Dost of the conventional arguments in favour of theism attempt to prove that God exists. In addition, there are a number of other arguments that make us consider the merits of becoming theists for some practical interest. Chief among these arguments are Pascal's wager and Kant's moral proof of God.
One could assume that the similarities between Pascal and Kant already end with the reference to practical interest. For while Pascal's wager is in favour of theism on the basis of our rational self-interest and the mathematics of infinite expected utility, Kant's argument turns us into theists in the face of a moral need. But there is, as I will argue below, a much deeper philosophical relationship between the two. For example, although Kant's argument for theism ultimately rests on our moral interests, at its centre is the requirement that we satisfy our self-interest by adopting beliefs of the sort that we should also adopt according to Pascal's wager: namely, that those who are "pleasing to God" will be rewarded in the afterlife.
The aim of my considerations is therefore to bring these two most famous practical arguments in favour of theism into conversation with each other, which will then lead to a reformulation of Pascal's wager in the light of the Kantian moral argument. As I will show, such a reformulation has a number of advantages, without violating the original character of the wager as a "decision under ignorance". This includes overcoming the objection of the many gods as well as the objections raised by Kant against the wager.
I begin with a brief overview of Pascal's wager and its decision-theoretical matrix. I then turn to Kant's moral proof of God and its philosophical relations to the traditional wager. From there, I will develop a moral version of the wager and clarify how it deals with the many gods objection. I will then discuss the evidentialist critique of such arguments. It is contained in the so-called Clifford's principle. According to this principle, it is always, everywhere and for everyone wrong to believe anything on the basis of insufficient evidence. I want to raise a number of concerns about this principle of evidentialism, including whether it lives up to its own standard or whether Clifford's principle is self-contradictory.
Pascal's bet
The usual decision-theoretical formulations turn the traditional bet into a "decision under ignorance". This means that we do not know or do not need to know the relevant probabilities, apart from the assertion that they are greater than zero. Accordingly, the decision matrix of the bet can be represented as in the box below.
The choice of theism results in either + ∞ (a heavenly afterlife of infinite value) if you are right, or if you are wrong (P1), the finite value that our mortal life usually has. The choice of atheism, if wrong (P3), to eternal suffering if someone believes in hell, or to a finite value if our existence ends with our own death. If the atheistic choice is correct (P2), it leads to the value that our existence in this world normally possesses.
Assuming that the probability of God's existence is > 0, the expected utility of choosing to be a theist is + ∞, while the expected utility of choosing not to be a theist is either - ∞ if one believes in hell, or a positive finite value if no continued existence after death is assumed. Thus, the wager provides a prudence argument in favour of theism: however improbable the existence of God may be, as long as it is not impossible, striving for it at the slightest chance of an infinite reward deserves preference over the decision not to believe in God.
Kant on Pascal's wager
Kant mentions Pascal by name about a dozen times. On at least one occasion he casually expresses his reservations about Pascal's apologetic strategy. The most revealing passage, however, can be found in the Metaphysics of moralswhere Kant writes about our duties towards ourselves. The passage begins with the remark: "The greatest violation of man's duty to himself, considered merely as a moral being [...], is the antithesis of truthfulness: the lie" (AA 6:429). As an example of an "inner" lie, he then cites "professing faith in a future world judge [...] in order to curry his favour at all costs" (AA 6:430).
Of course, it can be argued that Pascal does not mean that the bettor is lying to himself about his beliefs in order to gain God's favour. Rather, according to the usual interpretation, the purpose of the wager is to persuade free spirits to open the door to theism, initially by recognising that this might be in their self-interest, but with the ultimate aim of putting them on the path to a sincere faith. But even after this clarification, there are at least two further Kantian objections that can be raised against the wager.
The first objection is that the wager portrays religious consent as too subjective, too close to wishful thinking. The second objection is that the wager transforms our endeavour to become "pleasing to God" into heteronomy. As we shall see, these objections have also been raised against Kant's moral argument. Before I deal with the two objections, more should be said about the moral proof of God.
