1 Introduction

In a series of papers (Hetherington 1998, 1999, 2001a, b), Stephen Hetherington has defended the tripartite analysis of knowledge, the thesis that knowledge that p is nothing more than justified, true belief that p. His defence has recently come under attack (Madison 2011; Turri 2012) on the grounds that Hetherington’s analysis fails to adequately diagnose why Gettier cases convinced so many people that there was something wrong with the tripartite analysis. In this paper, I critically evaluate those attacks as well as Hetherington’s newest formulation of his defence (Hetherington 2011a, b, 2012a, b). It seems to me that Hertherington’s newest formulation is vulnerable to a modified version of Madison’s and Turri’s objection, a new reason for thinking that Hetherington’s analysis continues to fail to adequately explain away all the relevant intuitions. However, it also seems to me that Hetherington’s considerations lend support to a different, though also radical, thesis which can meet the objection. This thesis is what I call the Divorce thesis: the theory of epistemic justification is importantly independent of the theory of knowledge. My aims in this paper are thus twofold: the first is to present an objection to Hetherington’s provocative case; the second is to marshal reasons in favour of the Divorce thesis by showing how Hetherington’s new argument supports it.Footnote 1 The two aims are related, since if it can be shown that his argument supports the Divorce thesis, then we have another objection to Hetherington’s case (given that knowledge cannot be justified, true belief if the Divorce thesis is true), viz. it does not prove what it purports to prove.

I proceed as follows. In the following Sect. 2, I present Hetherington’s earlier argument in support of the tripartite analysis as well as Madison’s and Turri’s objections to it, highlighting crucial features that will be important in Hetherington’s later work (such as his formulation of Fallibilism). Then, in Sect. 3, after formulating Hetherington’s new argument, I argue that it is vulnerable to a new version of Madison’s and Turri’s objection. In Sect. 4, I present the argument for thinking that the data Hetherington has brought our attention to supports the Divorce thesis, and that the latter is not vulnerable to the objection levied.

2 Defence of the Tripartite Analysis (First Attempt)

Rather obviously, given the state of epistemology since 1963, defending the tripartite analysis adequately will necessitate showing that Gettier cases are not, after all, counterexamples to the analysis as well as adequately diagnosing why the overwhelming majority of epistemologists were duped into thinking that they were. Hetherington’s caseFootnote 2 can be rendered so as to essentially consist of the following two claims:

Knowledge is Fallible: S can know that p even if it is the case that S could easily not have known that p.

The Epistemic Counterfactuals Fallacy: From the fact that you lack knowledge that p in a close possible world(s), it is a fallacy to infer that you actually lack knowledge that p in this world.

Hetherington explicitly does not claim to prove that S can know in a Gettier case; instead he aims at providing a principled interpretation of Gettier cases under which they are consistent with the tripartite analysis. His idea is that it is a matter of degree how easily S could not have known that p, in his words, how “failable” S’s knowing that p is. We can thus interpret Gettier cases as cases where S’s knowing that p is very failable. Yet, to reason that because S’s ‘knowledge’ that p is very failable, S does not, after all, know that p is to commit the epistemic counterfactuals fallacy. So Knowledge is Fallible and The Epistemic Counterfactuals Fallacy, combined, give us a principled way of interpreting Gettier cases such that they are not counterexamples to the tripartite analysis as well as an explanation as to why epistemologists were so comprehensively duped into thinking that they were. I think it is worth mentioning, however, that the epistemic counterfactuals fallacy on its own is not meant to do all the diagnostic work. We need to explain why epistemologists make that mistake. And the explanation is that S’s knowing that p is very failable. Epistemologists fail to apprehend that they are making a mistake because in Gettier cases S lacks knowledge that p in very close (or very many, or many, close) possible worlds.

