A Non-Doxastic Fear of Hell: On the Impact of Negative Factors for an Agnostic Religious Commitment

: On the standard view, an agnostic might commit non-doxastically to religion because she wants to receive some goods, which might be either natural or supernatural in kind. I broaden the picture by showing how the agnostic must also take negative factors into account. Negative mundane factors should be avoided as far as possible by the agnostic, and in extreme cases, even at the price of giving up supernatural goods. Negative supernatural factors, like eternal torment, work di ﬀ erently. An agnostic who considers an eternity of su ﬀ ering in hell a live possibility might rationally make a religious commitment in order to avoid it. Non-doxastic religion is commonly conceived as requiring a pro-attitude. If fear can have the impact I suggest, we must broaden the picture to allow for a negatively based commitment as well. To make explicit the kind of attitude relevant here, I o ﬀ er an analysis of fear as a rational, non-doxastic attitude.


Introduction
In recent literature, the idea that a religious commitment does not require religious belief is steadily on the rise. According to religious non-doxasticism, what is vital is not that the subject believes p, but that she considers p an epistemic possibility. As long as this is the case, a religious engagement can be based on epistemically weaker attitudes than belief, like hope or acceptance (Alston 1996;Howard-Snyder 2013;Palmqvist 2019a). In this way, non-doxasticism makes religion available also for the agnostic. Taking it for granted that a non-doxastic engagement with religion is both psychologically possible and rationally permissible, I will discuss the factors motivating an agnostic to make such a commitment.
Echoing Pascal's famous wager, much has been said about what one might gain by a non-doxastic religious commitment. Sometimes the benefits are purely mundane, to be gained regardless of whether the propositional content of the religion in question is true or not (Schellenberg 2009). The focus, however, tends to be on the extraordinary benefits one might receive if betting on the true religion, so to speak. These benefits include salvation and eternal life (Schellenberg 2009, p. 114;Ballard 2017, pp. 223-25;Howard-Snyder 2017, pp. 56-68;Palmqvist 2019b). We can call the mundane benefits which do not require the truth of the relevant religious propositions for "natural goods," as opposed to the truth-dependent "supernatural goods." 1 While I am largely sympathetic to non-doxasticism, I think there is much left to be said about the rational considerations which might lead an agnostic to take on a religious commitment. Since the 1 main ambition of most proponents has been to show how non-doxastic faith can be rational and why an agnostic should consider it, they have standardly focused on the positive things one can gain by such a commitment. The negative factors which should also influence the choice have not received much attention, 2 which is regrettable since they too carry significant force. Expanding the terminology, I will make a distinction between natural and supernatural "bads." Avoiding the natural bads should ceteris paribus be considered just as important as acquiring the natural goods. Avoiding supernatural bads can be more crucial than achieving the corresponding goods, and there are even cases where the possibility of supernatural bads might force a rational subject into a religious commitment. Or so I will argue.
On the standard account of non-doxasticism, the subject needs a pro-attitude as a basis for her religious commitment, an attitude often analysed in terms of faith or hope. The introduction of significant negative factors makes clear that a non-doxastic commitment might also be rationally based on negative attitudes such as fear. This is a significant result. To make explicit what such negative non-doxasticism might look like if the details are spelled out, I offer a full-blown analysis of fear as a rational non-doxastic attitude.
Any discussion concerning what the unbeliever is rational in doing when some supernatural goods are at stake has a natural point of reference in Pascal's famous wager. 3 When discussing the impact of natural and supernatural factors on the unbeliever's choice, I will use a simplified decision-theoretical framework reminiscent of the one standardly employed in discussions on the wager. 4 Contrary to Pascal's original wager, which builds strictly on the principle of maximizing expected utility, I argue that when it comes to the bads, risk-aversion should often be preferred.
The following section contains a short introduction to religious non-doxasticism. It is followed by "A voluntary commitment" where the choice an agnostic must make to engage with non-doxastic religion is examined in detail. In "The impact of the natural bads" and "The impact of the supernatural bads," the negative factors the agnostic must consider are briefly presented, and their impact explained. The penultimate section, "Non-doxastic fear", introduces fear as a major non-doxastic attitude. The text ends with a summary.

Religious Non-Doxasticism
The core idea of non-doxasticism is that one can make a religious commitment even if one lacks religious belief. All that is needed is that the agnostic considers the truth of her chosen religion an epistemic possibility.
