Virtue Epistemology Strikes Back
Abrol Fairweather & Carlos Montemayor
1. Anscombe’s exhortation and epistemic situationism
In “Modern Moral Philosophy”, G.E.M Anscombe famously exhorted moral philosophers to avoid speculation regarding ethical norms until they could be grounded in an adequate moral psychology (Anscombe, 1958). Anscombe’s claim that moral philosophy is “not profitable” without “an adequate philosophy of psychology” places something like the following empirical constraint on a moral theory: An adequate moral theory must be empirically adequate, and empirical adequacy will be achieved through an empirically adequate moral psychology. We see this as applicable to normative theories generally, including epistemology. Call this broad demand for empirical adequacy in normative theory Anscombe’s
exhortation. Anscombe argued that Natural Law theory and Kantian ethics fail to meet the empirical demand, and looked instead to an ethics of virtues and vices.
Virtue theory appears an attractive way to heed Anscombe’s exhortation in ethics because it is capable of grounding norms about proper functioning in facts about human nature and human flourishing 1 . This appealing moral psychology became a significant reason for the resurgence of virtue ethics in mid to late twentieth century moral theory.
If virtue ethics has an empirically adequate psychology, there is reason to think that the prospects for virtue epistemology are good as well. However, a wide range of psychological research on trait attribution and rationality has chipped away at what appeared to be a solid empirical footing, thereby challenging the adequacy of virtue ethics on the very point that appeared to be a primary strength 2 . Philosophers such as Gilbert Harman (2000) have been led to question the very existence of character traits, and others like John Doris (2003) have denied their robustness and explanatory value. Character trait attributions enjoin predictive and explanatory commitments that simply fail too often to meet norms that require virtue manifestation for success . Situationist’s argue that Anscombe’s turn to virtue ethics fails precisely because it does not meet her own demand for empirical adequacy.
Multiple forms of the situationist challenge have now been developed in virtue ethics, as well as numerous virtue theoretic responses, and the debate in moral psychology continues to be a live one.
Only recently has the situationist challenge been applied to virtue epistemology. In two recent papers, Mark Alfano (2011, 2012) presents a thorough criticism of virtue epistemology from a diverse range of empirical results about rationality, inferential abilities and trait attribution. Alfano nicely frames the challenge as an inconsistent
1 Annas (2005) rightly notes that virtue theories in ethics do not need to be naturalistic in this way as seen in Stoic and religious accounts of the virtues.
2 The literature on relevant research is enormous, for some comprehensive treatments see Doris (2002), Miller (2008), Alfano (2011, 2012, and forthcoming)
triad: a) non-skepticism: most people know quite a bit; b) Virtue epistemology: knowledge is true belief acquired and retained through intellectual virtue; and c) cognitive situationism: people acquire and retain most of their beliefs through heuristics rather than intellectual virtues. The dilemma for the virtue epistemologist is that empirical adequacy will require accommodating the empirical work presented by situationists and thus will accept (c) to heed Anscombe’s exhortation. But the research shows that all too rarely will an agent meet virtue theoretic standards for epistemic success, and will thus be unable to account for (a).
Failing to meet the non-skepticism desiderata would be a normative inadequacy in virtue epistemology because any such theory will be unable to assign positive epistemic standings in a way that keeps pace with the frequency of human knowledge. Call this the challenge of epistemic situationism. We agree with Alfano that this normative requirement is a necessary desiderata for a theory of knowledge, but will argue that the right understanding of (b) can be made consistent with (a) and (c).
While we are confident that Alfano’s challenge can be met, it raises a pressing concern and extends Anscombe’s demand to virtue epistemology 3 . The examination of current research in psychology required for a virtue theoretic response should be profitable in further understanding the empirical basis of epistemic virtues, and is thus a valuable contribution to the literature on virtue epistemology as a whole.
However, Alfano’s argument must meet certain constraints to constititute a refutation of virtue epistemology. Focusing on (b) in Alfano’s triad, we say that any virtue epistemology will define knowledge as a form of “success from ability” (see
Pritchard, Greco, Sosa) 4 . This involves at least the following three commitments as necessary conditions for knowledge from epistemic virtue: (1) Epistemic Success – an agent must be reliably successful (2) Cognitive Ability – an agent must possess the relevant ability, competence or character trait (3) Etiological Epistemic Value – an agent’s epistemic success must be ‘from’ or ‘due to’ their cognitive ability, competence or character. The last condition requires that epistemic success must be related to the agent’s abilities in the right way for that success to constitute knowledge. This is required for any success to be sufficiently from ability. These conditions are intended to capture uncontroversial constitutive commitments for a virtue epistemology. Situationists like Alfano must undermine either the condition for epistemic success, epistemic ability or the etiological condition to claim any victory against virtue epistemology. If situationism cannot falsify any of the three, it has failed to undermine virtue epistemology.
The most obvious situationist target is the success condition. The research on rationality used by Alfano appears to show that agents often fail to achieve the
3 Guy Axtel (2009) has developed a response to situationism in virtue epistemology which differs from the approach taken here.
4 There are forms of virtue epistemology that do not aim to define knowledge, for example the ‘regulative’ virtue epistemology of Roberts & Wood (2007). We here pursue what Greco calls the classical project (2009)
outcomes expected of rational agents. Human beings are not successful enough according to virtue theoretic norms to support the frequency of knowledge attributions as captured in his (a) above 5 . However, we should not assume that epistemic success is rightly defined in terms of the models of ideal rationality employed in the research Alfano relies on. We argue that, properly understood, psychological research on rationality does not undermine a virtue theoretic account of epistemic success. With an improved theory of epistemic success in hand, we argue that Alfano’s triad is consistent.
Central to our argument will be an account of frugal virtues. This will be a reliabilist account, so it remains to be seen if responsibilist virtue epistemology can be defended in a similar way. We proceed as follows: Section 2 raises a normativity objection to Alfano in defending the success condition for epistemic virtue; Section
3 examines a challenge to the etiological condition available to situationists based on epistemic luck which is not fully explored by Alfano, but which may provide a different kind of threat from situationism; the concluding sections (4-7) defend an account of frugal virtues that adequately responds to epistemic situationism on empirical grounds. This account will be a reliabilist virtue epistemology with support from Gerd Gigerenzer’s research in the theory of bounded rationality, alternative interpretations of psychological findings from Daniel Kahneman and research on knowledge of syntax. We consider this a form of naturalized virtue epistemology and argue that it can withstand the challenge from situationism. To begin, we consider some connections between normative epistemology and empirical epistemology of the sort Alfano presses to frame an argument against the account of epistemic success needed for his argument that is independent of any commitment to virtue epistemology.
