Oppy’s case

In this section, I expound on Oppy’s views: first, about his misgivings with arguments or “derivations”; second, about his proposals concerning worldview elucidation and evaluation; and third, about his conditions for epistemic justification of philosophical beliefs.

Against derivations

Oppy’s paper “What derivations cannot do” (2015) starts with the following words:

Here is one standard format for a paper in philosophy of religion. (1) Provide some introductory remarks that form the background to the subsequent discussion. (2) State an argument – either with the conclusion that God exists, or the conclusion that God does not exist – in standard form. (3) Give a derivation which establishes that the conclusion of the argument is appropriately related to the premises. (4) Defend the premises of the argument, making more or less frequent appeal to the fact that the premises seem reasonable to the author. (5) Respond to objections, including, in particular, the charge that your argument is question-begging. (6) Conclude that your argument is a good or successful argument for its conclusion.

No paper that conforms to this format should ever see the light of day. (Oppy, 2015, 327)

Something must be very wrong with these arguments. What is it? The problem, Oppy argues, is that there is a mismatch between the way arguments work and what they are supposed to achieve. The goal of arguments, for Oppy, is “bring[ing] about reasonable belief revision”:

a good argument is one that succeeds – or perhaps would or ought to succeed – in bringing about reasonable belief revision in reasonable targets. The most successful argument would be one that succeeds – or perhaps would or ought to succeed – in persuading any reasonable person to accept its conclusion; good, but less successful arguments would be ones that succeed – or perhaps would or ought to succeed – in persuading a non-0% of reasonable people to accept their conclusions. (Oppy, 2006, 10)

But the evaluation of arguments always happens against the background of one’s other beliefs (Oppy, 2006, 8–10), which is the source of Oppy’s main problem with them. Distinguish two kinds of arguments: deductive and non-deductive. Deductive arguments consist of a series of premises, from which a conclusion can be derived: p1, p2, … pn ⊨ c. With Oppy, I will sometimes call deductive arguments “derivations.” Non-deductive arguments can also be seen as having a series of premises, but those, rather than entailing the conclusion, render it probable or reasonably acceptable: “it is reasonable – perhaps even most reasonable – to accept the conclusion of the argument on the basis of the premises” (2006, 10). We set aside non-deductive arguments for now, though it seems that Oppy’s analysis applies to them equally.

The goal of an argument for c then is to get the person who antecedently rejects c reasonably to change her mind. Here Oppy introduces a dilemma: does she antecedently accept all of the argument’s premises? If the answer is negative, then she will not be moved by the argument, and she need not be – no rational belief revision ensues. If she does accept the premises, then rational belief revision may occur (whether she rejects one of the premises or affirms the conclusion depends on further factors), but then the argument amounts to the mere demonstration of an internal inconsistency among her views.

One further consideration needs to be added to render the dilemma as worrisome as Oppy sees it. For any interesting and controversial proposition p, one can consistently reject at least one premise in every argument for p – which is just to say that interesting and controversial propositions are not tautologies. If they were, or if at least many philosophers suspected that they might be, we’d expect philosophy papers to contain logical derivations of comparable complexity to those found in mathematics papers – but they don’t (Oppy, 2011, 9).

From these considerations it follows that derivations can never bring about reasonable belief revision in people who have and can maintain a consistent set of positions – with Oppy’s slight overstatement, if they are “perfectly rational” (Oppy, 2006, 13). It is difficult to see, then, how we could look to arguments to help us decide whether some interesting and controversial proposition p is true. The most they could do is tell us whether our other beliefs entail it, yet with the full knowledge that others might reject those of our beliefs which serve as premises – and in the case of philosophy, we will often be unable to point to anything unreasonable in their doing so.

For worldview comparison

Thankfully, Oppy does not leave us entirely in this pessimistic situation. He proposes an alternative approach to deciding whether some proposition is true: worldview comparison (e.g., Oppy, 2016, 24–28; 2017; 2018, 17–21). This approach comes in two parts: worldview elucidation and worldview evaluation. In the following, I summarize the two.