To understand the basic thrust of non-doxasticism, it is helpful to contrast it against another form of belief-less religion, namely religious fictionalism. 5 Fictionalism treats religion as fiction, and religious life as a game of make-believe. On fictionalism, the literal truth of religious propositions is as irrelevant as the literal truth of Dostoyevsky's The Brothers Karamazov (Eshleman 2005; Le Poidevin 2019). On non-doxasticism, however, the epistemic possibility of religious truth is all-important. The non-doxasticist plays no games, she tries to interact and align with divine reality, 2 Jay (2014, p. 212) is the only author of which I am aware who mentions the possibility that the goods one can gain might be outweighed by other factors. 3 As noted by Schellenberg (2009, pp. 175-76), on non-doxasticism the possibility of making a voluntary choice can be assumed without controversy, thereby avoiding many of the problems besetting the original wager. Also, since on non-doxasticism we are only concerned with what the subject perceives as epistemic possibilities, the many gods objection loses its force too. 4 For a discussion of Pascalian wagers with a heavy emphasis on the decision-theoretical framework, see Sobel (2004, chp. 13). 5 It should be noted that some philosophers, like Jay (2014, p. 210), define fictionalism so broadly that it includes non-doxasticism. My arguments could easily be restated in Jay's terminology, but I do think such reformulation would result in a loss of clarity. It should also be noted that technically, what I call non-doxasticism could perhaps more accurately be called sub-doxasticism, due to its reliance on beliefs about epistemic possibility (Eklund 2017, p. 28). Fictionalism, on the other hand, is truly non-doxastic in that the approach does not include any beliefs at all. However, it is standard terminology in the field to use the term "non-doxasticism" to denote religiosity based on epistemic possibility, and I will stick to it to avoid misunderstandings. just like the believer. What sets her apart is that she lacks belief concerning the existence of such a reality-her religiosity rests on a weaker cognitive base.
Regarding the analysis of this cognitive base, how the details are spelled out and even what it is called differs substantially between accounts. While most accounts come in terms of non-doxastic faith, such faith is usually analysed as a composite attitude. 6 Arguably, it is the cognitive component of this composite attitude which is the real cognitive base of non-doxasticism. There have been many suggestions concerning the nature of this cognitive component, the most important identifies it with hope (Pojman 1986), acceptance (Alston 1996), imaginative voluntary assent (Schellenberg 2005(Schellenberg , 2009), or belief-less assuming (Howard-Snyder 2013. 7 Their differences aside, the attitudes suggested by proponents of non-doxasticism share a common core, consisting of two basic conditions. 8 It has recently been summarized as follows: For S to have a non-doxastic pro-attitude towards some state of affairs p, it is necessary that the following two conditions are met: ND1: S desires that p or judges that p is an overall good thing. ND2: S believes p to be epistemically possible. (Palmqvist 2019b, p. 560) 9 The first condition is about desire or positive evaluation. Since non-doxasticism is voluntary, it is generally acknowledged that the subject needs some pro-attitude to have any reason to take it on. It is this condition I challenge by introducing non-doxastic fear. 10 The second necessary condition concerns epistemic possibility. However, exactly what is meant by epistemic possibility varies with the account. I want to highlight the difference between a wide and a narrow sense of epistemic possibility, since this difference plays an important role in my arguments. In a wide sense, p being epistemically possible only means that we cannot conclusively rule p out. In other words, it is a matter of p being "neither known nor justifiedly believed to be false" (Schellenberg 2009, p. 8). The wide sense contrasts with the narrow, which can be spelled out in at least two different ways. One way is to say that p has a non-negligible chance of being true (McKaughan 2013, p. 113), another that p is a Jamesian "live possibility"-meaning that p is actually considered by the subject, it is "among the mind's possibilities" (James 2010, pp. 9-10). While it seems plausible to suggest that the most common reason that something is among the mind's possibilities is because one perceives it as having a non-negligible chance of being true, it is also the case that the conceptions can come apart for various reasons. To keep things simple, I will use the term live possibility to denote the narrower sense, but take it to mean a possibility which is both "live" in the Jamesian sense and in the sense of having a non-negligible chance of being true.