One might wonder whether Hetherington’s diagnosis can really explain why so many (i.e. the overwhelming majority of) epistemologists committed the epistemic counterfactuals fallacy. Is it really plausible that they were so duped by the mere fact of knowledge’s being very failiable in Gettier cases? Perhaps not, but I won’t press this issue any further, since clearly philosophers have (comprehensively) made obvious mistakes in the past and Hetherington’s analysis does at least give us some diagnosis as to how this might have happened. But, nonetheless, there is good reason to be sceptical that Hetherington’s diagnostic work succeeds. As John Turri has pointed out (Turri 2012), if we are not prone to making the mistake in cases where, like in Gettier cases, a subject has very failable knowledge, then something else needs to explain why we make the mistake about Gettier cases and not about these other cases of very failable knowledge. But, according to Turri, we are not prone to making the epistemic counterfactuals fallacy about other cases of very failable knowledge. Consider, for example, someone who by some incredible luck survives after being shot at by a firing squad. Having survived, she knows that she is still alive, and we seem to have no trouble in affirming that she does so. And we have no trouble affirming this even though in most close possible worlds she does not know that she is still alive, since she does not have the belief, and the belief would not be justified nor true. Why don’t we make the invalid inference here, if we do so readily when we consider Gettier cases? Hetherington’s account does not give us an answer to this question.

One could say, however, that this reply poses no dialectical burden on Hetherington, since his opponent would need to explain why we have no problem granting that in cases such as the above the subject in question has knowledge, but does not in Gettier cases. But perhaps his opponents do have the means to explain this. They can claim that knowledge is incompatible with only a certain kind of luck, what might be termed veritic luck—luck that S’s belief is true, given S’s evidence. It is not incompatible with the luck operative in cases like the above, evidential luck—luck that S ends up having evidence that her belief is true (Madison 2011 Footnote 3). The person that survives the firing squad is evidentially, but not veritically, lucky (since she was lucky to have the evidence that she is still aliveFootnote 4), and so we have no problem saying that she knows that she is still alive. But the subjects in Gettier cases enjoy veritic luck—e.g. it is veritically lucky that my belief that this barn in front of me is a real barn is true while I am in fake-barn country—and so do not have knowledge.

I think it is possible to persuasively oppose the distinction between veritic and evidential luck. But a much simpler reply would be merely to point out that the claim that knowledge is incompatible with veritic luck is to deny Knowledge is Fallible, once the sense of “S could easily not have known that p” has been correlatively disambiguated. That is, once we interpret it in the sense that S could not easily have known that p, relative to S’s evidence as regards that p. Of course, Hetherington’s opponents might prefer instead to deny that Knowledge is Fallible properly captures their fallibilist intuitions. For instance, the fallibilism that is at work in Gettier’s own paper is the following:

[The] sense of ‘justified’ in which S’s being justified in believing that p is a necessary condition of S’s knowing that p, it is possible for a person to be justified in believing a proposition that is in fact false. (Gettier 1963, p. 121).

So we might re-formulate the thesis that knowledge is fallible as follows:

Knowledge is Fallible*:

It is possible for S to have a justified, but false, belief that p.

Note that Knowledge is Fallible* is prima facie compatible with the thesis that knowledge is incompatible with veritic luck. So the fact that Knowledge is Fallible is at odds with the thesis that knowledge is incompatible with veritic luck does not mean that we must eschew fallibilism, since we can merely revise the definition of fallibilism to Knowledge is Fallible*. But also note that Knowledge is Fallible* looks like a gross misnomer. It tells us something about epistemic justification, not knowledge; it captures our fallibilist intuitions about justification, not knowledge. This will be important later on. For now though it should be conceded that Knowledge is Fallible* plus The Epistemic Counterfactuals Fallacy cannot do the diagnostic work required to explain away anti-luck intuitions brought on by considering Gettier cases. We do not make the fallacy when we consider non-Gettier cases where S knows that p very failably, and an explanation is available as to why Gettier cases are relevantly different to other cases of very failable knowledge that does not violate fallibilism (under some description).

In more recent work, Hetherington has offered a more compelling, sophisticated (more compelling, in part, because more sophisticated) explanation as to why Gettier cases have errantly led so many to deny the tripartite analysis (Hetherington 2012a, b). I now turn to evaluate that account.