2. The empirical turn, epistemic success, and normative adequacy
Recent epistemology has witnessed an explosion in value driven epistemology, leading some philosophers to declare a ‘value turn’ in the last decade of epistemic theory (Riggs 2003). The considerable literature on the value problem has improved epistemic axiology and value driven epistemology has been a catalyst for work on epistemically valuable states other than knowledge, such as understanding and wisdom 6 . For this very reason, it may be that virtue epistemology calls for a compensating empirical turn 7 . With continuing developments in psychology and
5 Flanagan (2007) responds to situationism in virtue ethics by arguing that the demonstrated correlation coefficient for traits is generally .3 to .4, which is actually quite high compared to chance and that behavioral predictions based on traits will substantially increase accuracy.
6 For recent work on epistemic value see Pritchard (2007) and Haddock et. al.
(2009)
7 see Henderson and Horgan’s The Epistemological Spectrum for a fine example of empirically driven epistemology, as well as Beebe (2010)
other relevant empirical work, epistemologists will regularly have to re-consider the empirical adequacy of their currently favored normative theories. Responding to epistemic situationism will require virtue epistemologists to do just this in light of the recent value turn, and we support Alfano’s project to this extent. However, the concerns of normative epistemology cannot be ignored in the process. We argue that Alfano’s position falls prey to an inconsistent quadrad , with the result that his position is normatively inadequate and his account of epistemic success must be rejected. Alfano does not overtly endorse any specific principle of epistemic success, but his argument clearly relies on norms of rationality from research programs in social psychology and behavioral economics. This use of rationality theory to set norms for epistemic success will be problematic for Alfano, irrespective of any further commitments to virtue theory in subsequent sections of this essay. We begin with some considerations on epistemic normativity that apply to any attempt at an empirically adequate virtue epistemology and then present an inconsistent quadrad for Alfano.
Suppose we have an epistemic theory that meets Alfano’s demand for empirical adequacy in hand. This empirical epistemology will still face normativity challenges and these have traditionally been troubling for naturalistic theories. The classic challenge to naturalism in ethics took shape around GE Moore’s Open Question
Argument (1903). Moore demanded that a naturalistic account of moral properties satisfy a stringent test for an adequate meaning analysis of the target normative properties such that for any natural property manifested by an agent or action, if it remains intelligible to ask whether the agent or act is nonetheless good or right, the account is normatively inadequate. Failing to meet Moore’s semantic demand is a worry because the prospects for constructing an adequate normative theory are dim without the necessary semantics for normative judgements. While developments in philosophy of language since Moore show some mistakes in his formulation of the argument, the general demand that a naturalistic theory must show itself to be normatively adequate continues to be significant in meta-ethics and the moral realism debate 8 .
In “Natural Facts and Epistemic Norms,” Carrie Jenkins examines how Moorean open question arguments extend to naturalistic accounts of epistemic norms. Any epistemic theory must account for a range of intuitively true normative claims such as “we ought to fit our beliefs to our evidence, justified beliefs are in good standing, blind trust in unreliable informants is wrong, it is irrational to believe an explicit contradiction, and so on.” (Jenkins, 2007) A naturalized epistemology will have to provide natural properties that allow for the derivation of these or related normative standings. Any theory that fails here cannot be said to provide a theory
8 For example, Brink’s (2001) ‘synthetic moral realism’ makes use of contingent a priori propositions to semantically ground normative properties in natural properties. See also Zagzebski’s (2010) use of natural kinds semantics in her recent
‘exemplarist’ virtue theory.
of knowledge. The Moorean worry in epistemology is that no set of natural properties can semantically ground the normative judgements essential to epistemic appraisal.
Jenkins rightly weakens Moore’s demand for conceptual reduction, but argues that, while neither a biconditional proposal nor a supervenience proposal will provide sufficient natural grounds, sameness of truth makers will. For Jenkins, if the fact that makes a normative epistemic claim true is the same fact as that which makes a natural claim true, an epistemologist has met Moore’s worry. In epistemology, she argues that natural facts about ‘probabilifying’ relations and facts such that contradictions are not true will be the very facts that make true many normative epistemic claims about what we ought to believe, when a belief is irrational or when a belief is in good standing. There appears to be a range of natural facts that provide some hope for meeting open question arguments in epistemology and thus grounding some form of realism about epistemic norms 9 .
While there are questions raised by Jenkins’ list of natural facts and her claim that identity of truth makers suffices to meet the Moorean worry, her proposal may constitute the beginning of a normatively adequate naturalistic epistemology.
However, the kinds of natural facts she identifies do not appear to ground the full range of normative standings necessary for a theory of knowledge. Specifically, she discusses no natural fact to account for the credit an agent receives for an epistemic success that is properly from ability. This would be required for meeting Epistemic
Open Question Arguments in the present context, and thus Jenkins’ proposal will not enable the derivation of all necessary normative epistemic standings. Even meeting her broad demand for sameness of fact, an epistemic theory will also need to distribute a wide range of normative standings to a wide range of cognitive activities and will require a more complete semantic structure for deriving normative standings from natural facts. There is hope that further inquiry into the empirical underpinnings of cognitive abilities and epistemic success will provide further insight into the natural facts that make attributions of credit for epistemic success come out true, though Alfano will presumably take a less optimistic view of these prospects 10 .
Another instructive example of a normativity challenge to an empirical epistemology is found in Kim’s (1998) objection to Quine’s naturalized epistemology. In his famous paper “Epistemology Naturalized,” Quine rejected
9 Cuneo argues for epistemic realism and that this implies a form of moral realism in his “parity argument” in The Normative Web. Heathwood (2009) argues against
Cuneo, but still accepts epistemic realism.
10 Normative standings for agent credit might be grounded in natural facts about proper functioning or desire satisfaction This sets up an interesting challenge the situationist might make that cannot be pursued in detail here, namely whether situationism blocks any adequate resolution to answering epistemic open question arguments vis-à-vis facts about normal functioning.
epistemology grounded in conceptual analysis, and exhorted epistemologists to provide a robust, ground-level role for empirical psychology in framing and answering issues in epistemology. One influential criticism of Quine’s account comes from Jagewon Kim’s (1988) well known objection that Quine’s epistemology is nonnormative. Since epistemology is essentially a normative discipline, Kim argues that
Quine has changed the subject by going empirical. Kim’s objection is of a piece with
Moore’s worry about the possibility of a normatively adequate naturalism in ethics.
According to Kim, Quine’s naturalistic epistemology would be woefully inadequate because it does not enable one to derive any normative epistemic standings.
Kim might ultimately be right that Quinian naturalized epistemology is normatively inadequate, but his argument has an answer from Quine. Quine’s full naturalized epistemology will be constituted by his favored empirical psychology and a semantics for epistemic norms. “Epistemology Naturalized” did not develop a theory of epistemic norms, but his subsequent work did, perhaps largely due to
Kim’s objection. It is now clear that epistemic normativity for Quine is seen as the
“technology of truth seeking” 11 . His view is that an empirical explanation of cognition achieves positive epistemic standing if it is “good technology” for the pursuit of truth and other epistemic endeavors. The norms of epistemology then become the norms of a certain kind of engineering, and this becomes an empirical inquiry through psychology, statistics and “heuristics generally”. While this account has been criticized and leaves much to be explained, it shows how the derivation of normative standings from empirical conditions can respond to Kim’s objection and aspire to normative adequacy. More would have to be shown to demonstrate normative adequacy for Quine’s account, but it is one way to pursue an epistemology that is both empirical and normative.
