RELIABILITY AND SOCIAL KNOWLEDGE- RELEVANT RESPONSIBILITY*

Knowledge seems to need the admixture of de facto reliability and epistemic responsibility. But philosophers have had a hard time in attempting to combine them in order to achieve a satisfactory account of knowledge. In this paper I attempt to find a solution by capitalizing on the real and ubiquitous human phenomenon that is the social dispersal of epistemic labour through time. More precisely, the central objective of the paper is to deliver a novel and plausible social account of knowledge-relevant responsibility and to consider the merits of the proposed combination of reliability and responsibility with respect to certain cases of unreflective epistemic subjects.

Reliabilism, like all good theories, has problem cases. In fact, a pure reliabilist theory of knowledge doesn't seem plausible given certain cases of unreflective epistemic subjects. As it happens, it is often thought that a satisfactory account of knowledge should combine objective and subjective standards of appropriateness for a true belief to be knowledge. Normally this is understood in terms of a combination of de facto reliability and epistemic responsibility. But the problem, with which philosophers have had a hard time, is to find a satisfactory combination (WILLIAMS, 2008) and this paper attempts to find one given a particular responsibilist approach that consists roughly in reflectively endorsing the knowledge-yielding procedures. More precisely, the main goal of this paper is to put forward a knowledge-relevant notion of responsibility that is suitable to deal with some problem cases about epistemically naïve subjects the pure reliabilist must face (but I won't presently suggest this combination can also cope with other problems the views combined must individually face). And I argue that such responsibility requirement needs to be (partly) framed at the social level (not merely at the Trans/Form/Ação, Marília, v. 38, n. 1, p. 187-212, Jan./Abr., 2015 personal level). So one way to sort some reliabilist problems out is, I suggest, by leaving behind an individualistic epistemology: more specifically, by adopting a social conception of knowledge-relevant responsibility that allows ordinary epistemically unreflective subjects who exploit socially endorsed belief-forming procedures to know.
The paper proceeds as follows. First, I present the case for the insufficiency of pure reliabilism about knowledge. A classic problem case concerning reliable but naïve subjects, made famous by Laurence BonJour, threatens the claim that reliability is sufficient for the knowledge-relevant normative status. I consider and rule out some pure reliabilist responses, some of which are thwarted by a novel case concerning knowing naïve subjects. Second, given a responsibilist dimension is required for knowledge, I show that the responsibilist condition that BonJour's remarks suggest, which demands the knower to reasonably take the belief-forming procedures to be reliable, won't do. In fact, the same knowing naïve subjects that trouble the pure reliabilist show that this condition is too strong. Third, I propose a particular combination of reliability and responsibility that copes with these problem cases as well as some variations. This hybrid succeeds because it exploits an anti-individualist modification to BonJour's responsibilist requirement that capitalizes on our comprehensive division of epistemic labour. Fourth, I pre-empt some potential misunderstandings and objections concerning this social conception of knowledge-relevant responsibility. Finally, as a means of conclusion, I draw a general moral.

PuRE RELIABILISm AND NORmAN'S CASE
Reliability of the local or global form (or a combination of the two) seems required for knowledge. 2 Here I take reliability to be necessary for knowledge and for simplicity's sake I shall just speak in terms of global reliability. This condition seems anyway desirable because views that don't take it into account seem to fail to be appropriately normative. If Sid responsibly decides to use Tea-Leaf-Reading practices to find out the truth, those unreliable 2 Local reliability is here understood as the reliability of a procedure with respect to a particular belief, so the procedure is locally reliable in that instance if the belief in question is true throughout a range of possible worlds. Global reliability is the more general reliability or truth-conduciveness of a belief-forming procedure. In this case, a procedure is globally reliable, roughly, if it tends to produce true beliefs (for present purposes this characterization will do). I shall employ the term 'procedure' throughout but I don't mean to differentiate between innate processes and acquired methods.
procedures do not seem to provide the knowledge-relevant normative status (GOLDMAN, 1986;GRECO, 2010).
Pure reliabilism, however, maintains that, very roughly, all that matters for the knowledge-relevant status of a true belief (absent Gettier problems and defeaters) is that the belief is formed by a procedure that tends to deliver true beliefs (GOLDMAN, 1979). 3 But this theory of knowledge doesn't seem plausible. A classic type of thought-experiment suggests that this is wrong: reliability isn't enough for such normative status.
This type of case is concerned with subjects with unusual but reliable cognitive faculties who lack evidence for or against the belief-forming procedure and the corresponding belief. Let me introduce a particular case (this is just one of several examples like it available in the literature-we consider others later on but for now we focus on this one). This is how BonJour (1985, p.41) describes Norman's case: Norman, under certain conditions which usually obtain, is a completely reliable clairvoyant with respect to certain kinds of subject matter. He possesses no evidence or reasons of any kind for or against the general possibility of such a cognitive power or for or against the thesis that he possesses it. One day Norman comes to believe that the President is in NYC, though he has no evidence either for or against this belief. In fact the belief is true and results from his clairvoyant power under circumstances in which it is completely reliable. 4 The crucial feature of the case is that, ex hypothesi, Norman doesn't possess either positive grounds or undefeated defeaters (of neither the rebutting nor the undercutting sort). That is, there is meant to be no "cognitive conflict" among the subject's beliefs, but a complete lack of grounds or reasons (FOGELIN, 1994, p.44). Anyway, even if Norman has a true belief that is reliably formed and so, according to pure reliabilism, knowledge, it seems that he doesn't know. And BonJour claims that the reason this is so is that it seems 3 I refrain from using the term 'epistemic justification' and instead speak more neutrally of the positive normative epistemic status required to turn true belief (Gettier problems and defeaters aside -I hereafter ignore the qualification) into knowledge, or more simply knowledge-relevant status. This helps us avoid confusions that can give us the impression that something is wrong with an account of knowledge because it fails to capture some phenomena associated with some sort of justification not required for knowledge. 4 BONJOUR (2003, p. 28) clarifies that the procedure, which delivers beliefs that occur spontaneously and forcefully to the subject, has so far eluded scientific investigators. epistemically irresponsible to hold a belief in such a way (whether or not he believes himself to have a clairvoyant power; 1985, p.42).
