THE MOOREAN ARGUMENT

To show that facts about justification are accessible, Smithies defends four theses having to do with propositional justification:

If these theses are correct, we do in fact have special access to facts about justification. With the justifications we have for our beliefs about the external world, it seems there is a gap between the best justifications we have to believe the facts are a certain way and the way those facts are. It also seems there is a gap between the facts as they are and our having justifications to believe they are that way. These theses, if correct, show that there is no similar sort of gap in the case of epistemic justification. Familiar externalist views (e.g., process reliabilism) imply that there can be a gap between having justification to believe one has a justification to believe p and having that justification and would certainly imply that one can have a justification to believe p without thereby having a justification to believe one has it. Thus, if an argument can be given that establishes that these theses are correct, it would effectively settle the question as to whether any of the familiar externalist views in the literature are adequate.

In support of this package of theses, Smithies offers the Moorean argument:

I shall argue that the Moorean argument for access internalism is unsound.

An important background assumption in this discussion will be a uniqueness thesis. Smithies maintains that for any proposition one will have justification to take one and only one attitude (i.e., belief, disbelief, or withholding) towards it at a time.Footnote ⁵ According to the uniqueness thesis, if one has justification to believe p, that excludes having justification to believe ~p and excludes having justification to withhold on whether p.Footnote ⁶ The uniqueness thesis plays an important role in the argument for access internalism. If there are counterexamples to the infallibility theses or self-intimation theses, the uniqueness thesis implies that one would have to have justification to have combinations of attitudes other than those sanctioned by his access internalist view. As we shall see, the relevant attitudes are not rationally co-tenable. Thus, if we accept MA3, we would have to concede that it would be impossible to have justification to have these attitudes. Since, however, one must have justification to have some set of attitudes or others, it turns out that the only combinations of justifications one could have are the combinations that the infallibility and self-intimation theses suggest.

While I take MA1 to be unproblematic, MA2 and MA3 deserve comment. It takes a bit of work to see why one should accept MA2. To determine whether MA2 is true, we have to think about which justifications one would have to have or lack to falsify one of the infallibility or self-intimation theses and then consider whether one could rationally have the attitudes one would have justification to have if one or more of these theses were incorrect.

Consider the positive infallibility thesis, PI. If PI is false, it is possible to have justification to believe that one has justification to believe p even if one does not have justification to believe p. I take it that PI is not the weak claim that one has some justification to believe p whenever one has justification to believe that one has justification to believe p, but is instead the stronger claim that one would have to have adequate justification to believe p. Smithies observes that if PI is false, one would have to have justification to have the attitudes expressed by one of the following:

1. 1. I have justification to believe p, but ~p.

2. 2. I have justification to believe p, but it is an open question whether p.

Smithies assumes, quite plausibly, that if one believes neither p nor ~p, the attitude one has can be expressed by saying that it is an open question whether p and also that one must always have justification to believe p, disbelieve p, or to neither believe nor disbelieve p. Thus, if PI is false, we must have justification to have the attitudes expressed by (1) or (2). If we grant that the attitudes expressed by (1) and (2) are not rationally co-tenable, MA3 says that it is impossible for a person at a time to have the justifications to have the attitudes expressed by (1) or (2). Thus, one would have to have justification to believe p after all. That's just what PI says.

We can run through the other theses, too, but I should note that I agree with Smithies that the attitudes expressed by the following all seem irrational:

1. 3. p and I do not have justification to believe p.

2. 4. p and it is an open question whether I have justification to believe p.

3. 5. I have justification to believe p, but ~p.

4. 6. I have justification to believe p, but it is an open question whether p.

5. 7. It is an open question whether I have justification to believe p, but ~p.

6. 8. It is an open question whether I have justification to believe p, and it is an open question whether p.

If the attitudes expressed by (3) are irrational, MA3 says that if one has justification to believe that one lacks justification to believe p, one lacks justification to believe p. NI would be vindicated. If the attitudes expressed by (3) and (4) are irrational, MA3 says that if one has justification to believe p, one has justification to believe that one has justification to believe p. Thus, PSI would be vindicated. Finally, MA3 says that if NSI is false, one must have justification to have the attitudes expressed by (5), (6), (7), or (8). If, however, Smithies is right that one cannot rationally have those attitudes, MA3 suggests that NSI must be true.