3 Defence of the Tripartite Analysis (Second Attempt)

In his most recent work (Hetherington 2012a, b), Hetherington has continued to maintain that if we are committed to fallibilism, we should not rule that Gettier cases are counterexamples to the Tripartite Analysis—against received wisdom, the subjects in Gettier cases have knowledge. What explains epistemologists’ manifest failure to recognise this? Hetherington’s new accountFootnote 5 is that it is a latent propensity toward infallibilism, unrecognised infallibilist intuitions. That is, many of us consciously assent to Knowledge is Fallible*, but yet harbour, latently, infallibilist intuitions that are manifested when we attend to Gettier cases.

And Hetherington argues that nothing else but infallibilist intuitions can explain why we rule out Gettier cases as cases of knowledge. The argument goes something like this: Gettier cases purport to show that in certain cases, S’s justified, true belief that p is not knowledge that p. So in all Gettier cases, S has a justified, true belief that p (else the putative counterexamples would miss the target). It follows that in all Gettier cases S has a true belief that p. So if S does not have a true belief that p, S is not in a Gettier situation. So our ruling that Gettier cases are not knowledge must be merely a reaction to the fallibility of the subject’s ‘justification’; that is, that being justified in believing that p does not entail that p is true. Why? Because if it is the case that when S’s “Gettiered” belief is false, S is not properly in a Gettier case anymore, then while the counterfactual situation we are presented with in order to show that Gettier cases are not cases of knowledge (where S’s belief is false) is usually ostensibly held to be a ‘Gettier’ case, really it is not. So there is nothing distinctive about the intuition brought about by considering Gettier cases beyond the recognition that S’s justification is fallible, no special Gettier intuition. When we intuit that Gettier cases are not cases of knowledge our “attention is caught—and held—by the marked fallibility; then by nothing further, nothing more distinctive of Gettier situations.” (Hetherington 2012a, p. 227). This is not to say that the belief in question cannot be false—that there is no possible world in which it is false—but, rather, that it cannot be false and yet be a Gettiered belief.Footnote 6 So it must be this infallibilist intuition that explains why people think that Gettier cases are counterexamples to the tripartite analysis. But the infallibilist intuition is clearly opposed to Knowledge is Fallible*—the claim that it is possible for S to have a justified, but false, belief that p; so we cannot be true to Knowledge is Fallible* while paying heed to this infallibilist intuition. So we must rule, then, that subjects in Gettier cases do know, after all. That is, eschew the infallibilist “Gettier” intuition.

Of course, the argument tells us that de jure people cannot consider Gettier cases to be counterexamples to the tripartite analysis; it does not rule out that de facto people have considered them so. Indeed it seems that many have done. So in fact Hetherington gives us a disjunctive explanation: when people judge Gettier cases to be counterexamples to the tripartite analysis they are either (1) falsely assuming that Gettiered beliefs can be false (qua Gettiered beliefs) or (2) mobilising latent infallibilist intuitions. What explains (1) is what Hetherington calls Gettier Partialism: an analysis that aims to “determine a verdict about a Gettier case on the basis of a description of part of the case, claiming to notice sufficiently revealing features of that part” (Hetherington 2012a, p. 221). That is, we consider a counterfactual situation where only part of the Gettier case we are considering remains the same, the part about the belief’s being true is left out. As we saw, though, this is fallacious, since a Gettier case that does not involve a true belief as its target is not really a Gettier case.

The disjunctive diagnosis is thus a comprehensive one, and as such, I would say, a plausible one. And Hetherington even tries to account for the infallibilist intuitions in (2). Here he cites Lewis’ claim that saying that knowledge is fallible sounds contradictory:

If you claim that S knows that p, and yet you grant that S cannot eliminate a certain possibility in which not-p, it certainly seems like you have granted that S does not know after all that p. (Lewis 1996, p. 549) [cited in Hetherington 2012a, p. 228].Footnote 7

So it is plausible to imagine that we have these infallibilist intuitions floating around latently. If Lewis had those intuitions, surely other philosophers have them as well, even if latently. But, unfortunately, Hetherington stops short at addressing the following concerns: what explains those infallibillist intuitions’ latency? What explains our having contradictory intuitions about Gettier cases? To make the explanation in (2) plausible, Hetherington has to assume that we fail to realise that we have the infallibilist intuitions, else we would never have so easily ruled that Gettier cases constituted counterexamples to the tripartite analysis. But why is it that we repress our infallibilist intuitions? Why don’t we repress, or fail to acknowledge, our fallibilist intuitions instead? And why do we repress either when we consider Gettier cases at all? The idea that we repress one of them seems somewhat ad hoc. And the choice of which one seems arbitrary.

