My religion preaches ‘p’, but I don't believe that p: Moore's Paradox in religious assertions

Abstract

In this article, I consider the cases of religious Moorean propositions of the form ‘d, but I don't believe that d’ and ‘d, but I believe that ~d’, where d is a religious dogma, proposition, or part of a creed. I argue that such propositions can be genuinely and rationally asserted and that this fact poses a problem for traditional analysis of religious assertion as an expression of faith and of religious faith as entailing belief. In the article, I explore the possibility of undermining these commonly held assumptions and argue that the assertability of religious Moorean propositions can be justified by an account of faith as an intention to form religious beliefs. In the end, I also consider the consequences of such a stance, especially concerning the debate on the ethics of religious belief and doxastic voluntarism.

Introduction

[T]he word ‘believing’ has wrought horrible havoc in religion . . . But if instead of ‘belief in Christ’ you would say: ‘love of Christ’, the paradox vanishes, that is, the irritation of the intellect.

(Ludwig Wittgenstein, Diary, 19 April 1937; in Klagge and Nordmann 2003, 247)

Is belief essential for religious faith? And, if so, can this belief be expressed in a flat-out assertion? Intuitively, the answer to both of these questions is a resounding ‘yes’. Every day, Muslims around the world state in shahada that there is no deity but God and that Muhammad is the messenger of God. Every Sunday Mass, Roman Catholics state in unison during the creed ‘I believe in God the Father almighty, creator of heaven and earth.’ Those statements seem to express the beliefs of religion's followers and hence may be labelled religious assertions.¹ The collective or individual act of a creed also serves another purpose: it is perceived as a confession or statement of faith, which is considered by many religions a necessary condition for inclusion into a religious community.

It is natural to think that these two functions are not quite distinct. When we describe religious communities (communities of faith), it is common to speak of ‘believers’. It is also natural to say that while Christians believe that Jesus was God, Muslims believe him to be a prophet. Both of those statements may be found in the holy books of the respective religions, and both may be (and have been repeatedly) asserted by religious scholars, leaders, saints, and ordinary members of the Christian and Muslim communities. Believing that Jesus was the son of God and asserting that during the creed seem to be part of what it is to be a Christian. Therefore, when Cath, being a Roman Catholic, asserts:

(1) The one God exists in three divine persons.

she expresses her belief that the one God exists in three divine persons, and that belief is at least in part constitutive of her religious faith² in that proposition.

We might try to capture this natural view of religious assertions in two plausible-sounding theses. The first would state that religious faith is doxastic – that is, that having faith in religious propositions (e.g. dogmas) entails believing them:

(Dox) If agent A has a religious faith that d, then A believes that d.

The truth of (Dox) is entailed by many analyses of religious faith, starting from Thomas Aquinas's and John Calvin's writings on the topic. Alvin Plantinga (2000) famously argued that religious faith is a kind of knowledge (and hence a kind of belief). Richard Swinburne in his detailed account characterizes faith as ‘the theoretical conviction that there is a God’ (Swinburne 2005, 138). As noted above, such analyses follow a natural way of speaking about religious faith as nearly synonymous with religious belief.

The second principle may be proposed to provide a link between assertions performed in religious contexts (i.e. having a religious content, justification, or being performed during a religious rite) and faith:

(FA) By asserting in a religious context that p, the speaker S expresses that they have faith that p.

(FA) captures the intuitive sense in which religious assertions (in line with the observation of their role in a creed or religious discourse) express faith, analogously to how ordinary assertions are said to express belief. If one supports (Dox), then one may also say that there is nothing special about religious assertions. They are just plain assertions expressing beliefs, which in religious contexts are simply constitutive of faith. As Linda Zagzebski writes, ‘[i]f I overhear you reciting your creed, I hear you making assertions that express your beliefs’ (Zagzebski 2012, 120).

As it becomes clear further on in the article, however, distinguishing these two principles is desirable. What may be noted right now is that, among different contemporary accounts of assertion, some may require some strong assumptions regarding faith that are not necessary. For example, if one is tempted to follow Williamson (2000, 243) and others in saying that knowledge that p is constitutive for asserting p, then one would need to commit oneself to the claim that religious assertions are unwarranted unless one stands in a privileged epistemic relation to their truthmakers, which cannot be easily said about most religious followers. Religious discourse seems to be special to the extent to which some have denied that in religious contexts one properly asserts anything. Instead, utterances of (1) have been analysed as having metaphorical meaning (Kenny 2004) or expressing practical recommendations (Santayana 1905; for an overview, see Scott 2022). In this article, however, I wish to proceed with an assumption that religious assertions are indeed possible and, hence, require special treatment among other assertions. A principle that links them with religious faith, clearly distinguishing them from other assertions, seems to me the most intuitive and easy to operationalize for present purposes.³

According to these two principles, we get an easy way of conceptualizing Cath's utterance of (1), which seems to conform to our pre-theoretic intuitions and linguistic practice. Given that (1) is an assertion – it was, say, spoken during a Sunday Mass sincerely and with full conviction – it expresses Cath's propositional faith (by [FA]) that only God exists in three divine persons. Since, by (Dox), propositional faith entails belief, we may say that Cath also believes that only God exists in three divine persons, and therefore she also expressed this belief by asserting (1).

