The Genealogy of Knowledge and Anti-Luck Virtue Epistemology

1 For Der Begriff des Wissens/The Concept of Knowledge, (ed.) S. Tolksdorf, (Berlin: de Gruyter, 2011). THE GENEALOGY OF THE CONCEPT OF KNOWLEDGE AND ANTI-LUCK VIRTUE EPISTEMOLOGY DUNCAN PRITCHARD University of Edinburgh ABSTRACT. In his seminal book, Knowledge and the State of Nature: An Essay in Conceptual Synthesis, Edward Craig offers a highly influential story which accounts for the nature of our concept of knowledge, a story which can be broadly cast as ‘genealogical’. I argue that while, on the face of it, this story seems to favour those robust virtue-theoretic theories of knowledge which completely analyse knowledge in terms of cognitive success that is due to reliable cognitive ability, on closer inspection this genealogical account of the concept of knowledge in fact favours a different kind of theory of knowledgewhat I call anti-luck virtue epistemologywhich analyses knowledge in terms of both cognitive ability and an anti-luck condition. 1. CRAIG ON THE GENEALOGY OF THE CONCEPT OF KNOWLEDGE In a highly influential work, Edward Craig (1990) has argued for a distinctive way of approaching the theory of knowledge, one that looks to what we might broadly speaking call the ‘genealogy’ of the concept of knowledge.1 His idea is that instead of conducting the traditional epistemological enterprise whereby we reflect on our ordinary usage of the concept of knowledge and then try to figure out how best to analyse the concept in light of this usage, we should instead begin by reflecting on what practical purpose the concept of knowledge might serve and proceed from that starting-point. To this end, Craig asks us to consider an imaginary society of cognitive agents who are very similar to us except that they lack the concept of knowledge. Like us, these agents need true beliefs in order to successfully navigate their environment, and they can either acquire such true beliefs via their own on-board cognitive resources (e.g., their perceptual faculties) or they can 2 make use of the on-board cognitive resources of others in the community by seeking them out as informants. Here is the question that Craig asks: what would prompt such a society which lacks the concept of knowledge to introduce it? Craig’s suggestion is that it would be very useful in such a community to have the conceptual resources to flag reliable informants. After all, if one is forced to depend only on one’s own on-board cognitive resources then one will be severely limited in the true beliefs that one can form about one’s environment. But if making use of others in the community as informants is to assist us in this regard then it had better be the case that these informants are reliable, since otherwise they could just as well be leading us astray rather than furnishing us with true beliefs. We could thus imagine a concept very like knowledgea kind of proto-knowledge conceptbeing employed for just this purpose. Call this proto-knowledge, ‘knowledge*’, and call anyone who possesses knowledge* a ‘knower*’. The idea is then that it would be useful to label good informants as knowers* with regard to the range of propositions which they are good informants about, and accordingly to label the accurate information that they offer on subjects about which they are good informants, knowledge*. So, for example, imagine that John lives on a hill and so has a particularly good view of what is happening in the valley below (and that he is generally truthful and helpful, etc.,). He would thus be a reliable informant when it comes to a range of propositions concerning what is happening in the valley. It would clearly be practically useful for us to flag the fact that John is a good informant in this regard, and we can do this by calling him a knower* as regards these propositions, in that his true beliefs in these propositions amount to knowledge*. Note that knowledge* is not yet the same as our concept of knowledge. For one thing, this concept only applies to other people’s true beliefs, while we also use the concept of knowledge to classify our own beliefs. In addition, in deciding whether an agent is a knower* we are only assessing how good an informant is relative to the actual circumstances that she finds herself ini.e., the ‘live’ error-possibilities that are in play in her environment. In the case of John just described, for example, all that is at issue is whether he has a good view of the valley and the ability (and inclination) to make use of this advantage. The salient error-possibilities when it comes to the question of whether he has knowledge* are thus such things as whether there is something in his environment which is obscuring his view (a heavy fog, say). Note, however, that our concept of knowledge treats a far greater range of error-possibilities as salient. In particular, it is also responsive to potential error-possibilities, even if they are not actual. So, for example, we do not treat someone as having knowledge if they formed their cognitive success is due to luck, even if in the actual circumstances that obtained there was in fact no impediment to the acquisition of true belief (we will consider an examplethe ‘Barney casewhich illustrates this point in §3). 3 Interestingly, however, we would expect the proto-concept which is designed to pick out reliable informants to evolve over time so that it begins to resemble our concept of knowledge. For example, we could imagine knowledge* ultimately being used to classify oneself and not just others, and the application of the concept being ‘stretched’ so that it is responsive to non-actual but potential error-possibilities (a process that Craig calls ‘objectification’). Over time, then, we would expect knowledge* to evolve into knowledge. Call this Craig-style story about the genealogy of the concept of knowledge the genealogical account. Many philosophers find the genealogical account very compelling, at least in broad outline.2 That is, while many will quibble over the details, I think most epistemologists would be inclined to grant the general thesis of the genealogical account that the fundamental point of the concept of knowledge is to flag reliable informants.3 My goal in this paper is not to defend the genealogical account, but rather to explore what consequences this widely endorsed view has for our thinking about knowledge since, as we will see, I think this proposal may well end up supporting a very different account of knowledge than many suppose. 