The difference between the wide and the narrow sense should not be underestimated. To appreciate it, consider the example of misplacing your keys. When you realize that you do not know where your keys are, some likely possibilities are bound to pop up in your mind, like "I must have left them in my car" or "they must be in my other trousers," etc. These are epistemic possibilities in the narrow sense. Epistemic possibilities in the wide sense are all the possible scenarios you cannot conclusively rule out, however unlikely. Like "I accidently dropped the keys in the garbage can, from where they were later picked up by an artist, incorporating them into his latest objet trouvé piece." 6 For a notable exception which builds on hope only, see Muyskens (1979). For a recent account which treats hope as an independent attitude complementary to faith, see Palmqvist (2019a). 7 Not all proponents of non-doxasticism are as interested in identifying its cognitive basis. For accounts with a focus on risk and action, see Buchak (2017) and McKaughan (2016). 8 This shared core coincides with the standard account of hope, which suggests that hope is the most basic non-doxastic attitude. 9 Palmqvist's account comes in terms of state of affairs, while many other accounts are propositional. Since I take it to be obvious that non-doxastic attitudes have a propositional content and that they are directed at some possible state of affairs, I think both approaches are feasible. 10 It has also been challenged on the ground that it is possible to have faith in a testimony that something bad has happened (Buchak 2014, p. 53

A Voluntary Commitment
There is a major difference between belief-based religion and non-doxasticism when it comes to voluntariness and choice. Since belief is an involuntary state, belief-based religious life is in an important sense involuntary too. 11 For instance, if you have a full-fledged belief that the eightfold path of the Buddha leads to salvation, your rational course of action is to try to follow it to the best of your abilities (unless, of course, you also believe in karmic reincarnation and are content with improving your karma this life, saving salvation for a future life). You really do not have much choice in the matter.
Unlike the believer, the agnostic always has a choice. Without any belief compelling her to pick up a religious life, she can always choose to remain uncommitted. Also, unlike the believer whose beliefs standardly restrict her to a single religious option, the agnostic who wishes to engage with religion can choose more freely. For her, every religious option which represents an epistemic possibility is at least prima facie available. However, if she is to remain rational, there are a great many factors she needs to weigh in when making her choice.
The voluntariness inherent in non-doxasticism means that the agnostic needs some reason to engage with religion. Without any such reasons, she will presumably remain uncommitted. This is where the natural and supernatural goods enter. 12 There is no doubt much to be gained from engaging with religion, even if one does not believe. The natural goods are earthly benefits one receives already in this life. Spiritual growth and moral awareness are stock examples here. 13 The supernatural goods are non-natural benefits the agnostic might receive if her chosen tradition turns out to be true, like eternal bliss in an Abrahamitic context or good reincarnation in a karmic tradition. 14 To facilitate discussion I will focus exclusively on maximally good supernatural goods like salvation, eternal life, moksha, nirvana, etc., leaving lesser supernatural goods like blessings out of the picture. I will assume that these greater supernatural goods are exceedingly more valuable than any natural goods.
As briefly mentioned, natural goods are truth-independent and supernatural truth-dependent. The achievement of the former is unproblematic and can be regarded as quite certain by the non-doxasticist. The supernatural goods, on the other hand, can only be hoped for or pursued in faith. The non-doxasticist's expectations regarding supernatural goods must therefore be on level with the probability she assigns for her chosen religious view. 15 If you have a weak hope that your religious view is true, you can only have a weak hope of receiving the supernatural goods associated with it.
While proponents of non-doxasticism have always been careful to show how an agnostic religious commitment can be motivated by natural and supernatural goods, the corresponding negative factors have not received much attention. This is unfortunate since a religious commitment can come with many negative consequences for the individual. The agnostic who considers a fictional engagement should not only think about what she might receive, she must also pay attention to what negative consequences she might face. And not only should she take into account the possibility of achieving 11 While doxastic voluntarism has few defenders left, it is sometimes suggested that belief can be voluntary in a weak sense.