Worldview elucidation

Take a proposition in whose truth one is interested: p. The goal of worldview elucidation is to extend p into a theory by adding to it other propositions relevant for assessing the truth of p; propositions that show up in arguments for p, and – more importantly – the negations of propositions that show up in arguments against p.

How expansive should such a theory be? Oppy is a bit ambiguous.Footnote 1 Perhaps he means to distinguish between two notions of worldview: an idealized and a practical notion. A worldview in the idealized sense would be a complete theory that takes positions on every interesting question in philosophy, science, and the humanities. It seems unlikely that we will ever develop a worldview in that sense. A worldview in the practical sense, on the other hand, is a theory that affirms or rejects a controversial claim p and takes positions on many issues relevant to assessing p.

Assuming, then, that one wishes to evaluate the proposition God exists, one might extend this proposition into a theory by taking positions on issues like the possibility of infinite causal regress, the existence of moral truths independent of human evaluation, whether some things are contingent, what God’s omniscience is supposed to involve, and whether the historical narratives found in religious scriptures can be trusted.

Worldviews built around some p can be quite diverse, both with respect to their extent and to their content: some p-worldviews do not take a position on whether q, others affirm q, and yet others deny it. (If we are to take the idealized sense of complete worldviews as fundamental – as Oppy seems to suggest (2017, 178) – p-worldviews that do not take a position on whether q should be seen as ambiguous or undecided between two comprehensive p-worldviews, one of which affirms q, the other of which denies it.) To find the truth about p – says Oppy – we need to find the best p-worldview and compare it with the best ¬p-worldview. Both steps require a way to evaluate or compare worldviews. This is what we turn to now.

Worldview evaluation

Oppy proposes a two-step process for evaluating worldviews. First, one assesses the internal consistency of a single worldview: if it entails contradictions, it must be rejected or revised (except, I suppose, if it endorses a paraconsistent logic).Footnote 2 This may not be trivial, as illustrated by mathematical research, where the search for logical entailments quickly becomes difficult to follow for the layperson. However (as mentioned in Against derivations), Oppy points out that philosophy of religion papers are not like mathematical papers: the former typically feature just a handful of premises and simple rules of inference that rarely exceed the complexities of quantified modal logic. This is arguably evidence that the best theistic and atheistic worldviews on the philosophical market today are usually thought to pass the internal consistency test. On the other hand, ontological arguments for God and atheistic arguments from the incoherence of divine attributes do seem to belong to this first step; but it is true that they aren’t the most popular arguments today.

The second step in Oppy’s process is a pairwise comparison of worldviews. Whenever comparing two worldviews, we split them into data and theory. Plainly, the data are what they agree on, and the two theories are what’s left. When comparing two p-worldviews, p is a datum and the theories are other details of p-worldviews; when comparing a p-worldview and a ¬p-worldview, p and ¬p end up in the opposing theories.

Equipped with two theories and a data set, one moves on to the assessment of the theories with respect to theoretical virtues. There are two of these: minimizing theoretical commitment and maximizing explanatory breadth and depth. These roughly correspond, respectively, to the prior probability of the theory and the conditional probability of the data on the theory. As Oppy points out, these two virtues often point in opposite directions: the more one packs into the theory, the more it can predict and thus explain, but the less economic it becomes. It follows that we need tradeoffs, but these will inevitably be contentious; therefore – says Oppy – we should expect plenty of rational “agreeing to disagree” concerning judgments of worldview superiority (cf. Oppy, 2016, 47).

Nevertheless, our best shot at assessing whether God exists is to develop the best theistic and atheistic worldviews we can (by developing many and comparing them), and then evaluating the finalists against said virtues. For Oppy, the winner is a version of atheism (2018, 48).

Epistemic justification

The previous section laid out how Oppy thinks we should go about assessing the existence of God in particular, and philosophical claims in general. In this section, I’ll briefly examine Oppy’s – closely related – epistemological views about rational justification and disagreement with respect to such claims.