So the above case suggests that reliability isn't sufficient for the knowledge-relevant status. Moreover it seems this case, given BonJour's understanding of it, is easily dealt with by adding some sort of responsibilist condition on knowledge. So this type of case is also thought to provide a main intuitive consideration to motivate some such requirement.

PuRE RELIABILIST FIxES AND NORm'S CASE
However some might complain there is no need to supplement pure reliabilism. They might say that our intuitions seem to vary in this sort of cases (I am not sure when they don't! 5 ). In fact, there seem to be other cases, such as the infamous naïve chicken-sexer, who is extremely reliable in differentiating male from female chicks but is oblivious to this, where intuitions seem clearly divided (PRITCHARD, 2005, p.183). So while some deny knowledge to the chicken-sexer, others attribute it, and still others don't have any firm or strong intuitions with regard to this sort of case. 6 But the point remains that our intuitions seem strong enough in some cases, such as Norman's, to suggest pure reliabilism isn't sufficient. 7 Anyway, if we were to introduce some further condition to capture Norman's case, we would also want it to help us explain why our intuitions are divided (and even ambivalent) in some cases like the chicken-sexer.
Of course the pure reliabilist can still attempt to explain away the denial of knowledge in Norman's case. After all, since we believe that clairvoyance doesn't exist (and so that it cannot be reliable), she can suggest that these intuitions are contaminated by our worldview. That is, even if we stipulate that these background beliefs are wrong, the pure reliabilist can suggest these stipulations don't insulate our intuitions from the beliefs that they are meant to neutralize (FOGELIN, 1994, p.45;ZALABARDO, 2006, p.143). The claim then is that our denials of knowledge in these cases are really the result of our inability to suspend certain assumptions.
Moreover the pure reliabilist can further suggest the negative and positive verdicts in the chicken-sexer case are the result of a bias against the existence (and reliability) of such a power and lack of it, respectively (although this alone won't explain the ambivalent verdicts). Leaving aside that this is a genuine ability, which we can acquire by methodical training and achieve near-perfect reliability (MARTIN, 1994), 8 it might be natural for some to think that some people can do it. Some might think that chicken-sexers can also differentiate sexes from another species and, like those who possess perfect pitch, they have an ability that most of us don't. These people then will be tempted to attribute knowledge to the chicken-sexer. But of course, the pure reliabilist can claim, those who think that it doesn't exist clearly won't.
But this sort of move would seem to suggest that our intuitions about a modified Norman case shouldn't be what they seem to be. Consider this case: Norm, under certain conditions which usually obtain, is a completely reliable clairvoyant. Moreover, she grew up in a community that relies on clairvoyance. Some members of the community have evidence for the existence and reliability of this power in the community and so for endorsing the community's practice. However she hasn't given these matters any thought and possesses no evidence for or against the reliability of such a power. One day Norm believes that the President is in NYC, though she has no evidence for or against this belief. In fact the belief is true and results from her clairvoyant power under circumstances in which it is completely reliable. Now, intuitively it seems Norm knows, and in fact pure reliabilism can easily explain why this is so (since she has a reliably-formed true belief ). 9 However, if our worldview contaminates Norman-type cases, there seems to 8 In fact, there are (very well-paid) enlightened chicken-sexers who are extremely reliable (the best ones being above 95% accurate). 9 Of course I don't expect everyone to share this intuition. If experimental philosophy has taught us something is that this might just be so (apparently there is even divergence about Gettier cases -e.g. BUCKWALTER; STICH, 2011). So I am under no illusion that everyone will share this intuition. But as Boyd and Nagel (2014) argue neither (systematic) variation within the population at large nor disagreement among philosophers show that intuitions aren't valuable evidence about the nature of knowledge. Moreover I take it we could explain away at least some negative verdicts in Norm's case, say, due to performance errors, such as losing track of some significant feature of the case. But regardless of that, the problem is to explain why the positive intuitions that some of us have are immune to the above alleged contamination. Just saying that in Norm's case the new information counters the effects of our contaminating beliefs won't do, since it isn't clear why this would be so. After all, in both cases we claim clairvoyance exists and is reliable, the only difference be no reason to think it wouldn't do the same in this case. That is, if our worldview contaminates Norman-type cases and since both cases seem to be the same with respect to this contamination, the contamination hypothesis seems to predict the denial of knowledge in Norm's case. So the pure reliabilist explanation as to why we deny knowledge in Norman-type cases seems to wrongly suggest that in Norm's case we shouldn't attribute knowledge. Given this intuition-contamination hypothesis wrongly predicts the denial of knowledge in Norm's case, I suggest our denial in Norman's case isn't due to this contamination. So the pure reliabilist can't diagnose BonJour's Norman case by saying that we are projecting our own worldview when we judge that Norman doesn't know that the President is in NYC. Moreover Norm's case also generates problems for other reliabilist responses.