2. FROM THE GENEALOGICAL ACCOUNT TO VIRTUE EPISTEMOLOGY On the face of it, the genealogical account would seem to favour those theories of knowledge which put reliable cognitive abilities at the centre of their theory. This is because a necessary condition of a reliable informant is surely that the informant forms her beliefs in the target propositions via her own reliable on-board cognitive abilities. Accordingly, if the fundamental purpose of the concept of knowledge is to flag reliable informants, then one would expect any viable theory of knowledge to make essential appeal to cognitive abilities. In the contemporary epistemological literature those theories of knowledge which make essential appeal to reliable cognitive abilities are known as virtue epistemologies.4 Thus we might conclude from the foregoing that the genealogical account lends support for virtue epistemology. In general terms, I think this is entirely correct, though since there are a wide range of virtuetheoretic proposals in this regard this still leaves an awful lot of room for further discussion. In particular, it will be a further question whether the genealogical account lends support for a particular form of virtue epistemology. Before we get to this further question, however, let us first explore in a little more detail the motivation (the implications of the genealogical account aside) for having an ability condition play a central role in one’s theory of knowledge. It is certainly the case that we have a strong intuition that knowledge demands cognitive ability. This can be brought out quite nicely by considering 4 cases where agents are forming beliefs which are guaranteed to be true but where this cognitive success in no way reflects the cognitive abilities of the agent concerned. In such cases we are strongly disinclined to attribute knowledge. Consider first the following case, which we will call ‘Temp’: Temp Temp forms his beliefs about the temperature in the room by consulting a thermometer. His beliefs, so formed, are highly reliable, in that any belief he forms on this basis will always be correct. Moreover, he has no reason for thinking that there is anything amiss with this thermometer. But the thermometer is in fact broken, and is fluctuating randomly within a given range. Unbeknownst to Temp, there is an agent hidden in the room who is in control of the thermostat whose job it is to ensure that every time Temp consults the thermometer the ‘reading’ on the thermometer corresponds to the temperature in the room.5 I take it we have a strong intuition that Temp cannot acquire knowledge in this fashion, since one simply cannot gain knowledge of the temperature of the room by consulting what is, unbeknownst to one, a broken thermometer. Interestingly, as this example demonstrates, this remains the case even if even the scenario is explicitly set up so that Temp is guaranteed to be correct in the beliefs that he forms by consulting this broken thermometer. The problem is that even despite this guarantee of cognitive success there is simply the wrong direction of fit between Temp’s beliefs and the corresponding facts, in that the facts are changing to fit with what Temp believes rather than Temp’s beliefs being responsive to the facts. This is because Temp’s cognitive success is nothing to do with his cognitive abilities and everything to do with a feature of his environmentviz., the hidden ‘helper’. In contrast, in genuine cases of knowledge the cognitive success in question is significantly related to the cognitive abilities of the agent.6 Consider also this second case, which we will call ‘Alvin’: Alvin Alvin has a brain lesion. An odd fact about the brain lesion that Alvin has, however, is that it causes the sufferer to form the (true) belief that he has a brain lesion. Accordingly, Alvin truly believes that he has a brain lesion.7 Given how Alvin is forming his beliefs he is guaranteed to be right. Clearly, though, Alvin does not have knowledge in this case, and the reason for this is that his beliefs are true despite his cognitive abilities and not as a result of them. Indeed, the Alvin case is an example of a cognitive malfunction, albeit a cognitive malfunction that just happens to be reliable. As such, the reliability of this belief-forming process cannot reflect the cognitive abilities of the agent concerned. Cases like this demonstrate that one of the core intuitions we have about knowledge concerns the fact that knowledge entails a cognitive success that is in some substantive sense due to the agent’s reliable cognitive ability. Call this the ability intuition. The genealogical account can explain why we have this intuition, since given that account we would expect our concept of 5 knowledge to be responsive to whether an agent is forming her beliefs in a way that reflects her reliable cognitive abilities. This is because a reliable informant about a certain subject matter will be an informant who is forming true beliefs about that subject matter via her reliable cognitive abilities. On this view, then, it is no wonder that we have the ability intuition. So the genealogical account can explain why we have the ability intuition, and thereby offer support for those virtue-theoretic accounts of knowledge which place reliable cognitive abilities at their heart. On the face of it, however, it also seems to offer support for a specifically robust rendering of virtue epistemology, whereby the appropriate employment of reliable cognitive abilities is not only necessary for knowledge, but