To understand Smithies’ argument, it is important to understand the difference between propositional justification and doxastic justification. The former has to do with what one has justification to believe whether or not one believes it. The latter has to do with beliefs that are justifiably held. On a standard picture of how these notions are related, doxastic justification requires propositional justification plus proper basing. To have justification to believe p, one might need evidence that lends adequate support to believe p but one need not have any view on whether p. To justifiably believe p, however, one must have a belief about p, have justification to have that belief, and to believe for good reason. This much seems uncontroversial. If you would be irrational to believe p whilst believing q, it would be impossible to justifiably believe p without believing q. Thus, in granting that the combinations of attitudes are irrational, we would have to grant that it would be impossible to justifiably hold the attitudes expressed by (1)–(8). The crucial question is whether this has to do with propositional justification.

If the infallibility and self-intimation theses are correct, the reason that it is impossible to justifiably believe the attitudes expressed by (1)–(8) is that it is impossible for a subject at a time to have justification to have these attitudes. One way to resist the Moorean argument for access internalism is to grant that the attitudes expressed by (1)–(8) are irrational but insist that this fact is not best explained by appeal to the self-intimation or infallibility theses. The natural way to do this is to argue that there is something about holding the attitudes themselves that prevents one from properly using the justifications that one has to justify one's beliefs. If, say, rationally believing p requires having a belief that's based on a good reason to believe p and having further attitudes would prevent one from basing one's belief on a good reason, we could explain why someone doesn't rationally believe p in spite of the fact that she has good reason to believe p. In the concluding section of the paper, I shall explain the relevant data (i.e., that it would be irrational to have the attitudes expressed by (1)–(8)) without appealing to the infallibility and self-intimation theses. If successful, this shows that the data are interesting but might not be of much use when it comes to trying to settle the internalism/externalism debate. Before I do that, I shall explain why we need an alternative explanation of the data. We shouldn't try to explain the data by appeal to the infallibility and self-intimation theses because they are subject to powerful objections.

The problematic premise in the Moorean argument is MA3. I think that it's possible to have justification to adopt attitudes that are not rationally co-tenable. Smithies disagrees. He maintains that the rationality of certain combinations of commitment should be explained in terms of facts about the structure of propositional justification.Footnote ⁷ To show that this is not the right order of explanation, it will help to show that the claims about the structure of propositional justification he puts forward to try to account for the Moorean absurdities are mistaken.

ON POSITIVE INFALLIBILITY AND NEGATIVE SELF-INTIMATION

According to PI, one cannot fail to have justification to believe p provided that one has justification to believe that one has it. While justification is not typically taken to be factive, PI says there are exceptions to the general rule. One can have justification to believe there is a bear where none is to be found, but not to believe there is a justification to believe where none is found.

One reason to be skeptical about PI is that it seems there can be rational disagreements about the justificatory significance of certain facts.Footnote ⁸ Suppose there is disagreement about the justificatory significance of some fact (e.g., a disagreement about whether it is rational for a party to a disagreement with a peer to be steadfast or about whether one can rationally believe lottery propositions without insider's information). The disagreement would be between one party who thought that someone had adequate justification to believe p under such and such conditions and another who thought that someone could not have adequate justification to believe p under those conditions. Might there be disagreements such as this where the parties to the disagreement are rational or are justified in their epistemological beliefs?