In the canon of Critique of Pure Reason Kant contrasts what he calls "conviction" with what he calls "persuasion". Here he describes conviction as a valid form of consent because it is based on an objective reason, and persuasion as an unauthorised form of consent because its reasons are only subjectively valid (A 820 / B 848). According to what has just been said, then, the error of the traditional wager consists essentially in its appeal to our self-interest as a merely subjective ground of belief. Can the same now be said of Kant's
moral argument?
In an important footnote to the Critique of practical reason Kant responds to his younger contemporary Thomas Wizenmann, who criticises Kant's appeal to faith as being no different from a lover's need to believe that his feelings are reciprocated. In his reply, Kant rejects the analogy of the lover and instead explains that, in contrast to the selfish and subjective need to believe something to be true, rational faith is "a need of reason, arising from an objective determinant of the will, namely the moral law" (AA 5:143). That is, while Kant sees faith as a matter of choice as well, the ground of choice is quite different. Instead of wishful thinking or some other subjective personal need, faith is rationally justified. The justification is from practical reason rather than theoretical reason, so for a person to reject theism would be morally wrong. As I will explain in more detail in the next two sections, (a) Kant's moral argument for belief in God and the immortality of the soul provides the argument for this objective ground of determination, and (b) the wager can perhaps be restaged to meet Kant's requirement that belief also needs an objective ground of determination.
Kant's moral argument
Kant argues in favour of the moral significance of faith in all three Reviews and in numerous shorter works throughout the critical period (since 1781). Behind this is Kant's assertion that there must be certain limits to knowledge in order "to make room for faith" (B XXX). Such limits are often referred to as the "restriction thesis" of transcendental idealism: We possess neither knowledge nor theoretical cognition of objects that lie beyond the limits of possible experience. While Kant otherwise holds that we should not assent to claims that have not been justified by "objective determinants" that can be convincingly communicated to other rational agents (A 820 / B 848), this rule is suspended (or, one might say, reinterpreted) when it comes to claims about the supersensible.
With regard to the "space" for belief freed up by the restriction thesis, Kant argues that there are valid non-epistemic reasons to believe in God. Which of these reasons is emphasised varies from text to text, but they all share the further thesis that finite rational agents face various "obstacles" to following morality. These include: Self-interest, inclination, despair due to our encounters with the evils of the world, and our interest in the efficacy of our moral endeavours. Thus, the general form of the moral argument can be represented as follows:
(1) We should obey the laws of morality.
(2) Nevertheless, various obstacles stand in the way of our moral resolve.
(3) These obstacles are easier to overcome if one believes that each person receives the happiness appropriate to his or her moral worth through God and in an afterlife.
(4) While in most cases it is wrong to assume beliefs for which there are no sufficient epistemic grounds, this principle does not apply to claims about objects that lie outside the limits of possible experience.
(5) So we have sound moral reasons to accept faith in God and life after death.
Now, one could object that by tying our moral obedience to the idea of a reward after death, Kant deviates from the kind of moral purity that he describes in the Foundation according to which we should act out of duty and not according to externally determined interests. This objection is reminiscent of one aspect of Kant's own critique of the wager in that, according to both, practical assent to theism is driven (in whole or in part) by the self-serving desire "to curry its favour at all costs" (AA 6:430).
Of course, there is a crucial difference between the traditional wager and the moral argument in this regard. As already mentioned, the wager as it is commonly understood would lead us to turn to theism out of self-interest. This is certainly a level or stage that can be found in the moral argument, but it is not the end of the story. Rather, Kant holds that human nature is such that our self-interested need for happiness cannot be extinguished. While the Foundation seems to paint a rather idealised picture of our actions, Kant's broader view is that we have a way of dealing with the conflicting influence of self-interest
have to find.