In this article I aim at challenging this simple picture of religious assertion and belief by pointing to cases of Moore-paradoxical assertions which might be, as I shall argue, made by doubting religious followers without irrationality. In the next section, I will present these examples and argue why accepting both (FA) and (Dox) leads to the conclusion that such assertions are signs of irrationality and why this conclusion is unwelcome. In the third and fourth sections I will consider the possibility of abandoning, in turn, (FA) and (Dox). I will argue that only abandoning (Dox) gives us the right predictions concerning religious Moorean assertions. I will propose Vahid's (2023) theory presenting religious propositional faith as constituted by an intention to form beliefs in accordance with a set of principles laid out by the respective religion as especially promising. In the last section I shall counter the ‘doxastic involuntarism’ objection to such theory raised also against other analyses of propositional faith.

Moore's Paradox in religious assertions

One of the most prominent puzzles intertwined with an analysis of assertion and belief is the so-called Moore's Paradox. As noted by George Edward Moore (1942; 1944), it seems that the following types of assertions (labelled consecutively ‘omissive’ and ‘commissive’ forms of the paradox; see Williams 1979; Sorensen 1988) seem ‘absurd’ or even ‘self-contradictory’:

(OMP) p and I don't believe that p (e.g. ‘It is raining, but I don't believe it is’).

(CMP) p and I believe that ~p (e.g. It is raining, but I believe it is not raining’).

But why do they sound so if they are not, in fact, contradictions? It is perfectly possible that I have a false belief (as stated by CMP), and even more so that I am not omniscient (OMP). I may with no absurdity say ‘p, but I didn't believe that p’ or ‘p, but she believes that ~p.’ What is so peculiar about the first-person present-tense assertion that makes it so problematic? The most straightforward answer, championed by many epistemologists and philosophers of language, is that the content of such assertions is impossible to be believed or known by the asserter, and the assertion of OMP or CMP expresses such impossible belief or knowledge (for solutions along these lines,⁴ see, for example, Hintikka 1962; Williamson 2000).

Consider now the two following cases:

(CATH, THE DOUBTING CATHOLIC)

Cath is a devoted Catholic who regularly attends Sunday Mass and asserts the content of the ‘Credo’. However, when she introspects in private, she realizes that the concept of the unity of the Divine Trinity seems to her inconsistent and incomprehensible. She is therefore disposed to assert the following:

(2) The one God exists in three divine persons, but I don't believe that the one God exists in three divine persons.

(BUD, THE JOYFUL BUDDHIST)

Bud is a devoted Buddhist who regularly reads Buddhist sutras and preaches the Four Noble Truths to his friends and family with conviction. However, when he introspects in private, he finds that he cannot fully believe that even such a joyful event as birth should be considered suffering, as opposed to what is said in the Noble Truth of Suffering. He is therefore disposed to assert the following:

(3) Birth is suffering, but I don't believe that birth is suffering.

As it is clear, both (2) and (3) share the omissive Moore-paradoxical form (where d is a religious dogma or a proposition with religious content):

(4) d, but I don't believe that d.

Do they sound as absurd and self-contradictory as traditional Moorean assertions? At first glance, they seem to be significantly different from the examples usually presented in demonstrations of Moore's Paradox. But given the standard epistemological assumptions and (FA) and (Dox) principles, the usual conclusion follows: both Cath and Bud should be thought of as irrational. To demonstrate that, consider that, by (FA), we take their assertion of the first conjunct of (4), d, to express their faith and the second to express belief. By (Dox), we may reason that by expressing their faith in d they also express their belief in d. Hence, they believe a Moorean proposition of the form: d & I don't believe d. Since plausibly, by the principle of negative belief infallibility,⁵ believing that one does not believe that p entails not believing that p, then Cath and Bud both believe (per their belief in the first conjunct) and do not believe that p (per their belief in the second), which is a contradiction. We need then to agree that they either knowingly hold a false belief (i.e.  they violate negative belief infallibility), which is a sign of irrationality, or that their assertions are insincere and do not, after all, express their beliefs. While some may not find this consequence troubling, I think that one should be cautious with classifying Cath and Bud as irrational or insincere. Certainly, both of them could be classified as doubting followers of their respective religions. They are nevertheless ready to assert the doubted dogmas. Should we classify such assertions as irrational?