The kind of disagreement we should focus on is a bare disagreement, one in which two subjects have incompatible attitudes about the issue at hand but do not know that there is a party who disagrees with them. This way, we can avoid having to take any stand on what to say about cases of mutually acknowledged disagreement between peers. Suppose the disagreement concerns the evidentialist view that states that one has justification to believe p iff p fits one's evidence. (Let's assume that p can only fit one's evidence if the evidence provides an adequate degree of support for p.) What might justify a belief in this evidentialist view? According to the evidentialist, evidence would justify that belief. Conee and Feldman suggest that the justification we have to believe philosophical views often comes from reflection on thought experiments and hypothetical judgments about them. Such judgments, they claim, provide a kind of inductive support for general principles much in the way that observations provide support for hypotheses about laws.Footnote ⁹ Just as the justification one has to believe something is a law will depend upon whether one has an adequate body of observational evidence and an absence of defeating evidence, the justification one has to believe the evidentialist view will depend upon the intuitive reactions one has to thought experiments and hypothetical cases and whether one has defeating evidence. Of course, whether one has this defeating evidence will depend (in part) upon the way that one reacts to the situations or thought experiments one confronts. Just as it should not be hard to imagine someone who considers an ample range of cases having reactions that would make it rational to embrace the evidentialist view, it should not be hard to imagine someone who considers an ample range of cases that would make it rational to embrace an alternative view. Consider a rival pragmatist view, one that says that it is sometimes permissible to believe p without evidence the evidentialist would take to be adequate (e.g., when the evidence supports p and ~p equally well, the subject knows that there is no possibility of uncovering further evidence, and the subject knows that believing p would further the subject's ends). While I think the view is mistaken, I cannot see how I can say, in keeping with the evidentialist view, that it's impossible for someone to rationally embrace such a view. To do that, I would have to deny that it would be possible for someone to have adequate evidence to believe the pragmatist view and yet the evidence seems to indicate that it often possible for philosophers to have adequate evidence for incompatible philosophical views.

If we grant that William can have sufficient justification to believe his pragmatist view, it seems that William might have sufficient justification to believe that someone has justification to believe p even when this person does not have evidence that supports p on balance. By hypothesis, William is wrong about this. We're assuming that one cannot justifiably believe p unless one's evidence supports p on balance. It seems that if the evidentialist view is correct about what it takes to be rational to believe, we have a counterexample to PI.Footnote ¹⁰ William has justification to believe that he has justification to believe p because he has adequate evidence to accept a philosophical view that entails that he has justification to believe p. William does not, however, have adequate justification to believe p because he does not have adequate evidence to believe p.Footnote ¹¹

This sort of case also seems to be a counterexample to NSI. By hypothesis, William lacks justification to believe p. According to NSI, he would have to have adequate justification to believe that he lacks this justification. He could only have adequate justification to believe that he lacks adequate justification to believe p if he had adequate evidence to support the belief that he lacks adequate justification to believe p. William has no such evidence. Indeed, all the evidence in his possession indicates (misleadingly) that he has justification to believe that he has justification to believe p. According to MA3, it would be impossible for William to have (i) justification to believe that he has justification to believe p and (ii) justification to believe that he lacks justification to believe p. These attitudes are not rationally co-tenable:

1. 9. While I have justification to believe p, I lack justification to believe p.

Thus, if it is possible to have sufficient evidence to adopt a mistaken view about justification or the requirements of rationality, we must reject both PI and NSI.

What would it take to block this sort of counterexample? The possibility of a counterexample seems to follow from these four rather mild assumptions. The first is the evidentialist assumption that a subject's evidence and the degree of support it provides determines whether one has justification to believe propositions about what it takes to have adequate justification (‘epistemic propositions’ hereafter). The second is the anti-skeptical assumption that one can have sufficient justification to believe epistemic propositions. The third is that the sufficiency of the justification one has for believing epistemic propositions does not require having entailing evidence to believe these epistemic propositions. The fourth is that it is possible for different subjects to have justification to believe incompatible epistemic propositions with an adequate degree of evidential support.

To deny the fourth assumption, it seems one must adopt the bizarre view that there is a limit on the degree of support for believing an epistemic proposition depending upon the truth or falsity of that proposition. It would be rather ad hoc to appeal to the idea that no matter how much evidence one had to believe a false epistemic proposition the degree of support it provided could never be a sufficient degree of support to provide a sufficient justification to believe that epistemic proposition in order to defend PI or NSI from objections. It seems that the second and third assumptions should stand or fall together since it seems rather plausible that many of us have justification to believe epistemic propositions only if the broadly inductive support from intuitions about cases can be adequate. To block this counterexample, it seems that Smithies would have to reject the second assumption or the first. Once we grant the fallibilist assumption, it is hard to see what principled reason there could be to insist that the fallible justification we have could never be enough to have justification to believe an epistemic proposition. That suggests that the only way for Smithies to block the counterexample is to deny the evidentialist thought that justification depends upon whether a belief fits the evidence, but not whether it also fits the facts. Since, however, it seems rather plausible that good evidential fit should be sufficient for justification, it seems that we have good reason to reject both PI and NSI.Footnote ¹²