Unlike the traditional wager, however, the Kantian moral proof of God is not driven by this interest, but by the more complex assumption that we should adopt a worldview for moral purposes according to which our self-interest no longer prevents us from morality. It is not that our moral interest is exhausted and self-interest takes its place to provide the energy needed to act (which would render our will impure or other-directed). Rather, by adopting a religious worldview, we are better able to break through our moral barriers because religion shows us a very concrete way to make the incessant pressure on our will disappear.
The moralisation of the bet
The gist of the traditional wager is that we should become theists out of our rational self-interest. In contrast, according to Kant's moral argument, we should become theists for the sake of our moral resolve. And yet there is an important level at which the two arguments overlap. For although Kant rejects the wager as a stand-alone argument for theism, his moral argument contains the same consideration as the wager: namely, that it is in our rational self-interest to give up happiness in this life for the sake of greater happiness in the next. To be clear, this is not the whole moral argument, but only one step in it, for it justifies theism not on the grounds that it is in our self-interest, but on moral grounds, because it gives us a way to lessen the influence of self-interest on our moral resolve.
Of course, Kant does not express the reference to self-interest equally in the logic of infinite utility, but this, we can say, is the maximisation of how theism finds a support in self-interest; and self-interest was the lever used by Pascal with his target audience, the freethinkers of his day, in mind. But now imagine repurposing the argument for a different audience: a community of non-theists committed to morality and looking for a way to control the obstructive effects of their self-interest. What might an argument for theism in the style of the wager look like, aimed at the latter group?
To this end, let us contrast the traditional bet with its moralised version. Here 'chooses1' as much as 'chooses on the basis of self-interest', 'chooses2' as much as 'chooses on the basis of morality' and 'TH' the theological hypothesis that one will receive an eternal reward in the hereafter if one lives "pleasing to God".
I will begin with a few brief remarks on the Mathematics of the above matrix. Firstly, it is worth noting that both in the traditional ('chooses1') as well as the moral ('chooses2') version can have the same expected utility. Second, while Kant's moral argument does not necessarily depend on the mathematics of infinite utility, he does connect our otherworldly happiness to the blissful state he describes as "complete good" (5:123). Ultimately, however, the decision table above is meant to illustrate what is intended by a moralised version of the wager: the same decision table, the same structure overall, merely a shift from one principle of choice to another.
Therefore, one may ask whether the moralised version of the bet differs from its traditional counterpart in any other way than by shifting wählt1 to choose2? For reasons of space, I won't go into this topic in depth, but will content myself with saying that much depends on what else is considered a building block of the traditional bet. But this is a contentious area. For according to both Pascal's account and the general view in the literature, the traditional wager seems to leave undefined what exactly is required to become "pleasing to God". Is it only the act of committing to theism? Is it a certain (contingent?) inner transformation in the bettor that results from this act? Are there any practices or rituals that are necessary for salvation, such as baptism or adherence to Sharia law?
This leads us to the three main denials of the validity of the wager: the so-called mixed strategies objection, the many practices objection, and the many gods objection. That is, as long as betting is viewed strictly as a decision under ignorance, we seem to lack the necessary means to determine which of the many different possible options the bettor should choose. I will leave aside the first two objections and focus on the many gods objection, since the solution I will offer to it can, in my view, also be applied to the first two objections. As I will show, the moralised version of the wager offers a solution to the usual aporetic consequences of the many gods objection.
The objection of the many gods
Let us first consider the standard formulation of the objection, according to which the behaviour required by one TH to become "pleasing to God" is condemned by the deity of another TH.
Suppose, for example, that TH1 is the traditional theological hypothesis and TH2 is the "invented" theological hypothesis of a deity who requires that we always wear purple slippers while in the house. Of course, a decision matrix that takes into account the many gods objection could theoretically extend to infinity if one considers all the theological hypotheses that someone could "invent". But for the sake of simplicity, I will limit myself to three (mutually exclusive) options, shown in the box below.
