Actually Knowing Stephen Hetherington The Philosophical Quarterly, Vol. 48, No. 193. (Oct., 1998), pp. 453-469. Stable URL: http://links.jstor.org/sici?sici=0031-8094%28199810%2948%3A193%3C453%3AAK%3E2.0.CO%3B2-M The Philosophical Quarterly is currently published by The Philosophical Quarterly.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/philquar.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
JSTOR is an independent not-for-profit organization dedicated to and preserving a digital archive of scholarly journals. For more information regarding JSTOR, please contact
[email protected].
http://www.jstor.org Mon Jun 4 08:30:39 2007
The Philosophical Qarterb, L70101. 48,~$5.193 LSXV 0031-Soy4
October rggS
ACTUALLY KNOWING
Few, if any, epistemologists accept this thesis: T . If a belief is true and well supported by evidence, none of which is false, then the belief is knowledge. The epistemological consensus regarding (T) is (and has long been) that it is clearly falsified by some simple counter-examples, specifically, some simple Gettier cases.' I shall argue that it is not so clear that (T) is falsified by those putative counter-examples, because it is not so clear that the cases are really Gettier cases. Epistemologists have assumed that the cases are ones where there is a lack of knowledge; I shall develop an interpretation according to which the cases are ones where there is knowledge.
(T) is routinely taken to be falsified by cases in which an epistemic subject relies on no false evidence, yet is Gettiered all the same. (Epistemic subjects are Gettiered when their belief is Gettiered. Their belief is Gettiered when it is true, and well supported by evidence, without being knowledge.) I shall concentrate on three well known cases - Carl Ginet's fake barns,' Gilbert Harman's political assassination3 and Richard Feldman's existential generalizati~n.~
' Edmund Gettier, 'Is Justified True Belief Knowledge?', dnahsis, 23 (1963),pp. 121-3. Although the example is Carl Ginet's, Alvin Goldman brought it to the attention of epistemologists in general, in 'Discrimination and Perceptual Knowledge', Journal ofPhilosop/y, 73 ('979, PP. 7719,L at P P 772-3. 7hought (Princeton UP, 1g73),pp. 143-4. Harman (see pp. 47-8, 151, 153)is one who does accept something close to (T). But his account is unnecessarily complicated, as I shall explain. ' 'An Alleged Defect in Gettier Counter-examples', Australasian Journal of Philosopb, 5 2 ('974). PP. 68-9. The Editors of 7728 Philosophical @tartub, 1gq8. Pulll~shedby Blackwell Publishers, 108 Cowley Road, Oxford oxq ijr, UK, and 350 blain Street, blalden, MA 02148,USA
454
STEPHEN HETHERINGTON
How do epistemologists try to justify the usual interpretation of the cases? All too often they call on nothing more than so-called 'intuition': 'Henry is looking at a real barn; yet there are many fake barns nearby, looking like real barns. So of course Henry does not know that he is seeing a barn. T o deny this is counter-intuitive." But that can sound dogmatic, even unphilosophical. Accordingly, often a fuller analysis is sought, such generic attempts to appeal to 'intuition' being supplanted by talk of luck: 'In Gettier cases, epistemic subjects, given the evidence upon which they rely and the context where that reliance is occurring, have reached a true belief only by luck. Henry is just lucky to be looking at a real barn, not a fake one, and hence lucky to have a t ~ u ebelief that he is seeing a barn. Clearly this luck is incompatible with his belief's being knowledge!' Yet we might ~7ishfor more depth than that, too. TVhat kind of luck is involved? (Or are we to answer this question just by falling back on 'intuition' again - e.g., as a way of justifying the use of 'clearly' in the previous paragraph's claim that the luck is clearly incompatible with knowing?) Rlaybe it is mere luck that a desk rather than a duck is in front of me at this moment. Does this prevent my knomzng that a desk is in front of me? If it does not, then why not? If the questioning gets this far, a counterfactual turn is likely to occur, as the epistemological story strives for a deeper understanding of itself. T o show how that occurs, I shall consider each of our cases in turn, highlighting the specific counterfactuals that contribute so much to the cases' standard interpretation. These counterfactuals are not necessarily part of the deepest, or the ultimate, justification of the cases' standard interpretation. But they are necessary at some level of that justification. They are not part of how all epistemologists explicitly and/or initially purport to justify the interpretation of these cases' epistemic subjects as lacking knowledge. Epistemologists need to call on such counterfactuals at some point, however, when trying to justify their interpretation of the cases. I . 'Thefake barns case. As Goldman himself observes (p. 778), his interpretation of the fake barns case the usual interpretation of the case, it has transpired - relies on a counterfactual claim about Henry. This is the kind of counterfactual in question: -
If Henry were to have been looking at one of the fake barns while driving along, he would unwittingly have been deceived (while having the same belief, 'I see a barn', along with seemingly similar evidence). " On the philosopllical use and episrelnic significance of 'intuitions', see G. Bealer, 'On the Possibility of Philosophical Knowledge', Philosophical Perspectives, 10 (19961, pp. 1-34, 0T h e Edltora of 7 7 ~ P / ~ ~ / u s u ~ h Qiarterii, ico/ 1998
455
ACTUALLY KNOWING
