1 Introduction

In the literature on the normativity of attitudes, it is commonplace to distinguish the right from the wrong kind of reasons. The right kind of reasons are said to bear a special relation to the correctness of the particular attitude; are said to be the only or at least the prime motivating reasons for those attitudes; and are commonly considered to play a crucial role in assessing the rationality of one’s attitude formation. According to the paradigmatic take on these right kind reasons, they are the kinds of considerations that are properly connected to the properties of the object at which those attitudes are directed, that is, they are object-given reasons of those attitudes.Footnote 1 However, several authors have expressed doubts about this latter claim: epistemic reasons, so the idea goes, are clearly the right kind of reasons of beliefs. Nevertheless, there are many features of epistemic normativity that cannot be accounted for in a theory which only makes use of object-given reasons of belief and disbelief.Footnote 2 In this paper, I rebut these challenges by closely examining the object-given reasons of the third doxastic stance – namely the suspension of judgement.Footnote 3 By introducing a new way of weighing reasons in doxastic deliberation, my account of suspension provides a neat picture of several features of epistemic normativity which are supposed to be in tension with such a framework of object-given reasons. Roughly put, there is a twofold weighing of reasons: on the one hand, reasons that favour belief are weighed against reasons that favour disbelief, while reasons that favour suspension are weighed against reasons that favour decidedness where belief and disbelief are both ways of being decided on an issue.

I start by outlining the distinction between the right and the wrong kinds of reasons of attitudes, as well as the idea of accounting for this divide in terms of object-given reasons (Sect. 2). Afterwards, I introduce five puzzles from the literature that plague the application of object-given reasons to an analysis of epistemic normativity (Sect. 3). Following these preliminary sections, I develop my rebuttal of these challenges. First, I provide an account of the third doxastic stance – the suspension of judgement – as well as an elaborate model of how the object-given reasons of suspension relate to the object-given reasons of belief and disbelief (Sect. 4). Second, I highlight that this picture of our doxastic options yields the necessary resources to account for the five puzzles of epistemic normativity in terms of object-given reasons (Sect. 5). However, before closing, I turn towards a remaining worry about the preceding account. While it is able to accommodate the potentially problematic features of epistemic normativity, an explanation is still required for why some of these features do not extend to practical normativity (Sect. 6). I argue that suspending intentions exemplifies a quite different normative profile than the suspension of judgement.

2 The right kind of reasons as object-given reasons

The distinction between the right kind of reasons (RKRs) and the wrong kind of reasons (WKRs) has its origin in the debate about buck-passing analyses of value properties.Footnote 4 According to this family of theories, one category of normative properties – namely values – can be reduced to a different dimension of normativity – namely normative reasons. Thus, roughly put, an object O has a valuable property P iff there are normative reasons for a subject to take up a P-specific form of attitude φ towards O. For instance, the thick value property of being admirable can be analysed as follows: being admirable is nothing but the fact that there is a normative reason for a subject to admire the object in question. However, simple buck-passing accounts are riddled with one decisive shortcoming. After all, prima facie there can be all sorts of normative reasons which speak in favour of admiring, and only a fraction of them has anything to do with admirability. For instance, a monetary incentive might speak in favour of admiring a given person, but this should not make them admirable. To circumvent this objection, buck-passers introduced the RKR–WKR distinction – where a WKR is any sort of consideration which gives rise to this WKRs problem. Thus, an object O has a valuable property P iff there are RKRs for a subject to take up a P-specific form of attitude φ towards O.

Scholars have proposed different ideas concerning how exactly one ought to flesh out this general, schematic distinction of RKRs and WKRs. A natural suggestion, with which I contend in this paper, is that RKRs should bear the proper connection to the object of one’s attitude and its value properties, or what has come to be known as object-given reasons (OGRs). After all, WKRs are considerations which are not indicative of the existence of such a valuable feature of the object. It must be noted that many diverging theories exist that specify the details of OGRs in slightly different ways.Footnote 5 For the purposes of this paper, one can neglect these differences. Since I conceive her account to be highly illustrative, I follow Hieronymi (2005) in specifying OGRs in terms of justifying answers to certain questions: rationally evaluable attitudes are sensitive to certain judgements, which can be understood as answers to characteristic questions. Admiration, for instance, is sensitive to the judgement ‘X is (very) admirable’, which is an answer to ‘Is X admirable?’ or ‘How admirable is X?’. An OGR is a consideration which can be alluded to in justifying an answer to this characteristic question. Paradigmatic WKRs like monetary incentives fail this test, as one cannot appeal to them in justifying one’s answer.

In his seminal work, Schroeder (2012b) proposes a slightly different take on the RKR–WKR distinction. Besides the role that RKRs are supposed to play in circumventing the WKRs problem, they also exemplify two further earmarks that make them interesting in a variety of ways, clearly exceeding the limits of buck-passing accounts. First, and less central to the purposes of this paper, the RKRs of an attitude φ can be used by a subject as a basis for their φ-ing. Only RKRs are considerations which can also be the content of the motivating reasons for the subject. Second, and more important for the discussion to come, only RKRs have a bearing on the φ-specific rationality of the subject’s attitude formation. While sensitivity to monetary incentives might be indicative of a form of practical rationality, the rational standards of one’s attitude formation seem to be unmoved by it. The clearest example of this earmark is the idea of epistemic rationality of doxastic attitudes. While monetary incentives might yield a reason in favour of a belief, which makes that belief practically rational, such reasons do not seem to have a bearing on its epistemic rationality. This point is, again, in stark contrast to evidence in favour of or against the believed proposition. If one is not sensitive to these kinds of considerations, one fails to be epistemically rational.

Now, the leading question that guides the following discussion is whether OGRs can be made to work as an analysans of RKRs even if one adopts Schroeder’s more sophisticated notion of RKR. I call the affirmative answer to this question the right kind as object-given thesis, or RKOG for short. Many proponents of OGRs as a solution to the WKRs problem do seem to endorse RKOG as well.Footnote 6 This is to be expected. After all, RKOG is an attractive thesis that ties one’s reply to the WKRs problem to a more general distinction between kinds of reasons for attitudes.