Even though belief is not subject to direct voluntary control, one can still influence one's beliefs indirectly, by taking the right kind of action. I will leave this possibility out, mostly because my interest is in epistemically rational religious commitment, and I very much doubt that one can gain a belief as envisioned by weak voluntarism without violating the demands of epistemic reason. 12 The categories should not be read as covering every positive reason. For example, one might commit because intellectually in love with the idea of God, and such reason is not easily reducible to the achievement of goods. 13 For the most thorough account of which I am aware of why an agnostic should engage in non-doxastic religion, see Schellenberg (2009, chp. 6-12). Schellenberg's account contains both natural and supernatural considerations. 14 Most philosophers working on non-doxastic faith are Christian, and they standardly presume non-doxastic faith to be saving faith. See Howard-Snyder (2017) or Alston (1996). 15 Unless she believes there to be other possible views which will grant these goods if true, and that her commitment is "close enough" to let her partake. For example, a non-doxastic liberal Protestant might believe that as long as some kind of theism turns out to be true, she will receive supernatural bliss.
supernatural goods, she must also consider if some negative supernatural bads might affect her if she makes the wrong choice. At this point, an objector might complain that the reasons for committing religiously I am considering have little to do with what is standardly identified as good reasons by the religious traditions themselves. To commit to achieve some goods might seem overly calculating and egocentric, and far removed from, for example, the love of God and the love of one's neighbour that are supposed to be the driving forces behind a traditional Christian religious commitment. In response to this objection, I have two major points to make.
First, I try to keep apart as much as possible the situation the unbeliever is in when making the choice to commit, and her situation after the choice has been made. I only address the former situation, and considerations which are proper in such circumstances. Justification which presupposes the view under consideration to be true cannot be invoked in a pre-commitment context, in order to support commitment. Such reasoning would be blatantly circular. However, after the choice has been made, the non-doxastic Christian will very much take part of Christian tradition, and she can then non-doxastically accept most traditional Christian reasons for commitment.
Secondly, some traditional reasons for a religious commitment are rationally unavailable to the agnostic. A Christian believer might commit since she trusts God, but an agnostic cannot rationally trust God since rational trust requires belief (she can at most non-doxastically accept that God is reliable). 16 To see this, consider that trusting implies expecting certain outcomes. If you trust me with your secret, you will expect me to keep that secret. If you think it is more likely than not that I will not keep it, you cannot be said to trust me. Likewise, if you trust God to bring you salvation, you must expect to receive salvation. As a non-doxasticist, you might hope for salvation, but you cannot rationally expect it since you only view God's existence an epistemic possibility.
Some basic decision-theoretical tools will be used to clarify the different choices an agnostic might face when considering non-doxastic religion. 17 I will assume that the agnostic should be guided by the principle of maximizing expected utility, except when explicitly arguing for some alternative principle (most often risk-avoidance). I will refrain from assigning concrete values to outcomes or calculating the utility in any exact manner. The points I wish to make are very basic and such calculations would only obfuscate them. Instead of determined probabilities, I rely on the distinction between epistemic possibility in the narrow and the wide senses. The former is conceived as being a "live possibility," the latter as being "very unlikely." As a starting point, I want to make explicit how the agnostic should prioritize in a situation with only goods. If the agnostic is unsure whether the possibilities she is considering are "live" or not, i.e., if she makes her choice under ignorance, choosing seems straightforward enough. Given that supernatural goods are always exceedingly more valuable than anything else, the dominant strategy would be to always choose an option which brings such goods, and if there are several such options, the one which brings the best natural goods in combination with the supernatural.
If the agnostic assigns some epistemic probability to the alternatives under consideration, it becomes what decision-theory refers to as a decision under risk. However, it seems plausible to suggest that her priority should still be to pick a view with supernatural goods, given their extreme value. Only if there are several views with supernatural goods do the probabilities become relevant. In such a case, the agnostic should pick an option that is a live possibility over an epistemic possibility in the widest sense, thereby maximizing her chance of receiving the supernatural goods. She should also strive to maximize the value of the natural goods she receives, but since the other two factors are more important, natural goods gets to play the role of tiebreaker: if there are several views around which are live possibilities granting supernatural goods, the agnostic should pick the one with the best natural goods.
We now turn to see how the agnostic's choice is affected by the introduction of natural and supernatural bads.

The Impact of Natural Bads
As a preliminary definition, I think of a natural bad as an aspect of adhering to a religious tradition which the subject evaluates negatively, prior to commitment. Unlike the supernatural bads, the natural bads are part of the package regardless of the truth of the religious tradition and have only natural causes. Arguably, this is a category with a rich and varied content. It can include social factors having to do with being religious in a largely secular society (like harassment), or factors having to do with the religious world-view (like viewing oneself as a sinner), or factors concerning moral and behavioural restrictions (like food regulations). However, I want to stress that what counts as natural bads is ultimately an individual matter, to a great extent determined by factors like background and personal disposition.