By no means are they unusual. Many Roman Catholics have doubts about the unity of the Divine Trinity or the reality of transubstantiation, as evidenced by lively medieval and contemporary theological discussions. In Buddhist and Hinduist traditions, the doubt (vicikitsa) in religious truths is a widely recognized issue and regarded as one of the primary obstacles to enlightenment or the right life (Jayatilleke 1963, 30–32, 369–401); however, it is not a condition which excludes one from the community of followers. This fact may be seen by some as of a sociological rather than epistemological importance. Its prominence in theology and religious creed seems to suggest otherwise – ‘doubt’ in these traditions is definitely not treated as a decisive reason to abandon faith, but an obstacle that one needs to overcome. What such traditions view as rational is not abandoning one's readiness to assert religious propositions or practise religious deeds, but acknowledging one's epistemic position and trying to change it through prayer, meditation, or participation in religious practice. Note that these actions are not, however, comparable to a process of rational inquiry; their aim, though it may be characterized as leading to a change in the follower's cognitive state, has little to do with ‘rationalization’. Though Cath and Bud are undoubtedly in an internal tension, this tension is not a sign of irrationality in their belief, but rather a conflict between the commitments of their respective religions and their current cognitive state.

We are then presented with the following trilemma: either (a) we regard Bud, Cath, and other doubting followers as irrational⁶ or not followers at all, or (b) religious assertions do not express faith ([FA] is false), or (c) religious propositional faith is non-doxastic ([Dox] is false). As noted above, it seems to me fairly intuitive that the first option is undesirable and we should at least try to explore the possibility of (b) or (c) being true before we accuse Bud and Cath of irrationality or insincerity. In the following sections, I will explore these two options and argue that only abandoning the assumption of the doxastic nature of propositional faith may provide a coherent description of Bud's and Cath's cases that does not find them irrational and accounts for the intuition that we observe an internal tension between their commitments and beliefs.

Weakening assertion

As noted above, religious assertions seem to be rather peculiar speech acts, and certainly they are not held to the same epistemic scrutiny as ordinary assertions. As early as in Jaakko Hintikka's (1962) seminal Knowledge and Belief, a Moorean religious assertion of the form ‘God is almighty, but I don't know that He is’ is called ‘more natural than the [Moorean assertion containing an empirically verifiable statement] because the purposes of religious discourse are not normally defeated by the speaker's failure to know what he is saying in the way the purposes of scientific (factual) discourse are thereby defeated’ (Hintikka 1962, 100). One may then try to stipulate that in religious discourse we are dealing with a lower threshold of subjective certainty required to make an assertion and that such threshold may be lower than the one required for a full-on belief.

A working proposal would then be this: let us suppose that in cases of religious assertion, a threshold n ∈ [0, 1] of credence assigned to a proposition that p is required for a speaker to non-defectively assert p (cf. Lewis 1976 for a view that high credence warrants assertion). Let us also suppose that, independently, the belief that p requires some credence m in p, such that m > n. We might follow Levi (1991), Greco (2015), and many others in insisting that this required credence is maximal (m = 1); some more complicated views might allow for lowering this threshold, either absolutely or according to the stakes in question or other context-related considerations (e.g. Weatherson 2005; Bach 2008). Under such assumptions, a speaker a is allowed to fully rationally assert the following (Cr_a(p) stands for a's credence in p):

(4′) a: d (because Cr_a(d) ≥ n, where n is an appropriate threshold for religious assertion), but I don't believe that d (because Cr_a(d) < m, where m is an appropriate threshold for belief).

The picture of our protagonists emerging from such analysis is a portrayal of people who are trapped in doubt that does not allow them to fully believe, and who hence by (Dox) have full propositional faith in the respective proposition held by their religions. The level of confidence they have allows them, however, to assert these propositions. This might make sense if we are ready to treat assertions less as expressions of the speaker's intentional states but rather as speech acts of commitment to defend certain propositions (after e.g. Brandom 1983) or as proposals to add a proposition to a common ground (after Stalnaker 1999). Bud and Cath may be presumably prepared to defend the content of their assertions and cite the relevant theological considerations, parts of scripture, and the like. It seems then that they fulfil the general social conditions for their utterances to be interpreted as assertions.