also (along with true belief) sufficient. After all, if it were to turn out that one’s virtue-theoretic account of knowledge invoked other conditions beside an ability condition, then so long as one accepted the genealogical account one would be faced with a puzzleviz., why did the concept of knowledge evolve out of the concept of knowledge* in such a way that it required a further condition over and above an ability condition? In order to make this concern vivid, suppose that one offered a theory of knowledge which incorporated both an ability condition and a further condition which is specifically invoked in order to deal with Gettier-style cases.8 This would be an understandable way to go, since Gettierstyle cases often do seem to involve agents appropriately employing their reliable cognitive abilities in order to attain a true belief, and yet such cases are explicitly designed to elicit the intuition that the cognitive success at issue is just too lucky to count as knowledge. Accordingly, it would seem that an ability condition will not by itself suffice, with true belief, for knowledge. Consider, for instance, the following example, which we will call ‘Roddy’: Roddy Using his reliable perceptual faculties, Roddy non-inferentially forms a true belief that there is a sheep in the field before him. His belief is also true. Unbeknownst to Roddy, however, the truth of his belief is completely unconnected to the manner in which he acquired this belief since the object he is looking at in the field is not a sheep at all, but rather a sheep-shaped object which is obscuring from view the real sheep hidden behind.9 Intuitively, Roddy is employing his reliable cognitive abilities entirely appropriately and coming, thereby, to form a true belief in the target proposition. And yet equally intuitively he lacks knowledge, in that it is simply a matter of luck that his belief is true, and luck of this sort seems entirely incompatible with knowledge possession. Thus, in order to deal with cases like that of Roddy, one might be inclined to supplement one’s virtue-theoretic account of knowledge with some sort of condition which excludes these cases, such as an anti-luck condition of some sort (i.e., that it should not be the case that one’s belief is only true as a matter of luck). The problem, however, is that from the point of view of the genealogical account it would seem to be just plain mysterious why we would have ended up with a concept of knowledge which treated knowledge 6 as being mostly about the employment of reliable cognitive ability, but which also demanded that a further codicil be met so as to avoid Gettier-style cases. Interestingly, a dominant trend in virtue epistemology has been to opt for a more robust rendering of this thesis which does away with the need to appeal to any further epistemic condition over and above the ability condition. Call any such view a robust virtue epistemology, in contrast to a modest virtue epistemology which also makes appeal to further (non-virtue-theoretic) epistemic conditions. Clearly one advantage of robust virtue epistemology is its theoretical simplicity when compared with its modest counterpart.10 Moreover, given the foregoing it would also seem that the genealogical account lends additional prima facie support to going down the simpler robust virtue-theoretic route.11 If robust virtue epistemology were a sound position, then this would be all to the good, but unfortunately, as we’ll see in the next section, this proposal faces some pretty stiff problems. We are thus faced with a puzzle, in that the genealogical account seems to favour robust virtue epistemology over modest virtue epistemology, and yet robust virtue epistemology seems unsustainable. Resolving that puzzle will be the task of §4. 3. CONTRA ROBUST VIRTUE EPISTEMOLOGY As just noted, it is without question that robust virtue epistemology is a proposal with many theoretical virtues. Unfortunately, it is also a proposal that faces some pretty stiff problems too. Before we get to the problems facing the view, however, we first need to explain what a robust virtue epistemology is, and in particular how it evades the problem posed by Getter-style cases that was noted earlier. In essence, robust virtue epistemologies grant that knowledge is more than just the conjunction of cognitive success (i.e., true belief) and reliable cognitive ability, since a mere conjunction of cognitive success and reliable cognitive ability will not exclude Gettier-style cases (as, indeed, we saw above). Nonetheless, they maintain that we do not need to introduce a further non-virtue-theoretic conditionsuch as a Gettier-excluding anti-luck conditioninto our theory of knowledge. Instead, what is required is simply that we specify the relationship that the cognitive success must bear to the reliable cognitive abilityviz., that the cognitive success should be because of the target reliable cognitive ability.12 The natural way to read the ‘because of’ here, I take it, is in causal explanatory terms, such that what is being claimed is that the cognitive success is primarily explained by, and hence primarily creditable to, the relevant reliable cognitive ability.13 So construed, this modification to 7 the view seems to comfortably solve the problem posed for modest virtue epistemology by the Gettier-style cases, since these cases do appear to involve a cognitive success which is precisely not primarily creditable to the reliable cognitive ability of the subject but rather explained by factors entirely outwith the subject’s cognitive agency. In the Roddy case, for example, what explains Roddy’s cognitive success is not his reliable perceptual abilities but merely the fact that there happens to be a sheep hidden from view behind the sheep-shaped object that he is looking at. So far so good for robust virtue epistemology, then. The problem comes once we start to consider other cases of knowledge-undermining epistemic luck beside the cases that fit the standard Gettier-style cases. In