An objector might claim that many alleged bads (like food regulations) only appear as bads from the outside, while they are experienced as important and meaningful by adherents to the tradition. If the subject is aware that this change of attitude will take place, it does not make any sense to take action to avoid these things. While I do not want to deny that one might stop to think of a religious phenomenon in negative terms after committing to a tradition, I think such change is more common on a full-fledged belief-based conversion. Since the agnostic who commits non-doxastically never starts believing in her chosen tradition, a re-evaluation seems less than likely. It is also clear that this objection only covers bads inherent to a tradition, and not those depending on the context, like social factors. So even if successful, the objection only delimits the category of natural bads.
What impact does the natural bads have on the choice of an agnostic considering which religious view to commit to? A straightforward suggestion is that they enter the considerations at the same level as natural goods, so that they too become tiebreakers between views where the achievement of supernatural goods is a live possibility. It also seems straightforward to suggest that they carry the same weight as the corresponding goods. That would imply that, for example, the positive value of moderately good goods can be cancelled out by the negative value of moderately bad bads. If so, it seems reasonable to suggest that we ceteris paribus should go for the view with the goods of the highest value after we have subtracted the negative value of the bads. All according to the principle of maximized utility. Pascal [1670Pascal [ ] (1958 famously claimed that supernatural goods have infinite value, and that they therefore should always be pursued as long as the cost is finite, no matter the probabilities involved. We can call this Pascal's principle: Pascal's principle: if the cost is finite and the chance of success > 0, supernatural goods should always be pursued.
The picture I have drawn this far is quite Pascalian: if there is a chance to gain the supernatural goods, such a view should always be preferred regardless of how unlikely, and we have not yet encountered a price that is too high. However, while Pascal's principle is certainly appealing, I am hesitant to accept it as a general rule. Consider a case where the natural bads are maximally bad, and there are no natural goods. Imagine a cult with the bizarre view, that only those who are tortured violently for decades can achieve salvation. Can mere epistemic possibility of achieving supernatural goods weigh more than a life of extreme suffering?
The natural bads of the torture-cult's religious tradition T 1 are as bad as it gets, but T 1 also claims to provide supernatural goods. Let us suppose the agnostic perceives the truth of T 1 a live possibility, and that it is the only tradition she considers epistemically possible.
Should the agnostic commit on this picture (Table 1)? The rewards are exceptionally high, and the chance of receiving them non-negligible. Pascal's principle would clearly suggest commitment. But would we really blame an agnostic who does not commit? Could such an agnostic be accused of being irrational? Or could being risk-averse in this case be at least as rational as maximizing utility?  (Table 2), which differs from the former only in that the tradition of the torture-cult is now considered an epistemic possibility in the widest sense. If the chance to gain supernatural goods is considered extremely unlikely, and the cost is a life of the worst possible kind, I am tempted to say that the price is too high, and I think the situation described provides an intuitive counterexample to Pascal's principle. At least I would not be willing to bite the bullet and try to defend the principle in the light of it. I suspect that the problem with Pascalian reasoning is that it assumes that supernatural goods should always be considered having infinite value. Even if supernatural goods have infinite value in an objective sense, a subject is not necessarily irrational for not assigning them infinite value when making a decision. A starving man might prefer a meal over a million dollars. To account for this, I have only claimed that supernatural goods should be considered exceedingly more valuable than any natural goods (this is so even in the torture-cult example, if we believe with certainty that decades of torture will bring eternal bliss, it seems a small price to pay).
Fortunately, torture-cults are few and far between. In real-life situations, I think Pascal's principle holds, and I take the idea that one should try to maximize one's chances of achieving supernatural goods to be reasonable. But only as long as there are no supernatural bads to consider.

The Impact of Supernatural Bads
Though not exactly at the forefront of contemporary theology, supernatural bads continue to be a part of many religious traditions. In Abrahamic religion, we have eternal damnation and hell, and sometimes purgatory and time-limited hell. Time-limited hells are also common in the karmic traditions, along with lots of other unpleasant ways to be reincarnated. We also have minor bads, like curses or temporal divine punishment. To facilitate discussion, I will approach the supernatural bads in the same way as the supernatural goods and only consider those which are maximally bad (like eternal damnation).