From the religious side, such a picture might also be supported by Swinburne's Bayesian analysis of propositional faith as the theoretical conviction of the existence of God and theoretical beliefs in related propositions such as that God is omnipotent, omniscient, and so forth. Swinburne (2005) argues that the very ascription of some positive probability to the existence of God may make one's choice to follow a religious creed rational – a position he labels ‘a pragmatic view of faith’ (Swinburne 2005, 147–151). Given that a certain person has some theoretically motivated credence that God exists and is such-and-such, then following the creed of an appropriate religion is motivated as a means to a probable end. Note that, according to the picture of Moorean religious assertions sketched above, one's credence might place one beneath the threshold of actual belief in God's existence, but may nevertheless (per Swinburne's argument) motivate one to act upon the respective religious creed, such as engage in God's worship and assert the propositions this creed contains. Although such a person cannot be said to have full propositional faith, since they do not have an outright belief in religious content, their assertions are, although Moore-paradoxical in form, fully understandable or even rational given their evidence.

There are, however, two important drawbacks to this approach. First, it requires from an agent the ability to make precise and transparent ascriptions of subjective probabilities to matters of one's religious creed, which may strike us as at best highly idealized and at worst impossible. Some philosophers in fact raised doubts that even approaching normal belief states as coming in degrees proportional to one's credence is highly improbable. As Terry Horgan writes:

[I]t is extremely implausible that epistemic agents normally have such quantitative degrees of confidence for all – or for most of the propositions that they can entertain and understand. . . . Much more plausible than this psychological myth about putative quantitative degrees of confidence is the contention that epistemic agents often have certain qualitative degrees of confidence in the various propositions that they can entertain and understand, but which they neither believe nor disbelieve. (Horgan 2017, 236)

Even if Horgan is wrong here about many cases of our ordinary epistemic practice, it certainly seems that the view that we ascribe precise subjective probabilities to propositions such as ‘God exists’ or ‘Jesus was a prophet’ and may be fully aware of these ascriptions is improbable. One might, of course, say that she thinks that God's existence is very likely or highly unlikely, but we do not often hear people saying that they are 89.2% sure that God exists. This observation points out that our solution requiring an ascription of probability to the religious proposition precise enough to place it between thresholds for assertion and belief may be nothing more than a highly idealized myth that has no use in the analysis of actual religious followers.

More importantly, however, toying with credence thresholds for assertion and belief has limits. Although Bud's and Cath's omissive Moorean assertions are rational, we might consider the following way of extending their story. Imagine that Cath, as described in the story above, finds in introspection that the unity of the Holy Trinity seems to her inconsistent and that, being a rational person, she takes internal inconsistency of a concept to be decisive evidence for its inapplicability. She concludes, therefore, that she not only does not believe that the one God exists in three divine persons, but also that she actively disbelieves ⁷ that very proposition. Given what we know about her, she is now disposed to assert the following:

(5) The one God exists in three divine persons, but I believe that the one God does not exist in three divine persons.

This sentence has the familiar commissive Moorean form:

(6) d, but I believe that ~d.

Now, (6) cannot be treated similarly to (4), since it would require meeting both a threshold m for the assertion of d and a higher threshold n for belief in ~d, which violates the principle that one's subjective probability of p and ~p being true must be complementary. Therefore, Cath cannot rationally assert (5), although her situation does not seem to be that different than previously. Her irrationality stems either from the fact that she takes an internal inconsistency as evidence for the falsity of a certain proposition (which seems perfectly rational) or from the fact that the provided interpretation is false.

How should we interpret our cases then? As long as we wish to keep the assertoric nature of speech acts of Cath's and Bud's indicative statements while simultaneously maintaining that faith requires belief, there seems to be not much more wiggle room. An expressivist about religious language, who thinks that Cath's utterance of (1) should be read as an expression of her affective (such as hope or appraisal), rather than cognitive attitude (such as belief) would be, of course, ready to accept this conclusion as well as denounce (FA); but then her utterance should be reclassified as another, non-assertoric type of speech act (perhaps an expressive akin to a congratulation, apology, or a wish). If we want to maintain that indicative religious sentences may be asserted, then the only possible way of avoiding (FA) is through lowering credential thresholds required for the act of religious assertion – and, as we have seen, this strategy does not work.

I think that the following observation could lead us to find a better treatment. Note that Cath does not seem to decide that she disbelieves inconsistencies or that the Holy Trinity is inconsistent, but only finds in introspection that she does.⁸ From this perspective, finding out that one does not believe that some specific dogma is true is not so different from finding out that one believes it to be false. Both of these introspective findings counter what one seems to be obliged to believe by one's faith. But does it necessarily mean that one can introspectively discover their lack of propositional faith? Intuitively, an act of faith is voluntary and conscious – it seems absurd to be unconsciously Christian or a Muslim against one's own will.⁹ Bud's and Cath's beliefs definitely do not come from the same realm as acts of faith, which seem to involve not only a cognitive but also a conative, volitional aspect. Therefore, it might be worth exploring whether a non-doxastic approach towards faith would not account better for the description of their mental state and lay out a probable account of religious assertion as an expression of faith, which both does not require belief and simultaneously is not merely an affective attitude.