Heeding both Anscombe’s exhortation and Moore’s worry, an adequate virtue epistemology will have to be empirically adequate (for Anscombe) and normatively adequate (for Moore). The constraints for a theory meeting (b) in Alfano’s triad will thus include maintaining the three constititutive commitements of virtue epistemology and the demands for both empirical and normative adequacy.
Meeting the full set of demands will require an empirically adequate psychology that also allows for an adequate derivation of normative epistemic standings 12 . The normative standings derived must also cohere with uncontroversial epistemic judgements such as the non-skepticism principle in Alfano’s triad, and must perform well enough in answering a range of ‘standard cases’ in epistemology. A complete account of normative adequacy is beyond the scope of the present paper, but this will not be necessary for the arguments made here. This semantic demand does not, of course, require that normative standings can be derived from empirical
11 See Kornblith (1993), Hylton (2007), Houkes (2002), Fairweather (2011) for discussion of Quine’s “technology of truth seeking” norm.
12 By ‘deriving’ we mean only that some semantic function will assign normative standings to empirically adequate descriptions of the cognitive states assessed in an epistemic theory.
explanations alone. That is most likely an impossible task, and for roughly the kinds of reasons given by Moore. But, it is also the wrong kind of demand. An empirical psychology requires a semantics to determine normative standings for a broad range of cognitive activities in order to do any epistemic work 13 .
We argue that the commitments needed for Alfano’s argument fall prey to a form normative inadequacy. While Alfano does not explicitly define or commit to a normative epistemic semantics, he clearly assumes that models of ideal rationality are the semantic devices that allow for the derivation of normative standings from empirical conditions. He then argues that the distribution of positive epistemic standings in virtue epistemology violate the non-skepticism principle because our actual cognitive endeavors will all too rarely meet virtue theoretic standards of success. Since non-skepticism is off the table, virtue epistemology must be rejected.
However, it is Alfano that falls victim to this line of argument. Conjoining ideal rationality semantics and situationist empirical psychology entails that no epistemic
theory can adequately assign positive normative status to actual human cognition.
Reliable inference is necessary for human knowledge in evidentialist, internalist, coherentist, foundationalist, and standard reliabilist accounts of knowledge, as well as virtue epistemology. But this means that the following quadrad is inconsistent:
(a) situationist psychology (b) ideal rationality semantics (c) ANY theory of knowledge and (d) non-skepticism. Since Alfano appears to accept (a-d), he must work his way out of this to have a coherent argument against virtue epistemology.
But this will be a challenging attempt. If he abandons (c), he is no longer doing epistemology. If he abandons (a), he presumably no longer has an empirically adequate epistemic psychology and the situationist is defeated. We agree with
Alfano that (d) is off the table. Thus, (b) appears to be the clear choice for elimination, Alfano’s epistemic semantics must be rejected. The problem arises primarily from his understanding of epistemic success. We defend an improved account of epistemic success in sections 4-6 that will solidify the success condition for virtue epistemology on empirical grounds.
3. The ability condition, environmental luck and situationism
Defending the success condition will not be enough to save virtue epistemology from situationism. A defining commitment of virtue epistemology is that cognitive success must be explained by a cognitive ability in the knower. This will require that agents possess abilities, and these will be some form of stable disposition to
13 The concern for normative adequacy can be seen in the debate over Goldman’s
“value T-monism” and challenges from Pritchard, Kvanvig and others who argue for some form of epistemic value pluralism. An empirical virtue epistemology faces these questions in a unique way because the axiology will be constrained by a commitment to a specific empirical psychology.
reliably bring about a certain outcome when the agent endeavors to. This is an issue in the metaphysics of dispositions, epistemic taxonomy as well as empirical psychology. A second requirement is that abilities must be sufficiently explanatory in an agent’s epistemic success. This is primarily an issue in the theory of epistemic agency and action. Bringing all of this to the table, the ability condition will be a nuanced and perhaps fragile commitment for virtue epistemology. It is also a high value target because, in conjunction with an etiological condition, it is responsible for carrying much of virtue epistemology’s success in responding to Gettier cases and other instances of epistemic luck, as well as providing an attractive solution to the value problem (Greco 2009).
Any situationist seeking to undermine virtue epistemology should certainly take aim at the ability condition. Alfano hints at how this might be done, but we expand this argument for the situationist below. The situationist might claim (a) that there are no cognitive abilities as conceived by virtue epistemology, or (b) cognitive abilities do not sufficiently explain cognitive success. John Doris (2003) has argued that agents only possess very narrow abilities, traits and dispositions – work place honesty, or honesty with family, rather than honesty simpliciter. This seems to undermine the attributability of broad abilities across wide ranging circumstances.
If the abilities relevant to virtue epistemology only come in this very broad sort, then virtue epistemology may run into trouble in the reification of abilities.
However, while virtue ethics inherits a historical commitment to a canonical set of classic traits, virtue epistemology is not so constrained. Epistemic virtues can be defined with a greater latitude than moral virtues, despite being tightly constrained by the value of truth and giving a prominent role to dispositions of the agent. Since virtues are most plausibly seen as clusters of abilities and dispositions by virtue epistemologists, there is no reason why an epistemic virtue cannot be defined as a set or cluster of narrowly defined Doris-abilities. The challenge to an ontology of abilities in virtue epistemology does not appear as strong as the challenge to classically defined character traits in ethics. Furthermore, the recent work on the metaphysics of abilities will continue to provide semantic and empirical support for a theory of cognitive ability that meets the needs of epistemologists.
However, the claim that abilities are non-explanatory may be harder to dismiss.
Alfano notes that when agents are successful outside of perception and other basic cognitive processes, it appears that success is often due to epistemically irrelevant but fortunate environmental conditions such as finding a quarter, truth maximizing ambient noise or receiving a compliment. While these results in social psychology initially applied to pro-social traits like helpfulness, Alfano can extend these worries to epistemology by challenging the etiological condition of virtue epistemology. If similar environmental factors make the difference between true and false belief, the agent’s abilities do not appear to be sufficiently explanatory of their success. If
Alfano can use empirical psychology to show that too many cases of virtuous outcomes are due to luck rather than the agent, virtue epistemology will not be able to connect an agent to their epistemic success as needed to satisfy the etiological condition.