It seems to me that Hetherington’s account continues to fail in its diagnostic responsibilities unless it can answer these questions. Else his account would have to retreat to the explanation in (1). This will impoverish his account considerably, given how far that explanation must go: can it really be the case that almost every epistemologist has made that mistake? Hetherington himself seems to put a lot more weight on the explanation in (2). I believe, however, that some of the considerations Hetherington offers us do provide us with the means by which to answer the diagnostic problem. However, in doing so, they support a different position than that perhaps envisaged by Hetherington; that is, I will argue, they do not support the thesis that knowledge is justified, true belief, but rather the thesis that justification is neither necessary nor sufficient for knowledge and thus that the theory of epistemic justification is importantly independent of the theory of knowledge.

4 Fallibilism and the Divorce Thesis

Consider the following thesis.

DIVORCE:

Epistemic Justification is neither necessary nor sufficient for knowledge.Footnote 8

I call the above DIVORCE because if it is true then the theory of epistemic justification is something separate to the theory of knowledge in the sense that it is not there with the purpose to tell us something about what knowledge is (i.e. the Divorce thesis). If the theory of epistemic justification is not to be redundant, epistemic justification must be something we want to find out about independently of what we find out about knowledge. It seems to me that the data that Hetherington brings our attention to supports DIVORCE and that if DIVORCE is true, we can remedy the diagnostic deficiencies of Hetherington’s account, namely that it cannot explain why we have contradictory intuitions about Gettier cases. Labelling the infallibilist intuition latent seems arbitrary: why is our infallibilism latent and our fallibilism not so? Why have we failed to appreciate our latent infallibilism?

If DIVORCE is true, here is what we can say: we do have both fallibilist and infallibilist intuitions but they are not in fact contradictory intuitions when we consider Gettier cases. This is because our fallibilist intuitions are directed at a different object than are our infallibilist intuitions. That is, we are fallibilists about epistemic justification (as captured in Knowledge is Fallible*), but infallibilists about knowledge. If epistemic justification is neither necessary nor sufficient for knowledge, we do not contradict ourselves by saying this as we would by saying that Knowledge is Fallible* and it is not the case that Knowledge is Fallible*. And if our intuitions are not contradictory, there is no need to posit any latency on the part of one of them, the postulation of which then requires explanation on pain of being otherwise arbitrary and/or ad hoc.

Of course, it is not enough that DIVORCE be convenient for it to be true. But the data that Hetherington has identified gives us good reason to think that DIVORCE is true. In Hetherington’s analysis what primarily explains why so many philosophers have been prone to conclude that the tripartite analysis is false is what he calls the Gettier Illusion…

…the illusion that one is assessing Gettier cases as a fallibilist should, when really one’s reaction bespeaks a residual infallibilism about what it is to know. It is a widespread illusion. (Hetherington 2012a, p. 227).