And Henry might, so easily, have continued his drive and been misled by a fake barn. So the subjective similarity of his present situation (where he is not deceived) to the counterfactual situation (where he is deceived) makes the actual situation one where he might as well be deceived, epistemically speaking. The epistemic result is the same in each situation. In each, he lacks knowledge of seeing a barn. The counterfactual lack of it implies the actual lack of it. (Alternatively, Harman would say that in the fake barns case Henry lacks knowledge by relying, at least implicitly, on reasoning which includes the false premise that there are no fake barns in the area. But that interpretation attributes to Henry a reflectiveness about the situation which, by hypothesis, he is not manifesting. I shall return to this kind of point late in $11 below, when discussing the existential generalization case.) 2. The political assassination case. Jill has read and believed a newspaper report, an accurate one, as it happens, of a political leader's being assassinated. However, she has not heard what others have heard - a (false) government announcement deying that there has been an assassination. The usual epistemological reaction is to deny that Jill knows that there has been an assassination. And that reaction relies, at some level, on a counterfactual like this: If Jill had heard the government announcement, she would have been unwittingly deceived. And Jill might, so easily, have heard the announcement. So the fact that, if she had heard it, her current evidence would not have been strong enough to make her belief knowledge implies that her current evidence al~eady fails to make her belief knowledge. The clear epistemic failing within the counterfactual situation infects the current situation. In each situation Jill lacks knowledge that an assassination has occurred. The counterfactual lack of it implies the actual lack of it. 3. ?he existentiul gene~alizationcase. Smith infers that someone in his office o ~ m as Ford; and he is correct, because Havit owns one. Smith's reasoning relies on the true existential generalization that there is someone in the office ~ 7 h ohas always been honest with him, who claims to own a Ford, and who has shown him a certificate of Ford okvnership. (The 'someone' happens to be Nogot.) And the usual epistemological reaction is that Smith's belief that someone in the office owns a Ford is not knowledge. That reaction, when pushed to justify itself, ~7illcall on some such counterfactual as this: If Smith had reasoned slightly differently in particular, if he had inferred the same conclusion from 'Nogot o~7nsa Ford', instead of from the existential generalization he did use - he would unwittingly have -
Q The Editots of 771~ Phtioiophacal Qua,leib, 1998
456
STEPHEN HETHERINGTON
been misled in his reasoning. He would have been using false evidence, decreasing his chances of deriving a true conclusion. And Smith might, so easily, have reasoned in that ~7ay.So he would have reached the same true conclusion ('Someone in the office owns a Ford') without realizing that he was being misled along the way. Hence his reaching that conclusion now, also without thinking that he is being misled, does not make his belief knowledge, even though it is true and he is not actually being misled. He might as well have been misled by his reasoning, epistemically speaking. The epistemic result is the same either way. The counterfactual lack of knowledge (if he had used false evidence in his reasoning) implies the actual lack of it (when he actually uses no false evidence).
The standard interpretation of those cases is that in each of them the epistemic subject - Henry, Jill, Smith - lacks knowledge. I explained in $1 above how in each case that interpretation relies on a pertinent counterfactual. Here I show why, given that explanation, the standard interpretation is not obligatory. That interpretation relies on what I shall call the epistemic countefactualsfallag. Epistemologists infer from the fact that Henry would be deceived if he were to continue his drive that he does lack knowledge. But this inference fails to distinguish between these two possible 'continuations' of the fake barns story: (a) Henry is never deceived by any fake barns (perhaps he turns the car around, heading home for a just-remembered social engagement); and (b) Henry is deceived by some fake barns as he continues on his way. Goldman's tale of Henry's travels does not tell us whether or not Henry continues on his way, actually being deceived. Is that significant? It is. Leaving the matter indeterminate allows readers to imagine the worst - which is exactly what epistemologists have done. By the same token, however, leaving the matter indeterminate does not require readers to imagine the worst which (as we are about to see) is why epistemologists need not have reacted to the case as they have done. Given (b), we are more likely to say that even now, before he has been deceived, Henry fails to know. But given (a), Henry is not in a situation which strikes me as being one where we have to deny him knowledge. Instead, (a) seems to give us a case in which it is at least as plausible to say only that Henry almost fell foul, epistemically speaking, of the fake barns: he did not fall foul of them, but it was a near thing. Maybe it is only a matter of -
0The Editors of n i r Philoiophiiol&art&,
,998
ACTUALLY KNOWING
457