This move has certain advantages for the situationist because they do not need to show wide ranging epistemic failure to undermine virtue epistemology. If situationism shows that cognitive success is all too often due to environmental factors rather than ability, these successes will not be knowledge. Neither will these successes be creditable to the agent, and the solution to the value problem has been lost as well. Even if we succeed in defending the success condition here, this challenge to the explanatory salience of cognitive abilities will be an additional burden on our argument.
As a first line of defense, virtue epistemologists can draw upon the well-developed literature on epistemic luck (See Pritchard (2011), Sosa (2007), Axtell (2001). Since virtue epistemology arose around concerns to give the agent a stronger role in making knowledge attributions true, one would expect that existing work on epistemic luck has anticipated or will at any rate suffice to respond to the situationist challenge on this score. The situationist might here appeal to
Pritchard’s arguments that virtue epistemology cannot adequately address cases of epistemic luck. A situationist can exploit Pritchard’s general strategy but rely on empirical research rather than thought experiments. This objection would then claim that empirical findings about the influence of epistemically irrelevant features of a situation create a form of knowledge undermining environmental luck. We present just the crux of Pritchard’s argument below and consider how an analogous argument based on Alfano’s research could be constructed to challenge the etiological condition 14 .
One crucial case for virtue epistemology is the familiar Barney case. Barney is in a field of masterfully crafted barn facades , with but one real barn. By chance, Barney looks directly at the one real barn and acquires a true belief about it. Most intuitions are that Barney’s belief is not knowledge because his belief so easily could have been false. Pritchard argues that we would nonetheless attribute a success from ability to Barney 15 . Because virtue epistemology will fall prey to this kind of knowledge undermining bad luck too often, he argues that a separate safety condition must be conjoined with the ability condition, A different case for the situationist is TrueTemp, who always gets the right answer because of environmental help. Again in this case intuitions are that Temp’s true beliefs are not sufficiently from ability because the environment is guaranteeing his success. In this case too much environmental good luck seems to undermine knowledge attribution. In both cases it appears that influence of good or bad luck in the environment an agent happens to be in will undermine their achievement of either a success from ability (True Temp) or knowledge (Barney).
14 An interesting question for Pritchard is whether the situationist would also challenge his safety condition for knowledge.
15 Sosa argues that we should attribute success from ability and knowledge in such cases as the agent’s abilities are the direct causal explanation of the success.
Alfano can point to the actual psychological findings of Khaneman and Tversky and many others to raise actual world luck objections to the safety of epistemic success
(as per Barney cases) and the explanatory sufficiency of ability in success (as per
True Temp cases). In the empirical research the environment rather than the agent plays a substantial explanatory role and this militates against attributing a success that is properly from ability. Many modal requirements on knowledge require actual world indexing to satisfy safety and normal functioning conditions. The situationist can argue that any such indexing must account for the fact that the actual world is beset by problematic forms of epistemic luck, but this can be incorporated into the antecedent and consequent conditions of the relevant dispositions or theories of proper functioning. Sosa argues that the actual world is a fortuitious place for faculties of perception and memory as well as basic inferential capacities but this accounts for the fragility of success and stability in a virtue. This is a disagreement about how lucky actual world cognitive success is and in what ways.
If empirical psychology shows that our actual cognitive lives are regularly susceptible to either kind of luck in unforeseen ways, this would seem to force at least a revision of the etiological condition in a number of virtue theoretic accounts of knowledge. This would also support Pritchard’s contention that cases of epistemic luck show the need for a requirement on knowledge outside of virtue epistemology. With the right account of real world luck in hand, an epistemic situationist might then claim victory over virtue epistemology without needing to provide the additional principle necessary for knowledge.
We discuss this challenge further in the final section of this essay, but an initial response is that the issue is really about the stability of an ability in the agent and this is the burden of the ability condition. This would require further work from virtue epistemologists on the nature of abilities and dispositions to show sufficient stability in the agent. The etiological condition has been a difficult area for virtue epistemology and situationism puts a new pressure on this commitment.
4. Bounded Rationality and epistemic virtue
In this section we defend an empirically adequate virtue epistemology for (b) in
Alfano’s triad. Normative adequacy will be addressed in section 5. In section 2 we argued that Alfano relies on norms of ideal rationality in ways that appear to lead to skepticism. Here we raise a different objection to this commitment. Classical models of decision making in economics and behavioral sciences assume that rational agents follow or conform to formal models of decision making such as expected utility theory or Bayesian probability theorems. These approaches to rationality assume that agents have highly idealized cognitive capacities. Utility maximization models require that agents know the prior and conditional probabilities of all relevant propositions, have preferences for all outcomes and update probabilities with changes in their evidence. Call this the assumption of full
rationality or unbounded rationality.
The research of Gerd Gigerenzer and others shows that consistent failure to manifest optimizing outcomes does not entail irrationality because real human decisions exhibit ‘bounded rationality.’ Real rational agents are limited in all sorts of ways which make optimizing rationality impossible because optimizing strategies are computationally intractable for bounded cognitive agents 16 . Gigerenzer argues that we are often more successful cognitive agents when we use fast and frugal heuristics than computational methods of reasoning. While bounded rational strategies are fast and frugal, they also include background information and contexts are taken into account in decision-making in ways that manifest epistemic virtues, rather than by allowing irrelevant information to impede successful reasoning. Frugal strategies of reasoning are also informationally rich in this way.
Gigerenzer (2008) argues that human beings have an adaptive toolbox for specific kinds of inferential tasks that include fast and frugal heuristics, evolutionarily designed to cope with a variety of epistemic challenges.
Evidence analyzed by Gigerenzer (which includes his own research and covers a range of experiments that goes back to the influential work of Piaget) reveals a rich variety of ways in which, for example, explicitly using the language of set inclusion, rather than probability, yields the rational response. He claims that the mistake was on the part of the theorists, not the subjects. Humans are rational. The problem is that psychologists, using the highly abstract and idealized scenarios that yielded probability theory (e.g., Dutch book arguments, abstract rules on conditionalization and total evidence, etc.) thought that these scenarios fully captured human rationality. The worry was that if human beings do not follow the rules of logic and probability, they would contradict themselves and systematically choose the wrong answer. It is now uncontroversial that human beings do not obey these rules systematically and yet manage to be fairly successful epistemically. Since the norms of ideal rationality are neither necessary nor sufficient for achieving the vast array of epistemic endeavors that should qualify as forms of human knowledge, virtue epistemologists can agree with Alfano that epistemic success is rarely due to that kind of ability
The research on bounded rationality is important for any theory of knowledge to consider. This is not just a question of whether to take a skeptical or optimistic stance toward the research. Human rationality is sophisticated, varied, and cannot be explained solely through the general and highly abstract constraints of logic and probability. Crucially, human beings depend on concrete epistemic environments, and have specific epistemic goals that require a range of cognitive abilities. Coping with uncertainty was never, in the evolution of the human species, an abstract exercise in logic and probability. This is the core idea behind bounded rationality.