particular, consider the following familiar case, which we will call ‘Barney’: Barney Using his reliable perceptual faculties, Barney non-inferentially forms a true belief that the object in front of him is a barn. Barney is indeed looking at a barn. Unbeknownst to Barney, however, he is in an epistemically unfriendly environment when it comes to making observations of this sort, since most objects that look like barns in these parts are in fact barn façades.14 Now Barney’s cognitive success is clearly lucky, given the environment he is in. After all, most of the things that look to Barney like barns in this locale are in fact barn façades, and hence he is very lucky to have formed a true belief in this scenario. On this score, the case is analogous to a normal Gettier-style case like Roddy. Where it differs, however, is in how the agent concerned really is seeing what he takes himself to seeviz., he really is looking at a genuine barn. The contrast with the Roddy case is instructive here, since it is vital to that case that Roddy is not looking at what he thinks he is looking at, even if the belief he forms as a result of what he sees happens to be true. This point is important since it makes a big difference to our assessment of what explains Barney’s cognitive success. For while we saw above that Roddy’s cognitive success is clearly not primarily creditable to his reliable cognitive ability, given that Barney really is seeing a genuine barn it does seem that his cognitive success is primarily creditable to his reliable cognitive ability. After all, and unlike the Roddy case, it is not as if there is something other than his reliable cognitive ability which is primarily creditable for his cognitive success; instead, it is the fact that he is appropriately using his reliable perceptual faculties in observing the barn which explains why he truly believes that there is a barn before him. But if that’s right, then it seems that merely appealing to a ‘because of’ relation will not suffice to ensure that one’s robust virtue epistemology can deal with all cases of knowledge-undermining epistemic luck.15 If this were not a serious enough problem for robust virtue epistemology, a second difficulty is waiting in the wings. Consider the following example, which will call ‘Jenny’: Jenny 8 Jenny gets off the train in an unfamiliar city and asks the first person that she meets for directions. The person that she asks is indeed knowledgeable about the area, and helpfully gives her directions. Jenny believes what she is told and goes on her way to her intended destination.16 According to standard views in the epistemology of testimony, Jenny gains knowledge in this case. Indeed, it is usually argued that if we are unwilling to ascribe knowledge in such cases where an agent trusts a knowledgeable informant, then we will have to endorse a widespread scepticism about much of our testimonial knowledge, given that so much of it is acquired in a similar fashion. The problem, however, is that it is hard to see how crediting knowledge to Jenny can be squared with robust virtue epistemology. After all, given that she forms her belief by, for the most part, trusting the word of another, Jenny’s cognitive success does not seem to be explained by her reliable cognitive abilities specifically at all, but rather by her informant’s. If that’s right, then it seems that robust virtue epistemology should deny knowledge in these cases, despite this being a counterintuitive result, and accordingly align themselves with a ‘sceptical’ view regarding the epistemology of testimony.17 Crucially, notice that this problem exacerbates the difficulty posed by the kind of knowledge-undermining epistemic luck in play in the Barney case. This is because while the Barney case shows that robust virtue epistemology is too weak a theory of knowledge, in that it wrongly counts cases of cognitive success as knowledge, the Jenny case shows that robust virtue epistemology is also too strong, in that it wrongly prevents certain cases of cognitive success counting as knowledge. Collectively, then, these two problems pull the proponent of robust virtue epistemology in opposing directions, since they call for both a weakening and a strengthening of the view. For this reason, it will be very difficult for the defender of robust virtue epistemology to offer a principled defence of the position which simultaneously deals with both difficulties.18 There is a good reason why robust virtue epistemology faces these challenges, and it concerns the fact that there is no appropriate rendering of the ability condition on knowledge which can deal with the quandary posed by knowledge-undermining epistemic luck. As the Barney case illustrates, even if you in effect ‘beef-up’ the ability condition by adding a ‘because of’ relation between the target cognitive success and the target reliable cognitive ability, it is still possible for the cognitive success in question to be subject to knowledge-undermining epistemic luck. But the problem is that in beefing up the ability condition in order to try to eliminate knowledgeundermining epistemic luck the proponent of robust virtue epistemology ends up with a theory of knowledge which is now unable to accommodate certain cases of bona fide knowledge, such as the testimonial knowledge at issue in the Jenny case. I suggest that what this demonstrates is the need for an anti-luck condition in one’s theory of knowledge in addition to an ability condition. Moreover, I maintain that this way of thinking 9 about knowledge is not an ad hoc response to the Gettier problem, but rather reflects the fact that these conditions are responding to two distinct master intuitions that we have about knowledge. The first is the ability intuition that we noted above. The second is the anti-luck intuition that knowledge requires cognitive success that isn’t due to luck. Now at first blush one might think that these two intuitions are essentially the same, in that for one’s cognitive success to be due to one’s