3 Five puzzles from epistemology

One recurrent line of attack on RKOG is its application to epistemic normativity. Epistemic normativity, so the idea goes, can be understood in terms of epistemic reasons, which, in turn, are nothing but RKRs of doxastic attitudes.Footnote 7 Given RKOG, one would expect that all features of epistemic normativity are explained by the OGRs of these attitudes. What are the OGRs of doxastic attitudes? In the case of belief and disbelief, most authors agree that the characteristic question of these attitudes is whether the (dis-)believed proposition is the case. Just like admiration, believing p is a judgement-sensitive attitude which answers the question ‘Is p the case?’.Footnote 8 Correspondingly, OGRs of belief and disbelief must be considerations which can justify such answers. The paradigmatic instance of such a consideration is evidence that supports p or non-p. After all, answering whether p is the case seems to depend on considerations which are indicative of p’s truth or falsity. For the purposes of this paper, I assume a weak form of evidentialism according to which only evidence is an OGR of belief and disbelief.Footnote 9 However, opponents of RKOG have pointed to five supposed features of epistemic normativity that exceed these OGRs and thus generate counterexamples – in that there must be some RKRs at work which are not reducible to OGRs.

First, and probably the least divisive feature on this list, epistemic normativity is said to balance in a prohibitive manner: if the evidence is tied – or even close to being tied – on some issue, neither belief nor disbelief should be rational options – both are prohibited.Footnote 10 As we will see in Sect. 6, this property of epistemic normativity is in stark contrast to the way in which practical normativity appears to behave. After all, in the case of ties between practical considerations, both options appear to be permitted. Consider, for instance, Buridan’s ass: a donkey stands between two bales of hay and has no reason to favour one bale over the other. Intuitively, if the donkey were to eat from either of the bales, this would count as rational behaviour. Thus, any account of epistemic normativity in terms of RKRs must account for this special feature of prohibitive balancing. However, given RKOG, all we seem to be left with is a tie between two options which, on the face of it, looks just like the tie in the example of Buridan’s ass. Therefore, we miss some sort of RKR against believing and disbelieving in the case of tied options, which would explain why they are prohibited.

Second, Schroeder (2021, p. 130 f.) points to bare statistical evidence and lottery propositions. One might argue that any theory of epistemic normativity has to accommodate the following two features of statistical evidence: statistical evidence is evidence and thus counts as an RKR in favour of belief and disbelief,Footnote 11 yet bare statistical evidence does not suffice to rationalise beliefs. At least in certain cases, beliefs that are only based on statistical evidence are not rational. Consider the following example which is prominent in legal epistemology.

Blue Bus

A car has been accidentally hit by a bus. 90% of the buses driving through the area are operated by the Blue Bus Company; the remaining 10% belong to the Red Bus Company. An inquiring observer learns these facts and comes to believe that the bus involved in the accident was run by the Blue Bus Company.

The inquiring observer in Blue Bus, according to the reaction of many scholars, is not rational in forming their belief.Footnote 12 However, given that reasons in favour of an option φ count as sufficient iff they are at least as weighty as the reasons against φ-ing, the proponent of RKOG is faced with the following challenge: given that the statistical evidence in Blue Bus is an RKR in favour of belief, there must be some RKR against believing which explains the absence of sufficient reasons. By construction, the only piece of evidence in favour of disbelieving is, again, the statistical evidence. However, it would be absurd to assume that the weight of the RKR based on the 90% chance that it was a Blue Bus Company vehicle is on a par with the weight of the RKR based on the 10% chance that it was a Red Bus Company one. Therefore, the RKRs in favour of disbelief are, in principle, insufficient for explaining the lack of sufficiency. Thus, again, proponents of RKOG seem to be short of an RKR against believing.

Third, Schroeder (2012b) argues that learning about forthcoming evidence or easily available evidence can function as an RKR against believing or disbelieving. Consider the following case. Detective D is investigating a murder. D has plenty of evidence that gardener G committed the crime. However, there are still some DNA test results that D is yet to learn about, which will be made available to D the next day. Now, according to the supposed intuition, learning that such critical evidence exists and is forthcoming appears to make believing that G is the murderer irrational.Footnote 13 Similarly, it appears to be irrational to form a belief about the weather based on yesterday’s forecast; instead, one ought to get out of bed and look out the window. Thus, the fact that one knows that some evidence would be really easy to access appears, again, to constitute an RKR against believing. Yet, neither the information about the forthcoming results, nor the fact about availability is itself evidence concerning the proposition in question. After all, the fact that DNA-results will be imminently available does not make G any more or less likely to be the murderer.Footnote 14 Therefore, whatever accounts for the irrationality of believing in these cases cannot be an OGR of belief or disbelief.

Fourth, while being a heavily debated subject, one might endorse pragmatic and/or moral encroachment as a feature of epistemic normativity.Footnote 15 Paradigmatically, defenders of encroachment theses allude to pairs of cases auch as the following (DeRose, 1992; Stanley, 2005):

Low Stakes

Hannah is driving home on a Friday afternoon and intends to deposit a paycheque at the bank just because it would be convenient. However, upon noticing long lines at the bank, she considers driving straight home and returning on Saturday morning instead. She remembers the bank being open on a Saturday in the past, and although she thinks to herself that it is possible that the bank changed its hours, she believes that it will be open and drives home.

High Stakes

Hannah is driving home on a Friday afternoon and intends to deposit an important paycheque. If she does not have enough money to pay her mortgage in her bank account by Sunday, her very livelihood is on the line. However, upon noticing long lines at the bank, she considers driving straight home and returning the next morning instead. She remembers the bank being open on a Saturday in the past, and although she thinks to herself that it’s possible that the bank changed its hours, she believes that it will be open and drives home.