ON POSITIVE SELF-INTIMATION

According to the positive self-intimation thesis (PSI), if one does have justification to believe p, one has justification to believe that one has this justification. To motivate PSI, Smithies appeals to a kind of Moorean test. It would be irrational to have the attitudes expressed by (1) and (2), so MA3 suggests that one cannot have justification to have these attitudes. By elimination, one can only have justification to believe p if one has justification to believe that one has it.

The Moorean test is unreliable. Once we see this, we can see that PSI is unmotivated. To show that the Moorean test is unreliable, I shall show that it supports theses about justification that generate a contradiction.

Consider the following:

To show that the Moorean test offered in support of PSI supports JK, consider what attitudes one would have justification to have if it were false. One would have to have justification to have this combination of attitudes:

1. 9. p, but I do not know that p.

2. 10. p, but it is an open question whether I know that p.

The attitudes expressed by (9) and (10) are irrational, so MA3 implies that one cannot simultaneously have justification to have these attitudes.Footnote ¹³ By elimination, one can only have justification to believe p if one has justification to believe that one is in a position to know p.

The trouble with JK is that it commits us to the following thesis:

Suppose that one has justification to believe p. According to JK, one has justification to believe that one is in a position to know that p. Suppose that one does indeed have justification to believe this (i.e., that one has justification to believe that one is in a position to know that p). According to JK, one has justification to believe that one is in a position to know that one is in a position to know that p. And repeat.

According to a weak version of the KK principle, one can only know p if one is in a position to know that one is in a position to know p. This KK principle is false.Footnote ¹⁴ We can use cases of inexact knowledge to show that this is so. We can know that the page one is reading is within a certain distance from one's face without knowing the distance. If the page one is reading is a foot from one's face, one can know that the page one is reading is between ten and fourteen inches from one's face without being in a position to know that it is twelve inches from one's face. One can only know that the page is not one inch from one's face. If one knows that the page is one inch from one's face, KK says that one can know that one knows that the page is not one inch from one's face. One knows that if the page is just one and a quarter inches from one's face, one could not know that the page was just one inch from one's face. (One knows that the margin of error would be too slim to know that the page was an inch from one's face if that was where the page was.) Thus, if knowledge is closed under known entailment, one can knowingly deduce that the page is not one and a quarter inches from one's face. The reasoning can be repeated to yield the conclusion that one knows that the page is not twelve inches from one's face. This, however, contradicts the assumption that this is just where the page is.

Suppose a subject knows that KK is false and that it is falsified by cases of inexact knowledge. Let us suppose that our subject knows p and knows that she is not in a position to have arbitrarily high orders of higher-order knowledge that p (e.g., she knows that the page is between eight and twenty inches from her face and knows that in these cases of inexact knowledge is where we can find counterexamples to KK). Because she knows p, she has justification to believe p. Because she knows that she is not in a position to have arbitrarily high orders of higher-order knowledge that p, she does not have justification to believe that she is in a position to have arbitrarily high orders of higher-order knowledge that p. It follows from JK+ that she has no justification to believe p. This contradicts the assumption that she knows and has justification to believe p. Thus, once we reject KK, we have to reject JK+. Since JK+ is a consequence of JK, we have to reject JK. Since the Moorean test says that JK is correct, this test is unreliable. Combined with the known failure of KK, the Moorean test generates contradictions.

ON SELF-INTIMATION

The self-intimation theses assert that the presence or absence of justification is self-intimating. According to PSI, one can only have justification to believe p if one also has justification to believe that one has justification to believe p. According to NSI, one can only lack justification to believe p if one happens to have justification to believe that one lacks it. The Moorean test that Smithies appeals to in order to motivate PSI and NSI would seem to suggest that one cannot have justification to have the attitudes expressed by this variant on (3):

1. 10. p, but I have no more/better reason to believe p than ~p.