Challenging doxasticism about faith

After exploring the option of weakening the (FA) assumption and finding its limitations, it seems right to think of challenging (Dox). In fact, this assumption has been challenged much more widely by philosophers such as William Alston (1996) and Robert Audi (2011), who analysed propositional religious¹⁰ faith as a form of commitment, connected with the possession of other propositional attitudes such as hope (Muyskens 1979) or trust (McKaughan 2016) instead of an outright belief. Considerations leading to such analyses were convergent with the intuitions elicited by Bud's and Cath's stories, that is, that the correct analysis of faith should provide space for doubt in religious content. Abandoning or modifying (Dox) may allow us to explain why Bud's and Cath's faith in d (expressed by their assertions) need not explicitly conflict with their lack of belief or an outright disbelief in d, and hence  why their assertions are not a sign of irrationality.

What theory of propositional faith should (or could) we then adopt? Answering this question is not as straightforward as it may seem at the start. Importantly, the consideration of Cath's extended case of commissive Moorean assertion should lead us to doubt that religious propositional faith that d does entail the lack of disbelief in d and, hence, any positive credence that d. If we agree that Cath may sincerely and warrantably assert (5) and that the first conjunct of her assertion expresses her faith, then she must have both faith that the one God exists in three divine persons and the belief that it is not the case. Assuming that having a belief that d entails at least some positive (that is higher than ½) credence in the truth of d, it follows that propositional faith indeed needs to be compatible with having a negative (lower than ½) credence in d. This goes against many non-doxastic accounts of faith (e.g. Howard-Snyder 2013, 2016; Jackson 2021a); indeed, this point has been called something on which even ‘non-doxasticists nearly universally agree’ (Mugg 2021, 3). The mentioned analyses of propositional faith as hope or trust do seem to support this condition since these propositional attitudes are usually taken to entail the lack of disbelief (Day (1969); cf. Pojman (1986), 170–171 for dissent). This condition, however, is not fully argued in the literature: many sources cite the apparent ‘incoherence’ (Jackson 2021a, 41) of having faith that d and a belief that ~d, while some point to the apparent dispositional character of belief that involves, for example, asserting that ~d or ceasing to act as if d were the case. As we see in the example of Cath, such a situation is not conceptually incoherent, nor does such disbelief need necessarily involve robust behavioural dispositions apart from a disposition to think that ~d during introspection. An agent may then have a rational faith that d guiding their behaviour while simultaneously consciously believing that ~d; disbelief does not seem to prevent propositional faith.

As noted at the end of the last section, religious faith, unlike belief, seems to be voluntary. Although many people do inherit their religious faith in the same way that they inherit or obtain many beliefs, it seems that one may intentionally change one's religious faith in the way in which one cannot intentionally change one's beliefs. A person may decide to become a Christian and join the community through baptism, or voluntarily cease to be one and leave the community through apostasy. While such cases seem to be appropriately described as changes of faith, they do not entail the respective change of belief since one does not directly elect that one will, from now on, believe or disbelieve the existence of a Christian God. The change in one's beliefs may also be gradual and take a long time, while matters of faith seem to be more spontaneous and tied to specific voluntary decisions.

It seems then that religious faith is intrinsically linked to a variety of intentional actions and may even be analysed as an intention itself. In a recent article, Hamid Vahid (2023) proposes the development of Alston's theory along these lines. He characterizes and defends an account of faith ‘in terms of the intention to do something, namely, forming the belief that p, where p denotes a theistic proposition’ (Vahid 2023, 6). The phenomenological considerations leading up to such formulation would not trouble us here; what is worth noting is that this account of faith allows us to accommodate to some extent its voluntary character. For if intentions guide actions and constitute plans, it seems natural to think that if some of our actions are voluntary, then intentions are also voluntarily elected (prima facie as opposed to desires and beliefs). This view, therefore, gives space for the possibility of voluntaristic faith divorced from the necessity of having a correspondent belief structure, especially since the intention in the question itself aims at changing it, not vice versa.

Such an account also fits nicely the observation that many religions come with a set of certain doxastic obligations (Jackson 2021b). For example, Roman Catholics are seemingly obliged to believe in the Divine Trinity as well as have different ethical beliefs following from the teachings of the Church like believing that lying during confession is morally impermissible. If religions are at least partly constituted by the specific set of such obligations, then Vahid's account of religious faith as an intention may be seen as explanatory; on this view, having faith as characterized by religion R is an intentional submission to the doxastic obligations constituting R, instead of simply believing the set of R's dogmas. This, of course, does not mean that to have faith (and be rightfully described as a religious follower) we always need to live up to these expectations or form the beliefs we intend to form. If the Pope announced tomorrow ex cathedra that hell does not exist, then it ought to be regarded as true by all Roman Catholics by the doctrine of papal infallibility; it would not mean, however, that all the Catholics believing in hell instantaneously cease to be Catholics at the moment of the Pope's announcement. Intuitively, this would only be the case if they did not intend to listen to the Pope and recognize his words as true, while the process of such recognition (getting oneself to believe that hell does not exist) may take a very long time.