If the probabilities of TH1 and TH2 jare each > 0 and P1 to P7 are equal to one or the other finite value, then, according to the objection of the many gods, the expected utility of the two upper lines is equal. Furthermore, to simplify the calculation, if we assume non-existence and not hell, in case someone is wrong, the top two lines each give + ∞ and the third line gives a finite value. This confirms theism over atheism. Of course, there seems to be no decision-theoretical solution as to which deity, religion or set of rituals someone should choose, so that the bettor lacks the means to determine which theological hypothesis best suits his interests. If we restrict ourselves to the content of the bet, it looks as if the bettor is in a fix because he cannot choose between a traditional theological hypothesis and a strange "invented" hypothesis about wearing purple slippers.
Although various strategies have been put forward in defence of the wager against this objection, none have proved satisfactory. They either violate the principle that the wager should be a decision under ignorance, they restrict the wager to a particular audience that does not consider competing theological hypotheses (which runs counter to the wager's goal of persuading the unbeliever), or they can block some hypotheses on formal decision-theoretic grounds, but not all. The moral wager can do better.
So what if we judge each TH not in terms of its utility for self-interest, but in terms of whether or not a TH is compatible with the moral interest? That is, what if the principle of choice chooses2 instead of choosing1 would be? Is there a way to refute the objection of the many gods with this shift, i.e. with a moralised version of the wager?
Let us first consider that choosing theism over atheism on the grounds that this choice is necessary to overcome various moral obstacles would not be compatible with such deities as the god of purple slippers. Since the moral wager is about aligning self-interest with morality, a TH would be chosen whose deity distributes rewards based on moral merit and not on some arbitrary act of piety. While self-interest could theoretically be satisfied by any theological hypothesis, since the moralised wager aims to satisfy self-interest in such a way that it overcomes an obstacle to moral resolve, the theological hypotheses under consideration must be narrowed down to those that involve our self-interest in a way that leads us to morality.
However, this would not be the case with the theism of the purple slippers. For even if such a bettor may forgo this-worldly happiness for the sake of future infinite reward, he has no reason to align his behaviour with morality. Non-moral "invented" hypotheses will probably even tend to perpetuate the primacy of self-interest over morality, and are therefore not suitable candidates for a type 2 choice. In other words, just as in a type 1 choice any TH that is not optimised for self-interest would be rejected by the traditional bettor on the basis of its decision principle, in a type 2 choice the moral bettor would accordingly reject any TH that does not fit the moral interest.
Instead of the "invented" hypotheses from the recent literature on Pascal's wager, what about the field of different religious traditions, Judaism, Christianity, Islam, Buddhism and so on? Which religion should we choose? According to many Christians, our "guilt of sin" is infinite and can therefore only be cancelled by participating in the atonement brought about by the crucifixion of Jesus. This may include baptism, receiving Christ "into one's heart" or other acts that lie beyond the realm of morality. Similarly, some Muslims believe that pious observance of certain rituals is necessary, and Buddhists may conceive of enlightenment in a way that is distinct from morality. So it seems that some aspects of the many gods objection remain. For example, a moral atheist or a Muslim might not be able to enter the Christian heaven.
But there is a Kantian rebuttal. One of the central motifs of Kant's Religion within the limits of mere reason states that all "true" religions, unlike "mere cults", agree with the only "pure rational system of religion" (AA 6:12). They are "vehicles" for moral truths and convey these truths "because of the natural need of all men always to demand something sensual and tangible [...] in addition to the highest rational concepts and reasons" (AA 6:109). Although religions also contain various commandments of piety, such characteristics of a religion are all "in themselves accidental" and "arbitrary prescriptions" (AA 6:105f.). They can contribute to the formation of a religious community, but no rule of piety (e.g. dietary regulations) and no ritual (e.g. baptism) are necessary to become "pleasing to God". Kant emphasises this point emphatically in the Fourth Part of the Religious scripturewhen he criticises the value and function of ecclesiastical practices: "Everything that a person thinks he can do to please God, apart from living a good life, is mere religious delusion and an after-service to God." (AA 6,170f.).