The residual infallibilism Hetherington has in mind here is surely this: S cannot have a false, justified belief.Footnote 9 But the infallibilism operative when we consider Gettier cases could instead be formulated like this: S knows that p only if S’s belief that p is true in all close possible worlds. Let us call this formulation of the inclination to infallibilism the Safety Intuition. That we have the Safety Intuition would in fact capture better the idea that the infallibilist intuition at work in assessing Gettier cases concerns what it is to know, as opposed to what it is to be justified in believing that p. As I mentioned earlier, Knowledge is Fallible* is somewhat of a misnomer, since it is directed at justification, not knowledge. Now, the Safety Intuition is often held to be an anti-luck intuition. But one way of making the point here is that if this is how to understand the anti-luck intuition, the anti-luck intuition looks like an infallibilist intuition. This is so despite the fact that it allows for error in remote possible worlds, since the fact that S’s belief is safe guarantees that S’s belief is true in all close possible worlds, and therefore in the actual world. So the material conditional that if S’s belief is safe then S’s belief is true, is true. In contrast, according to a notion of fallible justification, it is not true that S’s belief being justified guarantees that S’s belief is true in all close possible worlds, nor indeed that the material conditional that if S’s belief is justified then it is true, is true. So if it is correct to call the proposition that no justified belief is false ‘infallibilism’, then it is also correct to call the proposition that no safe belief is false ‘infallibilism’. When we consider Gettier cases, then, we rule that they are not cases of knowledge because the justified true beliefs in Gettier cases fail to have a property that entails that they are true in this world. We do not therefore need to consider any counterfactual scenario where the Gettiered belief is false, and we thus do not (pace Hetherington) become victims of Gettier Partialism.

One might object that I’m relying here on an overly strong conception of Safety according to which one’s belief is safe just in case it is true in all close possible worlds. Pritchard, for instance, sometimes defines Safety as follows (he calls this Safety III):

For all agents, φ, if an agent knows a contingent proposition φ, then in nearly all (if not all) nearby possible worlds in which she forms her belief about φ in the same way as she forms her belief in the actual world, that agent only believes that φ when φ is true. (Pritchard 2005, p. 163).

If this is how we should understand Safety, then it is quite possible for a belief to be safe but yet false in the actual world, since the actual world could be the world where the relevant remote possibility obtains.Footnote 10 However, Pritchard has since modified his account to (let’s call this Safety IV):

S’s belief is safe iff in most near-by possible worlds in which S continues to form her belief about the target proposition in the same way as in the actual world, and in all very close near-by possible worlds [my emphasis] in which S continues to form her belief about the target proposition in the same way as in the actual world, the belief continues to be true. (Pritchard 2007, p. 292).

Under this formulation, one cannot have a safe, but false belief that p—given that the actual world is one of the very close near-by worlds,Footnote 11 the material conditional if S’s belief is safe then it is true, is true.Footnote 12 The reason for the shift, I think, can be explained like this: in Epistemic Luck, Pritchard discusses the move from a weaker Safety principle “Safety II” to Safety III above. Safety II:

For all agents, φ, if an agent knows a contingent proposition φ, then, in most [my emphasis] nearby possibly worlds in which she forms her belief about φ in the same way as she forms her belief in the actual world, that agent only believes that φ when φ is true. (Pritchard 2005, p. 156).

Pritchard prefers Safety III to Safety II because Safety II cannot deal with the lottery puzzle. But an objection to Safety III is that rare possibilities can obtain in the actual world and in nearby possible worlds. Pritchard’s reply is to say that if a rare (knowledge relevant) possibility can obtain in very nearby possible worlds, then we rule that the agent does not have knowledge, but that if the possibility can obtain in worlds that are not nearby, then the agent does; as he puts it: “…even though the possibility at issue is rare, our intuition is, I take it, that the agent does not know…” (Pritchard 2005, p. 165). But if this point holds for Safety III, it also holds for Safety IV. Indeed, Safety IV handles the point better. So Pritchard ends up with no principled reason to prefer Safety III to Safety IV, and so he (unsurprisingly) ends up preferring Safety IV. Footnote 13 Further, Safety III cannot capture the intuition behind our ruling that Gettier cases are not cases of knowledge, if Hetherington is right that no Gettiered belief qua Gettiered belief can be false, since the belief’s being safe would not then entail its truth. That is, in order to rule that a Gettiered belief is not a case of knowledge because it contravenes Safety III, one would have to consider counterfactual situations where the target belief is false—and so, according to Hetherington, succumb to Gettier Partialism. Footnote 14

But the Safety Intuition does not tell us anything about whether S can have a false, justified belief. For all the intuition tells us, it might be the case that knowledge is safe, but justification not so. So if we interpret our infallibilist tendencies in terms of the Safety Intuition we can be infallibilists about knowledge and fallibilists about justification. And we do not then need to explain conflicting intuitions. Hetherington’s case for thinking that it must be infallibilist intuitions that make us rule subjects in Gettier cases as failing to know supports our driving a wedge between our intuitions about knowledge and our intuitions about epistemic justification.