According to proponents of encroachment, the difference in the stakes involved in these cases creates a diverging assessment of Hannah’s rationality. While her belief might be rational in the former case, it fails to be rational in the latter one. Provided that the evidence in both cases remains the same, a proponent of RKOG is, again, faced with a puzzle: what could possibly account for this shift in Hannah’s rationality? After all, there appears to be a need for some RKR against believing which is not evidence – that is, not an OGR of belief or disbelief.

Finally, following a more recent attack on fittingness-first theories of normativity, Lee (2022) points to the phenomenon of higher-order defeat to drive a wedge between RKOG and epistemic normativity.Footnote 16 Higher-order defeat differs from other forms of defeat by leaving the evidential support relation between the subject’s evidence and the believed proposition unaltered.Footnote 17 Consider the following case: after dwelling over some logical derivations, a subject S comes to believe that p entails q. However, S also learns that their drink was laced with a mind-altering drug that inhibits their capacity to perform logical derivations. According to some authors, if S were to continue to believe that p entails q, then S would be irrational. Therefore, one might again expect the higher-order defeater to create some RKR against believing.Footnote 18 Since learning about the drug in the water is neither evidence in favour of the fact that p entails q, nor evidence against it, the envisaged RKR cannot be an OGR of belief or disbelief.

To conclude, there are five issues that arise when a thesis like RKOG is extended to account for epistemic normativity: (1) prohibitive balancing; (2) bare statistical evidence; (3) future and easily available evidence; (4) pragmatic and moral encroachment; and (5) higher-order defeat. While issues (2) to (5) rest on contentious assumptions, I take them for granted as correct depictions of epistemic normativity. In the next section, I introduce a feature of the doxastic option space which the opponents of RKOG have neglected – namely the OGRs of the suspension of judgment, the third doxastic option – before returning to these five puzzles in Sect. 5.Footnote 19

4 On suspension and decidedness

In this section, I present the core tenets of my defence of RKOG against the aforementioned set of challenges. First, I argue that the third doxastic state of suspension is characterised by a different deliberative question than belief and disbelief (4.1). Therefore, it can be sensitive to different considerations, which provides a proponent of RKOG with the resources to rebut the five puzzles. This latter task is the subject of the subsequent section. However, before turning to this endeavour, my proposal must explain how RKRs of suspension, which are supposed to to answer a different question than belief, can nonetheless count as RKRs against believing (4.2). My solution to this issue relies on a novel conception of the doxastic option space. I argue that belief and disbelief are ways of being decided on an issue, and therefore, are sensitive to a secondary deliberative question (4.3). The resulting picture is a two-scale weighing model of epistemic reasons, on which each scale is concerned with one of the two deliberative questions (4.4).

4.1 The suspension of judgement and its deliberative question

As Schroeder (2012b) already notes as a common reply to his cases, a defender of RKOG might point to the OGRs of the third doxastic stance, namely the suspension of judgement. Since suspension is also a rationally evaluable attitude, it should also be judgement-sensitive like belief.Footnote 20 Given that suspension and (dis-)belief are in competition with each other – an assumption that I investigate further in the following subsections – an RKR in favour of suspension counts as an RKR against believing and disbelieving. Now, if suspension is connected to a judgement that answers a different deliberative question than belief and disbelief, a proponent of RKOG has some further resources to draw upon in handling the five puzzles: the missing RKRs against belief and disbelief stem from the OGRs of suspension.Footnote 21Footnote 22

Why should one believe that suspension is sensitive to a different question than belief and disbelief? Here, I provide a list of prima facie reasons to doubt that suspending on whether p is the case is connected to a judgement that answers ‘Is p the case?’. First, according to a widespread view, suspension is not directed at propositions but rather at questions. It is an interrogative attitude, rather than a propositional one (Friedman, 2013; Booth, 2014). One does not suspend that p; rather, one suspends on whether p. Hence, one might suspect that the characteristic question of suspension also differs.Footnote 23 Second, and relatedly, most analyses of the third doxastic stance entail that suspension should come with a different characteristic question. Consider, for instance, Friedman’s (2017, 2019) seminal theory according to which suspension is an attitude that combines open-mindedness on some question Q with a pro-attitude towards inquiring into whether Q. Such an attitude should be an answer to a question like ‘Is Q an interesting subject of inquiry?’. On the other hand, higher-order accounts of suspension regard the third doxastic stance to be linked to a higher-order cognitive state about one’s epistemic situation.Footnote 24 Again, on such an account of suspension, it should answer a question like ‘Am I in a good epistemic situation concerning p?’. Finally, every account in the literature which investigates the fittingness or correctness conditions of the third doxastic state seems to assume that these conditions differ from the ones of belief and disbelief, and that they resemble the correctness conditions of certain higher-order beliefs about one’s epistemic situation.Footnote 25 Thus, these views seem to come with a similar upshot as the higher-order accounts of suspension concerning its characteristic question.

However, the strongest reason to opt for an alternative characteristic question for suspension is a much more straightforward one based on a basic platitude about suspension. Whatever suspending on whether p might be, it amounts to a certain degree of neutrality as to whether p is the case. Now, being neutral as to whether p hardly counts as an answer to a yes-or-no question like ‘Is p the case?’.Footnote 26 Hence, one should not expect the judgement which corresponds to suspension to provide an answer to that question either. Therefore, I follow the lead of the previous pointers and assume that suspension answers a different question. While a full-fledged theory of suspension would go beyond the scope of this paper, I defend a, rough, characterisation of this deliberative question.