16 Adam Morton develops this approach to epistemic virtue in Bounded Thinking
(forthcoming, OUP)
Taking on a theory of bounded rationality might be a controversial move for a theory of epistemic virtue because of the frugality of the virtues. While heuristics are less demanding on an agent’s overall cognitive resources and are often more effective than the costlier Bayesian calculations in achieving the agent’s epistemic and practical aims, they also appear to be slothful and less than conscientious. The degree of sloth will of course vary depending on the kind of heuristic involved, and some epistemic virtues like conscientiousness might yet turn out to be cognitively diligent endeavors even on a frugal account of the virtues. However, this is really a worry for responisiblist accounts of epistemic virtue rather than the reliabilist account we defend here. We appeal to the bounded rationality research to meet the empirical adequacy demand on virtue epistemology and this will be accomplished here with a reliablist theory of frugal virtues.
One clear implication of Gigerenzer’s research is that unbounded rationality and epistemic success are two different things. Alfano appears to use unbounded epistemic norms in claiming that virtue epistemology cannot satisfy the nonskepticism principle in his triad. We understand epistemic virtues in terms of
epistemic success rather than (full) rationality. Once this distinction is made, there is no easy move from the evidence counting against ideal rationality to any conclusion about failures of epistemic virtue. This is a promising response for an empirically adequate virtue-reliabilism.
Further support comes from Gigerenzer’s interpretation of the ‘Linda the Bank
Teller’ case. Infamously, when asked whether, given a character description of
Linda, it is more probable that she is (a) a bank teller or (b) a bank teller and active in the feminist movement, 85% of the subjects answered (b), clearly committing the
“conjunction fallacy” and violating basic theorems of probability calculus.
Gigerenzer notes that subjects are required to use syntactic, content blind rules of reasoning where the values of the variables are not relevant to getting the answer right, and we typically do poorly in content blind reasoning. When the question is reformulated, we get very different results. If one asks ‘how many’ instead of ‘how probable’, research shows better, more rational results. When asked, out of 100 people that satisfy Linda’s description, how many would be bank tellers and how
many would be bank tellers and active in the feminist movement, subjects’ performance significantly improves, producing the right answer. This shows that different framing of logically equivalent information will get very different results in assessments of rationality, and the framing is thus playing a big role here. Relative to certain frames, people answer quite rationally. The failures of rationality in the psychological research are thus very local and do not support the claim of widespread failure that situationists use to push virtue epistemologists to skepticism.
The philosophical implications of bounded rationality research for our argument can be made in two ways. Bounded rationality research might show the need for a revised and improved understanding of human rationality, but we can still use rationality as a central norm for epistemology. This reading insists that rationality
should be defined as bounded rather than full or optimizing. We suggest a different way of expressing essentially the same point, put in terms of epistemic success rather than rationality. Grant Alfano that rationality will mean full rationality, but insist that epistemic virtue theory can grant epistemic success even when agents
regularly fail to manifest (full) rationality. Satisfying norms for epistemic success simply does not require satisfying norms of (full) rationality, and Gigerenzer’s research shows exactly why.
We defend frugal rather than optimizing virtues as the norm for knowledge. This is a turn away from epistemic rules in each of two senses: explanation and regulation 17 . Cognitive activity that manifests epistemic virtues does not have to be guided by, based on or explained by formal rules of rationality. We take this to be fairly uncontroversial as any such requirement will be subject to psychological plausibility worries and will have trouble accounting for ordinary cases of knowledge from perception or testimony. Or, one might grant that optimizing rules do not describe the process through which success is actually achieved, but still insist that they are normative because they prescribe the standards for measuring epistemic success. We argued above that optimizing rationality does not set right standards for measuring epistemic success.
5. Knowledge of syntax and frugal virtues
An important example of a frugal virtue can be taken from research on knowledge of syntax. This is a paradigm of formally constrained knowledge, which is considered to be one of the clearest cases of exclusively human epistemic capacities.
Knowledge of syntax requires the manipulation of information according to strictly formal rules. Children have epistemic skills that allow them to learn any language based on these rules. The modal robustness of these skills is extraordinary. A vast amount of research in psychology, neuroscience and linguistics aims at explaining this robustness.
18 Specifically, scientists have tried to understand how it is possible for infants to learn a language given the incredibly diverse contexts they are in, the impoverished stimuli they are exposed to, etc.
The abilities that produce knowledge of syntax are, like perceptual skills involved in perceptual knowledge, remarkably stable dispositions. Although knowledge of syntax is highly formal (in the same sense in which Bayesian rationality is formal, e.g., it depends on formal rules that specify correct and incorrect linguistic structures), humans manifest such knowledge at a very early age, and they do so
17 See Greco (2009) for an extensive discussion of rule following. An extreme case of non-rule governed perceptual knowledge is imagined by Reid, who considers a being that moves right from sensory stimulation to perceptual belief without any intervening sense data or sensory representation.
18 See Jackendoff (2003), Chomsky (1986 and 1987) and Hornstein (1984).
reliably and without conscious effort or monitoring. Infants do not need classes of universal grammar and the rules of syntax in order to distinguish the syntactic components of (in many cases poorly constructed) utterances of a language. They are certainly not introspecting on these rules, or accessing evidence that could justify them to parse an utterance in terms of subject and predicate. What the infant is doing is highly complex and the whole discipline of linguistics is devoted to uncover the intricate rules that underlie the structure of language. But the infant performs this incredible epistemic task in a perception-like fashion.
This is a case of a formal, uniquely human capacity that our account can accommodate perfectly well. The epistemic abilities involved in knowledge of syntax are as good an example of our view as are perceptual skills. Infants use fast, reliable, modally robust epistemic skills to achieve knowledge of syntax, and succeed in learning any language. These skills are not susceptible to introspection, but this may actually explain their modal robustness across a vast amount of epistemic scenarios, and should not be used as an objection against their status as stable epistemic dispositions.
The situationist may insist that even the most robust epistemic dispositions can be easily disturbed by very easy manipulations of the stimuli, and this might threaten the anti-luck and safety intuitions needed for virtue epistemology. In response, we would like to provide an illustration of why although information processing may always be disturbed under laboratory settings, this by no means threatens the stability of epistemic dispositions.
For instance, some information interferes with the speed and accuracy of behavioral responses, but this by itself does not entail that the abilities involved are unreliable.
In the Stroop task, the interference between inclinations (the automatic inclination to read a word vs. identifying a color) does not entail that the capacities involved are unreliable because of alleged context sensitivity. The capacities to read and detect color are incredibly reliable across subjects in many conditions. Interference only shows that having two inclinations affects processing. As mentioned, any virtue conceived as a stable disposition will be disturbed or “masked” under some conditions. But being disturbed in non-standard situations is just part and parcel of
being a disposition.
