coherence and explanatory integration from the agent’s situated perspective. This notion is implicit in what Sosa calls his "virtue perspectivism," which we may categorize as a form of what often goes in the literature as "perspectival internalism" (Schmitt, Haack). There is a needed place for a perspectival or internal coherence constraint on knowledge. But there is a sharp tension between it and the objective orientation of reliabilism: because it views reliably formed beliefs as neither necessary nor sufficient for belief sanctioned by the subject’s perspective, perspectival internalism seems to ill-fit with reliabilism (Schmitt, p. 5). No wonder then that one finds such a sharp distinction between subjective and objective justification in some reliabilist thinkers like Alvin Goldman. This tension may be resolved, though, in mixed accounts that achieve a satisfactory level of stability. Greco, for instance, is at pains to argue the "In cases of knowledge, objective reliability is grounded in epistemically responsible action."

There is a problem about whether responsibility-constraints can find a place in virtue reliabilist accounts, if these accounts identify the agent’s perspective on belief acquisition only with "a sort of inner coherence" (Sosa) or with an agent’s own standards or beliefs "about reliable belief-forming processes" (Goldman of 1986). Kornblith appears correct in insisting that a responsibility or "action-theoretic" constraint on knowledge is also needed in addition to reliability and internal coherence constraints. It is needed partly in order to render reliability and internal coherence constraints consistent with one another. Thus Zagzebski’s acceptance of and attempt to build off of Kornblith’s model is intriguing. For our purposes it shows that virtue responsibilism and virtue reliabilism have a strong bond in the shared goal of developing a "mixed" externalist account of knowledge and justification. On the mixed account, the epistemic agent (1) has beliefs formed by reliable processes, (2) is cognitively well-integrated and ‘in a position to know’ because properly affected or attuned to her environment, and (3) seeks to acquire virtuous habits or dispositions (possibly both for their own sake and) as a means of appropriately basing her beliefs on good reasons.

It should be noted in closing this section that Julia Driver and Jonathan Dancy both claim that compromise solutions are philosophically unstable. Neither wants an approach to ethical and epistemic evaluation that might be thought an unhappy amalgam of consequentialist and aretaic ideas, and again, neither wants to divide the field by taking very different approaches in the two fields. Ironically, this leads the former to a thoroughgoing consequentialist ethics (in line with process reliabilist views in epistemology), while it leads the latter to reject what he calls virtue consequentialism and to adopt "strong" virtue-theoretical views in both ethics and epistemology. It is of course tempting to play their views on the rejection of compromise solutions off against one another. But the serious point here is that the problem of stability is present not only for a virtue theory itself, but for mixed accounts of justification, which surely are compromise solutions to the internalist/externalist opposition over epistemic justification. This problem of stability deserves further examination in both contexts, that of what constitutes a sound and viable virtue theory, and that of what constitutes a satisfactory mixed account of justification.

We have seen that there is a shared goal among virtue epistemologists to develop a mixed externalist account of epistemic justification. While the goal is shared, the methods for achieving it, as we also saw, are often a matter of dispute. I believe that through focussing on the mixed account of justification as a shared project, the specific divergences we discussed can be mediated, leading to the significant broadening and merging of research interests among those doing work in this field.

Virtue epistemology has unique advantages, relative to other approaches, as the basis for a reconstruction in philosophy. The concept of intellectual virtue usefully explains both what it means to have beliefs formed by reliable processes, and what it means for agents to be responsible and well-motivated in their acquired cognitive habits and dispositions. To the extent that a mixed account must understand subjective justification in terms of "responsibility" and "proper motivation," it relies integrally upon what used to be called the language of ethics. Therefore the development of a mixed account in epistemology is at the same time a step in the direction of the further goal of a unified theory of value across epistemology and ethics.

In its broader context, I have characterized virtue epistemology as one side of an integrated account of the human being as both reasoner and valuer. It is in this broader context of virtue theory (rather than virtue epistemology or virtue ethics, per se) that we can bring the Fact/Value problem back into focus. Virtue theory addresses this problem by responding to the fragmented state of reason that John Dewey, a pragmatist who opposed the predominance of logical empiricism during much of his lifetime, presents in The Reconstruction of Philosophy as the central problem for modern life:

Restoration or reconstruction in philosophy, according to Dewey, is achieved by integration and cooperation. Integration is a theoretical concern, cooperation largely a practical one. Restoration itself is a response to perceived fragmentation. Epistemology in twentieth century Anglo-American thought has followed a path with many significant turns. Leaning upon a sharp distinction between the language and science and the language of ethics in the predominant logical empiricist tradition, twentieth century philosophers constructed an artificial divide between ethical and epistemic evaluation, a divide which it remains for the present century to bridge. The sharpness of this divide and the persuasive definitions of scientific and ethical discourse that supported it—tokens of their manner of responding to the Fact/Value problem—were often celebrated as confirmation of the scientific mind. Yet its result, in human terms, is precisely that crisis Dewey characterizes for us a fragmentation of human reason. Since Dewey’s period the residual effects of fragmented reason continue to be felt through the various radical "turns" proposed for philosophers to take over the past quarter or half-century, and in epistemology specifically, through the unsustainably sharp conflicts between internalism and externalism, and between non-naturalism and eliminative or reductive naturalism.

The concept of a virtue has both descriptive and prescriptive senses, which together, if not conflated or used in an intuitionistic manner, provide normativity a "comfortable home" within a pragmatic-naturalistic account of knowledge and understanding. As I have endeavored to show, contemporary virtue theory, with one leg firmly rooted in classical philosophy and the other in contemporary pragmatic naturalism, provides resources to fuel this endeavors.