I maintain that the deliberative question governing suspension should be concerned with one’s epistemic situation. It should not be about the truth or falsity of the proposition in question; rather, it should concern some epistemic propriety of my standing towards answering the deliberative question of belief and disbelief. This basic idea strikes me as quite plausible and is backed by most of the considerations that I alluded to above, such as higher-order views of suspension or work on the fittingness of suspension. The only line of resistance to this broad idea that I was able to identify stems from Friedman’s inquisitive attitude view of suspension. After all, the degree to which a question is interesting and merits being inquired into differs from an assessment of my epistemic standing towards said question. While a complete rebuttal of this concern goes beyond the scope of this paper, I provide a quick sketch of how one should dismantle it.Footnote 27 First, as has been recurrently noted in response to Friedman’s theory, the third doxastic state does not always come with such an openness to further inquiry (e.g., Ferrari & Incurvati, 2022). Sometimes, we remain doxastically neutral towards a question without putting it on our research agenda. Thus, second, if the deliberative question governing suspension should hold for both pro- and anti-inquisitive instances of suspension, then it should refer to an issue common to both. Third, the question that is common to either variety of suspension appears to be nothing but the proposal under investigation – that is, some propriety of my epistemic situation. For the purposes of this paper, I assume that something in the ballpark of ‘Am I in a position to tell whether p?’ is the correct depiction of the deliberative question of suspension. As I argue in Sect. 5, such a proposal entails OGRs of suspension which bear some independent plausibility.

4.2 A looming problem

Before turning towards the investigation of these OGRs of suspension, however, the following challenge must be answered, which constitutes an apt reply to the strategy of defenders of RKOG as I have laid it out thus far. Given that suspension and (dis-)belief answer to different deliberative questions, why is it that they are even in the appropriate form of competition with each other? An OGR of belief and disbelief justifies answers to ‘Is p the case?’, while an OGR of suspension supports an answer to ‘Am I in the position to tell whether p?’. According to the envisaged manoeuvre, OGRs of suspension should constitute RKRs against believing. Yet why should reasons that justify a negative stance on the latter question have any impact on how one replies to the former question? Consider a different pair of attitudes connected to diverging questions, such as admiring Ghandi and believing that there are 10 pencils on one’s desk. An OGR favouring this admiration does not constitute an RKR against the belief in question – thus, why should things be any different in the case of suspension and belief? Hence, without providing a theory of how the reasons of suspension bear on the rationality of belief, an appeal to a third doxastic state, which happens to be sensitive to different considerations, should strike any opponent of RKOG as an ad hoc manoeuvre to account for the five puzzles.

Therefore, to proceed with the strategy, one has to amend RKOG with some theory of how answers to different deliberative questions can still have a normative bearing on each other. I presume that there are different ways in which one can accommodate this challenge. For instance, one might point to the apparent disanalogies between pairs of suspension and belief and pairs of admiration and belief like the one above. While having both of the latter attitudes does not create any normative conflicts, suspending judgement on whether p while believing p gives rise to clear normative tensions (Friedman, 2017, 2019). An individual who forms both of these attitudes appears to be structurally irrational.Footnote 28 However, in the remainder of this section, I present a different approach to the connection of (dis-)belief and suspension which is able to circumvent the deep waters of structural rationality.Footnote 29 Instead, my preferred account can be seen as a slight extension of RKOG itself. It provides a somewhat deeper analysis of why suspension and belief are related, which explains why OGRs in favour of the former are RKRs against the latter as well as why there is a normative tension in having both attitudes.

4.3 Ways of being decided

To introduce my solution to the abovementioned challenge, it makes sense to draw attention to the following considerations: suspension, as described thus far, is an attitude that is sensitive to a negative judgement, answering the question of ‘Am I in a position to tell whether p?’. However, is there also an attitude that is sensitive to an affirmative judgement on that question? That is, is there any antagonistic attitude opposed to suspension in the way that belief is opposed to disbelief? Given that suspension as the third, neutral doxastic state can be seen as a form of undecidedness concerning the deliberative question governing belief and disbelief, it appears only natural to look out for some attitude of decidedness with regard to that question, which comes with a commitment to the affirmative judgement of ‘I am in a position to tell whether p.’. Postulating the existence of such a fourth doxastic option of decidedness would neatly fit into the picture of doxastic deliberation that I have envisaged up to this point.

However, assuming such an additional, fourth doxastic attitude of decidedness strikes me as quite contentious. It is clearly not in consonance with the state of the literature in epistemology. There, the consensus appears to be that the doxastic option space only has three members ({believe; suspend; disbelieve}), rather than four. Hence, even if there might be such a sui generis attitude of decidedness, models of doxastic deliberation seem to be able to do without it.Footnote 30 This first, negative observation leads me to a second, positive one, which can be regarded as the focal point of my new account: instead of requiring any distinct attitude of decidedness, our attitudes of belief and disbelief function as ways of being decided. Decidedness on the question of whether p can be regarded as a disjunctive state, which is either manifested by believing p or by disbelieving p. Therefore, qua ways of being decided on the matter, belief and disbelief come with a commitment to the deliberative question of suspension in that they are sensitive to an affirmative judgement, answering the question ‘Am I in a position to tell whether p?’. Thus, belief and disbelief play a dual role in deliberation. They are sensitive to two kinds of judgement; consequently, they constitute answers to two deliberative questions.

Such a dual role of belief and disbelief would lend a defender of RKOG all the necessary resources to accommodate the puzzles from the previous subsection. First, since belief and disbelief, qua decidedness, commit one to an affirmative judgement on ‘Am I in a position to tell whether p?’, they must be in rational tension with suspension. Just as belief and disbelief are in competition with each other since they are related to opposed judgements on one and the same deliberative question, so too are suspension and (dis-)belief. Second, and relatedly, an OGR in favour of suspension would be an OGR against the disjunctive option of decidedness, and, given RKOG, it should count as an RKR against being decided on the matter. Furthermore, a reason against a disjunctive option φ-or-ψ should entail a reason against performing either φ or ψ. Thus, the RKR against decidedness explains why an OGR in favour of suspension would constitute an RKR against belief and disbelief. Therefore, both open issues generated by attributing a divergent deliberative question to suspension can be solved by viewing belief and disbelief as ways of being decided.