This point is crucial to understand why virtue epistemology is unscathed by the situationist challenge. The situationist generally grants that perceptual knowledge is unchallenged by their evidence (thus their challenge has a limited scope, and affects only inferential reasoning). We are construing virtues as perception-like dispositions to form true belief. This, we claim, is not an ad hoc way of characterizing dispositions tailored specifically to respond the situationist objection.
Rather, this characterization is the most natural way of understanding epistemic dispositions in the light of the most recent evidence in psychology.
Knowledge of syntax is not, therefore, an isolated case that happens to comply with our characterization of virtues as stable epistemic dispositions. All forms of inferential and formal reasoning can be so characterized. Consider knowledge of logic. We have the capacity to reason according to modus ponens and this capacity is part of a set of stable dispositions to draw deductive inferences that are truthpreserving. One may actually say that these dispositions constitute what we mean by deductive inference.
19 If this is the case, then one could not know the meaning of what a deductive inference is without having such stable epistemic dispositions. It is a truism that basic deductive reasoning (for example an application of modus ponens) can be achieved without explicit understanding of such rule and that these dispositions, like those underlying knowledge of syntax, are remarkably stable.
Demanding an explicit understanding of the rules for deductive reasoning increases cognitive demands, and although we can be trained to have such explicit understanding, this is not a necessary condition to have the stable dispositions that are implicit in our capacity to identify these rules. More importantly, requiring such explicit understanding is open to traditional objections against accessibilism and deontological accounts. Thus, it seems that the best strategy is to characterize these fundamental rules for deductive reasoning in accordance with our perception-like model.
20
The situationist seems to face a new dilemma. Either we posses stable epistemic dispositions that allow us to identify valid rules for deductive inference or we don’t.
If we do, then situationism is false. If we don’t, it is not clear how we are able to understand what we mean when we talk about, for instance, modus ponens. For it is not clear that highly unstable and easily disturbed capacities would help us succeed in specifying what we mean in every situation by the fundamental rules (modus ponens, modus tollens, etc.). Thus, it would not be entirely clear that we mean the
same fundamental rules when we characterize a piece of deductive reasoning as modus ponens or something else. The situationist needs to explain why the psychological evidence would have such a dramatic result and this strongly suggests that situationism is in trouble. Obviously, the easy way out of this dilemma is to affirm that situationism is false, which is what we propose.
21
Regardless of this dilemma, the foregoing discussion shows that stable dispositions are the right conceptual tool to characterize epistemic virtues, and once this characterization is in place, one can demonstrate that the empirical evidence is not only fully compatible with virtue epistemology, but actually supports it. Just as every human being has the capacity to achieve knowledge of syntax and her success
19 See Boghossian, 2000.
20 Notice that this is quite different from having a conscious-intellectual “seeming,” which is one way of defining intuitions.
21 This is a concrete way of making a point suggested to us by Lauren Olin in conversation, which is that relativism is much more troubling in the epistemic case, as compared to the moral case. If we are right, situationism is also a lot more
implausible in the epistemic case.
depends on stable epistemic dispositions (which are attributable to her), knowledge of logic is also achievable for any human being, independently of the explicit understanding of the rules of basic logic. The empirical evidence confirms the stability of these capacities, and the skeptical interpretation of the evidence concerns cases that show exclusively how one can manipulate them.
So what happens when we go wrong in the Linda case? As was suggested, these are cases of cognitive interference, like the interference in the Stroop effect. The result of the interference in cases like the Linda experiment, however, is not decreased cognitive speed, but the rapid choice of one interpretation (which is pragmatically informed) over another (a strictly logical and abstract one). As mentioned, this interference does not entail lack of stable epistemic dispositions for inferential reasoning, because such interference can easily be understood as a case of masking.
The reminder of the paper argues for the empirical plausibility of our proposal.
6. Reinterpreting the findings by Kahneman and Tversky?
Meeting Moorean worries about normative adequacy will require our theory to adequately derive normative epistemic standings from our empirical commitments.
In this section, we examine different interpretations of relevant empirical research to further secure normative adequacy.
The standard interpretation of evidence from Khaneman and Tversky would seriously challenge virtue epistemology by showing that human rationality is highly unstable and remarkably unreliable. On this interpretation, human rationality produces false belief in a vast amount of circumstances and is inordinately susceptible to the presence of irrelevant stimuli. The only plausible way to demonstrate that the standard interpretation is compatible with virtue epistemology is by showing that the standard interpretation still has room for stable forms of reliably produced belief, and that these are based on abilities that are impervious to trivial and irrelevant stimuli. One can actually find remarks encouraging this project by Kahneman himself (2011). At the beginning of his recent book, Kahneman (2011) argues that there are two systems for reasoning.
More accurately, he says that these two systems (systems 1 and 2) are useful fictions that capture two broadly similar ways in which the brain engages with problem-solving and truth-evaluation in a variety of situations. Kahneman describes experiments suggesting the existence of a fast, flexible and unreliable system that in many cases trumps a slow, consciously demanding and reliable system. The fast system evolved to respond quickly to either urgent or typical situations, and is responsible for much of our success as a species. The slow system is more cautious and, instead of responding quickly, considers evidence and examines the nature of problems slowly and carefully.
Kahneman says it is a mistake to associate human rationality with system 2 alone.
The truth is that system 1 calls the shots very frequently, and this is perfectly
compatible with properly functioning cognitive systems. This suggests the account of epistemic virtues defended above. The use of fast and frugal belief forming processes, including heuristics, is perfectly compatible with epistemic virtue read as proper cognitive functioning rather than ideal reasoning. Situations where normatively irrelevant stimuli trigger fast and unreliable responses appear to impugn the stability of system 2. While manifesting virtues, as traditionally understood, requires situational stability, it appears that system 2 cognitive processes all too easily fall prey to system shifting stimuli. However, this does not show that system 2 is unreliable. This must be determined by cases where system 2 is put to use, but Khaneman explains our persistent errors as the result of shifting systems.
The situationist may yet claim that system 2 is not stable in the agent because of the prevalence of shifting. If system 2 is not stable in the agent, the worry for the virtue epistemologist will be that the agent might not be reliable overall and they will not have sufficiently stable dispositions to meet the etiological conditions and the situationist’s worry about luck and safety. But, no serious account of virtues claims that they must be so stable that they will manifest irrespective of circumstance, manifesting virtues always requires a certain level of cooperation from the environment. System 2 can still be read as stable when the conditions for its proper functioning obtain, and in such cases it will also be reliable. The prevalence of shifting in extra-normal conditions does not threaten the relevant stability or reliability of system 2. These relevantly stable system 2 abilities could then be used to ground a range of important normative standings in virtue epistemology.