If one believes in the existence of supernatural bads, taking precaution to avoid them is only natural. But what if one only regards them as epistemically possible? How should the epistemic possibility of supernatural bads guide the agnostic's non-doxastic commitment? Unlike natural bads, which only repulse the unbeliever, supernatural bads actually serve as reasons to commit to the view they are associated with. Unlike natural bads, they do not counterbalance the goods but add to their strength as independent reasons. For example, you might consider adopting some traditional form of Islam in the hope of achieving salvation. If you also consider it an epistemic possibility that what this tradition says about hell is true, you will not be deterred. On the contrary, you get a new, strong reason to commit: the avoidance of eternal damnation.
When it comes to supernatural bads, risk-avoidance seems a sound strategy. It even seems reasonable to weigh the avoidance of supernatural bads higher than the achievement of supernatural goods. Suppose an agnostic regards two religious traditions T 2 and T 3 as live possibilities. However, she thinks T 2 is significantly more likely to turn out to be true than T 3 . To keep things simple, let us say that she finds T 2 three times as likely as T 3 . While T 2 only promises supernatural goods, T 3 also threatens the non-committed with supernatural bads. To make things simple, let us suppose that we are talking about heaven and hell, eternal bliss and eternal torment (Table 3). Even though there are individual factors involved, like how much a subject values freedom from pain or her inclination to take risks, it seems at least as rational to go with T 3 as T 2 . To see this point, consider the following analogous example concerning earthly pain and suffering. You partake in a peculiar lottery and get to choose between two tickets. Ticket A has a probability of 0.3 of giving you the billion-dollar jackpot, while the probability of winning on Ticket B is 0.1. However, if you do not buy Ticket B and B is drawn, not only will you lose everything you own, you will also be horribly tortured for the remaining years of your life. Would you dare not to buy Ticket B? Presumably not. And how much more horrible is not a Dante-style inferno of eternal pain than only a lifetime of torture?
If avoiding supernatural bads is more important than achieving the goods, this also suggests that if the choice is made under ignorance rather than risk, avoidance of supernatural bads should be the dominant strategy. That is, if we get to choose from T 2 and T 3 , and we have no clear conception of the probabilities involved, going for T 3 seems highly advisable.
When is it rational to adopt T 3 in order to avoid the possibility of eternal damnation, if the choice is made under risk? I think the difference between a wide and narrow conception of epistemic possibility plays an important role in this regard. Presumably, one is always rational in trying to avoid a negative live possibility. However, it does not seem rational to act to avoid a negative possibility in the wide sense. Acting to prevent a possibility only because one lacks certainty or outright belief that it will not obtain is standardly considered irrational. We tend to treat fear of flying as irrational even though a crash is epistemically possible in the wide sense, as we do the fear of dogs, even though in rare cases, it does happen that dogs kill people. It seems safe to conclude that being guided by fear that highly unlikely scenarios should turn out to be true is not a viable (or healthy) strategy. So, for it to be rational to let the possibility of supernatural bads guide your decision, you need to view their actuality as an epistemic possibility in the narrow sense. This contrasts against when it is rational to commit to receive supernatural goods, where only possibility in the wider sense is needed.
We can call the idea that one should be risk-averse when it comes to supernatural bads for the dominance of fear principle: Dominance of fear principle: If there is a live possibility that a tradition containing supernatural bads is true, it should always be pursued.
Of course, it could be the case that the agnostic has multiple live fears. What if she regards two traditions as live possibilities, and these traditions happen to be exclusivist when it comes to salvation and damnation? In such case, we simply return to the other relevant factors. Do both views provide supernatural goods? If so, which comes with the most natural goods and least natural bads?
An objector might complain that basing one's non-doxastic commitment on fear of supernatural bads seems rather uncheerful, and should one really let fear guide one's religious life? I cannot but concur that the results of this section are slightly discomforting. However, one must not conflate what is rational with what is the most praiseworthy approach. Obviously, being brave in the sense that one does not let oneself be ruled by fear is often more commendable than acting on fear. However, that does not mean that acting on one's fear cannot be the most rational choice. Unlike hope and faith, attitudes that often tend to be both rational and praiseworthy, acting on fear can be rational without being praiseworthy. So, while one might have moral concerns about basing one's religiosity on fear, it is hard to see any obstacle for its rationality.