Before moving on to the details of how to conceive of the weighing of reasons on such a bifurcated picture of doxastic deliberation, some qualms remain that my proposed account faces: how can belief and disbelief be related to two deliberative questions? Relatedly, how can they be sensitive to two diverging sets of OGRs – one related to the question ‘Is p the case?’ and the other to ‘Am I in a position to tell whether p?’? Stated succinctly, I need to provide some more motivation for how something like the abovementioned dual role of belief and disbelief could be possible.Footnote 31

While my appeal to a dual role of attitudes might strike one as mysterious, I take it to be the natural consequence of translating a much less mysterious idea in the literature on normative reasons into an RKOG framework. Here, I refer to the idea that at least some normative reasons are contrastive: a consideration R might be a normative reason to φ rather than ψ, but not a reason to φ rather than χ.Footnote 32 When applied to epistemology, authors such as Snedegar (2017) and Tucker (forth.) have defended the idea that epistemic reasons must be conceived of in this fashion. A consideration does not simply favour believing or suspending. Instead, it supports believing rather than suspending, or believing rather than disbelieving. Now, if one conceives of these RKRs in terms of RKOG, one faces a prima facie difficulty in accommodating this contrastivist idea. If a consideration supports the judgement on ‘Is p the case?’ to which belief is sensitive, then why should that not count as a reason in favour of belief, regardless of whether one compares it to disbelief or suspension? However, this shortcoming can be lifted if one turns to my dual-role view of some attitudes: a contrastive reason to φ rather than ψ must be an OGR that pertains to the deliberative question A to which both φ and ψ are sensitive qua being related to incompatible judgements on A. However, this does not entail that the same consideration will also be an OGR that favours φ over χ if these two attitudes are related to incompatible judgements on a different deliberative question B. To state this less abstractly, in deliberating about whether to believe rather than disbelieve, one seeks to answer one question – namely ‘Is p the case?’ – and in deliberating about whether to believe rather than suspend one is answering a different one – namely ‘Am I in a position to tell whether p?’. To conclude, while it is a substantial commitment of my account, the mere possibility of having an attitude be sensitive to two diverging deliberative questions and sets of OGRs does not strike me as overly problematic.Footnote 33

4.4 The resulting picture

Based on the aforementioned account of doxastic deliberation, I conceive it as being concerned with two distinct questions and two prima facie independent sets of OGRs that bear on those questions. Thus, to get a grasp of the upshots of my outlook, one must use a two-scale weighing model. Each scale corresponds to one of the two deliberative questions and the corresponding set of OGRs. On scale A, concerning the question ‘Is p the case?’, the OGRs of belief and disbelief are weighed against each other. Following the assumptions from the previous section, the result on scale A is purely determined by the evidence that concerns the proposition in question. Hence, the new components which my theory of suspension and decidedness adds do not alter the result on this scale. However, next to this traditional picture of weighing epistemic reasons, there is also a scale B, which is concerned with the question ‘Am I in a position to tell whether p?’. Here, the OGRs of suspension and decidedness are the important factors. To put this picture into the idiom of contrastive reasons, scale A concerns the reasons to believe p rather than disbelieve p (and vice versa), while scale B concerns the reasons to {believe p or disbelieve p} rather than suspend on whether p (and vice versa).

Now, to derive the overall status of either of these options, I propose the following, natural suggestions: suspending on whether p is rational iff suspension does not lose on scale B – that is, the OGRs to suspend are at least as weighty as the OGRs to be decided on whether p. Believing p, on the other hand, is rational iff believing does not lose on scale A and decidedness does not lose on scale B. Analogously, disbelieving p is rational iff disbelieving does not lose on scale A and decidedness does not lose on scale B. Therefore, an OGR in favour of suspension can make believing p impermissible: it just has to outweigh the reasons in favour of decidedness on scale B.Footnote 34

5 Revisiting the five puzzles

In this section, I delve into the question of what kind of considerations justify answers to ‘Am I in a position to tell whether p?’. Specifically, I investigate what the OGRs of suspension and decidedness might be and, thus, which factors determine the outcome on scale B. I organise my thoughts on this matter in such a manner that accounting for these OGRs makes solutions to the five puzzles from Sect. 3 apparent. First, I present some general thoughts regarding how my account is able to handle the phenomenon of prohibitive balancing (5.1). Second, I argue that it is an independently plausible position that live error possibilities are OGRs of suspension (5.2). This observation grants a proponent of RKOG the required resources to handle the behaviour of statistical evidence, cases of future and easily available evidence, as well as pragmatic and moral encroachment. Finally, I propose that higher-order defeat might work quite similarly to undercutting defeat in that it acts as an enabler or modulator concerning the OGRs relevant on scale B (5.3).

5.1 On the balance of evidence

My novel approach to the analysis of prohibitive balancing rests on the following claim: evidence concerning p – the OGRs at work on scale A – are not themselves at work on scale B.Footnote 35 Instead, I wish to argue that scale B is solely occupied with certain facts about the evidence. That certain facts about the evidence are RKRs of suspension is an intuitively plausible idea, and one that has been recurrently noted in the literature. Roeber (2016) and Schmidt (2023), for instance, have both explained the prohibitive balancing of epistemic normativity by treating the fact that the evidence is tied as such an RKR. Furthermore, the account of suspension presented in the previous section shares this prediction. After all, one seems to be perfectly able to justify a negative stance on ‘Am I in a position to tell whether p?’ by pointing to the fact that one’s evidence concerning p is tied. I propose that one should think about the OGRs in favour of decidedness in an analogue fashion. Just like its antagonist, one can only justify a positive stance on ‘Am I in a position to tell whether p?’ by pointing to certain facts about the evidence, rather than to the evidence itself. The most straightforward fact about the evidence that would justify the disjunctive option of decidedness is plausibly that one’s set of evidence is decisive (i.e., the fact that there is a large swing on scale A).