However, the findings on the relationship between systems 1 and 2 speak against this defense of epistemic virtue for the following reasons. System 2 is extremely energy consuming and, actually, lazy. These two aspects of system 2 are captured in a variety of experiments in which the quick and erroneous epistemic deliverances of system 1 prevail over the more stable epistemic processing of system 2. This overpowering of system 2 by system 1 is certainly not one that can be overcome by virtuous training (the evidence seems to overwhelmingly suggest this is just how we are “wired”). Kahneman himself acknowledges that we cannot overcome some of the troubling biases that guide our decisions and actions. The prevalence of system
1 is, therefore, a crucial claim of the standard interpretation. What occurs when people make decisions and judgments based on system 1, when they should respond based on system 2, is not a “battle among equals” between good and bad epistemic reasoning. Rather, it is a rigged game in which system 2 is almost always willing to give up. This may be good for a lot of practical reasons, but it is bad for epistemic virtue. So this response to situationism is inadequate. The only option is to reject the standard interpretation, because it entails situationism. Instead, we seek to clarify how the findings on systems 1 and 2 are no threat to virtue epistemology by appealing to an alternative interpretation.
Since system 1 overpowers system 2 in many situations (situations that are not really atypical), one must show that the frugality and seeming unreliability of system
1 does not preclude the existence and development of epistemic virtues. Is system 1 really so bad? As mentioned, Gigerenzer thinks not. If he is right, then a virtue epistemology based on frugal responses is certainly possible. One needs to keep in mind that Kahneman uses the word ‘system’ as a useful fiction. There are a variety of epistemic processes that satisfy different epistemic goals, according to bounded rationality. For example, the Linda case is one in which virtues that work perfectly well when information is given in clearly numerical expressions (such as set inclusion) do not work well under circumstances where the language invokes likelihood.
This can be explained by the fact that likelihood invokes pragmatic considerations.
Suppose that the frugal process interprets the information about the Linda case as follows: Linda is a bank teller, but given that she participated in the feminist movement, it is more likely that she is a bank teller that is active in the feminist movement, rather than one who is not at all involved in the feminist movement. This frugal response is not irrational. One would be missing very important information about Linda if one ignored that she may be currently involved in the feminist movement. The frugal response insists on not missing this crucial information about
Linda.
The abstract system for mathematical calculations and probability works well when no loss of information is guaranteed. But here it seems that one may loose something important by giving the “right answer.” The right answer is actually too unsophisticated to deal with the full range of epistemic interests Linda has given her complex set of practical interests. The “right” answer may not be the epistemically virtuous answer in such cases.
The contexts in which epistemic goals are satisfied always include standards for evaluating the appropriate response. It is very likely that, given the type of epistemic environments one generally encounters, practical interests will frame not only the standards for satisfactory response (see Greco 2009)) but also the criteria for adequate knowledge attribution (see Stanley 2005). In any case, it seems that a virtue epistemology built on frugal (as well as “slower”) epistemic virtues, understood in this way, fits nicely with recent theories of epistemic virtue and knowledge attribution, even if not with formal accounts of full rationality. Moreover, one can provide a conceptual characterization of these virtues in terms of stable and successful dispositions, thereby satisfying the basic theoretical and empirical considerations for a theory of epistemic virtue.
7. Epistemic virtue, stability and frugality
We have argued for virtue-theoretic accounts of epistemic success, empirical adequacy, and the ability condition. This leaves the etiological condition and normative adequacy. It is necessary for our account to meet the etiological
condition in order to constitute a virtue epistemology and to achieve full normative adequacy. If we can defend the etiological condition here, we have an empirically and normatively adequate virtue epistemology that is consistent with nonskepticism and situationism. The challenge to the etiological condition from environmental luck was discussed in section 3. Here we extend that discussion to secure the etiological condition and full normative adequacy. Pritchard’s case against robust virtue epistemology requires showing that Barney’s true belief from ability is neither safe nor sufficiently from ability. In order for Alfano to have an analogous argument, he will have to show that situationism demonstrates that virtue-theoretic epistemic success can still be unsafe or not sufficiently from ability.
We argue below that situationism undermines neither.
Suppose we have an agent (Alice) that happens to be in an ideal epistemic environment. Suppose that Alice is in a room that has been constructed to be epistemically ideal, she is alone in her room thinking about bank tellers and feminists, and her environment lacks all of the ‘epistemically irrelevant features’
Alfano notes that lead us astray – no mood depressors, problematic ambient noise, information framed in misleading ways, etc. Let’s say that when Alice sits down to think about bank tellers she correctly applies basic principles of Bayesian probability calculus, correctly applies modus tollendo ponens, etc. Moreover, assume that when Alice is in ideal epistemic environments, these inferential methods are reliable for her. Suppose that lurking outside of Alice’s room are all and only environments replete with just the problematic ‘epistemically irrelevant’ variables that lead to false beliefs - mood depressors, pay phones without dimes, no one that will compliment her, etc. If Alice were to go outside her room to do the same thinking she would arrive at false beliefs because of these environmental variables. Assume that Alice randomly decided to go think in her room rather than outside, but all too easily could have decided to go outside to do her thinking.
When Alice arrives at true beliefs in her room, the virtue reliabilist will be inclined to attribute the manifestation of an intellectual virtue to Alice, and, given that she gets it right, to attribute knowledge. Alfano can say that Alice’s success in her room is due to environmental good luck rather than her reliable abilities, and her true belief also appears to be unsafe. The worries about Alice’s true belief will be twofold. We might attribute her success to good environmental luck (as in True Temp cases), or we might say her success is unsafe (as in Barney cases). In either case,
Alfano can argue that virtue epitemology fails to satisfy intuitive requirements for knowledge.
This argument overlooks the role of ‘normal conditions’ for the manifestation of an ability or competence in both cases. Sosa is careful to distinguish the role of normal conditions for assessing any performance (2007, pg. 103) . Evaluations of the success and explanatory salience of an ability are restricted to the conditions that are normal or better for its manifestation. Sosa’s requirement for knowledge is that an ”Acceptance of a deliverance constitutes knowledge only if the source is reliable, and operates in its appropriate conditions, so that the deliverance is safe, while the
correctness of one’s acceptance is attributable to one’s epistemic competence.”
(ibid.) This leads him to say that the Kalediescope believer that is deceived by a jokester in the wings is yet truly manifesting a success from ability when the jokester does not to intervene. There is a direct causal connection between the agent’s abilities and their true belief, even though the success was environmentally lucky. For Sosa, the belief is still sufficiently from ability when the jokester does not intervene. According to Sosa, the competence was operating in its normal conditions, even though it might easily have been in non-normal conditions and given an unsuccessful deliverance. The same commitment to normal conditions leads Sosa to claim that the kaledieiscope believer’s belief is also safe, again parting company with Pritchard. The reason is again that in normal conditions for the exercise of the relevant competence, the belief would not too easily have been false.
The failure of an ability when not in conditions that are “normal or better” does not count against the relevant reliability, stability or explanatory salience of the ability.