Thus, the moral bettor may choose a religion for the appeal of its particular creeds and practices, but what matters about a religion is its assistance in one's moral endeavours. Therefore, there is no aporia here, since the choice is ultimately about adopting a religious perspective that helps one morally, not about the mutual exclusivity of traditional doctrines or practices that fuel the many gods objection.
The formal reason why the traditional bet is susceptible to the many gods objection (as well as the mixed strategies and many practices objections) is that there is no intrinsic connection between the bettor's goal (self-interest) and the behaviour required to obtain the reward. The moralised bet, on the other hand, offers such a connection. For if the reason for the choice is to promote morality, the behaviour required for the reward must reflect this. If Pascal had designed the wager for righteous atheists rather than freethinkers, we can imagine that he would have proposed the moralised wager instead.
But what about other objections? I have already discussed, in addition to the objection of the many gods, Kant's own criticism of the wager as intellectually dishonest, as well as the accusation often levelled against the moral argument that it taints Kantian morality with a dose of external determination. In the following, I will discuss another objection, perhaps the best-known objection to practical arguments in favour of faith in general, the principle of the English mathematician and philosopher William Clifford.
Clifford's principle
I repeat once again the Clifford's principle: "It is always, everywhere and for everyone wrong to believe anything on the basis of insufficient evidence." There are two points I would like to reflect on. The first is whether Clifford has presented a satisfactory argument for his principle, more precisely whether his argument is self-consistent or not; and secondly, even if Clifford's principle is applicable to the majority of cases, Clifford's own justification for it perhaps allows for more exceptions than he noted.
As the title of his essay suggests, Clifford presents his principle not as a fact or a state of affairs, but as a normative assertion. But if it is always and everywhere wrong to believe something without sufficient evidence, this should also apply to normative assertions. So how does Clifford argue in favour of his principle? Actually, his argument is consequentialist and relies on the expectable effects of belief for reasons other than evidence. As an example, he cites a ship that sinks - despite the owner's "will to believe" that it will not sink. Another example is the unjust persecution of people on the basis of unfounded accusations. Clifford draws the general conclusion that if our society developed a habit of "believing on unreasonable grounds" (Clifford 1877: 173), the negative effects would tend to increase. He further adds that if we "lost the habit of scrutinising and examining things", society would be doomed to "relapse into savagery" (Clifford 1877: 174).
I think it's easy to see how Clifford's concerns are borne out when we look at times in history when a population has accepted certain dogmas that stem more from ideology than empirical study and that have prevailed through groupthink or other psychological mechanisms against the norms of critical thinking.
But does Clifford's "ethics of faith" remain coherent in itself? The form of the argument is empirical and consequentialist. Clifford mentions various (real and hypothetical) examples of unevidenced beliefs and considers the implications. But how good is his own evidentialism here? His approach is anecdotal. Based on a few examples, he attempts to provide a universally valid argument against unevidenced beliefs. This certainly deserves to be criticised as a hasty and even one-sided generalisation. What if, on the other hand, it were true that practical beliefs in some areas of life improve the effects of one's endeavours? And what if a rule were established that narrowed down practical beliefs to an area where such beliefs tended to improve rather than worsen the outcome?
Think of William James' classic example that you are more likely to successfully jump over a precipice if you believe you can do it. Accordingly, there are many studies that show that a "positive attitude" improves the outcome of treatment by around 5 to 10 per cent. When we compare subjects with similar diagnoses (and eliminate confounding factors), optimism has been shown to improve the rate of treatment success. So why not limit the areas where practical faith is permissible to where such psychological mechanisms are effective? Indeed, we can use Clifford's own evidentialist and consequentialist reasoning to support the use of practical faith in such cases.