As we have seen, Oppy thinks that the failure of arguments to bring about a consensus prohibits us from judging them as successful. What then, one wonders, is the standard of justification that Oppy’s takes his atheistic belief to meet? He proposes the following answer:

Some people claim that, in order to have a justified belief on a controversial philosophical question, you need to have an argument: that is, you need to have a derivation of your controversial philosophical opinion from other (perhaps less controversial) claims that you accept. It should be obvious by now why I am inclined to suspect that this kind of claim is seriously mistaken. In order to have a justified belief on a controversial philosophical opinion, you ought to have developed a best theory that embeds the belief in question, and you ought to have satisfied yourself that that best theory is not trumped by an extant best theory that embeds the denial of that controversial philosophical belief. But you simply do not need derivations with the controversial philosophical claim as conclusion in order to do these things. (Oppy, 2015, 330)

Oppy’s criteria for justification of belief in a “controversial philosophical opinion” (p) are, then, twofold:

  1. 1.

    Having developed a worldview W that embeds p.

  2. 2.

    Having satisfied oneself that W is not trumped by any extant W’ that embeds ¬p.

Both reason and a slightly later passage from Oppy (2019a, 127) suggest a stronger alternative for (2):

  1. 2.

    Having satisfied oneself that W trumps every extant W’ that embeds ¬p.

Since 2’ refers to the W introduced in (1), I take it that (1) and (2’) can faithfully be combined into the following principle:

Justification: One’s belief in a controversial philosophical opinion p is justified if and only if one has satisfied oneself that a worldview one has developed that embeds p trumps every extant worldview that embeds ¬p.

Why “if and only if” – why is this condition both necessary and sufficient? The necessity is shown by Oppy prefacing the two components of the condition with “you ought to”.Footnote 3 The sufficiency is shown by the following sentence’s denial of the necessity of derivations, and the general context where Oppy’s point is denying that something beyond worldview comparison is necessary for justified philosophical opinions.

Though I take Justification to be faithful to Oppy’s point in the quoted passage, it’s not straightforwardly harmonious with some other things he says. Specifically, he is much more lenient about the criteria for rational or epistemic permissibility (e.g., Oppy, 2019, 11, 13), which, on a deontological conception, coincides with justification. Oppy seems, then, to be committed to some other notion of justification which, judging by Justification, is still internalist. In the broader literature, the most readily available and plausible alternative conception is the one concerning “adequate grounds… sufficiently indicative to the truth of p” (Alston, 1985, 71).

The upshot is that justification is severed from the norms of belief revision, which Oppy (understandably) takes to correspond to rational permissibility and obligatoriness. If one holds one’s belief in a rationally permissible way, then belief revision is not called for – not even suggested. This would mean that one can have beliefs upon which one has no rational way to improve (including decreasing confidence or suspending belief), yet one knows to be unjustified. (I take up some issues with this account in Section II.1.2.)

A dilemma for Oppy

This section argues that Oppy’s account of epistemic justification endorses foundationalist and coherentist elements in an inconsistent way. First, I will argue that Oppy’s account commits him to the following two theses:

Truth-conducive reasons: Reasons for belief that aren’t truth-conducive can’t justify.

Thoroughgoing coherentism: Justification for philosophical opinions depends entirely on one’s beliefs about the comparative theoretical virtues of comprehensive worldviews.

Second, I will argue that these two theses jointly imply that there cannot be any justified “philosophical opinions,” contrary to Oppy’s position.

Exegesis

I now turn to Oppy’s writings to argue that we should take him to be committed to Truth-conducive reasons and Thoroughgoing coherentism.

Truth-conducive reasons

I take it that Truth-conducive reasons entails that if some claim is to count toward the justification of a belief, that claim has to constitute pro tanto reason to think that the given belief is more likely to be true than its negation. There are three lines of evidence for Truth-conducive reasons in Oppy’s writings.

  1. 1.

    His discussion of pragmatic arguments for religious belief. Oppy denies that pragmatic arguments – arguments that aim to show the desirability of theism – can rationally give reason to believe and says that “beliefs can only be properly responsive to truth-conducive reasons” (2011, 18) and that “the goal of belief is to track the truth” (2011, 21).