Sosa’s reading of the etiological condition would block Alfano’s attempt to exploit
Pritchard’s rejection of robust virtue epistemology. Situationist cases do not undermine the explanatory saliency of abilities or the safety of epistemic successes because they take cognitive agents outside of normal conditions for evaluating the abilities in question. Kahneman shows that the stimulus in situationist cases appears to shift the psychological states of the subjects in important ways, so we are not testing the reliability or stability of the same rational methods or abilities when we introduce situationist environmental variables. Utilizing Pritichard’s argument against robust virtue epistemology thus comes further under fire because the situationist cannot show that the same method would be in use inside the room and outside the room, or that Alice’s psychological processes are held constant.
The situationist might respond by arguing that when the method (or relevant psychological states) shift so easily, this shows that the method is not stable in the agent, and that this is still enough to undermine knowledge attributions. If an ability so easily becomes unavailable to an agent, this would appear to militate against explaining reliable success in terms of such a fragile ability. The question of what to count as ‘normal conditions’ for the exercise of Alice’s abilities is central here. If
‘Alice-outside-the-room’ is still Alice in normal conditions for the exercise of the relevant abilities, the situationist may have shown that she could all too easily have believed falsely. However, if we place ‘Alice-outside-the-room’ outside of normal conditions, we are then inclined to see the experimental conditions in the situationist literature as too distant to undermine safety or stability.
If an ability is not available to an agent in environments known to make that ability unavailable (as in situationist cases), this does not count against the relevant stability of the ability. Situationists may simply be illuminating characteristic masks of our cognitive abilities. If ambient noise and the like lead to bad epistemic outcomes as regularly as Alfano suggests, this may simply show that we need to define ‘normal conditions’ for the exercise of an ability or trait in a way that excludes their characteristic mask. All abilities and dispositions have characteristics
masks, and we typically see these failures to produce virtue relevant outcomes as non-culpable failures. The situationist’s experimental conditions are then analogous to cases like the tornado, or the lights going out, or being under water, and we can still attribute overall proper functioning to the agent despite that kind of failure. The situationist literature may indeed provide useful information about normal conditions for cognitive abilities and characteristic masks, but virtue epistemologists can build these results into the requirements for knowledge without any fundamental change in their position.
References
Alfano, M. (2011). Expanding the situationist challenge to responsibilist virtue epistemology. Philosophical Quarterly. Vol. 62 Issue 247. 223-249.
Alfano, M. (2012). Expanding the situationist challenge to reliabilist virtue epistemology. Forthcoming.
Annas, J. (2005) " Virtue Ethics: What Kind of Naturalism ?", in Stephen M. Gardiner ed., Virtue Ethics , Old and New, Cornell UP, 2005.
Anscombe, GEM (1958). “Modern Moral Philosophy”, Philosophy 33, No. 124
Boghossian, P. (2000) “Knowledge of Logic” in P. Boghossian and C. Peacocke, New
Essays on the A Priori, Oxford University Press, pp. 229-254.
Brink, D. (2001). Realism, Naturalism, and Moral Semantics" Social Philosophy &
Policy 18. 154-176.
Cuneo, T. (2007). The Normative Web: An Argument for Moral Realism, Oxford
University Press.
Chomsky, N. (1987) Generative Grammar: Its Basis, Development and Prospects,
Studies in English Literature and Linguistics, Kyoto University.
Chomsky, N. (1986) Knowledge of Language, Praeger.
Episteme, Vol. 2, No. 1, pp. 39-55, 2005.
Gigerenzer, G. (2008) Rationality for Mortals: How People Cope with Uncertainty,
Oxford University Pres.
Graham, P. (2011) ‘Epistemic Entitlement’ Nous, online January 2011.
Greco, J. (2010) Achieving Knowledge: A Virtue-Theoretic Account, Cambridge
University Press.
Greco, J. (1992). Agent reliabilism. Nous, 13, 273-296.
Haddock, H., Millar, A., & Pritchard, D. (2009). Epistemic Value , ( eds .), Oxford
University Press
Harman, G. (2000). The nonexistence of character traits. Proceedings of the
Aristotelian Society, 100: 223-226.
Harman, G. (1999). Moral philosophy meets social psychology: Virtue ethics and the fundamental attribution error. Proceedings of the Aristotelian Society, New Series
119, 316-331.
Heathwood, C. (2009) Moral and epistemic open question arguments. Philosophical
Books 50 (2009): 83-98.
Henderson and Horgan (2011) The Epistemological Spectrum. Oxford University
Press.
Hornstein, N. (1984) Logic as Grammar, MIT Press.
Houkes, E. (2002) "Normativity In Quine's Naturalism: The Technology Of Truth
Seeking". Journal For General Philosophy Of Science 33.
Hylton, P. (2007). Quine . Routledge 2007.
Jackendoff, R. (2003) Foundations of Language: Brain, Meaning, Grammar, Evolution,
Oxford Univesity Press
Jenkins, CI (2007). Epistemic Norms and Natural Facts. American Philosophical
Quarterly.
Kahneman, D. (2011) Thinking Fast and Slow, Farrar, Straus and Giroux.
Kahneman, D. & Tversky, A. (1973). On the psychology of prediction. Psychological
Review, 80, 237-251.
Kamtekar, R. (2004). Situationism and virtue ethics on the content of our character.
Ethics, 114, 458-491.
Kelp, Christoff (2009). Pritchard on Knowledge, Safety and Cognitive Achievements.
Journal of Philosophical Research 34, 51-53
Kim, J. (1988) "What is Naturalized Epistemology?" Philosophical Perspectives 2:
381-406.
Kornblith, H. (1993). Epistemic Normativity. Synthese.
Miller, C. (2003). Social psychology and virtue ethics. The Journal of Ethics, 7:4, 365-
392.
Moore, GE (1903). Principia Ethica, Cambridge: Cambridge University Press, 1903
Pritchard, Duncan (2007). The Nature and Value of Knowledge.
Pritchard, D. (2007). Recent work on epistemic value. Mind. Volume 44, Number 2,
Quine, W. (1969). Ontological Relativity and Other Essays. New York: Columbia
University Press.
Riggs, W. ( 2006). The value turn in epistemology. in New Waves in Epistemology,
(eds.) V. Hendricks & D. H. Pritchard (Ashgate)
Roberts, R. & Wood, J. (2007). Intellectual Virtues. Oxford: Oxford University Press.
Sosa, E. (2011). Knowing Full Well. Princeton: Princeton University Press.
Sosa, E. (2007). A Virtue Epistemology. Oxford: Oxford University Press.
Snow, N. (2008). Virtue as Social Intelligence: An Empirically Grounded Theory. New
York:Routledge.
Sreenivasan, G. (2002). Errors about errors: Virtue theory and trait attribution.
Mind, 111, 47-68.
Stanley, J. (2005) Knowledge and Practical Interests (New York, NY: Oxford
University Press).
Zagzebski, L. (2010). Exemplarist virtue theory. In Battaly (ed.), Virtue And Vice:
Moral and Epistemic, pp. 39-55. West Sussex: Wiley-Blackwell.