chance that, by having driven home when he did, Henry was in situation (a). Still, as I said, once he is in (a), it seems to be a situation where he does not fail to know. In the following analogous situation, if scientist Harriet, interpreting her data, infers a true and well warranted conclusion c, this is a plausible candidate for (fallible) scientific knowledge. Still, there are many competing conclusions she might have drawn (conclusions equally well warranted even if false) which are located in the 'nearby' logical space. And many of these, if she had continued thinking about the issue, as in fact she did not, would have deceived her into not concluding that c. Must we deny that c is knowledge because these competing hypotheses exist? I hope not. Yet those competing hypotheses stand to c as the fake barns stand to Henry's belief that he is seeing a barn. And the original version of the fake barns case does not tell us which of these situations, (a) or (b), obtains. So we should not assume, when presented with the case's original version, that Henry fails to know. The fact that if he had continued his drive, then he would have been deceived, should not be assumed to entail that he lacks knowledge now, before all of that has occurred. All that we have to infer is that he is close to being deceived (and hence close to lacking knowledge in that clear-cut way). Depending on what else occurs, he might, or might not, lack knowledge (of seeing a barn) at this stage of his journey. Again, though, epistemologists do infer, and unhesitatingly so, that Henry lacks knowledge. Nevertheless we have seen enough already to appreciate the hastiness of that inference. T o understand the case's full epistemic potential, we need an interpretative option other than the usual one. So I shall outline a portrayal of Henry as hauing the knowledge which the usual interpretation denies him. This might generate a somewhat revisionary view of knowledge; but part of my thesis is that, in so far as the usual interpretation of this case (and hence the usual view of knowledge) relies at some stage on the kind of counterfactual I have identified, that interpretation is not the only coherent way to assess Henry's epistemic situation. The usual interpretation is not entailed by that counterfactual; this leaves the door ajar for an alternative interpretation (as follows, of this case and of our other ones). We saw, a moment ago, that the fact that Henry would have lacked knowledge if he had continued his outing and been deceived does not entail that he lacks knowledge now, when he is not being deceived. His only undeniable epistemic failing is a counterfactual one, not an actual one. Admittedly, there is some temptation to interpret him as actually lacking knowledge. But it is easy to misjudge the worth of that temptation, which is rather like the misguided temptation to deny that an action is generous, when our ground for that denial is that the person who performed the Q The Editors of 7he Ptztioiophtcnl Qunrferb, 1998
45%
STEPHEN HETHERINGTON
action is rarely generous in other situations, even in extremely similar ones, even in situations which he and/or we would regard as being qualitatively identical to the one where the action occurred. That denial would be mistaken, because even if the person is not generous, this does not entail that the particular action is not generous. Although the fake barns do pose an epistemic threat to Henry (perhaps to his general epistemic trustworthiness in this area), this does not entail that they deprive him of knowledge right here, right now. As Stewart Cohen says, of another situation, why should S fail to know the table is red when he sees it under perfectly good perceptual conditions, when he does not possess any defeaters? Surely not because his feeble reasoning powers would make it difficult for him to appreciate a defeater were he, contrary to>ct, to possess one."
All that is immediately epistemically worrying in Henry's situation, on my interpretation, is that he has almost been deceived in such a way as to lack knowledge of seeing a barn. Almost to lose a race, however, is not to lose it. Henry has not been deceived, and this is a reason for not regarding him as lacking knowledge either. (But this is not to say that a person's having a true belief that p - not being deceived when having the belief that p - is sufficient for knowing that p. I shall explain in §IV below why that worry bypasses my interpretation.) Would we describe as lacking knowledge someone who spent a lifetime almost being deceived, never being deceived? I doubt it. Here is a story with an analogous moral, about the go-year-old man on his deathbed ... joyous and relieved over the success of his deception. For go years he has affected courtesy, kindness, and generosity - suppressing all the malice he knew was within him while he calculatedly and artificially substituted grace and charity. All his life he had been fooling the world into believing he was a good man. This 'evil' man will, I predict, be welcomed into the Kingdom of Heaven.'
The same is true, mutatis mutandis, of the assassination case. Jill has certainly come close to believing what the other people who heard the government announcement are believing, and hence to being misled as they are being misled. But, on the interpretation I am sketching, Jill only almost lacks knowledge of the assassination. If she left the country immediately after reading the newspaper, does the fact that if she had remained, and had heard the announcement, she would have been deceived imply that she lacks knowledge now, as she sits in an aeroplane speeding away from the country, never to hear the announcement? This is like the possible G
1
Knowledge, Context, and Social Standards', G~nthese,73 (1987),pp. 3-26, at p. 23 fn. 11.