ability as the ability intuition demands is surely for it to not be due to luck, and for one’s cognitive success to not be due to luck as the anti-luck intuition demands is surely for it to be due to one’s cognitive ability. Closer inspection, however, reveals that they make distinct demands on one’s theory of knowledge. We have already see that one can satisfy the ability intuition in a robust fashion and yet still fail to satisfy the anti-luck intuitionthat was precisely the moral of the Barney case discussed above. We can also illustrate how one can satisfy the anti-luck intuition without thereby satisfying the ability intuition by appealing to the Temp case that we outlined in §2 in support of the ability intuition. For recall that what was significant about that case was that the agent concerned was actually guaranteed to be cognitive successful, given how he was forming his beliefs. Accordingly, it cannot possibly be the case that his cognitive success is a matter of luck. Nonetheless, Temp does not acquire knowledge in this case, and the reason for this, as we noted, is that his cognitive success is not due to his reliable cognitive abilities at all, but rather entirely due to the assistance of a hidden helper.19 Given that these two intuitions impose distinct demands on one’s theory of knowledge, it follows that one will need distinct epistemic conditions in order to accommodate them. We are thus led to a theory of knowledge that has both an anti-luck and an ability condition, a view that I call anti-luck virtue epistemology. Since it incorporates an ability condition this view is a form of modest virtue epistemology, albeit one which gives equal weight to both the anti-luck condition and the ability condition (i.e., it does not regard the anti-luck condition as being a mere codicil which is tacked-onto one’s theory of knowledge in order to deal with certain problem cases involving knowledge-undermining epistemic luck). I defend this proposal at length elsewhere,20 but it ought to be clear from the foregoing what advantages such a view has. To begin with, notice that if you have an anti-luck condition in your theory of knowledge then you can thereby deal with those troublesome epistemological cases which trade on knowledge-undermining luck, including both standard Gettier-style cases like Roddy and nonstandard Gettier-style cases like Barney. For another, your ability condition will deal with those epistemological cases which trade on the ability intuition, such as Temp and Alvin. And since one has the anti-luck condition to appeal to, one does not need to beef-up one’s ability condition in order to try to eliminate knowledge-undermining luck. Accordingly, one does not make the 10 mistake made by robust virtue epistemology of setting the bar for knowledge too high and thereby excluding certain bona fide cases of knowledge, such that present in the Jenny case. In short, with both conditions working together in one’s theory of knowledge, one is able to satisfactorily deal with a wide range of problem cases in epistemology. There is, however, one point on which anti-luck virtue epistemology seems to be at a major theoretical disadvantage, and that concerns the motivation for the view. In particular, while I think it is undeniable that we have these two intuitions about knowledge and that they make distinct demands upon us, a worry still remains about why our concept of knowledge should incorporate these two discrete epistemic conditions. The puzzle seems to be exacerbated once one reflects on the plausibility of the genealogical account, for while that offers a cogent rationale for why we might have the ability condition, it does not seem to explain at all how a separate anti-luck condition may have come about. If that’s right, then it is a fairly serious strike against anti-luck virtue epistemology. More generally, we seem to be faced with a dilemma here. On the one hand, insofar as we accept the genealogical account then we seem to be led towards endorsing a form of robust virtue epistemology, a position which we have found to be untenable. On the other hand, insofar as we opt for the alternative anti-luck virtue epistemology which can avoid the problems facing robust virtue epistemology, then we seem to be forced to reject the highly intuitive genealogical account. Fortunately, as we will see in the next section, there is a crucial ambiguity in the genealogical account, and once this is brought to light this account does offer a good explanation of why our concept of knowledge should have two separate epistemic conditions of this sort. On closer analysis, then, far from favouring robust virtue epistemology, the genealogical account in fact favours anti-luck virtue epistemology. Thus, the dilemma just set out is in fact entirely illusory, in that one does not need to choose between anti-luck virtue epistemology and the genealogical account. 4. RETHINKING THE GENEALOGICAL ACCOUNT The ambiguity that I have in mind in the genealogical account concerns the very notion of a reliable informant. In one sense, it can mean an informant who possesses a reliable cognitive ability with regard to the target subject matter (and who is willing to sincerely communicate what she believes, something that we will take for granted in what follows). In another sense, it means an informant whom one can rely on (i.e., whose information will not lead you astray). Now one might naturally think that this is a distinction without a difference, in that informants who possess 11 reliable cognitive abilities in the sense just specified are thereby informants one can rely on, and vice versa. Closer inspection, however, reveals that first appearances are deceptive on this score. In order to see this, we just need to notice that it can be appropriate to rely on an informant who is forming her true belief via an unreliable cognitive ability, and also that it can be inappropriate to rely on an informant who nevertheless is forming a true belief via a reliable cognitive