Non-Doxastic Fear
The introduction of supernatural bads does more than just complicate the agnostic's choice of religious tradition. On the standard account of non-doxasticism, the subject needs a pro-attitude to engage with religion. I have argued that a non-doxastic commitment can also be based on a strongly negative attitude towards p. In the previous section, I discussed the rationality of making a commitment based on such an attitude. I will now focus on the attitude itself, which is fear of the type "S fears p". 18 Although fear is a subject philosophers touch upon occasionally, 19 not many have provided a proper analysis. A notable exception is J. P. Day (1970Day ( , 1998, who treats fear as a parallel but opposing attitude compared to hope, a kind of negative "mirror-image" of the positive attitude. According to the standard account, hope can be analysed in terms of desire and epistemic possibility. 20 Following the mirror-image approach, we get an analysis of fear in terms of aversion and epistemic possibility: 'a fears that p' is true if and only if 'a is averse in some degree, however small, that p and a believes that it is probable in some degree, however small (e.g., 1/1000), that p' is true. (Day 1970, p. 369) The aversion-condition is clear enough. The condition about believing that something "is probable in some degree, however small" expresses the idea that p must represent an epistemic possibility, in the wide sense. This requirement is all-important, since it is this dependency on epistemic possibility rather than belief which qualifies fear as a non-doxastic attitude. Reformulating Day's analysis in these terms, we get: F1: S is averse that p. F2: S believes p to be epistemically possible.
However, it seems obvious that these conditions are not sufficient for fear. F1 seems all too weak. You might well feel aversion against possibilities you do not fear. For example, you might feel aversion against a colleague, and due to an upcoming reorganization, you might have to share your office-space with her. Does that necessarily make you fear that possibility? While perhaps fear is within the range 18 It might be suggested that "dread" would be a more suitable term for the attitude one can base a negative religious engagement on. While this is mostly a matter of terminology, I think we should preserve the intuition that fear is the opposite of hope, and since hope is the basic pro-attitude, it makes considerable sense to label the basic negative attitude my attempted analysis is aimed at as "fear." However, I do think that this basic attitude might be developed into something stronger, just like hope might be developed into faith, and I suggest we reserve the label "dread" for such a stronger, negative attitude. 19 See for example Kendell (1978) for an account on fear and fiction, or Svendsen (2008) for a cultural analysis in terms of fear. 20 One should note that the standard account of hope is questioned since it seems unable to rule out despair. For the most influential contemporary analysis of hope that builds on such criticism, see Martin (2014). For a defence of the standard account, see Milona (2019).
of possible responses, frustration and annoyance seem at least as likely reactions. Or, to take another example, you might feel aversion against stupidity without in any way fearing that new people you meet might be idiots (which is always an epistemic possibility before you get to know someone). What F1 is missing is that fear normally implicates danger. Not necessarily a real danger, since we can easily be wrong about what is dangerous and not, but a perceived danger. Also, the danger must be relevant. You do not fear a poisonous spider if it is located on the other side of the globe, but you do fear it when it is in your bed or crawling across the face of your infant. To be relevant, the danger must be potentially harmful for the subject or something or someone the subject cares about. We can capture this by saying that to fear p, the danger it presents must be personally relevant.
F1 harm : S judges p to be potentially harmful in a personally relevant way.
Since our interest lies in rational belief-less religion, and since we have concluded that it is irrational to fear something that is only an epistemic possibility in the wider sense, we need to rework F2 as well. Redesigning F2 to capture the narrower notion of epistemic possibility, we get: F2 rational : S believes p to be a live possibility, having a non-negligible chance of being true.
While I take it that F1 harm and F2 rational are both necessary for fear, and that they adequately capture its rational, action-grounding aspects, they do not represent a complete analysis. We also need to add an emotional component. Clearly, a robot incapable of having any feelings might still satisfy F1 harm and F2 rational , and that will not do. But how should we conceptualize the emotional component of fear? As a first step, I think it is important that the emotional component must depend on the judgement of danger. We do not want someone who feels dread independently of their assessment of p as dangerous to satisfy the condition.
F3: The judgement that p is dangerous incites an emotional response in S.
However, not any emotional response will do. Think of the fearless daredevil who laughs in the face of danger. Clearly, if your emotional response is excitement, you do not fear p. The emotional response needs to be negative, in the sense that it is uncomfortable, a feeling you would prefer to be without (at least if you consider it in isolation, without raising the question whether it could be helpful).
F3 negative : The judgement that p is dangerous incites a negative emotional response in S.