' Willard Gaylin, 'What You See Is The Real You', in T. Cooley (ed.), The dVorton Sampler, 2nd edn (New York: W.W. Norton, 1982),pp. 160-2, at p, 161. O The Editors of Tne Phrlosophiiai Qraitnb, 1998
ACTUALLY KNOWING
459
continuation (a) of the fake barns case described earlier, in which Henry departs, never seeing a fake barn. Jill has some luck, gaining good evidence for a true belief and then leaving, never actually hearing the announcement. She is only ever counterfactual& deceived; if she had stayed, she would have been deceived. Still, you might object, even when Jill leaves the country, never hearing the announcement, her counterfactual epistemic limitations remain. It remains true that she would have been deceived, and easily so. And this, you conclude, is why she lacks actual knowledge of the assassination's having occurred. My reply is that counterfactual epistemic limitations can be just that counterfactual. The fact that Jill would have lacked the knowledge does not entail that she does lack it. It is beguiling to focus on the seemingly most relevant respect in which Jill is like the other people - the fact that she would be misled, just as they are, if she were to hear the government announcement. But we should not overlook the vital respect in which she dfiers from the others - the fact that she is not in the situation where she would be misled. MTe must not link the actual and the counterfactual too closely, or we shall confuse lacking knowledge with almost lacking it. I have no precise criterion ready to hand for this difference between knowing and almost knowing. (Nor, as far as I know, has anyone else.) But the difference must exist. And reflection on these cases is a good way to begin to understand the difference. Thus, for instance, Jill had some epistemic luck in not hearing the announcement. Yet this need not be taken to imply that she had a pertinent epistemic lack - that she therefore failed to know that there had been an assassination. Instead, there is conceptual space for us to say that Jill was lucky to have that knowledge. She was lucky to have that belief, which happens to be knowledge. The usual interpretation is that, because of the luck in her having the belief, it fails to be knowledge. As before, though, that interpretation is optional, because the counterfactual which is assumed to ground it at least fails to entail it. We have the opportunity to develop an alternative interpretation ofJill's epistemic situation: She almost failed to know - because she was almost in the position other people were in, hearing the announcement as they did; hence she was almost deceived as they were. Nevertheless she did escape that fate, and, given the accuracy of her belief and her evidence for it, knowledge was her reward. Analogous remarks apply to Smith, in the existential generalization case. Feldman does not tell us why Smith favoured the inferential path he did. % The Editors of nie Pliilo~opliiialguarterb, 1gg8
460
STEPHEN HETHERINGTON
Maybe he reasoned via his existential generalization ('There is someone in the office who has always been honest with me, who claims to own a Ford, and who has shown me a certificate of Ford ownership') rather than via the claim that Nogot owns a Ford, because he was deliberateb being cautious. He inferred a similarly general conclusion 'Someone in the office owns a Ford'. Perhaps this, too, reflected caution on his part. If that was why Smith reasoned as he did, he should be credited with knowing. At any rate, we are not forced to view Smith as lacking knowledge. He might have reached the same conclusion via 'nearby' false evidence (e.g., by thinking that Nogot owns a Ford) without being aware of that evidence's being false. But this entails only that he would have been deceived in his reasoning - whereas in fact he has not been. I am not saying that therefore we have to see him as knowing. However, I am urging a way of thinking about knowing which concedes that Smith has come close to being unwittingly deceived, without conceding that therefore he lacks knowledge. That way of thinlung also allows us to dispose of the following common interpretation of the case: Even if Smith has not explicitly called on a false premise, he would do so if he were to be asked to defend his conclusion in more detail. In that sense, he is relying on false evidence after all. (And that is why he lacks knowledge.) This interpretation epistemically evaluates Smith within his actual context by imagining his being in a dgerent epistenlic situation, where he engages in explicit and epistemically significant reflection which he does not explicitly pursue in his actual context. The supposed epistemic significance of that imagined reflection is that Smith would be deceived in that other situation. This is because he would be calling on false evidence, hence on evidence which is not knowledge itself. And, other things being equal, no belief is inferential knowledge if it relies on evidence which is not knowledge. The classic epistemic regress principle tells us this. Perhaps some special situations are exceptions to that principle; however, none of them is present in these cases. But that common interpretation is misguided. Indeed, it is one more manifestation of the epistemic counterfactuals fallacy. We must not assume that whenever a person has not reflected on some specific false proposition upon which he could easily have reflected, then he might as well be doing so, epistemically speaking (and therefore that he does lack knowledge). If we are to assess a person's epistemic status at a specific time, we must respect exactly how reflective he is being, rather than how reflective he might easily be. After all, whenever a person fails to reflect on a true proposition which O The Editors of Tlie Philoiophidnl auorterb,
1998
ACTUALLY KNOWING
461