However, one might question the restriction of RKRs in favour of decidedness to facts about the evidence. Is the first-order evidence itself not also an RKR of decidedness? If one receives evidence that p, can one not become decided rather than suspend for that reason?Footnote 36 While this observation might seem compelling on first sight, one should also consider the following phenomenon: countervailing evidence does not act as a rebutting defeater of the supposed RKR to become decided, but as an undercutting one. That is, if one would also receive the same amount of evidence that not-p, there would not be any reason to become decided anymore. This result looks odd on the view that the first-order evidence itself is a reason to become decided: even if we add countervailing evidence to the picture, the first-order evidence does not cease to exist. Yet, this result fits perfectly fine with my view that only facts about the evidence constitute RKRs of decidedness. After all, by adding countervailing evidence, the facts about the evidence (e.g., its decisiveness) change. Thus, even in simple cases in which one just receives some evidence that p, I take it that the RKR for which you become decided is a fact about that evidence and not the evidence itself.

If these observations are on the right track, one has a very straightforward approach for accommodating prohibitive balancing. If the evidence on scale A is close to being tied, the OGRs in favour of suspension outweigh the OGRs in favour of decidedness on scale B. Thus, in light of the theory presented in the previous section, both ways of being decided become impermissible. This picture of prohibitive balancing is, I argue, significantly simpler than its competitors and bears some intuitive plausibility. The central component which makes such a neat explanation of prohibitive balancing possible is the somewhat novel idea that evidence in favour of p – the paradigmatic RKR of beliefs – is not directly weighed against the RKRs of suspension. This shift in conceiving of the weighing of epistemic reasons makes my approach immune to two problems that plague its rivals and require them to opt for increasingly complicated pictures.

First, several authors have expressed doubt regarding the idea that one can allude to any such fact about the evidence as an additional epistemic reason.Footnote 37 The following is a prima facie detrimental objection to the appeal to any such metareason concerning the balance of one’s first-order evidence: the balance of evidence can only be a derived reason, and thus, it should not make any difference to a weighing of reasons which already makes use of those considerations from which the derived reasons are derived, namely the evidence supporting p and evidence supporting non-p. Adding facts about the evidence to one’s weighing of epistemic reasons amounts to an illegitimate form of double counting. However, the usage of facts about the evidence in my two-scale model is completely immune to this challenge. After all, at no point do I weigh the fundamental reasons – evidence concerning p – against facts about the evidence – that it is balanced or decisive. Instead, both sets of considerations are separated and each is weighed on its own scale. Thus, by not having evidence itself amount to a reason on scale B, I am able to circumvent this severe problem.

Second, accounts of prohibitive balancing recurrently struggle to accommodate two diverging prohibitive balancing scenarios. After all, the evidence concerning p can be tied, either, in cases in which one is in the possession of close to no evidence, or in cases in which much evidence is accumulated on either side. In short, the weight of the reasons in favour of suspension should be flexible enough to make plausible predictions concerning both scenarios.Footnote 38 However, if – as in my account – evidence that p is not directly weighed against suspension, one does not face any problem in accommodating the different ways in which the evidence might be tied. Regardless of the total amount of evidence at work on scale A, the weighing of reasons on scale B – which explains prohibitive balancing – remains the same. After all, it is only the swing of scale A that matters for scale B, not the total weight of evidence on either side.

5.2 Live error possibilities as reasons to suspend

In this subsection, I wish to draw the reader’s attention to a different sort of fact about the evidence – namely the fact that it is able or unable to exclude some salient error possibility. I use the term ‘live error possibilities’ (LEPs) to denote an error possibility which, on the one hand, is made salient by some contextual factors, and, on the other hand, is not ruled out by one’s evidence. Just like being balanced and being decisive, the existence of an LEP is a fact about one’s evidence. Furthermore, there is good reason to believe that this different sort of fact about the evidence is also at work on scale B. After all, in doxastic deliberation, pointing towards an LEP seems to have the envisaged effect: it rationalises the third doxastic stance of suspension over belief and disbelief. I regard this to be an independently plausible datum. Furthermore, it sits well and thus supports the idea that suspension is governed by a question along the lines of ‘Am I in a position to tell whether p?’. After all, an LEP can justify a negative judgement on that question.Footnote 39

Suspension’s general sensitivity to LEPs entails some interesting upshots. First, it explains the odd behaviour of bare statistical evidence. In Blue Bus, learning about the distribution of buses in the area makes one error possibility salient which is in principle not ruled out by the statistical evidence: perhaps the bus involved in the accident was one of the 10% of vehicles that are not part of the Blue Bus Company.Footnote 40 Therefore, there is an OGR in favour of suspension which can explain why the evidence in question is not sufficient for rationalising a belief. After all, if the LEP in question makes suspension win out on scale B, believing ceases to be a rational option. It is important to note that this behaviour of statistical evidence does not preclude it from sometimes rationalising beliefs. This is in consonance with much of the literature on statistical evidence, where scholars have attempted to explain why some statistics appear to be ill foundations for rational beliefs while others appear fine.Footnote 41 Provided that statistical evidence can be more or less capable of making possibilities of error salient, one can explain this fact in terms of the existence of LEPs – and thus, by pointing towards the OGRs in favour of suspension.

Second, I maintain that intuitions regarding cases of future and easily available evidence are also best explained in terms of LEPs: learning about the availability of good evidence on a question only constitutes an RKR against believing insofar as it makes certain possibilities of error salient which are not yet excluded by one’s evidence.Footnote 42 Consider the example of detective D from Sect. 3, who learns about forthcoming DNA results. This information should make the possibility of the DNA results defeating D’s case against the gardener G salient; and by extension, some error possibilities of how G did not commit the murder should also become salient, such as the possibility that the butler framed G. A similar story can be told about easily available evidence. The fact that I know that getting out of bed and looking out the window can easily generate a defeater for my evidence that it will rain today makes the error possibility that the weather forecast was wrong salient.Footnote 43

Third, suspension’s sensitivity to LEPs can also yield an explanation of pragmatic and moral encroachment. The idea that the stakes involved in paradigmatic cases of encroachment broaden the range of salient error possibilities is one standard approach to this supposed aspect of epistemic normativity.Footnote 44 In the pair of cases from Sect. 3, Hannah does not behave rationally in High Stakes because she has not yet ruled out some salient error possibility. The possibility that the bank changed its opening hours since she last checked is made salient due to the practical aspects of the context. In Low Stakes, Hannah is not required to exclude such a distant possibility. Therefore, in High Stakes, an LEP generates an additional OGR in favour of suspension which, in turn, explains why she counts as irrational in forming her belief.