In this reading, Cath's and Bud's assertions of the sentences ‘birth is suffering’ and ‘the one God exists in three divine persons’ may be thought of as expressing their intention to form the belief of such form (by [FA]). On a more general level, these intentions are derived from their guiding intention of forming beliefs that accord with the teachings of their respective religions (Buddhism or Catholicism). Since they may fail to fulfil this intention and recognize themselves to fail by introspection, their assertions of (2), (3), or (5) are perfectly fine. Note that by characterizing faith in this way, we are also giving a unique and fertile account of religious assertion,¹¹ explaining some potentially troubling characteristics of religious discourse (as noted in the quote by Hintikka).

Even though the fact that an assertion may express intentions may counter some analyses of this speech act, such as views positing the existence of a uniform constitutive rule of assertion (such as Williamson's) or traditional taxonomy of speech acts that take assertions to necessarily express beliefs, note that many other available analyses of assertion, which define it in terms of commitments formed by the asserter (mentioned in Stalnaker's and Brandom's analyses), need not, unlike other expressivist accounts of religious statements, contradict this claim. Note, that when Bud asserts that birth is suffering in a religious context, it still makes sense to require from him some explanation or defence of his claim when challenged (as per commitment views). The fact that he is obliged to fulfil these commitments in a conversational context is perfectly in line with him not currently believing that d; given that he intentionally takes upon himself a commitment to believe that d, he is nevertheless required to defend d against challenges and provide positive reasons for d's truth.

The threat awaiting such analysis of faith is accounting for a long-standing problem coming from the debate on the ethics of belief (Jackson 2021b). If we are not in control of our beliefs (belief is involuntary), we cannot be obliged (via the ought-implies-can principle) to change them without being presented with appropriate evidence or blamed for possessing the beliefs that we have. Similarly, most philosophical analyses of intention agree that we cannot intend to perform actions we believe to be impossible (or at least it would be irrational to do so). Although Vahid's proposal paints a picture of faith as non-doxastic – that is, not entailing occurrent belief – it still requires from the agent the potential ability to form beliefs in accordance with chosen religious principles. This, in turn, seems to contradict the thesis of doxastic involuntarism and would make possessing such intentions impossible to execute and hence  irrational to possess.

The problem seems to be especially tied to the discussion of Bud's and Cath's doxastic state. On the one hand, their examples force us to accept doxastic involuntarism. If we were in direct control of our beliefs, then there is no reason why Bud and Cath should not simply decide to believe the dogmas they respectively found themselves to disbelieve by introspection. On the other, however, it seems that this stance disallows the use of independently plausible analysis, which provides us with a tempting interpretation of them. Agreeing that we are not in control here threatens the voluntaristic component of the intentional analysis of faith and therefore an explanation of how one may have faith in something one disbelieves.

In the next section, I analyse in more detail accounts of doxastic voluntarism to determine whether any of them allows an interpretation of the intentional character of faith that saves the assumptions that voluntarily forming one's beliefs in accordance with the religious creed is possible and that belief possession is not in the direct control of the agent.

Religious doxastic voluntarism

A natural retreat for many theoreticians wishing to argue for some voluntary leeway in belief without committing to full-blown voluntary control over beliefs is to argue for an indirect form of doxastic voluntarism. This stance is in some readings even trivially true. If I wish to believe that I attended Mass next Sunday, I may simply attend it and make sure that I will remember this event vividly. There is no reason, however, to suspect that we may produce any good evidence for our religious belief in this way. As we noted before, it seems doubtful also because, presumably, most of the possible evidence for religious hypotheses is coherent with non-religious alternatives. It seems clear then that an indirect version of doxastic voluntarism based on the activity of gaining or producing relevant evidence seems undesirable for our purposes.