For example, another main sort of response has been to focus on the no-defeater condition. The pure reliabilist might question the plausibility of the stipulation that there are no defeaters. Only simple pure reliabilism ignores the possibility of epistemic defeat (cf. GOLDMAN, 1979), so if there are defeaters available, she can explain the denial of knowledge in Norman's case. But even if it is hard for some to envisage BonJour's description of Norman as lacking defeaters (GOLDMAN, 1986, p.112), it isn't clear there is a defeater available to Norman given his evidence (even when not believing himself to be clairvoyant). 10 After all, Norman might not have, say, any beliefs about the epistemic worth of out-of-the-blue deliverances and might even lack the data to come up with them. He certainly seems to be extremely unreflective and would be psychologically very different from us. Granted, the case is underdescribed in this respect, but it's meant to be designed so that Norman doesn't have defeaters, so we should fill in the details accordingly. But leaving that aside (as well as whether a no-defeater condition can be couched in reliability-friendly terms- BONJOUR, 1985;GRUNDMANN, 2009), can the presence of defeaters be the difference between Norman's and Norm's verdicts? The answer seems negative. Again, it seems that Norman and Norm are the same with respect to defeaters and, as long as they throw contrary verdicts, the existence or lack of defeaters cannot explain the difference.
In fact the only difference between these cases seems to be the existence of an established practice that is endorsed by the community. So it seems a nobeing the community's endorsement of a rooted practice. They would need to explain why this stipulation counters the contamination. 10 GOLDMAN (1979, p.351) rightly restricts the evidence to be exploited to previously acquired evidence. defeater condition, or for that matter any response that doesn't latch on to that difference, won't help pure reliabilism. 11 Such a negative condition wouldn't be strong enough to deal with this sort of case. And, given that reliability seems to need supplementation, the difficulty for those who attempt to offer it is to introduce a positive condition that isn't too strong either. That is, they need to avoid capturing the denial of knowledge in Norman-type cases by introducing a necessary condition that rules out knowledge in cases of unsophisticated epistemic beings, such as Norm.

BONJOuR'S mISDIAGNOSIS AND REFLECTIVE ENDORSEmENT
A responsibilist condition is normally put forth. 12 This condition aims to capture the perspectival epistemic dimension that is naturally thought to be missing in Norman's case. Indeed, BonJour thinks Norman's case not only helps us draw the negative conclusion that reliably-formed true belief isn't sufficient for knowledge, but also see that some sort of responsibilist condition is needed. In particular, he thinks Norman-type cases show that, for a belief to be appropriately normative, the knower needs to at least reflect critically on the sources of her beliefs. As he says, "[p]art of one's epistemic duty is to reflect critically upon one's beliefs, and such critical reflection precludes believing things to which one has, to one's knowledge, no reliable means of epistemic access" (1985, p.42). In particular, the knower should at least reasonably judge the belief-forming procedures to be reliable from within her system of beliefs (1985, p.50, 123).
BonJour's necessary responsibilist requirement for knowledge is meant to stop the belief being true by accident from the subject's perspective (1985, p.43; 2003, p.27). In other words, this perspectival appropriateness stops it from being, in a sense, lucky. And according to BonJour, this luck precludes knowledge. Norman, he would say, doesn't know because he doesn't pursue the truth responsibly: he believes "blindly." But this responsibilist requirement seems too demanding. It suggests that unsophisticated epistemic beings, such as Norm, don't know. Again, the main difference between Norm and Norman is that Norm engages in an established practice endorsed by her epistemic community. Importantly, Norm isn't meant to have positive reasons for the procedure's reliability. Norm, like Norman, "[…] has never checked empirically to see whether any of [the beliefs in question] are true, nor has the potentially available empirical evidence for the truth of any of the specific claims and in consequence for her general reliability been supplied to her by others" (2003, p.28). So BonJour's responsibilist condition cannot help us capture Norm-type cases.
In fact most ordinary subjects would fail this condition (even if allowed to satisfy it tacitly; 1985, p.50), so requiring the knower to have reflective access to her epistemic situation seems to over-intellectualize knowledge. 13 This responsibilist condition to reflect critically on the sources of our belief that is exclusively framed at the individual level is overly strong. So BonJour's diagnosis of Norman-type cases seems mistaken, since the sort of individualist responsibilism he adopts is intuitively too demanding. The issue doesn't seem to be about the subject's belief being epistemically appropriate from her point of view.
Anyway, given reliability is necessary but not sufficient for the knowledge-relevant status and that Norman-type cases legitimately motivate a responsibilist condition, we still seem to need some sort of admixture of reliability and responsibility. After all, as suggested, it is natural and common to advocate that a true belief needs to be appropriate along both objective and perspectival dimensions in order to count as knowledge. 14 The problem, however, is to find a satisfactory combination of reliability and responsibility.
As seen, we cannot, as BonJour does, understand epistemic responsibility as demanding the knower herself to reasonably take the procedure exploited to be reliable. 15 It seems clear that knowledge needn't be a reflective success of the subject: knowledge seems to require responsiveness to the world but not necessarily to reasons of the subject. 16 Having said that, we needn't give up BonJour's idea that we require some reflective endorsement of the knowledgeyielding procedures. And, significantly, once suitably modified, this approach to responsibility, as we shall see next, can be exploited by the reliabilist to overcome the problem cases. So let me first suggest the modification and then show how it can cope with those cases and some variations.

SOCIAL KNOWLEDGE-RELEVANT RESPONSIBILITY
As seen, the personal reflective endorsement of procedures (that is, one reasonably taking them to be truth-conducive) is implausibly demanding. But the social analogue of the personal reflective endorsement isn't. This social condition doesn't require each knower to have positive reasons for the truthconduciveness of the knowledge-yielding procedures. We can all exploit, as we often do, different procedures rooted in the epistemic community, for which 14 Indeed, this is the "standard way of looking at things" (GRIMM, 2011, p.90). 15 There are other ways in which one might attempt to capture this knowledge-relevant responsibility.