To make things clearer, here is how my argument works so far (granting Knowledge is Fallible* as a background assumption):

  1. 1.

    Infallibilist intuitions are behind our not ruling Gettier cases as cases of knowledge, else we are just making a mistake due to Gettier Partialism. [PremiseFootnote 15].

  2. 2.

    The infallibilist intuitions in (1) can be interpreted as intuitions directed at knowledge (as in the Safety Intuition)Footnote 16 or as intuitions directed at epistemic justification. [Premise]

  3. 3.

    If we interpret them as intuitions directed at justification, we postulate inexplicable, contradictory intuitions. [Premise]Footnote 17

  4. 4.

    If we interpret them as intuitions directed at knowledge, we do not postulate inexplicable, contradictory intuitions. [Premise]

  5. 5.

    An account that does not stipulate inexplicable, contradictory intuitions is better than an account that does. [Premise]

  6. 6.

    We should interpret the intuitions behind our not ruling Gettier cases as cases of knowledge as infallibilist intuitions about knowledge [From (1–5)].

But how do we get from (6) to DIVORCE? (6) does not entail DIVORCE, since knowledge could be defined as justified, true belief plus some safety condition meant to address the Safety Intuition. Let’s call this account of knowledge JTB + . If (6) is true, justification is fallible (if we have intuitions that it isn’t, it is because we are mistaking infallibilist intuitions about knowledge, for infallibilist intuitions about justification). Now, it is a platitude that an infallibilist condition on knowledge is going to be stronger, when it comes to ruling out veritic luck, than a fallibilist condition. But that is the crucial premise that gets us from (6) to DIVORCE. For if the ‘+’ condition in JTB + is doing all the anti-luck work, then what is justification doing in the analysis of knowledge? It cannot be there to do anti-luck work, on pain of being redundant.Footnote 18 Of course, justification might do some anti-luck work (as a part of a conjunct, perhaps), but since it is work that the ‘+’ condition also does, the fact that it does so does not stop it being made redundant if it is there as an anti-luck condition.

In light of this, there are two options available: (a) we simply accept that justification is neither necessary nor sufficient for knowledge, or (b) motivate the justification condition via other means than the anti-luck, infallibilist intuition. Only option (b) seems to give us any chance of opposing DIVORCE. What else might motivate the justification condition? We might say it is motivated by the Deontic Intuition, Footnote 19 viz. the intuition that we ought to (have a duty or obligation to) have epistemically justified beliefs. But this intuition cannot motivate the thesis that justification is necessary for knowledge, since we might have an obligation to have epistemically justified beliefs even if the definition of knowledge does not contain a justification condition. So it would have to be a slightly different intuition that does the motivating work, perhaps the intuition that S knows that p only if S has a right to believe that p. Put differently, the intuition that S has to deserve to believe that p if S is to know that p.Footnote 20 Call this the Knowledge is Deontic Intuition. But if the Knowledge is Deontic Intuition motivates a justification condition for knowledge, then we also have the Deontic Intuition. At least is would be very odd to have the intuition that only knowledge, i.e. not justification, is deontic. But if we have the Deontic Intuition, we have the means to explain away the Knowledge is Deontic Intuition. That is, we can claim that it is a misapplication of the Deontic Intuition, which is merely about justification, explained by the long history of the idea that knowledge entails justified belief. We think that justification is deontic and mistakenly assume that therefore knowledge must be deontic too.