5.3 Higher-order defeat revisited

The idea that an appeal to suspension and its OGRs can help us to understand the nature of higher-order defeat is already proposed by Lord and Sylvan (2021). According to them, treating higher-order defeaters as reasons to suspend judgement is a plausible position in the debate about higher-order evidence and is in line with their analysis of what it is to suspend judgement. However, I propose a slightly different connection between higher-order defeat and the OGRs of suspension, which is based on the following two observations.Footnote 45

First, while higher-order defeat leaves the evidential support relation intact, this does not mean that it does not work like an undercutting defeater in other regards. Consider the case of the laced drink from Sect. 3. My evidence on the logic puzzle is still very good and the way in which this reason bears on the question ‘Does p entail q?’ is not altered. Furthermore, my evidence is quite decisive. However, the degree to which my evidence on whether p entails q is decisive does not seem to speak in favour of an affirmative answer to ‘Am I in a position to tell whether p?’. Since I received a higher-order defeater in the form of information about my laced drink, the decisiveness of my evidence no longer appears to be a reason in favour of decidedness.Footnote 46 Thus, concerning the weighing on scale A, higher-order defeat does not have any effect. Nevertheless, on scale B it acts as a disabler or attenuator in the sense of Dancy (2004, Chap. 3): the information about my drink is not itself a reason against being decided, but it undermines or modulates the support relations in which other reasons stand to this option.Footnote 47 Therefore, just as in the case of prohibitive balancing, an individual who is subject to higher-order defeat might lack any positive reason in favour of being decided on an issue.

What remains to be demonstrated to accommodate the objection, is that there are also some OGRs in favour of suspension in these cases. This brings me to my second observation, which is that pointing to LEPs seems to be an appropriate way to justify suspension in cases of higher-order defeat. The possibility that there is an instance of p being true without q being true is clearly ruled out by my evidence on whether p entails q. Yet, given the higher-order defeater at work in these cases, this error possibility seems to be live nonetheless. The information about the laced drink strikes me as re-enabling those once-excluded possibilities of error. Thus, next to being a disabler or attenuator of the OGRs in favour of decidedness, there is a good case to be made that higher-order defeat also functions as an enabler or intensifier of some OGRs in favour of suspension. Therefore, in cases of higher-order defeat, there should be a special preponderance of OGRs in favour of suspension and, given RKOG, an additional RKR against believing.

6 Permissive balancing and suspending intentions

The previously specified account is able to defend RKOG against the objections that stem from epistemic normativity by extending the set of RKRs against believing. Yet, the following pressing question remains: how does RKOG fare with regard to other attitudes? In particular, one might worry that RKOG overgenerates RKRs against intentions. After spelling out this objection in detail (6.1), I point towards a reasonable and independently motivated rejoinder on the side of RKOG (6.2). I argue that the desired difference between doxastic and practical deliberation can be explained by a difference in the kind of attitudes at work.

6.1 A remaining worry

As I mentioned in Sect. 3, prohibitive balancing is supposed to be a feature that is peculiar to epistemic normativity. By contrast, practical normativity is said to balance in a permissive manner in that two options that are equally supported by one’s reasons both emerge as permissible. Now, one might consider the previous account of suspending judgement to raise an issue when applied to conative states or intentions. According to some authors, just as there is a third, neutral doxastic state, there is also a third, neutral conative state, namely the suspension of intentions.Footnote 48 Just as one can remain undecided about how the world is, one can also remain undecided with regard to what to do. Thus, prima facie, the suspension of intentions should receive the very same treatment as the suspension of beliefs. However, if this point holds up to scrutiny, there is a case to be made that practical normativity exemplifies the very same kind of prohibitive balancing as epistemic normativity.

The rationale for this conclusion is as follows: intentions, just like beliefs, are judgement-sensitive attitudes.Footnote 49 Instead of being sensitive to considerations that bear on the question ‘Is p the case?’, OGRs of intentions justify answers to the question ‘Is it right to φ?’.Footnote 50 For the sake of simplicity, I only consider two-option cases; therefore, intending to φ and intending not to φ are the only relevant attitudes at play. Given the assumption above that the doxastic and conative attitudes work in parallel, intending to φ and intending not to φ are also ways of being decided with regard to a particular choice. Qua manifesting decidedness, intentions also answer a secondary deliberative question – namely ‘Am I in a position to tell whether it is right to φ?’. The other relevant attitude, which is sensitive to a negative judgement on this question, is the suspension of intentions. Thus, practical deliberation can also be characterised by a two-scale weighing model: on scale A, intending to φ competes with intending not to φ, and the relevant OGRs are considerations about the rightness of φ-ing. On scale B, however, the relevant OGRs should again be facts about one’s evidence, such as the degree to which one’s evidence that φ-ing is right is decisive, or LEPs concerning the rightness of φ-ing.

Now, one might argue that this model of practical deliberation is able to provide the correct predictions in the case of Buridan’s ass. After all, more than one option can be a right thing to do. Thus, the donkey is in a position to determine whether eating one bale of hay rather than the other is a right thing to do. Therefore, it has good reasons to be decided on the matter and, thus, suspension does not interfere with permissive balancing.Footnote 51 However, consider the following scenario:

Miners

After an earthquake, a group of miners is trapped in a mine, which water is now entering. The miners are either in shaft A or shaft B. If the team of rescuers on the surface does not flood one of the shafts in due time, the miners are not going to survive. However, the rescuers do not know which shaft the miners are in.