One is by requiring some kind and measure of voluntary control, as opposed to reflective control, over our beliefs (e.g. FELDMAN, 2001;GINET, 1985). But this option does not seem promising either (ALSTON, 1988;cf. WILLIAMS, 1973). Of course there are further notions of epistemic responsibility available (e.g. HIERONYMI, 2008;WILLIAMS, 2008). But considering these options would take us too far afield and here I have the more limited aim to show that the reflective endorsement conception of epistemic responsibility can be modified to capture the considered data. 16 Indeed, the fact that "knowledge attributions can be underwritten by a believer's reliability, even when the believer is not in a position to offer reasons for the belief " can be seen as "the Founding Insight of reliabilism" (BRANDOM, 2000, p.99). So we qua reliabilists cannot accept the above understanding of epistemic responsibility. But, anyway, this is an insight because, regardless of one's sympathies, no such reflective responsiveness to the world seems required in the case of knowledge. See e.g. AYER, 1972;GOLDMAN, 1979;LEWIS, 1996; we lack positive reasons, as long as the reflective endorsement is met at the social level, as Norm-type cases suggest.
So I propose we adopt an anti-individualist approach to the notion of epistemic responsibility as reflective endorsement of the knowledge-yielding procedures and capitalize on the real and ubiquitous human phenomenon that is the social dispersal of epistemic labour through time. This new approach then corrects the unfortunate individualist simplifications of much current mainstream epistemology, which are unsuitable for theorizing about knowers who are members of social communities and so seem to be stopping us from making progress (DE BRASI forthcoming;FRICKER, 2010;GOLDBERG, 2010), by emphasizing the cooperative and interactive aspects of knowing.
Indeed, our epistemic reliance on others needn't be limited to instances in which one exploits an inter-personal knowledge-yielding procedure, such as some kind of testimony. 17 It can be, and I suggest is, much more pervasive since, although the knower needn't possess the positive grounds for the endorsement of the procedures that she and other members of the community rely on, someone does. These procedures are, I shall say, socially endorsed: in the sense that some subject or, more likely, a group of subjects of the knower's epistemic community have undertaken through time the positive epistemic work for the endorsement of the community's established procedures (we can call these subjects 'epistemic experts' 18 ). The clearest example of subjects who appreciate the rationale behind our belief-forming procedures and who would also promote their revision through time if regarded necessary, is that of regulative or ameliorative epistemologists, whose job is to actively engage in the project to remedy the deficiencies of our epistemic practices (to increase their truth-conduciveness). In particular, the social endorsement, I suggest, is sometimes the product of some sort of epistemic policing that prompts the correction or perfection of inadequate procedures (e.g. BRUNER, 2013;GOLDMAN, 2011a). And this social endorsement means that one, as a member of the community, is permitted to exploit any of its rooted knowledge-17 Testimony has rightly received copious attention recently -e.g. LACKEY; SOSA, 2006 -but I suggest it is not the most epistemically interesting social phenomenon. 18 These are subjects who appreciate the rationale behind our knowledge-yielding procedures. But notice no single epistemic expert needs to be able to reflectively endorse all procedures. So, in this sense, many of us are likely to qualify as epistemic experts (given we can and are likely to personally endorse some procedures), not merely regulative epistemologists and the like (see below). Perhaps it's a bit too grandiose to call some of these subjects 'experts' given their potentially limited subjectmatter and the relatively easy way in which they can reach the endorsement-more on this below. yielding procedures even in the absence of personal positive reasons for their endorsement.
This anti-individualist approach yields the following reconciliation of de facto reliability and epistemic responsibility as reflective endorsement of the knowledge-yielding procedures: (as a first approximation and ignoring Gettier-cases, defeaters and factivity) S knows that p iff (i) S's belief that p was formed by a reliable procedure, AND EITHER (ii.a) the procedure is reflectively endorsed by some member(s) of S's community and (ii.b) reliance on the procedure is a standard or established practice of the community, OR (iii) S herself reflectively endorses the procedure.
To be clear, the social endorsement makes two demands: (ii.a) and (ii.b). 19 Below we look at cases, such as Mr. Truetemp's, where just one demand is met and consider our reactions to them. But what matters now is that these two demands are to be met if social endorsement is to make it permissible for a subject to know when exploiting a procedure she doesn't reflectively endorse. 20 Moreover, although the responsibilist condition can be met either at the personal or social level, in our case this responsibility is likely to be borne by the epistemic community. This then is another way in which the knower normally depends on other epistemic subjects.
But of course some might not find it natural to think of this knowledgerelevant responsibility as a requirement that can be satisfied at the social level. This, I suggest, might just be a remnant of the strongly individualist orientation of epistemology (KITCHER, 1994), which the Cartesian ideal of epistemic autonomy (metaphorically put, that the epistemic agent ought to stand on her own epistemic feet) seems to have helped establish. According to this picture, knowledge, for example, is taken to be an essentially private and personal achievement and so its responsibilist condition needs to accommodate this. But, since relying on others seems to be cognitively fundamental for beings like us (at least testimony should show this much- BURGE, 1993;CODE, 1987), any investigation into human knowledge should be at odds with this Cartesian ideal and its accompanying individualist framework. As 19 The notion of a standard or established practice is quite vague, and more would need to be said about it. Having said that, for present purposes, our intuitive understanding of what counts as some such practice will do. Furthermore, notice this vagueness might not be objectionable if it helps us explain divergent and ambivalent verdicts (see the chicken-sexer case below). 20 Notice furthermore that given this reasonably taking the procedure to be truth-conducive needn't entail knowledge, its social endorsement, just like the personal one, can satisfy the knowledge-relevant responsibilist requirement without worries about infinite regresses (cf. KORNBLITH, 2012).