would have been well within his intellectual reach, do we assess his epistemic status in terms of how easily he could have reflected on that proposition? Are we that generous in our epistemic evaluations? Does a student's exam answer constitute knowledge that p because he is close to citing the vital truth which would support his claim that p? Of course it does not: at best, we say that his answer almost constitutes knowledge that p on his part at that time. Similarly, in part because Smith is being less reflective (in particular, by overlooking some easily available false evidence), he has some knowledge which he would have lacked if he had been more reflective in the way suggested by the common interpretation. So his having that knowledge is somewhat fortunate, a touch lucky. Even so, on my interpretation he still has the knowledge. Analogously, a person could know that p, in part because he has never reflected on some associated sceptical possibilities - and this is so, even though it would be easy for him to have reflected on those possibilities, and even though if he were to reflect on them he would fail to know that p (for instance, because he would not retain his belief that p). As ever, we must be vigilant if we are to avoid exemplifying the epistemic counterfactuals fallacy. Part of the interpretative difficulty with this case is that Feldman, like Harman and Goldman (not to mention most other epistemologists) with their own cases, leaves open possibly significant details in his epistemological story. For instance, is Smith deliberately being cautious in his reasoning? Exactly what is in his mind? As epistemologists in general would agree, Smith should be denied knowledge if his conclusion really means, for him, only that Jl'ogot owns a Ford. And it can seem reasonable, when trying to assess Smith's epistemic status, to attribute to him either (i) that very reading of his conclusion, or (ii) the belief that Nogot owns a Ford (as an important but unstated component in his reasoning). Yet we must not read into the case what someone other than Smith would think, or even what we would think, in his situation: to attribute (i) or (ii) is gratuitous. It is also the kind of attribution to which philosophers are regrettably prone. It interprets an epistemic subject as exercising a level of self-awareness and reflectiveness in short, a level of rationality - that philosophers themselves routinely believe they routinely exercise (and which they believe that they would exercise if they were in the epistemic subject's situation). We are potentially in the presence of projection. We should interpret Smith in accord, first and foremost, with what he clearly does, rather than with what we might like to think that he is doing 'behind' what he clearly does. We should not assume that Smith is thinking in either of those ways, (i) or (ii), about Nogot, since (a)we are not told that he is; and (b) on the face of it, he is talking more cautiously, less specifically. 0The Ehto1.s of 7lz Phil~rophrcalQtorferb, 1998
462
STEPHEN HETHERINGTON
As with Henry and Jill, then, we have the chance to interpret Smith as knowing; we are not forced, by the kind of counterfactual to which we are standardly directed, to interpret him as lacking knowledge. So my alternative interpretation says that he has the knowledge even while almost lacking it. He happened to follow a non-misleading inferential path (one which led to a true conclusion, using only true evidence), just as Henry happened to be reacting only to a real barn and just as Jill happened to be reacting only to accurate evidence. Like Henry and Jill, too, Smith might easily have been misled without realizing it. However, in fact he was not and its being counterfactually true that he would have been, if he had reasoned differently, does not entail that in fact his epistemic situation is one where he lacks knowledge. Part of the larger picture behind my interpretation is a fallibilism about knowledge. And, surprisingly, that is already enough to distinguish it from the usual interpretation. The latter denies Smith knowledge on the ground that he could so easih have called on false evidence rather than on the wholly true evidence which in fact he used (and on the ground that he would not have realized that he was calling on some false evidence rather than only on true evidence). In that way, according to the usual interpretation, there is luck in Smith's calling only on true evidence rather than on the equally accessible false evidence. So, concludes the usual interpretation, Smith's evidence fails to give him knowledge. But that conclusion follows only if the usual interpretation is presuming this principle: Other things being equal, one knows that p only if it is not true that one might easily not have known that p. And this is an infallibilist principle, since it denies that knowledge can just happen to be present. Because it was quite possible for Smith to have reached for the false evidence (being deceived by it), it was quite possible for him not to have known. Accordingly, in fact he does fail to know. It is ironic that most epistemologists will accept this interpretation of Smith while also claiming to be failibilists about knowledge. In fact, it is not simply ironic; it reveals a fundamental confusion on their part. For instance, it is part of why epistemologists have been unable to agree on a non-sceptical analysis of Gettier cases (and hence why 'the Gettier problem' has proved to be so problematic). After all, it is standard for epistemologists to observe that Gettier cases arise if we accept that knowledge that f can rest on evidence which does not entail p - that is, if we accept that knowledge can be fallible in that way. Now most epistemologists who accept that knowledge can be fallible analyse its fallibility in no more depth than this - and then they expect to understand Gettier cases in a way % The Editors of
nit Philoiop/iicnlQurttrb,
1998
ACTUALLY KNOiYING
463
that also sees no more fallibility than this in fallible knowing. However, my present point is that, in a fuller sense of epistemic fallibilism, the usual view of the cases rests, at some level, on an infallibilism about knowledge - the kind of infallibilism described in the previous paragraph. (The usual view denies knowledge to the episteinic subject within the Gettier cases, and this denial, I have argued, rests on what is really an infallibilist principle, even if not a standardly discussed one.) Consequently, if we are to understand Gettier cases, and hence to solve 'the Gettier problem', without denying that there is fallible knowledge, then we should reject the usual interpretation of these cases. T o that end, I am offering the following fuller conception of fallible knowledge: One knows that p even when one might not have done so - indeed, even when one might easily not have done so. For short: one knows luck$ that p. Equivalently: one knows that p even while almost not doing so. A necessary condition of one's understanding Gettier cases without denying that there is fallible knowledge is one's accepting this conception of fallible knowledge. Superficially, it sounds correct to deny that there can be knowledge which is lucky. But that response is infallibilist - and if we are infallibilists, we are conceiving of knowledge in a way that does not really allow Gettier cases to be anything other than a demonstration of how knowledge is absent when fallibility is present. If so, then it is confused for us to seek any fallibilist solutions to 'the Gettier problem'. Accordingly, I take it, the usual infallibilist response, that of denying that one can know luckily, is misguided.