  2. 2.

    His comments on the reasons for (his atheistic) belief. Oppy expresses his preference for atheism in the following way: “when I make the best evaluation that I can of all of the relevant considerations, I come down on the side of the claim that there are no gods” (2018, 48). He also says that “[a]ll minimally rational atheists suppose that best theistic big pictures are inferior to best atheistic big pictures” (2019a, 127). To be a rational atheist, it seems, is to believe that one’s reasons favor atheism over theism.

  3. 3.

    His analysis of the reasons for steadfastness in disagreement. Oppy seems to agree that a consideration in favor of a view cannot equally support that view’s negation when he argues for steadfastness in disagreement based on one’s first-order reasons: “My grounds for believing that I am right and you – my doxastic peer – are wrong are just my grounds for believing as I do” (2010, 198). This clearly assumes that one’s grounds for believing as one does are grounds for believing that claims incompatible with one’s beliefs are false.

For Oppy himself, then, the grounds or considerations that drive one reasonably to believe a view are grounds or considerations for thinking that the view’s negation is false – which is the claim made by Truth-conducive reasons.

On the other hand, Oppy gives a story of two judges that seems to point in the contrary direction (2010, 192–193). Here, both judges have a 95% track record of judging some kind of cases correctly. In 90% of the cases, they agree; in 10%, they disagree. Oppy argues that in the controversial 10% of cases, one could be rational (depending on how one favors gaining true beliefs over avoiding false ones) in accepting the verdicts of one judge rather than suspending belief, since the former strategy gets one 5% more truth. If one were to do that, one’s reasons for accepting Judge 1’s verdict V in a controversial case would appear to be that (1) there’s a 50% chance that the verdict is correct and that (2) the alternative of belief suspension is unattractive. Neither reason favors V over ¬V, yet Oppy believing V on their basis is claimed to be rational.

What should we make of this apparent contradiction? I see two options. The first is to concede the contradiction and argue that the overly permissive part should be rejected. The other is to explore whether I misconstrued something in Oppy’s approach. The case of the two judges does seem open to an alternative interpretation. Maybe what rationality is supposed to be permissive toward is the higher-order choice between methods or strategies one of which yields 10% belief suspension and the other 5% truth and 5% falsehood. It could still be true that the first-order reasons on which one bases one’s beliefs need to favor those beliefs over their negations; it’s just that one’s strategy for obtaining and evaluating reasons is not statistically reliable. The more precarious, 5% true / 5% false policy should then still furnish one with plausible-looking reasons to believe those propositions that turn out to be false. This interpretation absolves Oppy of contradiction and renders him an unambiguous advocate of Truth-conducive reasons.

Thoroughgoing coherentism

Showing that Oppy is committed to Thoroughgoing coherentism will take a bit more work, though at one point he seems plainly to affirm something similar explicitly: “every belief stands or falls by coherence with the rest” (2006, 9). But here I want to show how this follows from his account of justification for philosophical opinions (Epistemic justification):

Justification: One’s belief in a controversial philosophical opinion B is justified if and only if one has satisfied oneself that a worldview one has developed that embeds B trumps every extant worldview that embeds ¬B.

Though ‘having satisfied oneself’ is not standard epistemological vocabulary, fulfilling it presumably involves reasonably holding the following belief:

Trumping: B-embedding worldview W trumps every extant W’ that embeds ¬B.

When and why might it be reasonable for a believer in B to believe Trumping, by Oppy’s lights? The answer is certainly not “when she has a good argument,” both because controversial philosophical positions rarely if ever have arguments for them that are good by Oppy’s standards, and because he denies the necessity of such arguments for justified belief: ”In order to have a justified belief on a controversial philosophical opinion, … you simply do not need derivations with the controversial philosophical claim as conclusion” (2015, 330). Two other options are left: non-propositional entitlement or justification and worldview comparison. I consider both in order.

1. Nonpropositional entitlement or justification. Perhaps, having considered the pros and cons of W and varieties of W’, it just seems to one without argument that Trumping is true. This would constitute an appeal to intuition or insight. What would Oppy make of that?