This seems to be a variant of (3) as (3) has to do with a justification one has and (10) has to do with a reason that one has. It seems that one can have a reason to believe p and better reason to believe p than its negation without thereby having any justification to believe p. The reason might be too weak. One might have views on which one can have a justification to believe p where that is understood as having a right, entitlement, or permission without having any reason to believe p, so I would not want to assume that the impossibility of having justification to have the attitudes expressed by ‘p and I do not have justification to believe p’ established that it would be impossible to have justification to have the attitudes expressed by (10). It seems that we should be able to use the Moorean test directly to determine whether it would be possible to have justification to have the attitudes expressed by (10) and the test indicates that one cannot.

Can one have justification to believe that one has better reason to believe p than its negation without having any reason to believe p? I suspect that this is possible. While I think it is quite plausible that whenever one has justification to believe p, one has some reason or other to believe p, I do not think that whenever one has justification to believe that one has justification to believe p one must have justification to believe p. Suppose, however, that I am wrong about this and right in suggesting that whenever one has justification to believe p, one has some reason to believe p. Let us further suppose that this reason consists of a piece of evidence that supports p. This assumption, when combined with NSI and PSI suggests that one cannot have a justification to believe p unless one has some evidence that one has this justification and that one cannot lack a justification to believe p unless one has some evidence that one lacks this justification.

Now, consider the vexed case of the lottery proposition. I don't know whether it is possible to know that one will lose without insider's information. I don't know whether it is possible to justifiably believe that one will lose without insider's information. There are important epistemic facts that I'm ignorant of and it seems that my ignorance might be quite widespread. Let ‘s’ be the proposition that I would have sufficient justification to believe a lottery proposition on the basis of the probabilistic grounds I have. Let ‘l’ be the proposition that the ticket I hold will lose. I honestly cannot think of any reason to accept or reject s. It seems both plausible and (more importantly) possible that I might have no reason at all to believe s and no reason at all to believe ~s. If I might speak on behalf of those who suffer in a similar state of ignorance, I think we should suspend judgment on l and on s.

On the supposition that one can only have justification to believe something if one has better reason to believe it than its negation (J → R), it seems we have a powerful objection to the idea that facts about justification are self-intimating. Suppose, as seems quite possible, I do not have better reason to believe s than ~s and do not have better reason to believe ~s than s.Footnote ¹⁵ Either s is true or ~s is true. If ~s, it seems I must lack justification to believe l. If so, NSI implies that I have justification to believe that I lack justification to believe l. If so, J → R implies that I have better reason to believe ~s than s. This contradicts the assumption that I don't have better reason to believe ~s than s or to believe s than ~s.Footnote ¹⁶ On the other hand, if s, it seems I must have justification to believe l. If so, PSI implies that I have justification to believe that I have justification to believe l. If so, J → R implies that I have better reason to believe s than ~s. This contradicts the assumption that I don't have better reason to believe ~s than s or to believe s than ~s.

Insofar as it seems possible that one might have no better reason to believe ~s than s or to believe s than ~s and it seems quite plausible that one always has better reason to believe a proposition than its negation when one does indeed have justification to believe that proposition, it seems that justification is not self intimating in the way that PSI and NSI suggest. That one lacks justification to believe something (e.g., l?) might be something that one cannot properly take any view about because one lacks the sorts of reasons that might rationally support a view about that matter. The lack of a justification doesn't always draw attention to itself. Moreover, it should be possible for one to have a justification to believe something (e.g., l?) even if one lacks the sorts of reasons one would have to be able to bring to bear to have any justification to believe that one's justification for believing is adequate.

ACCOUNTING FOR IRRATIONALITY

In light of the objections to PI, PSI, and NSI, we have good reason to doubt that the irrationality of having the attitudes expressed by (1)–(8) is down to the fact that it would be impossible to have propositional justification to have these attitudes. Thus, I think we have good reason to doubt Smithies’ suggestion that the irrationality of the relevant combinations of attitudes is “best explained in terms of more fundamental facts about the nature of propositional justification”.Footnote ¹⁷ We should seek an alternative explanation.