There are, however, other, non-evidential strategies one might employ to obtain a relevant belief. In religious contexts, the indirect form of doxastic voluntarism has been defended implicitly in the form of advising different practices aimed at acquiring or maintaining religious beliefs. As we have seen in the opening paragraphs, nearly all religions consider the emergence of doubt in religious teachings, perhaps even as radical as plain disbelief, as a common threat, but a threat that can be combated by relevant activities. One may think here of Pascal's famous advice from Pensées to ‘[f]ollow the way by which [other followers] began: by acting as if they believe, taking the holy water, having masses said’ (Pascal 2005, 49). Similar considerations may be found in other religious traditions: in Buddha's Dīgha Nikāya (22), where Buddha teaches how to remove the five hindrances (sensory desire, ill will, sloth, restlessness, and doubt) by appropriate techniques of meditation, or in the Muslim teachings, as evidenced by the following quote from Al-Ghazâlî :

[I]f you understand what it is to be a prophet, and have devoted much time to the study of the Qur'an and the Traditions, you will arrive at a necessary knowledge of the fact that Muhammad (God bless and preserve him) is in the highest grades of the prophetic calling. Convince yourself of that by trying out what he said about the influence of devotional practices on the purification of the heart – how truly he asserted that ‘whoever lives out what he knows will receive from God what he does not know’ . . . When you have made trial of these in a thousand or several thousand instances, you will arrive at a necessary knowledge beyond all doubt. (Al-Ghazâlî 1996, 67)

As we might see in the above paragraph, even an appropriately understood assertion in a religious context may be conceived as a tool for getting oneself to believe a religious proposition. Though such considerations may be thought of as highly speculative, it is clear that repetition as Al-Ghazâlî conceives it is presented here as both a source of knowledge and a remedy for doubt – that is, a way of acquiring belief and a remedy for disbelief. Reciting the Credo (as prescribed by the Catholic Church), meditating upon appropriate mantras (as advised by Buddha), or repeating Muhammad's words (as suggested by Al-Ghazâlî) in a declarative mood may indeed play a crucial role in strengthening one's faith and maintaining religious convictions. Therefore, it may be appropriate to think of such assertions as both expressions of religious faith and acts aiming at fulfilling one's intention to form appropriate religious beliefs.

If we grant the possibility of success to these strategies, we might get a clear account of Cath's and Bud's epistemic situation and explain the apparent tension in which they remain without accusing them of irrationality. Both Cath's disbelief in the unity of the Divine Trinity and Bud's lack of belief that birth is suffering stand in tension with their intention to believe that the one God exists in three divine persons or that birth is suffering despite the actions they undertake to convince themselves of these (e.g. reciting the Credo or reading Sutras, taking part in collective religious activities). Therefore, we may characterize them as having akratic beliefs,¹² namely beliefs ‘one believes one should not have’ (Chislenko 2016, 669). Their situation in this respect resembles, for example, believing that our plane will crash despite knowing that it is wildly implausible or believing propositions with racist or sexist content despite one's sincere and evidence-based anti-racist and anti-sexist approach.

Such cases, however, differ importantly from these and other standard examples of akratic beliefs in that usually the cases of akrasia in belief concern believing something contrary to consciously possessed conclusive evidence, because of involuntary features of belief – their connection with involuntarily held emotions, and prejudices or encountered and internalized stereotypes. Actions of someone who intends to form a belief that the plane will not crash or that black people and women are equally as intelligent as white men may be similar in type to the ones prescribed to Cath and Bud by their religions. Such akratic believers should presumably gather and expose themselves to more evidence, undertake psychotherapy, etc.; these actions, however, would be perfectly rational, or at least not irrational, since they would be guided by evidence supporting the beliefs they intend to form. But similar considerations would not transfer easily to practical reasons. Although I might have a practical justification for believing that the Earth is flat (e.g. this belief may be valued by my community and family members), it certainly seems irrational of me to try to convince myself of that by watching YouTube videos, reciting books written by Flat Earth supporters, and trying to forget the basic principles of physics.

This line of critique had also been aimed at such voluntarist approaches to religious propositions. In response to Pascal's advice, J. L. Mackie writes: ‘Deliberately to make oneself believe, by such techniques as he suggests – essentially by playing tricks on oneself that are found by experience to work upon people's passions and to give rise to belief in non-rational ways – is to do violence to one's reason and understanding’ (Mackie 1982, 202). Even if we agree that such indirect control of one's beliefs is possible, the argument would accuse such practice of being irrational or at least intellectually dishonest. Let us examine whether we have any grounds for describing our protagonists and others this way.

As stipulated before, the cases of Bud and Cath do not seem to involve any kind of conclusive evidence – neither their belief nor their faith is grounded in considerations of matters of fact. In this sense, the case of someone who tries to convince themself of the reality of the Divine Trinity is different from me convincing myself that the Earth is flat; neither Bud nor Cath need to delude themselves about the evidence they have. If we accept the premise that religious dogmas are not accepted based on evidence because of their evidential ambiguity, we have no reason to accuse people who believe or disbelieve them of being epistemically erroneous or irrational.¹³ As Stephen Davis writes in his interpretation of Pascal, considering prudential reasons for a belief in a certain proposition seems to be epistemically permissible if the truth of this proposition cannot be established by reason or empirical investigation (Davis 1991, 29–30; cf. Jackson 2019). If no evidence rationally forces me to believe that ~d, then I have no specific reason to trust that my currently possessed belief that ~d is actually correct; therefore, it is rationally permissible to act in the way that causes me to believe that d.