ability. First, consider a potential informant who possesses a reliable cognitive ability as regards a certain subject matter but who is in an environment in which there exists a misleading defeater, one which you know about, but the prospective informant does not, and one which moreover you are unable to defeat. An example might be an informant who is a reliable barn detector, but where you have been given a misleading ground (e.g., false testimony from a good source) for supposing that the informant is in barn façade county. Given that this is a misleading defeater, the informant is in fact a reliable informant about the relevant subject matter. But given also that you know about the misleading defeater, and are aware that you are unable to defeat that defeater, would you be able to rely on this informant? Surely not. The converse point also holds. In particular, we can imagine a case where there are compensating factors in play, known only to us, which mean that we can rely on the information presented to us by an informant even though this information is not the product of the informant’s reliable cognitive abilities. Imagine, for example, an informant who thinks that they have clairvoyant powers, but in fact is mistaken on this score (and we know this). Suppose further that we also know that this informant’s wife is a very powerful person who wants her husband to continue to believe that he has this power and hence does what she can, where possible, to make sure that events turn out in the way that her husband predicts. Finally, suppose that we know that the informant’s wife can fix the result of any horse race. With this knowledge in hand, the testimony of the informant regarding who will win tomorrow’s horse race would certainly be information that one could rely on, even though the informant’s true belief in this regard is in no way the product of a reliable cognitive ability. In general, what is key to both of these kinds of cases is the role that luck is playing. In cases in which the informant’s relevant cognitive abilities are reliable but where we are nonetheless unable to rely on the information she provides, the problem is that a dose of bad epistemic luck is cancelling out the good epistemic luck that our informant possesses the relevant reliable cognitive abilities (and thus is in this sense a good informant). In the case described above, for example, this bad epistemic luck is the presence of the undefeated misleading defeater regarding the barn façades. In contrast, in cases in which the informant lacks the relevant reliable cognitive abilities but is nonetheless providing us with information that we can rely on, a dose of good epistemic luck on our part is cancelling out the poor epistemic luck that our informant lacks the relevant 12 reliable cognitive abilities (and thus is in this sense a bad informant). In the ‘clairvoyant’ case described above, for example, this good epistemic luck is our knowledge of the compensating factors in play. With this point in mind, it ought to be clear why this ambiguity in the idea of a reliable informant explains why the concept of knowledge that evolves from the proto-concept will generate both the anti-luck and the ability intuition. For as the range of cases which the concept of knowledge is meant to apply to widens, so the distinction will open up between good informants who are reliable and good informants that we can rely on, and we would expect the concept of knowledge that results to respect both sides of this distinction. In particular, examples where an agent possesses the relevant reliable cognitive abilities but where the presence of epistemic luck means that we would not be able to rely on this agent qua informant would not be counted as cases of knowledge. Similarly, those cases in which an agent forms a true belief in an epistemically friendly environment—such that any true belief so formed would not be subject to epistemic luck—would not be counted as cases of knowledge so long as the agent concerned failed to exhibit the relevant reliable cognitive abilities (even though we could rely on this agent qua informant). In short, the concept of knowledge that results will both (i) disallow cases of true belief as knowledge where the belief isn’t appropriately due to the relevant cognitive abilities on the part of the agent, and (ii) disallow cases of true belief as knowledge where the truth of the belief is substantively due to luck. A very plausible and popular story about the genealogy of the concept of knowledge thus lends support to anti-luck virtue epistemology after all, despite first appearances. In fact, if I am right that the goal of picking out reliable informants is ambiguous in the way just described, then contrary to the prevailing wisdom on this score, this ‘just so’ account of the concept of knowledge actually favours anti-luck virtue epistemology over rival proposals, such as robust virtue epistemology. 5. CONCLUDING REMARKS I was very open right from the off that my purpose here was not defend the genealogical account, but merely to note its plausibility and consider what implications it had for our thinning about knowledge. I have now argued that, despite first appearances, such a proposal in fact lends support to the account of knowledge that I favour as opposed to a competing account. Nonetheless, the point remains that I have not offered a detailed defence of this claim, and so one might naturally be concerned that this ‘result’ is hostage to whether the genealogical account 13 stands the test of time. While this worry is, strictly speaking, entirely sound, I think we should be wary about overstating it. For one thing, let us not forget that there is as yet no competing proposal to the genealogical account available in the literature, and hence no alternative standard against which to evaluate the merits of anti-luck virtue epistemology on this score. Accordingly, it is hard to avoid taking the genealogical account as the default view. But a more important issue here is that the