A very similar case can be made with respect to alternative ways to motivate the justification condition for knowledge. First of all, suppose we just have a brute intuition that justification is necessary for knowledge. Perhaps we can cash this out like this: in considering cases where a subject has a safe but unjustified belief that p, we have the intuition that the subject does not know that p. That we have this intuition can only be the case, however, under certain analyses of epistemic justification—that is, where it is not merely an anti-luck condition. Perhaps, then, we can think of justified belief to be epistemically rational belief, or epistemically responsible belief.Footnote 21 But, again, assuming that another condition is doing all the anti-luck work in the analysis of knowledge, it would be very odd (not to say incoherent) to have the intuition that responsible or rational belief is necessary for knowledge and not have the Knowledge is Deontic Intuition. And once we have the Knowledge is Deontic Intuition, we also have, as I mentioned, the Deontic Intuition and so a way of motivating talk of justification in epistemology, independently of its figuring in an analysis of knowledge. And we also have a way of explaining away the intuition that justification is necessary for knowledge—i.e. the latter is just a misapplication of the Deontic Intuition explained by the long history of the idea that justification is necessary for knowledge.Footnote 22

Another motivation might be the following. Knowledge is not only incompatible with veritic luck, but it is also incompatible with another kind of luck, reflective luck. According to Pritchard (2005), one suffers from reflective luck when from the agent’s point of view it is lucky that one’s belief turns out to be true.Footnote 23 Knowledge is incompatible with the presence of reflective luck, we might think. For example, a reliable chicken-sexer, who takes her beliefs to be lucky guesses, does not know the sex of the chickens she examines, despite the fact that her beliefs regarding their sex are systematically accurate. One might then suppose that a justification condition is motivated such as to take care of this kind of luck, since clearly the modal anti-luck conditions fail to eliminate it (given certain disambiguations concerning what we mean by “systematically accurate” here). But, once again, it seems very hard to have the intuition that knowledge is incompatible with reflective epistemic luck without having the Deontic Intuition. Here is how Pritchard goes on to explain the anti-reflective luck requirement: “…we don’t just want agents to be forming beliefs in such a way that we can rely on the truth of those beliefs, we also want agents to be cognitively responsible for their beliefs, and this is only possible if they form beliefs in such a way that are more than just safe” (Pritchard 2005, p. 184).Footnote 24 Notice how Pritchard articulates this requirement on belief in a way that is independent of how we might define knowledge, such that the intuition behind the idea that knowledge is incompatible with reflective epistemic luck turns out just to be the Deontic Intuition. And that intuition, together with the explanatory work that the long history of the idea that justification is necessary for knowledge can do, will again give us all we need to defend DIVORCE.

5 Conclusion

None of the above proves DIVORCE to be true, to be sure; but it does lend credence to it. DIVORCE gives us the resources to explain what is happening when we attend to Gettier cases more successfully than does Hetherington, since we do not have the gargantuan task of explaining why everyone has made the mistake of thinking that Gettier cases are not cases of knowledge. Everyone did not make a mistake. Our infallibilist intuitions attest to that. Knowledge is not justified true belief. What we have to explain is how we can think this while respecting our fallibilist intuitions about epistemic justification (articulated in the misnomer Knowledge is Fallible*). Simple. Justification is not necessary for knowledge. We are tricked into thinking there is a problem because we assume that justification is necessary for knowledge. But we think justification is somewhat important, nonetheless. It seems weird to leave it out of epistemology altogether. The Deontic Intuition explains this, but the Deontic Intuition is independent of the claim that justification is necessary for knowledge. So DIVORCE accommodates all of our intuitions here, without having to qualify any as repressed in order to do so.Footnote 25 And it makes sense of the data that Hetherington has, rightly, identified: that it is infallibilist intuitions that are at the root of our feeling that Gettier cases pose a challenge to the tripartite analysis.

The tension that Hetherington really identifies, I think, is between these two theses: Justification rules out veritic luck. Justification is fallible. What we should give up, then, if we are fallibilists, is the thesis that justification rules out veritic luck. And this makes further plausible the claim that epistemic justification is something that we should be interested in independently of how we define knowledge, if it is something we should be interested in at all. But that does not mean that we should give up on the thesis that knowledge is infallible. And because knowledge is infallible, it must be something other than justified, true belief.Footnote 26