Unfortunately, there are other negative emotions that one might get as a response to perceiving a danger. S might start to hate p for endangering what she cares about, or she might become sad. What sets fear apart from other emotions? Well, if you fear that p, you will feel that you should act in some way to avoid the harm p might bring. The most basic response is probably some evasive manoeuvre, like running away, but you might also try to neutralize the threat that scares you, which might require you to be brave (on my analysis, brave is something you are even though afraid, it is not the same thing as being fearless). The important thing is that fear prompts you to take defensive action. Obviously, you do not actually have to take defensive action to fear, you just have to feel like you should.
F3 defence : The judgement that p is dangerous incites a negative emotional response in S, prompting S to take defensive action.
Combining F3 defence with F1 harm and F2 rational , we can say that for S to rationally fear p, the following three conditions must be met: 21 F1 harm : S judges p to be potentially harmful in a personally relevant way. F2 rational : S believes p to be a live possibility, having a non-negligible chance of being true. F3 defence : The judgement that p is dangerous incites a negative emotional response in S, prompting S to take defensive action.
This account of fear does not mirror the standard account of hope. Although it has sometimes been suggested that an action-condition should be added to the analysis of hope (Muyskens 1979;Pojman 1986), the standard account lacks an equivalent to my third condition. It should also be noted that on the standard account, hope only requires epistemic possibility in the widest sense, as compared to the narrow sense employed in my analysis of fear. Furthermore, the first condition in the analysis of hope is about desirability and positive evaluation rather than a judgement about p being beneficial. If my analysis of fear is correct, these differences suggest that either the mirror-approach advocated by Day or the standard account of hope itself must be mistaken. I suspect that the first alternative is correct, and that we should be hesitant to accept the mirror-approach. Especially since it seems unrealistic to suggest that one could only hope for beneficial things, while it is trivial that one can only fear what is harmful. However, it is beyond the scope of this article to take a definite stand on the matter.
Finally, I want to address a possible objection. It might be questioned if we really need to include fear in our account of non-doxasticism. Could not the claim that S commits to p out of fear of some supernatural bads be restated as the claim that S commits to p in hope to avoid these bads? And if so, is not fear superfluous to the account? Should we not prefer a simpler account, with as few non-doxastic attitudes as possible? I have two things to say to this objection.
First, a person who deems the existence of both supernatural goods and bads to be live epistemic possibilities, and who engages non-doxastically in some religious view to achieve the goods and avoid the bads, is typically basing her commitment on some combination of fear and hope (and maybe faith as well). Following the objector's suggestion would make it hard to properly appreciate such a complex commitment.
Secondly, the possibility of re-describing cases of fear in terms of hope depends upon the feasibility of the mirror-approach. If fear and hope are not parallel counterparts, it is not possible to claim that cases of the former can always be restated as cases of the latter. As we have seen, the mirror-approach is not uncontroversial and the analysis of fear here under consideration gives us reason to question it.

Summary
There are several negative factors the agnostic must take into account when considering a non-doxastic religious commitment. To facilitate discussion, I have made a crude distinction between natural and supernatural bads, negative counterparts to the natural and supernatural goods commonly discussed when assessing the rationality of a belief-less engagement with religion.
The natural bads are drawbacks of a religious commitment the agnostic will want to avoid, while maximizing the goods. I have suggested that in extreme cases, the avoidance of natural bads could prevent a commitment aimed at receiving supernatural goods. Standardly, however, as long as the existence of supernatural goods or bads remains an epistemic possibility, natural goods and bads only influence the agnostic's choice as tiebreakers.
The supernatural bads work together with the supernatural goods, and strengthen the case for the view they are associated with. As long as the existence of some supernatural bads is conceived as a live possibility, avoiding these bads should be the agnostic's top priority. If no possibility with supernatural bads is considered live by the agnostic, her choice should be guided by the achievement of supernatural goods as long as a view containing such goods remains epistemically possible.
Non-doxastic religion does not need to be based on a pro-attitude. If one commits non-doxastically to avoid supernatural bads, one is basing one's commitment on a negative attitude. The non-doxastic approach needs to be able to accommodate for that, and I have suggested that this can be done by introducing non-doxastic fear. To give a taste of what kind of fear a negative non-doxastic engagement might be based on, I have provided an analysis which combines viewing a danger as a live epistemic possibility with an emotional response prompting defensive action.