success of anti-luck virtue epistemology in dealing with cases that other popular theories of knowledge struggle with itself offers supporting grounds for the genealogical account, at least once we recognise, as I have argued here, that the notion of a reliable informant which is central to the genealogical account is crucially ambiguous. We thus have a highly plausible account of why we have the concept of knowledge that we do and a highly plausible account of knowledge itself which are mutually supporting. It is not then as if we have simply taken the genealogical account as read and then motivated anti-luck virtue epistemology as a result. Rather, we have independently motivated anti-luck virtue epistemology and then shown how it accords with the genealogical account, properly construed. In this sense, then, the genealogical account now enjoys additional theoretical support that it lacked when we began this project.21 14 REFERENCES Axtell, G. (1997). ‘Recent Work in Virtue Epistemology’, American Philosophical Quarterly 34, 410-30. Chisholm, R. (1977). Theory of Knowledge, (2nd ed.), Englewood Cliffs, New Jersey: Prentice-Hall. Craig, E. (1990). Knowledge and the State of Nature: An Essay in Conceptual Synthesis, Oxford: Clarendon Press. Fricker, E. (1995). ‘Telling and Trusting: Reductionism and Anti-Reductionism in the Epistemology of Testimony’, Mind 104, 393-411. Fricker, M. (2007). Epistemic Injustice: Power and the Ethics of Knowing, Oxford: Oxford University Press.  (2010). ‘Scepticism and the Genealogy of Knowledge: Situating Epistemology in Time’, Social Epistemology, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 51-68, Oxford: Oxford University Press. Gelfert, A. (2011). ‘Genealogies of the Concept of Knowledge and Their Cognitive-Ecological Dimension’, Episteme 8, 67-82. Goldman, A. (1976). ‘Discrimination and Perceptual Knowledge’, Journal of Philosophy 73, 771-91. Greco, J. (1999). ‘Agent Reliabilism’, Philosophical Perspectives 13, 273-96.  (2000). Putting Skeptics in Their Place: The Nature of Skeptical Arguments and Their Role in Philosophical Inquiry, Cambridge: Cambridge University Press.  (2003). ‘Knowledge as Credit for True Belief’, Intellectual Virtue: Perspectives from Ethics and Epistemology, (eds.) M. DePaul & L. Zagzebski, 111-34, Oxford: Oxford University Press.  (2007). ‘The Nature of Ability and the Purpose of Knowledge’, Philosophical Issues 17, 57-69.  (2008). ‘What’s Wrong With Contextualism?’, The Philosophical Quarterly 58, 416-36.  (2009a). Achieving Knowledge, Cambridge: Cambridge University Press.  (2009b). ‘The Value Problem’, Epistemic Value, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 313-21, Oxford: Oxford University Press. Greco, J., & Turri, J. (2009). ‘Virtue Epistemology’, Stanford Encyclopaedia of Philosophy, (ed.) E. Zalta, http://plato.stanford.edu/entries/epistemology-virtue/. Henderson, D. (2011). ‘Gate-Keeping Contextualism’, Episteme 8, 83-98. Kappel, K. (2010). ‘On Saying that Someone Knows: Themes from Craig’, Social Epistemology, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 69-88, Oxford: Oxford University Press. Kornblith, H. (2011). ‘Why Should We Care About the Concept of Knowledge?’, Episteme 8, 3852. Kelp, C. (2011). ‘What’s the Point of ‘Knowledge’ Anyway?’, Episteme 8, 53-66. Kusch, M. (2009). ‘Testimony and the Value of Knowledge’, Epistemic Value, (eds.) A. Haddock, A. Millar & D. H. Pritchard, 60-94, Oxford: Oxford University Press.  (2011). ‘Knowledge and Certainties in the Epistemic State of Nature’, Episteme 8, 6-23. Kvanvig, J. (1992). The Intellectual Virtues and the Life of the Mind, Savage, MD: Rowman & Littlefield. Lackey, J. (2007). ‘Why We Don’t Deserve Credit for Everything We Know’, Synthese 158, 345-61. Lane, M. (1999). ‘States of Nature, Epistemic and Political’, Proceedings of the Aristotelian Society 99, 211-24. Neta, R. (2006). ‘Epistemology Factualized: New Contractarian Foundations for Epistemology’, Synthese 150, 247-80. Montmarquet, J. (1993). Epistemic Virtue and Doxastic Responsibility, Lanham, NJ: Rowman & Littlefield Plantinga, A. (1993). Warrant: The Current Debate, Oxford: Oxford University Press. Pritchard, D. H. (2005). Epistemic Luck, Oxford: Oxford University Press.  (2007a). ‘Anti-Luck Epistemology’, Synthese 158, 277-97.  (2007b). ‘Recent Work on Epistemic Value’, American Philosophical Quarterly 44, 85-110.  (2007c). ‘The Value of Knowledge’, Stanford Encyclopædia of Philosophy, (ed.) E. Zalta, 15 http://plato.stanford.edu/entries/knowledge-value/. (2008). ‘Radical Scepticism, Epistemic Luck and Epistemic Value’, Proceedings and Addresses of the Aristotelian Society (suppl. vol.) 82, 19-41.  (2009a). ‘Apt Performance and Epistemic Value’, Philosophical Studies 143, 407-16.  (2009b). Knowledge, London: Palgrave Macmillan.  (2009c). ‘Knowledge, Understanding and Epistemic Value’, Epistemology (Royal Institute of Philosophy Lectures), (ed.) A. O’Hear, 19-43, Cambridge: Cambridge University Press.  (2009d). ‘The Value of Knowledge’, Harvard Review of Philosophy 16, 2-19.  (2010). ‘Anti-Luck Virtue Epistemology’, manuscript. Pritchard, D. H., Millar, A., & Haddock, A. (2010). The Nature and Value of Knowledge: Three Investigations, Oxford: Oxford University Press. Sosa, E. (1991). Knowledge in Perspective: Selected Essays in Epistemology, Cambridge: Cambridge University Press. —— (2007). A Virtue Epistemology: Apt Belief and Reflective Knowledge, Oxford: Clarendon Press. —— (2009). Reflective Knowledge: Apt Belief and Reflective Knowledge, Oxford: Clarendon Press. Williams, B. (2002). Truth and Truthfulness: An Essay in Genealogy, Princeton, NJ: Princeton University Press. Zagzebski, L. (1996). Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge, Cambridge: Cambridge University Press.  (1999). ‘What is Knowledge?’, The Blackwell Guide to Epistemology, (eds.) J. Greco & E. Sosa, Oxford: Blackwell.  16 NOTES 1 I say that the proposal is ‘broadly speaking’ genealogical because while Craig does consider the question of what practical purpose might have conceivably prompted us to acquire such a concept, he is not offering a historical account of how this concept came into being. 