Now to the crux of our discussion of Pascal and Kant: what about objects that lie outside the realm of possible experience? Clifford might argue that the only correct position is that of agnosticism. To the extent that he bases the justification of his principle on expected utility, of course, it doesn't take much to conclude that the decision matrix of the traditional wager tells us that theism is the better choice. After all, the point is that its TH brings an infinite expected utility. Both the moralised wager and Kant's actual moral argument assume that our moral strength is increased by theism. Surely this is a result that Clifford should honour.
One of Clifford's concerns, of course, is that religious assent is part of a more systematic weakening of our epistemic scruples. For if we are told to accept something of the highest importance on blind faith, this can just as easily encourage a distrust of reason and a habituation to accepting what is said by religious (and political) authorities. But it is precisely here that Kant's distinction between the possible objects of knowledge and opinion on the one hand and belief on the other is so important.
As mentioned above, Clifford generalises from a handful of examples to a universal principle. This conclusion does not itself obey Clifford's principle and thus raises the question of what exceptions there might be. What Kant holds is, first, that once a limit to knowledge (at least in possibility) has been established, there is "room" for belief; and, very importantly, he holds, second, that it is not as if anything takes that place. There are both "adequate" and "inadequate" grounds for belief, beliefs based on "merely subjective grounds" and beliefs based on the "objective ground" of the moral law. Beliefs based on subjective grounds are to be rejected because Kant also wants us to develop epistemic virtues. On the contrary, using the moral argument, Kant argues that there is an objective reason for practical consent in this case.
Accordingly, (a) Kant provides an argument for the value of theism; on the basis of Clifford's interest in expectable utility, (b) the argument may well fulfil Clifford's own requirements for "adequate" reasons; and most convincingly, I think, (c) instead of the arbitrariness that Clifford is afraid of (mainly because of an inaccurate understanding of James's "will to believe"), we have Kant's narrowing of belief to only those objects that are consistent with our interest in overcoming moral obstacles.
This point is decisive for Kant's attitude towards religion as a whole and is the guiding theme of his Religion within the limits of mere reason. For by examining which Christian doctrines have "unity" or "compatibility" with moral religion (AA 6:13), Kant actually provides a map of these doctrines and clarifies which of them are conducive to our moral interest and which are not. In this way, Kant evaluates the doctrines of original sin, the reception of grace, the incarnation, vicarious satisfaction, the role of ecclesiastical rituals, the function of prayer, and so on. Some of these doctrines prove to be compatible with what he calls "true" moral religion, while others are more akin to a religion of "favour seeking" or a "mere cult".
In this way, Clifford's practical concerns, and with them his practical argument against belief, are blocked without evidence. Only those objects or propositions qualify for the "place" of faith that (a) do not fall within the realm of evidentialist enquiry and (b) correspond to the most appropriate of all reasons, namely our interest in overcoming the obstacles to moral observance and our pursuit of moral perfection.
Conclusion
I have pursued three goals: first, to outline Kant's explicit and implied objections to Pascal's wager; second, to discuss some conceptual connections between the traditional wager and Kant's moral proof of God; and third, to present what I think is a novel and interesting sister to the traditional wager that even Pascal would have liked. According to this, the apologetics of the wager was not directed at freethinkers, but at those whom Kant had criticised in the Critique of judgement called "righteous" atheists (AA 5:452). Furthermore, the moralised version of the wager has been shown to have a number of advantages over its traditional counterpart: (a) it seems less "acquisitive" if religious consent, rather than self-interest, is based on our moral betterment; (b) the moralised version can address Kant's concerns about the traditional wager; and (c) the moralised wager provides a way of dealing with the many gods objection without compromising the argument as a "decision under ignorance".
Finally, I have argued that Clifford's own argument in favour of evidentialism is inherently inconclusive, since its consequentialist framing leaves room for practical belief, contrary to Clifford's intention. For although many of us would agree that evidentialism is intuitively appealing and is the appropriate norm in at least some contexts, it would be wrong to recognise it as the universal norm for belief in the absence of a better argument than the one Clifford gives. (Translation: Georg Sans SJ)