Any alternative explanation would need to explain why certain combinations of attitude are not rationally co-tenable without assuming that it would be impossible for someone to have justification for those attitudes. A natural suggestion is that the attitudes themselves can sometimes help to explain the irrationality of the combinations. Consider the case of intention. Once one takes out a bowl to make the kids a snack, one might have all the justification one needs to fill the bowl with ice cream or to fill the bowl with pickles. Although one has a justification to intend to fill it with ice cream and a justification to fill it with pickles, it would be irrational to intend to serve the kids pickles in the bowl while intending to serve them ice cream in the bowl. The irrationality of the combination of intentions is not down to the impossibility of having sufficient justification to have these intentions. It only becomes irrational to form one of these intentions when the other intention has been formed and is not being revised. Similarly, one possibility to consider is that the irrationality of a combination of beliefs is not always down to the impossibility of having sufficient justification to have these beliefs. Perhaps the irrationality of adding a belief to one's belief set is down to the fact that one has an attitude such that the subject has justification for both attitudes.

There seems to be two ways for an agent's attitudes to contribute to the irrationality of the agent's attitudes. If some subject's beliefs are rationally held, the subject must have adequate reason to hold the attitude and the attitude must be held on the basis of this reason. The subject's attitudes might defeat the justification provided by a reason that would otherwise be an adequate reason. It could do this by undermining the support the reasons would provide or by proving some overriding reason that conflicts with the justification. Alternatively, it might prevent the subject from properly basing her belief on such a reason.Footnote ¹⁸ If the attitude itself plays one of these roles then holding the attitude in place could make it irrational to add certain beliefs to one's belief set even if one would have adequate justification for the relevant belief if only one did not have the attitude.

In anticipation that someone might offer an alternative approach to understanding the irrationality of having the attitudes associated with (1)–(8), Smithies suggests that the natural line to take is to treat a subject's attitudes as defeaters. He thinks that even if a subject's attitudes constituted a kind of defeater, these attitudes would constitute the wrong kind of defeater to do the relevant explanatory work. He draws a distinction between doxastic defeaters and propositional defeaters. Whereas the former defeat doxastic justification, the latter defeat propositional justification. He seems to think that a doxastic defeater will (typically?) not double as a propositional defeater. In having doubts about the adequacy of a justification that is, let's assume, perfectly adequate to justify believing p, he concedes that it might be unreasonable to believe p whilst one harbors these doubts. Under these conditions, the doubts might function as a doxastic defeater, but he doubts that they would function as a propositional defeater. Why is that? He denies that they would function as propositional defeaters because, “the most reasonable course of action would be to retain one's belief that p, while abandoning one's conflicting higher-order attitude”.Footnote ¹⁹

The claim that an attitude that might constitute a doxastic defeater won't double as a propositional defeater rests on a claim about what the most reasonable course of action would be when one has this defeater. It's not obvious that Smithies is right about what's most reasonable to do in the relevant circumstances.Footnote ²⁰ Let's revisit a kind of case that's alleged to cause trouble for PI and NSI. Let's suppose that one learns that a friend's only reason to believe that the ticket she's about to toss into the trash is worthless is that the probability the ticket will lose is so low. In considering the case at hand, you start to realize that you have serious doubts about whether these probabilistic grounds constitute a good enough reason to believe outright that the ticket will lose. You suspect that suspension of judgment might be in order. In light of such doubts, I would think that if you persisted in the belief that the ticket will indeed lose, your present combination of belief, doubt, and suspension is not ideal. Even if the grounds you had were sufficient to justify your belief, the doubts that you know entertain seem to make suspension of judgment and a reassessment of the situation a perfectly reasonable one. In acknowledging that the suspension of judgment is itself perfectly reasonable in light of the doubts now entertained, it does not seem like we're taking a stand on whether there is sufficient grounds for believing lottery propositions. There might be, there might not be, but whether there are or not, it seems that the doubts make the suspension reasonable. These doubts are not hyperbolic and they are not pathological. What seems unreasonable to do is simply dismiss the doubts out of hand for no apparent reason and retain belief or form a belief. That seems like an unreasonable move even if the grounds had initially were themselves perfectly adequate.Footnote ²¹

I see no good reason to think that the attitudes that constitute perfectly good doxastic defeaters cannot also constitute perfectly good propositional defeaters. The suggestion to the contrary is based on a questionable claim about what's most reasonable to do in light of ‘conflicting’ higher-order attitudes. It is not easy to determine whether the attitudes themselves might do double duty as doxastic and propositional defeaters or whether they might only threaten to defeat doxastic justification, so the reader might rightly be skeptical of my suggestion that these attitudes can defeat propositional justification. It is worth repeating that we have already examined various principles having to do with propositional justification and found that they are deeply problematic. If our choice is between one view that says that higher order attitudes function as propositional defeaters and another that faces the objections discussed above, it would be reasonable to choose the first view rather than the second.