Epistemologists mistakenly regard the standard interpretation of those cases as mandatory, less an interpretation than a datum (so to speak). As $11 showed, that is because they place undue emphasis on various counterfactual truths, each of which has this logical form: I.
If the epistemic subject had done X, then he would have been misled as regards p. (He would have gained a false belief as regards p - either mistakenly accepting or rejectingp, as the case may be.)
If Henry had continued driving and had looked at a fake barn, he would have mistakenly thought that he was seeing a barn. If Jill had heard the announcement, she would have mistakenly thought that there had been no 0T h e Editors of T h e Phalosophica! Quortnb, 1998
464
STEPHEN HETHERINGTON
assassination. And if Smith had reasoned slightly differently, he would have mistakenly thought that Nogot owns a Ford. The standard interpretation then infers from (I) the following conclusion: 2. The epistemic subject in fact lacks knowledge that p (or knowledge that q, if his believing that q would have depended on his believing that p), in the event of his believing that p upon doing X, hence while being misled as regards p. Henry fails to know that he is seeing a barn (we are told).Jill (it is assumed) does not know that there was an assassination. And Smith (it is said) lacks knowledge that someone in the office owns a Ford - where his believing that someone there owns one would have depended on his believing that Nogot owns a Ford (if he had reasoned differently, by forming the latter belief, instead of the existential generalization upon which he did rely). But $11 above explained why, in each of our three paradigms, the relevant instance of the inference from (I) to (2) is invalid. The generalization of those three specific explanations of invalidity is this: The inference from (I) to (2) is invalid because an epistemic subject who has not done X (e.g., who has not been misled by a fake barn) need not lack a specific piece of knowledge just because he would lack it if he were to do X. It is easy to forget this fact, particularly given the alacrity with which some recent epistemologists have tried to bind a person's lacking knowledge to various counterfactual (or, more generally, subjunctive) truths about him. (Robert Nozick is probably the pre-eminent proponent of this trend.') Part of what I am arguing is that such analyses are mistaken. We need an intermediate option, a looser link between knowing and counterfactuals. We need a way to analyse the epistemic 'in-betweenness' of our cases - a way to understand how it could be, in each case, that the epistemic subject does not lack knowledge even though we are sore& tempted to think that he does. The counterfactual failings of Henry, Jill and Smith are evidence for us of their lacking knowledge. But that evidence, being inconclusive (and we should never forget that our judgements, even - or especially - on such matters, are defeasible), is compatible with there also being good reasons to regard Henry, Jill and Smith as having knowledge instead. I have pointed to some such reasons. Thus although each of our epistemic subjects does have knowledge, it would have been easy for any of them not to have done so. All we need add to our conceptual repertoire for discussing these cases is the idea of an See his Philosophical Explanations (Harvard UP, 1981), pp 0The Editors of ne Philosophical Quorierb, 1998
172-227.
ACTUALLY Kh-OWING
465
epistemic subject's not lacking knowledge, yet almost clearb doing so. Henry, Jill and Smith do not lack knowledge but almost do so - indeed, almost clearb do so. This standardly leads epistemologists to assume that each of these epistemic subjects does lack the knowledge in question. But it is no less reasonable to infer that the counterfactual lack of true belief on the part of Henry, Jill and Smith implies only their counterfactual lack of knowledge. T o assume that a counterfactual lack of true belief implies an actual lack of knowledge is to exemplify the epistemic counterfactuals fallacy.