On the one hand, he emphatically rejects intuition or insight as unnecessary for steadfastness in disagreement (2010, 193–194) and irrelevant in worldview comparison (2015, 327; 2019c, 10–11). On the other hand, he argues that one may often reasonably continue to believe what one finds intuitive when the experts don’t agree (2019c, 11):

if you hold opinions where there is no established convergent expert opinion, then, even if you have none but intuitive support for your opinions, it may be that you have no reason to reconsider. In matters of philosophy, politics, and religion, it is hard to see any good reason why experts have greater entitlement to hold particular beliefs than those who are not experts.

One promising way to harmonize these two claims is to distinguish between a kind of epistemic entitlement and epistemic justification (cf., roughly, Wright, 2004): say that one is entitled to a belief if one holds it and has no reason to reconsider it; and one’s belief is justified if one has (good) reason to think it is true (cf. Epistemic justification). One might then be entitled to an unjustified belief when one is not aware of any truth-conducive strategy for belief revision, as Oppy argues in cases of widespread disageement:

given that the goal of belief is to track the truth, we must take account of the fact… that there is no method open to us for aggregating our beliefs that we have reason to believe will increase the probability that we have true beliefs… where expert credences are distributed everywhere from zero to one, there just is nothing that recommends any particular revision of the credence of any given expert in the light of the credences of all of the other experts. (Oppy, 2011, 21)

Would Oppy accept this distinction between entitlement and justification? He often acknowledges another distinction between rational permissibility and rational obligation (e.g., Oppy, 2019c, 13). Permissibility and entitlement seem to be the same idea; but justification and obligation aren’t, because he believes that theistic and atheistic belief can be justified but there is not rational obligation to believe either. We would then have three increasingly strict levels of positive epistemic status: entitlement or permissibility, justification, and obligation. For Oppy, intuition is sufficient for the first; nothing feasible is sufficient for the third; and justification, at least on controversial matters, seems to require worldview comparison.Footnote 4

There is a looming inconsistency: (1) belief one is merely entitled to seems to be supported just by intuition; (2) Oppy rejects the rationality of holding one’s intuitions to be more reliable than those of others – so, where intuitions diverge, he wouldn’t seem to approve of treating them as truth-conducive reasons; (3) yet, as we saw in the previous section, Oppy says that beliefs are oriented toward the truth, so that it is only permissible to hold them based on truth-conducive reasons.Footnote 5 This apparent inconsistency does not stem merely from my reconstruction, but largely from claims made by Oppy himself.

I suspect that Oppy makes a mistake both here and with respect to the judges case discussed in the previous section. The inference from the truth-tracking goal of belief and the unavailability of a knowably truth-conducive method of belief revision to reasonable entitlement is invalid. It does not follow that if the goal of belief is to track the truth, then only truth-conducive methods of belief revision can be rationally obligatory. When one has no reason to think that one’s belief is true, then, given the truth-tracking goal of belief, it is entirely possible that there is an obligation for one to suspend belief even though that cannot increase the number of true beliefs one has.Footnote 6

The alternative to the proposed harmonization is to suggest that Oppy is inconsistent and overly permissive in the passages quoted above, contradicting Truth-conducive reasons. I consider the consequences of rejecting Truth-conducive reasons in Section III.1. Either way, there is little consistent basis in Oppy’s writings for Trumping being justified (or one being entitled to belief in it) nonpropositionally.

2. Worldview comparison. One might then suggest that the conditions for justified belief in Trumping are completely analogous to the ones for justified belief in B. These would turn out to be the following:

1T. Having developed a worldview WT that embeds Trumping.

2’T. Having satisfied oneself that WT trumps every extant WT’ that embeds ¬Trumping

How might belief in the claim included in (2’T) – that WT trumps every extant WT’ that embeds ¬Trumping – be justified? The answer, again, must be either non-propositional entitlement or justification – which Oppy may disfavor for reasons mentioned above – or worldview comparison, with a vicious regress looming in the latter case.