Jonathan Kvanvig says, "[…] we should never begin to think that the deepest epistemological questions concern the isolated intellect" (1992, p.177). 21 Accordingly, this proposal suggests that some knowledge-relevant status of one's belief can be satisfied at the social level. The proposal depends on a division of epistemic labour to be in place, where some members of the community endorse the procedures for others. Now there are different sorts and models of divisions of epistemic labour (GOLDBERG, 2011;KITCHER, 1993;MULDOON, 2013). Here I want to distinguish between those cases in which one depends epistemically on another subject or on a group of subjects (that extends through time within the community) and in which one is or is not aware of such dependence. Of course the most familiar sort of division of epistemic labour, the testimonial case, normally involves dependence on a subject and awareness of it. But, in our case, there is no need for a single subject to do all the epistemic work for the endorsement of any one procedure and for the knower to be aware that she is relying on someone (more on this below). In fact, in many cases, it is likely that a number of members of the epistemic community collaborate through time in this endeavour 22 and the subject isn't aware that she is relying on them, just as in Norm's case.
Moreover, it is also likely that in many cases not only are there social checks on procedures exploited, but also learning environments engineered so to scaffold the acquisition of new (better) procedures. In this way we can transcend the "quick and dirty" modes of thought left to us by evolution. And examples of this "perfecting" of procedures can most easily be found within formal institutionalized practices, such scientific and historiographical ones, where methodological changes are common and explicit. Anyway, the knower could benefit from a network of individuals of her community, such as regulative epistemologists, which are distributed in space and time (including those beyond the living!), to satisfy the knowledge-relevant responsibility that cases such as Norman and Norm seem to require. 21 It's highly desirable not to do so, since "[…] one is reminded here of the attempt to do ethics by beginning with 'desert island' cases; even if such cases are possible, it is absurd to think that we can come to be enlightened about the nature of the moral life we share by focusing on such cases. Just so in the epistemological case: divorcing epistemological concern from the realities of social interaction generates an epistemology built on answers to questions as relevant to the life of the mind as 'desert island' cases alone in ethics" (1992, p.178). See also SOSA, 1991, p.190, WELBOURNE, 1986 First, it is likely that, in some cases, no one human could probably do all the work that must be done to fulfil the endorsement (consider the development of scientific or historiagraphic procedures). Second, it is very likely that no one human actually performs all the cognitive tasks even if they could.

NORmAN, NORm AND VARIATIONS
One way of appreciating whether the above modification is the right kind of move consists in considering problem cases, such as those of Norman and Norm. This approach seems able to explain those cases and other ones designed to test it. So let us consider them.
Starting with Norm's case, this kind of responsibilist requirement allows us to make sense of her as a knower (even if she doesn't possess any positive grounds for endorsing the procedure). After all, she employs a reliable procedure that is reflectively endorsed by her community and relied upon as a standard practice. This social endorsement makes it permissible for Norm to know. So Norm does know, because she satisfies (i) and (ii.a-b). This, however, isn't to say there isn't something epistemically deficient with Norm. Indeed, her belief is, in a sense, lucky; but this deficiency, contra BonJour, doesn't preclude knowledge when social endorsement is present.
On the other hand, Norman doesn't know, since neither (ii.a-b) nor (iii) are satisfied. Or so we assume, given the case is silent about social endorsement (but see fn.4). That is, given this lack of details, it is natural for us to fill them in in accordance to our worldview (i.e., we take Norman to belong to an epistemic community like ours); in which case there is no social endorsement of clairvoyance in Norman's community. And given that in this case the responsibilist condition won't be met at the personal level, we can see how the verdict in Norman's case is likely to be negative.
Social endorsement of the knowledge-yielding procedure is taken to be absent in Norman's case but not in Norm's, hence the different verdicts although both subjects lack positive grounds for their powers. Moreover, we can make sense of the divergent intuitions in the chicken-sexer case. Depending on how one fills in the relevant details (i.e., whether one thinks that there is social endorsement or not 23 ), one's verdict changes. For example, some might regard the naïve chicken-sexer as exploiting a socially endorsed procedure and so attribute knowledge to her. That is, they might tacitly assume a case more like Norm than Norman. We can then explain the variability of intuitions (including the ambivalent verdicts, if one isn't sure how to fill out the story).
This modified condition can also deal with slightly different thoughtexperiments of naïve epistemic subjects with "strange" powers where some subject other than the knower endorses the procedure but there is no established doxastic practice in the community, such as Keith Lehrer's classic Mr. Truetemp example (1990, p.163-164): Mr. Truetemp undergoes brain surgery by an experimental surgeon who invents a small device which is a very accurate thermometer and a computational device capable of generating thoughts. Unbeknownst to Truetemp, this tempucomp is implanted in Truetemp's head. He is only slightly puzzled about why he thinks so obsessively about the temperature, but never checks a thermometer to determine whether these thoughts about the temperature are correct. He accepts them unreflectively, another effect of the tempucomp. Thus, he thinks and accepts that the temperature is 104 degrees. It is.
It seems clear Mr. Truetemp doesn't know. Our account can easily accommodate the verdict since there is neither personal nor social endorsement of the procedure. Even if the surgeon can reasonably take the procedure to be truth-conducive (ii.a), there doesn't seem to be a relevant rooted doxastic practice in the community (ii.b). Now consider this community wide Truetemp case from Weinberg et al. (2001): The Faluki are a large but tight knit community living on a remote island. One day, a radioactive meteor strikes the island and has one significant effect on the Faluki-it changes the chemical make-up of their brains so that they are always absolutely right whenever they estimate the temperature. The Faluki are completely unaware that their brains have been altered in this way. Kal is a member of the Faluki community. A few weeks after the meteor strike, the changes in his brain lead him to believe that it is 40 degrees. Apart from his estimation, he has no other reasons to think that it is 40 degrees. And it is 40 degress.
Again, it seems clear that Kal doesn't know. And we can explain this too. In this case the whole community has the ability. And we can imagine they all engage in this practice (ii.b). But there is clearly no reflective endorsement of the method by any member of the community (ii.a). And so no Faluki knows in this case. So the fact that the whole community engages in a reliable practice is clearly not good enough either.