Here is a possible objection to my interpretation: It collapses the distinction between knowledge and true belief by allowing an epistemic subject knowledge that p because he has a true belief that p (and because it is on& counterfactually that he is deceived as regards p). One reason for including a counterfactual element in an analysis of knowing is to prevent our allowing lucky and confident guesses, say, to be knowledge. Is it not badly mistaken, therefore, to remove that counterfactual element from one's analysis of knowing? The worry is misconceived. At most I am letting Henry's, Jdl's and Smith's counterfactually knowing fall with their counterfactual lacks of true belief. And so I should. Because knowledge entails true belief, a belief's being false entails its not being knowledge - and its counterfactually being false entails its counterfactually failing to be knowledge. (But even here I am conceding more to the objection than is strictly warranted. In the existential generalization case, Smith is not counterfactually lacking knowledge that p because of his counterfactually having a false belief that p. He counterfactually lacks knowledge that q because of his counterfactually having a false belief that p and because his belief that q counterfactually depends on his counterfactually falsely believing that p. Here p is 'Nogot owns a Ford', and q is 'Someone in the office owns a Ford'.) Still, I am not allowing that Henry, Jill and Smith have knowledge simply because each has a true belief (and hence is not actually being deceived). I am presenting an interpretation of them as having knowledge - but the interpretation's reason for this is not just that they are not being deceived. Henry, Jill and Smith have knowledge because each has a true belief which is much more than that - a true belief which is well supported by evidence, none of which is false. (This does not leave counterfactuals with no epistemological contributions to make. T o mention but two examples: perhaps they 0The Editors of Thr Philorophicol @ ~ o ~ i n1998 b,
466
STEPHEN HETHERINGTON
help to determine whether an epistemic subject really believes that p, or how well he is using his evidezce. T o say more about these matters is beyond this paper's scope, though.) If you remain unconvinced, here is a simpler supposed Gettier case: Russell looks at a clock, sees that it registers the time as 11.00, and infers the true belief that it is 11.00 a.m. He has often relied on this clock, knowing it to be reliable. Nevertheless, the clock happened to stop at 11.00p.m. the previous night. Does Russell know that it is 11.00 a.m.? No.g If the situation had been the same, but with the world ending immediately afterwards, would Russell have died not knowing what time it was? I think that he would have known. The inclination to think that he would not have known is due, in part, to assessing his actual thinlung's epistemic import in terms of how he would have thought a few minutes later, as he became suspicious of the clock's lack of temporal progress. No such assessment is conceptually mandatory. Hume would remark on a metaphysical independence here; there is also an accompanying epistemic independence.
Each of our three main cases is over twenty years old. Yet the same pattern of epistemological thinking still appears. I11 Linda Zagzebski's virus X/virus Ycase, Dr Jones is aware of the strong correlation between virus X and her patient's symptoms.10She infers that the patient has X. In fact the patient's symptoms are due to an unknown virus. Nevertheless, Xis also present, although it arrived too recently for it to have generated any symptoms. Zagzebski assumes that Dr Jones' diagnosis ('The patient has X'),although justified and true, is not knowledge. In other words, this is meant to be a Gettier case. But the case admits of the same treatment, mutatis mutandis, as Feldman's. Zagzebski's standard interpretation of it relies on a counterfactual like this: If Dr Jones had reasoned slightly differently - in particular, if she had reached the same diagnosis via the premise that the patient's symptoms were caused by X, instead of via the premise that they were strongly correlated with X - she would unwittingly have been misled in her "he example, from an unpublished paper by John Bigelow, is adapted from Bertrand Russell, Hunzan Knowledge: its Scope and Linzits (London: George Allen & Unwin, 1948),pp. 170-1. 'O 'The Inescapability of Gettier Problems', The Philosophical Quarterly, qq (~ggq), p p 65-73, at p. 71. For brevity, I shall s i m p l the ~ case a little. 0The Editors of Tne Phiiorophtcoi ~uorierb,1998
ACTUALLY ImOWING
467
reasoning. She would have been using false evidence, lowering her chance of deriving a true conclusion. And Dr Jones might, so easily, have reasoned in that way. So she would have reached the same conclusion ('The patient has virus X') without realizing that she was being misled along the way. Hence her reaching that conclusion now, also without thinking that she is being misled, does not make her belief knowledge, even though it is true and she is not actually being misled. She might as well have been misled by her reasoning, epistemically speaking. The epistemic result is the same either way. The counterfactual lack of knowledge (if she had used false evidence in her reasoning) implies the actual lack of it (when she uses no false evidence in her reasoning). That is a putative explanation, involving a counterfactual, of why Dr Jones' diagnosis is not knowledge. But it is optional. Here is an alternative interpretation of her diagnosis' epistemic status, an interpretation at least as plausible as the usual one: confronted by her initial evidence, there were different inferential paths Dr Jones could have follo~ledto reach that diagnosis. Some of them would have involved her being misled (using false evidence); the one she chose did not. Like our other epistemic subjects, then, Dr Jones has knowledge - even while almost lacking it. She happened to follo~la