In this case, only one of the two options appears to be the objectively right thing to do. However, the rescuers are not in a position to tell whether flooding shaft A, for example, is the objectively right option. Their evidence is balanced, and just as in the doxastic case, suspension should win out on scale B. Therefore, the rescuers ought to suspend their intentions, and both intending to flood shaft A and intending to flood shaft B are irrational. To conclude, if the preceding line of thought is on the right track, my defence of RKOG with regard to epistemic normativity appears to commit me to a problematic position on balancing cases of practical normativity.

6.2 Solution: suspending intentions differs from suspending beliefs

A defender of my take on doxastic deliberation could employ various strategies to avoid such a problematic outcome. After all, several assumptions in the above rationale can be subjected to doubt. For instance, one might question whether intentions aim at objectively rather than subjectively right actions. In that case, the miners case would merit the same treatment as Buridan’s ass.Footnote 52 Alternatively, one could deny that there exists something like the suspension of intentions, or that suspending intentions does not amount to a judgement-sensitive attitude at all.Footnote 53 However, my preferred response to the challenge is to keep intentions and beliefs aligned and allow for the existence of a judgement-sensitive attitude of suspending intentions, but to deny that suspending intentions and being decided on what to do answer to the same kind of question as suspending judgement and being doxastically decided. As I demonstrate below, this position is not ad hoc and bears some independent motivation.

I provide a rough draft of how I view the suspension of intentions: in contrast to suspending judgement, suspending intentions is not connected to a judgement answering the question ‘Am I in a position to tell whether φ-ing is a right thing to do?’. Instead, suspending intentions is linked to the following much more practical question: should I commit to a plan or form an intention regarding some choice of course of action? The same holds true for conative decidedness. Thus, the OGRs on our practical scale B are practical considerations about whether one should reach a decision on some issue. Such a shift in the normative profile of suspension and decidedness yields a straightforward account of Miners. The balance of evidence concerning which action is the right one is not necessarily a reason in favour of suspending one’s intentions. Quite the opposite, the practical need to commit to some course of action given the time constraints of the situation might be weighty OGRs in favour of decidedness. Thus, practical normativity balances in a permissive fashion.

One objection that might be raised against such a view of suspending intentions is already raised by Schroeder (2012b, p. 476): given that we make suspension sensitive to practical considerations, how can we avoid the emergence of a WKRs problem. Following Kavka’s (1983) toxin puzzle, the standard position on the WKRs of intentions is that they are practical considerations that favour the intentions to act rather than the act itself. After all, a monetary incentive to intend to drink a toxin does not make the act of drinking the toxin any more or less right. However, given my depiction of the question governing practical decidedness and suspension, such monetary incentives appear to make a difference: a monetary incentive to be decided on whether one should drink a toxin falls under the OGRs of decidedness and, given RKOG, should be an RKR to be decided.

My response to this concern consists of two parts. First, it is important to note that an OGR on scale B is not an OGR on scale A. Thus, allowing for incentives to be decided by manifesting either the intention to φ or the intention not to φ as RKRs in favour of becoming decided does not entail that a monetary incentive to intend φ is a RKR in favour of intending φ. Therefore, my account does not open the floodgates to WKRs of intentions. The only Kavkaesque scenarios that it allows for look quite different from the original toxin puzzle. Here, a billionaire might offer you a monetary incentive to decide whether you plan to drink the toxin tomorrow. It is perfectly possible to take the billionaire’s money and make the rational decision not to drink the toxin.

Second, and in stark contrast to Schroeder’s assessment, I take this result to be a feature of my theory rather than a bug: counting such monetary incentives as RKRs to come to a decision is the correct prediction. After all, paradigmatic RKRs to become decided concerning a choice or to suspend one’s intentions look quite similar to these monetary incentives. Above, I alluded to the time constraints of a given situation: given that postponing one’s decision comes with some practical downsides, one ought to decide sooner rather than later. These practical downsides might pertain to the action under investigation, such as the pressure to act in Miners, but they might also be related to the success of some other action of yours, or even actions of other people in cases of coordination. In fact, I do not consider any stark difference to exist between these kinds of practical considerations and the monetary incentive of the billionaire envisaged above. Therefore, instead of constituting an objection to my account, reflection on the RKRs of practical decidedness brings some independent motivation of my position to light.Footnote 54

7 Conclusion

I consider the contribution of this paper to be threefold. First, it introduces a new, innovative approach to weighing epistemic reasons and yields a natural picture of how suspension and (dis-)belief rival each other as doxastic options. The twofold weighing of reasons that concern different questions strikes me as a more adequate model of our deliberation than any of the simple weighing pictures in the literature, which run into all sorts of issues by cramping suspension, belief, and disbelief onto the same scale. Second, by making use of this picture of suspension, one is able to accommodate all of the supposed features of epistemic normativity, which are said to cause trouble for RKOG. The OGRs of suspension and decidedness can explain prohibitive balancing, the behaviour of statistical evidence, pragmatic and moral encroachment, the significance of forthcoming evidence, and higher-order defeat. Finally, as the previous section has demonstrated, the two-scale weighing model can plausibly be extended to the realm of practical deliberation without generating a form of prohibitive balancing.

In closing, I wish to flag some routes for further research, which the foregoing discussion has left open. First, while this work has highlighted the potential behind the additional resources of the two-scale model, many of the details, especially concerning the reasons on scale B had to remain vague. Thus, it remains an avenue for future research to investigate whether there are any additional RKRs to be found on this scale, as well as how those RKRs interact.Footnote 55 Second, the final result of this paper is also a clear gateway to a further analysis of practical deliberation. There might be many puzzling features of practical normativity which can also be resolved by pointing to the OGRs of suspension and decidedness. Therefore, there might be further theoretical payoffs down the line. Third, the inquiry of suspension of other forms of attitudes is still very much an open field. Hence, it would be interesting to investigate whether one can extend the present picture of suspension to these attitudes as well.Footnote 56 Furthermore, given that I argue that suspending intentions differs from doxastic suspension in a crucial way, it remains to be seen whether the suspension of any other attitudes is more in line with the former or the latter.