We can explain the irrationality of the attitudes expressed by (1)–(8) without having to appeal to the self-intimation or infallibility theses.Footnote ²² Recall that if PI is mistaken, it must be possible to have propositional justification for the attitudes expressed by:

1. 1. I have justification to believe p, but ~p.

2. 2. I have justification to believe p, but it is an open question whether p.

For reasons discussed above, I think it is possible to have justification to have this combination of attitudes. Thus, I must explain the irrationality of these attitudes without appeal to PI. Because one has the attitude expressed by the first conjunct in (1), one takes it that one has a justification that settles the question whether p. (Doubts about whether p are, in part, doubts about whether the reasons one has to believe p settle the question correctly and these are doubts about the adequacy of the justification.) In taking it to be so, one takes it that if ~p, one is mistaken about p.Footnote ²³ Because one has the attitude expressed in the second conjunct, one takes it that one would be mistaken if p. In combination, one would take it that one would be mistaken about p whether p or ~p. This is why the combination is absurd. The attitude expressed by the first conjunct of (2) commits one to p. The attitude expressed by the second conjunct is that of withholding. If one withholds on whether p, one cannot be mistaken about p whether p or ~p. Thus, someone with the attitudes expressed by (2) takes it that they cannot be mistaken about whether p if ~p and is mistaken about whether p if ~p. This combination is absurd.

On the approach that Smithies prefers, the irrationality of having the attitudes expressed by (1) and (2) has to do with the fact that one cannot have justification to have these attitudes. I've already raised doubts about whether it is impossible to have justification for these attitudes. There is a further question to consider. Is the lack of justification really the right thing to explain the irrationality of the attitudes? There seem to be cases in which two subjects respond differently to a single body of evidence (e.g., the testimony of a speaker or of the senses) where (i) it seems that the justification these subjects has is adequate for believing but (ii) these subjects’ responses are both perfectly rational. For example, having heard the testimony of a witness that is sufficient to justify belief, we might have one subject judge ‘She told me p, so it must be p’ and another that judges ‘She told me p, but it still is an open question whether p’. If both responses to the single body of evidence are rationally acceptable, it seems the irrationality of the relevant combination of attitudes is not due solely to the fact that one cannot have justification for these attitudes. If Smithies is right, one of these subjects either believes without adequate justification or suspends without adequate justification. Thus, he would have to explain why these failures don't constitute the kind of irrationality that he chalks up to believing without justification.Footnote ²⁴

On the alternative approach I prefer, no such explanation is needed because facts about propositional justification do not shoulder the explanatory burden. The subject's attitudes play a crucial role in accounting for the irrationality of those attitudes. What makes it irrational to have the attitudes expressed by (1) and (2) is not the mere lack of justification to have them but the appearance of a lack of a justification to have them. If one lacks justification to believe p but does not appreciate that this is so, one is not irrational in the way that one would be if one lacked justification to believe p and believed while acknowledging that one lacked this justification. Why should these attitudes play this role?

A natural way of understanding the notion of rationality is in terms of what makes sense given the subject's perspective.Footnote ²⁵ It might be rational to serve your guests a petrol and tonic. It would make sense to serve them petrol if the bottle believed to contain gin actually contains petrol and one's intention is to serve them gin and tonic. If a subject aims to, say, believe only what she knows, the attitudes it would be rational for her to have are the attitudes it would make sense for her to have in light of this aim and her perspective.Footnote ²⁶ If two subjects happen to have similar evidence for p but only one suffers from doubts about the adequacy of that evidence, I think it might make good sense for one of these subjects to believe p and for the other to suspend judgment. It would not make sense for someone who aimed to believe only what she knew to believe in spite of these doubts for she would essentially be willing to risk violating the knowledge-standard for no good reason. Likewise, it would not make sense for the subject who harbored no doubts to refrain from believing in light of what would strike many of us as being adequate grounds.