non-misleading inferential path (to a true conclusion, using only true evidence), just as Smith did, and just as Henry happened to be reacting only to a real barn, and just as Jill happened to be reacting only to accurate evidence. Like them, Dr Jones might easily have been misled without realizing it. However, in fact she was not - and its being counterfactually true that she would have been, if she had reasoned differently, does not entail that in fact she lacks knowledge. Presumably there is luck in Dr Jones' having follo~ledthe inferential path she did follow. Yet this does not entail that, once she is following that path, it cannot lead to knowledge. At worst, it entails a respect in which she almost failed to know. As Zagzebski emphasizes, Dr Jones has excellent evidence. Somewhat luckily, it led to a correct diagnosis. But, as I observed at the end of $11, the kind of luck involved is that with which - necessarily - fallibilism is compatible. (Zagzebski, too, remarks at pp. 72-3 on the possibility of allowing knowing to involve an element of luck, as well as on the link between fallibilism and the possibility of there being Gettier cases.) As fallibilists, we should accept that reasoning in one way rather than another can make the difference between gaining inferential kkwledge and not doing so - even when reasoners might easily have reasoned, like Dr Jones, in either way. The fact that they might be unaware of how close they were (by following a different inferential path) to not gaining the knowledge is neither 0The Editors of ?he Phrloiophicol Qarlerb, 1998
468
STEPHEN HETHERINGTON
here nor there. Even internalists do not make knowing depend on one's being aware of how close one was to not knowing, or (conversely) of how securely one does know. Nor will many epistemologists say that knowing that one knows is required if one is to know. My interpretation allows Dr Jones knowledge of the patient's having X, without implying that she knows her diagnosis to be knowledge. She can know even while almost not having known - regardless of whether she knows that she almost failed to know. A person's being in that situation makes it easy for an onlooker, even an epistemological one, to be mistaken in believing that the person is not in that situation. More precisely, it is easy to mistake knowing that p while almost not knowing that p for not knowing that p. There is luck involved in Dr Jones' reaching the correct diagnosis by wholly correct reasoning; and we (when examining her situation from 'outside'), unlike her, are aware of this. So it is natural for us to deny her knowledge. But I have argued that we may interpret the case, no less coherently, as one where there is - but almost is not knowledge. Again, Dr Jones knows - even though she might easily not have done so, and even though she would not have been aware of not doing so. Does all of this reveal a respect in which knowing is at least somewhat analogous to abiding by a convention? There might be luck, some arbitrariness, in Dr Jones' initial choice of a line of reasoning. But once this choice has been made, knowledge can result (just as an action can be conventionally correct, in spite of some arbitrariness in the original formation of the convention). In contrast, the usual epistemological interpretation seems to treat knowing as unable to withstand such arbitrary beginnings or foundations - so that if any luck was needed to reach one's true belief, the belief fails to be knowledge. As we saw at the end of $11, though, the usual interpretation is really a kind of infallibilism - not allowing that one knows that p, if it would have been so easy for one not to know that p. If we are fallibilists, we should accept that one can be lucky to know that p - for example, following a particular path of' reasoning, even if it is a less reflective one. Knowing has at least this muchfragili& - contingency. There can be close possible worlds where a person's knowledge that p disappears (even in circumstances in which it might have become more secure instead), without its therefore disappearing from this world. But a fallibilist should allow that knowledge can be fragde, or contingent, or lucky, in that way.
We have found that supposed Gettier cases whose epistemic subjects use no false evidence need not be treated as Gettier cases after all. In order to O The Editors of TiiePiitlorophicol Quarierb, 1998
ACTUALLY KNOWIKG
469
interpret them as cases where the epistemic subjects lack knowledge, we need to call on putatively explanatory counterfactuals. But those counterfactuals do not entail the knowledge-denials they are routinely assumed to ground. So, in so far as those counterfactuals are needed to ground those denials, we have some logical licence to reject the denials - to interpret the epistemic subjects as having the knowledge they are standardly assumed to lack. Part of that interpretation, of course, needs to explain why so many of us are nevertheless so tempted to deny those epistemic subjects knowledge. My claim is that we are misled by the fact that in each case the epistemic subject almost - indeed, almost clearly - lacks knowledge. We are insufficiently attuned to the phenomenon of a person's almost lacking knowledge. There is logical space, therefore, to conclude that thesis (T) is true (see p. 453 above). That is, we may coherently claim - even in the face of these cases - that it is sufficient, for one's well supported true belief's being knowledge, that one's support for it contain nothing false. Epistemologists have embraced too speedily, too willingly, a denial of that truth. Doing so has prevented their knowing what knowledge is."
University OfNew South Wales
" Thanks to anonymous referees for The Philosophical Quarter4 for very helpful criticisms of earlier versions of this paper. I presented an even earlier version at hfonash University, and I am grateful to those who contributed to the stimulating discussion on that occasion.
O The Editan of 7he Phtlorophicol Qunftah. ,998