2 Although Craig’s proposal has been highly influential within contemporary epistemology, until quire recently there have been relatively few published discussions of it. For some useful discussions of this proposal, see Lane (1999), Williams (2002), Neta (2006), Greco (2007; 2008), Fricker (2007; 2010), Kusch (2009; 2011), Kappel (2010), Gelfert (2011), Henderson (2011), Kelp (2011) and Kornblith (2011). 3 For two sceptical treatments of the genealogical account, see Kelp (2011) and Kornblith (2011). 4 Note that this is to construe virtue epistemology in rather broad fashion, albeit in a way that I think will be familiar to many readers. For example, this characterisation of virtue epistemology would treat Plantinga’s (1993) ‘proper function’ account of knowledge as a kind of virtue epistemology, even though he has explicitly disavowed this description of his view. Clearly, though, Plantinga is here thinking of virtue epistemology as involving more than just the idea that the employment of reliable cognitive abilities is central to the acquisition of knowledge. For some of the key defences of virtue epistemology, see Sosa (1991; 2007; 2009), Kvanvig (1992), Montmarquet (1993), Zagzebski (1996; 1999) and Greco (1999; 2000; 2009a). For two very useful overviews of the literature on virtue epistemology, see Axtell (1997) and Greco & Turri (2009). 5 I put forward, and discuss, the Temp case in a number of places. See, for example, Pritchard (2009b, ch. 2; 2010) and Pritchard, Millar & Haddock (2010, ch. 3). 6 There is, admittedly, a certain amount of ‘noise’ in this example. After all, it is not as if Temp is playing no part at all in the cognitive process that results in his cognitive successfor one thing, his perceptual abilities are presumably functioning as they should be as part of his checking of the reading on the (broken) thermometer. Even so, I think we can easily see that whatever cognitive ability Temp is displaying in this case it is irrelevant to his cognitive success. Nonetheless, if one prefers a ‘cleaner’ example then the obvious way to do this is to make the epistemic helper a benevolent demon of some sort. For example, we could imagine an agent forming beliefs in an epistemically terrible fashionby tossing coins, saybut who is guaranteed to be cognitively successful because the benevolent demon will reliably ensure that whatever our hero believes is true. As with Temp, then, there is the wrong direction of fit between belief and fact, in that the facts are responsive to what the agent believes rather than vice versa. 7 This case is due to Plantinga (1993, 199). 8 This is just the sort of view that Greco (1999; 2000) proposes in his early work on virtue epistemology. 9 This case is adapted from one offered by Chisholm (1977, 105). 10 Another theoretical advantage is its apparent ability to be able to offer a compelling account of the distinctive value of knowledge. See, for example, Greco (2009b) for a clear statement of this putative advantage of the view. For critical discussion of this general claim, see Pritchard (2009d; cf. Pritchard 2008; 2009a; 2009c) and Pritchard, Millar & Haddock (2010). See also Pritchard (2007b; 2007c). 11 Greco (2008; 2009a) is one proponent of robust virtue epistemology who has explicitly argued that the genealogical account lends support to this kind of view. 12 For the key virtue-theoretic proposals of this sort, see Sosa (1991; 2007; 2009), Zagzebski (1996; 1999) and Greco (2009a; cf. Greco 2003; 2007a; 2008). 13 This is the line taken by Greco (2008; 2009a), for example. In contrast, Zagzebski (e.g., 1999) treats this relation as an undefined primitive, while Sosa (2007; 2009) understands it in terms of the manifestation of a power. For reasons of space, and because I discuss them at some length elsewheresee, in particular, Pritchard (2009a; 2010) and Pritchard, Millar & Haddock (2010, chs. 2-3)I will not be exploring these alternative proposals here. 14 The barn-façade case was first put forward in print by Goldman (1976), who credits the example to Carl Ginet. 15 Elsewhere, I diagnose this inability on the part of robust virtue epistemology to handle the Barney case in terms of a more general failure on the part of robust virtue epistemology to recognise an important distinction between two types of knowledge-undermining epistemic luck, what I call ‘intervening’ and ‘environmental’ epistemic luck (where the kind of epistemic luck specifically at issue in the Barney case is of the latter variety). For more on this distinction and its application to robust virtue epistemology, see Pritchard, Millar & Haddock (2010, ch. 2) and Pritchard (2010). 16 This case is adapted from one offered by Lackey (2007), albeit to demonstrate a slightly different point. 17 Despite being counterintuitive, this view of the epistemology of testimony—known as reductionism—is not without its adherents. See especially Fricker (e.g., 1995). 18 I explore in detail elsewhere how proponents of robust virtue epistemology might respond to these problems, and why such responses are unsatisfactory. See especially, Pritchard, Millar & Haddock (2010, chs. 2-4) and Pritchard (2010). 19 One moral of cases like Temp is thus that an anti-luck epistemology which did not incorporate an ability condition would be doomed to failure. For discussion of anti-luck epistemology, see Pritchard (2005; 2007a). 20 See especially, Pritchard, Millar & Haddock (2010, chs. 2-4) and Pritchard (2010). 21 Thanks to David Bloor, Georgi Gardiner, Axel Gelfert, Mikkel Gerken, Sandy Goldberg, Alvin Goldman, Peter Graham, John Greco, David Henderson, Klemens Kappel, Hilary Kornblith, Martin Kusch, Christoph Kelp, Ram 17 Neta and Shane Ryan for helpful discussion on related topics. Special thanks to Stefan Tolksdorf for detailed comments on an earlier version. This paper was written while I was in receipt of a Phillip Leverhulme Prize.