Recall that the motivation for NI was that one cannot rationally adopt the attitudes expressed by the following:

1. 3. p and I do not have justification to believe p.

I think it is possible to have justification to have these attitudes (i.e., believing p and believing that one lacks justification to believe p), but not possible to rationally hold these attitudes simultaneously. It is not obvious, for example, that anyone whosoever has adequate justification to believe that peer-disagreement defeats justification for believing p thereby lacks justification to believe p in a case of peer disagreement. Maybe if the conciliatory epistemologists are wrong about the significance of acknowledged peer-disagreement, their original justification remains perfectly adequate. Just as mistakes about the law do not inculpate, perhaps mistakes about justification based on good evidence do not automatically defeat the evidential support one has for one's first-order belief. Why, then, aren't the attitudes expressed by (3) rationally co-tenable? In taking oneself to lack justification to believe p, one takes oneself to be in no position to know that p. Thus, it would not make sense for someone with this view about one's own epistemic position to believe p.

The motivation for PSI was that the attitudes expressed by (3) and (4) are not rationally co-tenable:

1. 4. p and it is an open question whether I have justification to believe p.

We have already addressed (3). As for (4), Smithies chalks up the irrationality to the fact that it doesn't make sense to believe p while withholding on whether one has justification to so believe. I agree. If one's aim is to believe only what one knows, say, it would not make sense to believe p while acknowledging that it is an open question whether one has justification to so believe as one would acknowledge that the facts that determine whether believing p fulfills one's aim are beyond one's ken. It might make sense to risk believing what one does not know if one had some further aim, but it seems there is no further aim to give us overriding reason to adopt the risk. This seems to be sufficient to explain the irrationality of the attitudes expressed by (4). We have seen already why we wouldn't want to explain (4) by appeal to PSI.

As for the motivation for NSI, it has to do with the irrationality of having the attitudes expressed by the following:

1. 5. I have justification to believe p, but ~p.

2. 6. I have justification to believe p, but it is an open question whether p.

3. 7. It is an open question whether I have justification to believe p, but ~p.

4. 8. It is an open question whether I have justification to believe p, and it is an open question whether p.

With (5) it seems that someone who takes herself to have justification to believe p takes herself to be committed to p. It does not make sense for someone to be committed to both p and ~p. Similarly, it does not make sense for someone to take herself to be committed to p and to lack such a commitment. Thus, it seems that the attitudes expressed by (6) should be irrational. As for (7), the subject seems to take herself to be committed to ~p. If the commitment made sense, she would take herself to know ~p. (Remember that “p but it is an open question whether I know p” does not express a rational combination of attitudes.) In taking herself to know this, it would not make sense from her perspective how she could have adequate justification to believe p. That question should be closed. As for (8), I think Smithies is right that it is irrational to withhold whilst one takes it to be an open question as to whether one has justification to do so.

One way to account for the irrationality of the relevant attitudes is to bracket contentious claims about propositional justification and focus on what it makes sense for a subject to believe given the aim or the goal of believing only what one would thereby know. Smithies challenges this approach on the grounds that it cannot explain why the attitudes expressed by these are irrational:

1. 11. p and it is an open question whether I know p.

2. 12. I know p and it is an open question whether p.Footnote ²⁷

He thinks that these cases are problematic because he seems to think that those who would appeal to a knowledge norm or goal would explain the irrationality of these in terms of whether it would be possible to know both conjuncts. He thinks such an approach cannot work.

It is surprising that he thinks that this approach would not work for (12). It is an open question whether p iff one neither believes p nor disbelieves p. If one knows that one knows p, one knows p and so believes p. If one knows that it is an open question for one whether p, one knows that one does not believe p. One cannot know both that one believes and that one does not believe p.

As for (11), one who believes (11) believes that it is an open question whether one knows p or not. By the subject's own lights, it would not make sense for them to believe p given the aim believing only what one knows. Because the subject does believe p, the subject's own commitments will not make sense to her. Thus, the commitments are not rationally co-tenable.