There may be a strange solution to the looming vicious regress: if W and WT are identical. In the following, I will argue that they need to be. Recall that in the idealized sense, worldviews are supposed to be comprehensive so that any two non-identical ones contradict each other (Oppy, 2017, 175–177). W and WT would therefore have to be either contradictory or identical. Take first the possibility that they are contradictory. The first interesting cases here are if the two contradict on B (WT denies it) or Trumping (W denies it). Both cases seem self-defeating: how could a worldview confused enough to deny B but endorse the claim that a B-embedding worldview trumps all ¬B-embedding ones play an important role in justifying B-belief? Similarly, how could W play an important role in justifying B-belief if it embeds ¬Trumping – that is, the claim that some ¬B-embedding worldviews are equal or superior?

The other option, still considering the scenario where W and WT contradict, is that they contradict not on B or Trumping – they both embed these – but on some other point. In this case, it is hard to see the relevance of WT for the rationality of belief in B. W already embeds B and Trumping. WT might turn out to be better if one performs the kind of worldview comparison recommended by Oppy (cf. Worldview evaluation), and the B-believer might then choose to adopt WT over W. But what role could considering a contrary worldview play in justifying Trumping given that one already accepts it?

On the other hand, if W and WT are identical, then Trumping turns out to be part of W: it is part of the worldview to claim its own superiority. Note that this would get around the regress worry by embracing a kind of circularity or, minimally, self-reference. It turns out, then, that if belief in a controversial philosophical claim is to be justified by worldview comparison, the preferred worldview embedding the given claim needs to include a self-referential superiority claim.

One might object that a lot depends on my employing the idealized notion of worldview. Are we any better off if we try to parse Oppy’s argument in terms of more realistic, less expansive worldviews? I doubt it. On the one hand, the regress problem would resurface: we’d need a third worldview to support the belief that WT trumps the ¬Trumping-embedding worldviews, a fourth worldview to support the superiority of that third worldview, and so on ad infinitum. On the other hand, it is hard to imagine that any decent and reasonably well worked-out worldview that embeds Trumping doesn’t embed B; and that any reasonably well worked-out worldview that denies B embeds Trumping. So, a very close link remains, and the separation is at best unstable: upon some reflection and expansion, the two worldviews may fuse.

Contradiction

Having argued that Oppy is committed to Truth-conducive reasons and Thoroughgoing coherentism, I will now argue that these two jointly imply that there cannot be any justified philosophical opinions. My argument takes the following form:

  1. 1.

    If Thoroughgoing coherentism, then beliefs about the comparative theoretical virtues of comprehensive worldviews are the only source of justification for philosophical opinions.

  2. 2.

    If beliefs about the comparative theoretical virtues of comprehensive worldviews are the only source of justification for philosophical opinions, then these beliefs constitute at least some of one’s reasons for one’s philosophical opinions, and if those reasons cannot justify those opinions, then those opinions cannot be justified.

  3. 3.

    If Thoroughgoing coherentism, then the reasons constituted by beliefs about the comparative theoretical virtues of comprehensive worldviews are not truth-conducive.

  4. 4.

    If Truth-conducive reasons, then non-truth-conducive reasons cannot justify philosophical opinions.

  5. 5.

    Therefore, if Truth-conducive reasons and Thoroughgoing coherentism, then philosophical opinions cannot be justified.

In the following, I argue for the premises in some detail.

Premise 1. Recall that Thoroughgoing coherentism claims exclusive significance with regard to justification for beliefs about comparative theoretical virtue. It thus involves an internalist picture: one mental state is justified by another. Note also that the exclusive claim rules out any nonpropositional or nondoxastic mental states playing any justificatory role.

Premise 2. This premise is an awkward bridge between the different languages of the two Oppyan theses: one is put in terms of reasons and the other in terms of justification. The assumption that when a set of beliefs serves as the only source of justification, then those beliefs are one’s reasons for belief seems fairly innocent. I see only one plausible way to question it, which invokes the distinction between propositional and doxastic justification.