Finally, consider cases of "strange" doxastic practices which are reliable but don't rely on "strange" powers. These are cases where the doxastic response to a certain input isn't the one we would normally have (so, again, these examples are slightly different from the above ones). Here are a couple of (modified) cases by Alvin Plantinga (1993, p.42, 62-63): Paul is such that when he is appeared to in one sense modality, he forms beliefs appropriate to another; this is due to a brain lesion. When Paul is aurally appeared to in the church-bell fashion, he finds himself with a powerful impulse to believe there is something that is appearing to him in that fashion, and that that thing is orange. He doesn't know about this quirk in his epistemic equipment, and his lack of awareness is in no way due to dereliction of epistemic duty. Add that as a matter of fact nearly everything that makes this church-bell sound in question is orange. When Paul catches a glimpse of a bird (without seeing its colour) and hears it make that sound, he forms the belief that there is something appearing to him in the churchbell fashion and that it is orange, which is in fact the case.
Although there is some subjective (sensory) experience on which the belief is based (and which might be seen as evidence 24 ), I take it we want to deny knowledge to Paul. And we can easily make sense of this denial since there seem to be a lack of both personal (iii) and social endorsement (ii.a-b) of the procedure. But now suppose: God (or evolution) had designed human beings in a different fashion. These creatures are by nature such that when appeared to in the churchbell fashion, they form the belief that they are appeared to that way by something that is orange. Imagine further that although these beings are often appeared to in that orange fashion, they inhabit a planet on which they seldom (if ever) visually perceive that an object is orange; atmospheric conditions make that for the most part impossible. Add that as a matter of fact nearly everything on this planet that makes the church-bell sound in question is orange. Now imagine there is a certain common but rarely visible orange bird that makes the church-bell sound. When Pauline catches a glimpse of this bird (without seeing its colour) and hears it make that sound, she forms the belief that there is something appearing to her in the church-bell fashion and that it is orange. Now, although Plantinga seems happy to grant knowledge in this case (as his account predicts), I think we can expect most of us to deny knowledge to Pauline. The fact that Divine design is involved doesn't seem to make a difference, since anyway Pauline's and everyone else's conception of their epistemic position is the same as in the previous scenario. 25 And we again can easily make sense of the denial since, if it is nearly impossible for these subjects to confirm this connection between sound and colour (ii.a), social endorsement of the practice, let alone personal (iii), seems extremely unlikely (though not completely impossible, so ambivalent verdicts might not be rare). 26 Again, the whole community engaging in the practice isn't enough for social endorsement since there doesn't seem to be any reflective endorsement.
So these cases suggest that the factors picked out by our condition are crucial to knowledge (as opposed to other factors, such as sensory experience and Divine design 27 ). Indeed, the fact that the modified responsibilist condition can accommodate our intuitions, as well as explaining the ambivalent and divergent intuitions, in the above problem cases, strongly suggests this strategy's plausibility.
Admittedly, I offer no independent argument for it. So one might worry this is just an ad hoc stopgap. But here I am merely interested in suggesting a plausible alternative combination of de facto reliability and knowledge-relevant responsibility (as reflective endorsement) given that one seems required. And, on this combination, the endorsement of the epistemic community's procedures by some of its members can satisfy the responsibilist requirement on knowledge. That is, this anti-individualist modification allows us to capture the sort of epistemic responsibility BonJour has in mind. And it would be irresponsible for someone to exploit some procedure that is neither personally nor socially endorsed, in the sense that some positive epistemic status crucial for knowledge would be missing. 25 Again what seems to make the difference is some sort of reflective endorsement. Consider the case of Paulie, who is just like Paul except that Paulie has been able to collect data that reasonably suggests the reliability of the procedure. Although there is no Divine design (or proper functioning of the faculty), I take it we want to say Paulie knows, given the personal endorsement of this reliable procedure. 26 COMESAÑA (2010, p.590-1) is committed to the non-ambivalent verdict that Paul and Pauline know. And, although I agree that any account "[…] that entails that alien doxastic practices can never give rise to justification and knowledge is guilty of epistemic chauvinism" (2010, p.591), ours clearly isn't guilty of it, as Norm's case suggests (see also below). 27 The cases of Norman and Norm and of Pauline and Paulie (see fn.25) suggest this, respectively. We must take advantage of our pervasive division of labour, characteristic of human society, and shift the burden to those epistemic experts who reflectively endorse the rooted procedures that we exploit as members of the community if we are to satisfy this perspectival dimension on knowledge. Otherwise put, we need to transfer this epistemic responsibility to others in order to cope with its demand. So given this epistemic interdependence, not all of the non-objective epistemic conditions on knowledge need to concern the cognitive states of the knower herself. In particular, the knower needn't possess positive grounds for exploiting the knowledge-yielding procedures as long as they are socially endorsed. The reliabilist then should hold that the perspectival epistemic dimension that contributes towards the knowledgerelevant status of a true belief can depend on properties and actions of subjects other than the knower.