Note that similar considerations may apply to moral or aesthetic considerations. If moral truths such as ‘murder is wrong’ or ‘treating other people as means to an end is impermissible’ cannot be established by reason or evidence, they may nevertheless be practically necessary to adopt regardless of one's beliefs. A person who finds themself disbelieving that murder is wrong may nevertheless strive to convince themself of it without irrationality; in fact, they seem to be morally justified and even obliged to do so.

The above considerations show that indeed the examples of religious Moorean assertions do find their justification within the contemporary accounts of religious faith, while one's indirect voluntary control over one's religious and moral beliefs required by these accounts has been anticipated by many important religious thinkers and leaders. What is worth noting here is that these accounts come from a plurality of traditions and therefore might be seen as universal considerations on the matter of religious – not only Christian, Buddhist, or Muslim – faith.

Conclusion

What is the conclusion of these considerations? In my view, it should be regarded as threefold. First, and most straightforwardly, it provides an analysis of a neglected group of cases that are religious Moorean assertions. Focusing on them is important from the perspective of an analysis of religious discourse and faith, as well as more general considerations in the philosophy of language and epistemology. That is because presenting the possibility of rationally asserting Moorean propositions in religious contexts seems to pose a twofold problem: it either stands as a counterexample to many solutions of Moore's Paradox based on the otherwise plausible accounts of assertion and belief or it requires a change in thinking about the specific characteristics of religious speech and faith by philosophers of religion. As I showed, there are plausible scenarios in which Moorean propositions may be asserted by religious followers having doubts about the propositions they are obliged by their creed to believe. The possibility of such scenarios was shown to be incompatible with holding two widespread assumptions: that assertions made in religious contexts express faith and that faith ought to be characterized as entailing belief. In such cases, one is therefore presented with the following trilemma: either one counterintuitively needs to argue that such scenarios demonstrate the irrationality or insincerity of doubting followers, or one needs to abandon one of these two plausible assumptions.

The second aim of this article was to investigate the consequences of rejecting these two views to prevent the conclusion that doubting religious followers are irrational or insincere. I argued that although one may present a motivated modification to the view that religious assertions express faith by postulating a gap between confidence thresholds needed to rationally assert and believe religious propositions, such a proposal comes with significant costs and should be regarded as unsatisfactory. This is primarily because such an option assumes the possibility of precise quantitative ascriptions of subjective probability to the religious propositions, and it fails to accommodate cases of commissive Moorean assertions, which are also intuitively possible in religious contexts.

In the last two sections of this article, I explored the possibility of challenging the doxasticism about faith in a way that could save the rationality of the above-mentioned doubting followers. I argued that the appropriate way of characterizing faith, which allows both omissive and commissive Moorean assertions, needs to outline its voluntary character, which ought to be contrasted with the involuntary character of belief. I showcased relevant contemporary interpretations of faith as an intention to form religious beliefs (Vahid 2023) and demonstrated how such an account may do justice to the intuition that faith may bring certain doxastic obligations. I also showed how it can save the rationality and sincerity of doubting followers but at the same time describe the internal tension they may find themselves in by using the notion of epistemic akrasia (Chislenko 2016). In the end, I argued that such a stance needs an appeal to indirect doxastic voluntarism, and I showed how needed versions of this stance may be found in Christian, Buddhist, and Muslim theological considerations.

What should we regard as a wider consequence of these considerations for the philosophy of language or religion? I think that the most obvious point is that religious discourse requires a lot more attention and context-specific treatment in the philosophy of language than it receives at the present moment. The above considerations clearly show that the peculiar epistemic context of religious faith raises problems for the analysis of religious speech acts in terms usually employed by philosophers of language and epistemologists. I argue that religious assertions should be analysed solely as expressions of faith, but not belief; the possibility of rational Moorean assertions, in my opinion, clearly shows that such assertions need to be clearly distinguished from assertions uttered in other epistemic contexts. Furthermore, these considerations clearly show that the traditional endeavour of analysing religious speech, experience, and faith in epistemological terms of ‘belief’, ‘credence’, and the like seems infertile and, as described in the opening quote from Wittgenstein, brings havoc instead of needed clarity. But if one is ready to provide a more open and fine-grained characteristic of it, then I think one might eventually be freed of ‘the irritation of the intellect’.