This distinction is not an entirely straightforward one. It is supposed to distinguish between the grounds available for someone that would justify one’s belief and the grounds on which one actually believes. This distinction renders intelligible the claim that one has justification to believe something one does not actually believe by interpreting said justification to pertain to the proposition, not the doxastic state (cf. Turri, 2010).

Applying the propositional–doxastic distinction, the objector might say the following: what follows from Thoroughgoing coherentism is that the propositional justification for philosophical opinions is exhausted by beliefs about the comparative theoretical virtues of worldviews, but one’s reasons for belief concern one’s doxastic justification – the reasons based on which one actually believes. A similar distinction might be invoked to say that even though the beliefs that in fact play a role in justification are the ones about the theoretical virtues of worldviews, one might wrongly take other beliefs into account when weighing what to believe, thereby treating those as reasons.

These distinctions might be sound; the objector might be right. But I don’t think the objection is relevant. This is because Premise 2 is not about all reasons, just the ones with justifying potential; and Thoroughgoing coherentism makes it clear that no beliefs other than ones concerning the relative theoretical virtues of worldviews play a role in justification. In other words, Premise 2 does not claim that beliefs about worldviews constitute all of one’s reasons for belief; it claims that they constitute all of one’s reasons that are candidates for being justifying reasons.

One worry remains: is it obvious that the way beliefs justify other beliefs is by being (taken as) reasons for them? It seems to me that in internalist contexts, the two are often used as synonyms or at least equivalents (e.g., Hasan & Fumerton, 2022). Still, one may sense an intuitive difference. The strongest way to formulate the intuitive objection is to say that a support relation between propositions counts as a reason only if one has recognized it at some time – while one version of internalism, access internalism, denies that this is a requirement for justification. Even if this is right (I really don’t know whether it is), Oppy’s wording – one “ought to have satisfied” oneself about the superiority of one’s preferred worldview – seems to require one’s recognizing the connection, thereby qualifying the superiority-belief as a reason.

Premise 3. The meat of the argument, the most controversial contention, lies here. Comprehensive worldviews take positions on everything relevant to their assessment: the facts, the existence and properties of competing worldviews, the standards of theory evaluation and how the given worldview fares in their respect etc. Thus, the belief that a worldview is superior to its alternatives is not a further judgment or recognition one brings to the consideration: rather, it is just one step among many in cataloguing the worldview’s claims. Accepting Oppy’s criterion of justification, a worldview’s commitment to its superiority is a mere aspect of its – epistemological or doxastic rather than logical – consistency; if one were to judge that the given worldview does not trump its alternatives, one would contradict the worldview.

To see the implications of this, consider the moment of judging a worldview superior to the other after some time of undecided reflection. If this judgment is justified, then, given Thoroughgoing coherentism, it is so in virtue of being embedded in a worldview. I’ve argued that this must be the very worldview under consideration. But in virtue of what does that worldview have the power to confer justification on one’s judgment? Well, the Oppyan answer would be that that power derives from one’s judgment that the worldview trumps to its competitors. As best I can tell, we’ve arrived at a very tight vicious circle of justification: justified belief in p (worldview superiority) is supposed to justify belief in q (the worldview), but justified belief in p requires justified belief in q. This is clearly unacceptable.

One wonders if this premise is a version of the alternative systems objection to more mainstream coherentism. There, the concern is that coherence is not truth-conducive because very different (and contradictory) systems can be highly coherent. There is some resemblance to the issue here: it is true that very different worldviews can embed the claim that they trump the others. But for that to be clearly problematic, I had to appeal to the observation that Thoroughgoing coherentism rules out any additional support for the superiority claim beyond its inclusion in some worldview. I take it that coherentism isn’t usually committed to the ideally comprehensive nature of worldviews as Oppy is, and therefore this version of the problem might be unique to his approach.

Premise 4. This premise is only included to make the argument clearly valid; it is merely a close paraphrase of Truth-conducive reasons.

In sum, the premises seem well-motivated; it isn’t clear to me how one might plausibly avoid the force of the argument. I take it that the conclusion’s consequent is unacceptable, and at any rate far from what Oppy would want to endorse. In the concluding section, I shall therefore consider the options before the person who decides she has to deny one of the Oppyan principles.