The reliabilist would end up with a hybrid view where (leaving aside Gettier problems and defeaters) the objective component for the knowledgerelevant status is satisfied by de facto reliability and the perspectival one is satisfied by either of two sub-components: a social or a personal responsibilist sub-component. To repeat, one component requires that the procedure P exploited (whether it is a humdrum procedure, such as perceptual and testimonial procedures, or a "strange" one) is reliable and the other component requires that P is either socially endorsed by the knower's community or personally endorsed by the knower herself (or both). 28 This is how the reliabilist can adequately deal with cases of knowing naïve subjects, such as Norm, and non-knowing ones, such as Norman. To 28 Let me briefly point out that this strategy has some interesting consequences within other debates in epistemology. Let me give you two examples. First, with regard to the internalism/externalism debate, we can see that Norman-type cases needn't motivate, as it is often thought (BONJOUR, 1985;FOGELIN, 1994;GRECO, 1990;PRITCHARD, 2005), epistemic internalism about knowledge: very roughly, the view that some positive dimension of the epistemic status required for knowledge is 'internal' in some sense to the knower (ALSTON, 1986;BONJOUR, 2010;VAHID, 2011). After all, given this individualist characterisation of internalism, our responsibilist condition fails to qualify as internalist (since positive reasons are not required of the subject for her to know). So it seems that one can have a responsibilist account without it being internalist (cf. PRITCHARD, 2012). Second, it even has repercussions within debates in social epistemology, since the positive reasons component of reductionism about testimony is meant to be motivated by cases in which we are unwilling to attribute knowledge to a hearer when she accepts testimony while lacking, like Norman, positive reasons about the source (and the belief ). In order for these conditions to hold, we are meant to imagine that the testifier is an alien about whom the hearer doesn't know anything (LACKEY, 2006, p.167). But, given the above, the reason we might deny the hearer knowledge isn't merely because she lacks positive reasons but because taking testimony from aliens isn't socially endorsed either.
repeat, without this anti-individualist modification that makes the burden of epistemic responsibility significantly less hard to endure, we couldn't normally meet this type of responsibilist requirement and make sense of our denial and attributions of knowledge in cases like Norman and Norm, respectively. So individualism about epistemic responsibility can be profitably rejected.
But I'm not here suggesting this is the only way the reliabilist can deal with these cases (although I think it probably is the best available hybrid, since Norm's case seems to generate problems to individualist ones 29 ). And of course I have provided nothing like a detailed account of either social or personal endorsement (let alone consider other problems the views combined must individually face). But, as mentioned, here I just want to explore the plausibility of this anti-individualist strategy to deal with the above problem cases about naïve epistemic subjects the pure reliabilist must face.

PRE-EmPTING OBJECTIONS
As a way of clarifying the social knowledge-relevant responsibilism suggested and indicating some further work needed to flesh out the proposal, let us consider four worries that some might raise given the above.
The first one is that this approach still seems to be too demanding for epistemic loners, such as Congenital Crusoe: an individual who is left on his own in an island from infancy. 30 The approach has no problem accommodating Normal Crusoe (since this is an individual like Norm, who is part of a community, although currently isolated from it). However, in the case of Congenital Crusoe, it seems that the responsibilist condition cannot be satisfied at the social level (ii.a-b) and is unlikely to be satisfied at the personal level (iii) for all the procedures that we would think provide him with knowledge.
The first thing to note is that Congenital Crusoe can have knowledge (in fact, the idea of him having knowledge isn't senseless -cf. WELBOURNE, 29 The two main competitors, which require that the reliable belief-forming procedure includes some evidence (where the subject used that evidence to form the belief -COMESAÑA, 2010) or the belief fits with the subject's current evidence (GOLDMAN, 2011b), seem to face the problem of accounting for Norm's positive verdict. If this sort of synthesis is meant to capture Norman's case, given Norman's apparent lack of evidence and Norm's analogous situation, this move denies knowledge to Norm (see also fn.26). 30 Make all the relevant assumptions to make the case as plausible as possible. Perhaps we need to think of Congenital Crusoe as a sort of Wolf-Boy.
1983, p.84) but, according to the above worry, his knowledge is likely to be counter-intuitively restricted. Now, assuming this is correct, it is clear that if, as suggested, we aren't to prioritize cases of isolated intellects when investigating the phenomenon of human knowledge, then it is clearly an option to explain away these attributions of knowledge. 31 And this can be easily done by invoking some sort of projection or loose-talk explanations. For example, when Congenital Crusoe sees a boat (in certain conditions), we assume he knows that there is a boat (conceptual issues aside). In this particular case, the relevant projection would seem very natural: after all, it is very natural to assume he does endorse the procedure (especially if imagined as an adult) or not to realize he doesn´t belong to one´s community. But, although different construals of the cases, as seen before, can be epistemologically significant, I take it that we can explain in the different cases why the folk would make some attributions of knowledge to Congenital Crusoe even if wrong.
A second worry is this: in cases where condition (iii) does not come into play, (ii.a) makes it clear that it is not enough for the procedure to be an established practice of the knower's community; it is further required that the epistemic experts have undertaken the positive epistemic work for the endorsement. But this work might be too demanding. For example, the positive epistemic work for justifying modus ponens is very hard and has attracted considerable philosophical attention. Certainly there have been many communities in which nobody has successfully completed that epistemic work (and perhaps ours is one such community). But still seems that subjects in those communities can come to know things by modus ponens.
This worry rightly raises the issue as to what exactly is the positive epistemic work required for the endorsement of procedures: the reasonably taking the procedure to be reliable. After all, the social endorsement, like the personal one, cannot be too demanding and, as things stand, one might think this could be so. Although I am here only attempting to motivate the further work that will fill in those details, let me give you a sense of what I roughly have in mind. What this epistemic work requires of us qua non-sceptics is some sort of appreciation of the procedure's truth-conducive, such as an induction whose conclusion is that the procedure is reliable. Since the data for the track record of some procedure's success can easily be available (see fn.13), this sort of epistemic work shouldn't make the social (nor the personal) endorsement hard to satisfy. 31 See in particular fn.21. Also the bizarre nature of cases such as this might facilitate this move.