What to reject?

At this point, I depart from the strict confines of internal critique. Assume one is convinced by my argument that the combination of Truth-conducive reasons and Thoroughgoing coherentism has untenable skeptical consequences. What should one do – what is the best alternative epistemological picture? In this section, I briefly consider the implications of rejecting Truth-conducive reasons – yielding what I call epistemological relativism – and those of rejecting Thoroughgoing coherentism – whose negation is, broadly speaking, foundationalism. I shall argue that the latter is preferable.

Epistemological relativism

What would the implications of rejecting Truth-conducive reasons be? I suggest that a radical kind of epistemological relativism would ensue, where one recognizes that one’s beliefs have no better claim to being true than incompatible alternatives, and where rationality either isn’t involved in belief formation or is completely reoriented toward, say, pragmatic and normative considerations. This is more extreme than the kinds of epistemological relativism one might find in the literature, which claim that the relevant epistemic norms are relative to a community or a belief system (Lynch, 2019). Perhaps it is indeed relative, in the sense of not publicly arbitrable, whether Vipassana meditation really provides insight into the truth of the no-self doctrine of Buddhism; but presumably those who take it to do so take it to furnish them with a truth-conducive reason for said doctrine. The kind of epistemological relativism we’re considering here denies the privileged relevance of such reasons.

I strongly doubt that this kind of epistemological relativism is feasible in psychological and conceptual senses. If I develop some kind of doxastic attitude toward a proposition p and do so for reasons I recognize are not truth-conducive, then either I engage in some form of self-delusion, or the doxastic attitude I take myself to develop is not one I take to be truth-aimed. But belief is a truth-aimed attitude: that’s what distinguishes believing that p from other propositional attitudes such as imagining that p and pretending that p (Fassio, 2015). This seems to be such a central feature of belief that this kind of epistemological relativism would likely involve a change in subject: we would no longer be talking about our beliefs and the norms guiding belief formation, but about some other doxastic states – preferences, fancies, wishes, perhaps. In the worse scenario, we would call these beliefs; in a clearer world, we’d just bracket the issue of belief or agree to suspend belief concerning philosophical issues.

It may be the case that something like the latter situation is taking place among metaphilosophers. I’m referring to the “acceptance without belief” approach to philosophical theories, which takes its inspiration from anti-realist attitudes toward scientific theories (Beebee, 2018, 20–22; Barnett, 2019). I don’t want to critique this position other than to say that if one is still concerned to know what one should believe, consulting truth-conducive reasons might be a psychological and conceptual inevitability.

Foundationalism

The other option is to reject Thoroughgoing coherentism. The claim is somewhat complex, so there are many ways to deny it. However, to avoid reproducing the problems that riddled it, preferable alternatives would need to reject the idea that epistemic evaluation concerns comprehensive worldviews, because, as we have seen, this collapses the evaluative judgment with the object of evaluation. Not to collapse evaluative judgment with the object of evaluation would seem to involve either infinitism or foundationalism. Infinitism is the view that if a belief is justified, it has an infinite regress of supporting beliefs; foundationalism is the view that some beliefs are justified in a way that does not require any further beliefs. Infinitism is widely considered a marginal and implausible view, though it has been defended occasionally.

On the other hand, foundationalism – currently the majority view among philosophers (Bourget & Chalmers, forthcoming, 12) – seems to allow for the rehabilitation of arguments, at least in principle. This is because foundationalism posits nonpropositional justification, and nonpropositional justification can serve as an epistemological symmetry-breaker between a proposition and its negation, even if both are embedded in consistent worldviews. It would follow that arguments having nonpropositionally justified claims as their premises (or having them somewhere along the support structure for their premises) are epistemologically privileged over arguments with premises lacking nonpropositional justification, even if those too are coherently embedded. The details depend on the specifics of each version of foundationalism, and the practical failure of philosophical arguments remains to be explained (cf. Pruss, 2007, Békefi, 2023) – but a way around Oppy’s formal critique of “derivations” is opened.