A third worry, again ignoring (iii), might be that if the knower exploits a procedure because it is socially endorsed, then it looks like all cases of knowledge will in fact involve the personal endorsement of a procedure: namely, the procedure of trusting the epistemic experts. Nevertheless this is clearly not what the above approach suggests. But, if it is sufficient just that the knower exploits a procedure that, as it happens, is socially endorsed, then the approach seems to make knowledge implausibly extrinsic, one might think. Imagine, for example, that Norman goes camping and that, while he is out there, the epistemic experts complete the epistemic work and so clairvoyance becomes endorsed by some members of Norman's community. All of the sudden, one might think, Norman's clairvoyant beliefs are knowledge, although he is completely unaware of all of this throughout his trip.
The above case is indeed odd but fortunately the approach suggested doesn't commit us to the claim that Norman, after such reflective endorsement (ii.a), knows since it is not at all clear that Norman would be exploiting an established practice of the community (ii.b), just like Mr. Truetemp doesn't. Of course, more needs to be said with respect as to what counts as an established practice of the community (see fn.19). For example, it might be that a practice counts as established only when there are a series of mechanisms in place to check the use of the procedure in the community. The point anyhow remains that there is room to deny that in the above case Norman knows.
The fourth and final worry that we shall consider is that even if we admit that social endorsement is an epistemically relevant factor, there is still a question about whether the "location" of that factor is in the subject's context. In other words, whether we should be relativizing the social endorsement to the knower's community. The above suggests that we should, but consider the following case: S 1 and S 2 are members of different epistemic communities but both exploit the same evidence and procedure to arrive at the belief that p. It might then seem a bit odd to suggest that S 1 knows that p but S 2 doesn't since "S 1 knows that p but S 2 doesn't while sharing procedure and evidence" can sound odd to our ears. But it can be true if the location of social endorsement is in the subject's context (after all, S 1 's community might endorse the procedure while S 2 's doesn't).
It might sound odd to hold that S 1 knows that p but S 2 doesn't when they share evidence and procedure, but that can be explained as a failure to note that something is missing in S 2 's case (given that we normally consider subjects belonging to the same epistemic community): the relevant endorsement. Still it can be true given that one but not the other might "inherit the right" to exploit the procedure. Moreover, if the location of the social endorsement is not in the subject's context, we end up with a form of epistemic chauvinism that seems undesirable, since we couldn't hold that cognizers with alien abilities know (see fn.26). And of course I think we want to say that Norm knows, even though the self-proclaimed clairvoyants of our community don´t.

CONCLuDING REmARKS
We saw that the pure reliabilist must face certain problem cases about epistemically naïve subjects and that BonJour's natural responsibilist requirement for the knower to reflectively endorse the belief-forming procedure isn't adequate to deal with them. I suggested an anti-individualist modification to this responsibilist requirement that captures our intuitions about these cases. On this view, knowledge is compatible with lack of personal endorsement as long as there is no lack of social endorsement, as in Norm's case. Indeed, moving away from an individualistic framework allows us to make sense of knowing naïve subjects, such as Norm with regard to clairvoyance and some of us with regard to other capacities. It is the social element of chance that precludes knowledge in the naïve subject cases, as in the cases of Norman, Mr. Truetemp and Paul. So a responsibilism of a social strand can come to the reliabilist's rescue. In other words, on the assumption of a broadly reliabilist epistemology, it seems that that the right way to accommodate a responsibility requirement as reflective endorsement is by allowing such requirement to be met at the social level: that is, by letting go of the old paradigm that urged us to understand epistemic responsibility in terms of cognitive processes of the knower.
Moreover, if what I say is correct, we have shown the importance of social factors for knowledge. In other words, because of the division of epistemic labour with regard to the responsible endorsement of procedures that this proposal suggests, knowledge has a pervasive social dimension. So this social implementation of responsibility renders the proposal an instance of a social epistemology. Now, the social epistemology promoted by this proposal should not be merely understood as, to borrow a building metaphor from Martin Kusch (2011), adding a new storey to the epistemological house (but not laying any new foundations) or building a new wing to the house, as most contemporary social epistemologies do. Instead most of the epistemological edifice can be social epistemology.
This social knowledge-relevant responsibility allows us to do justice to those realities of social interaction that are overlooked in traditional epistemology. But while recognizing this ubiquitous social dimension on knowledge is epistemically important, it is equally vital to avoid the excesses associated with the nowadays-fashionable relativist and constructivist approaches that many 'end-of-epistemology' and postmodernist enthusiasts support (FRICKER, 1998;KORNBLITH, 1994). That would be an illadvised response to the deficits of mainstream epistemology. But, although these approaches have been thoroughly criticized (e.g. BOGHOSSIAN, 2006), no plausible substitute has been advanced. This proposal provides a sound alternative that preserves some core tenets of traditional epistemology, such as objective normative standards and truth. It affords us a more wideranging social epistemology, but one that, by the reliabilist's lights, still is "real epistemology" (GOLDMAN, 2010).
So we can start to make sense of oft-perceived radical claims by proponents of the "strong programme" and other "revisionist" social epistemologies, such as the claim that "[…] the social component is always present and always constitutive of knowledge" (BLOOR, 1991, p.166), without the oft-perceived excesses of these views. We can also avoid the deficits of Cartesian epistemologies, which hold on to individualist considerations and their accompanying asocial conception of epistemic subjects. This proposal then is a more radical conception of social epistemology, which is much more in the spirit of a truly social epistemology (TALBOTT, 2002), while remaining "real epistemology", hence providing us with an innovative and important viewpoint on epistemological discussions. And the moral to be drawn from the above, I take it, is that one can make progress in epistemology if one doesn't neglect, or lose sight of, our epistemic interdependence, though much work is left to do is one is to live up to this promise. 32 32 I would like to thank audiences of the 3 rd Annual KCL/UNC Conference: Social Epistemology and Epistemic Responsibility and the Workshop on Alvin Goldman´s Individual and Social Epistemology, and in particular to Matt Kotzen and Alvin Goldman, for their comments.